Part 4 History - Cornell Universityusda.mannlib.cornell.edu/usda/AgCensusImages/1992/02/History.pdf · AC92-S-4 Volume 2 SUBJECT SERIES Part 4 History Census of Agriculture U.S. Department

AC92-S-4

Volume 2SUBJECT SERIES

Part 4

History

Census of Agriculture

U.S. Department of CommerceEconomics and Statistics AdministrationBUREAU OF THE CENSUS

This history was written in the Policy Office by Michael A. Hovland , AgricultureCensus Historian, under the general supervision of William F. Micarelli, Acting Chief Historian.

The following individuals provided valuable information and/or critically reviewed the manuscript:Linda G. Allen, Roger Beinhart, Demosthenes Birbilis, John A. Blackledge,Carol V. Caldwell, Quentin C. Coleman, Irving W. Copeland, Tommy W. Gaulden,Douglas A. Hartwig, Linda J. Hutton, Kent C. Hoover, Charles L. Kendall, Ronald D. Lewis,Karen G. Meyers, Joseph Miller, Debbie A. Norton, David I. Peterson, Elaine Richards,Joseph T. Reilly, Jane Dea Sandusky, Robert T. Smith, and Ewen M. Wilson, Agriculture andFinancial Statistics Division (formerly Agriculture Division); Beverly E. Battle, Christopher L.Berbert, Douglas J. Miller and George E. Pierce, Economic Planning and Coordination Division(formerly Economic Planning Staff); and Brian Greenberg, Statistical Research Division.

The staff of the Administrative and Customer Service Division,Walter C. Odom, Chief, providedpublication planning, editorial review, design, composition, and printing planning and procurement.Bernadette J. Gayle provided publication coordination and editing.

Acknowledgments

AC92-S-4

Volume 2SUBJECT SERIES

Part 4

History

Census of Agriculture

U.S. Department of CommerceMichael Kantor, Secretary

Economics and Statistics AdministrationEverett M. Ehrlich, Under Secretary

for Economic Affairs

BUREAU OF THE CENSUSMartha Farnsworth Riche, Director

Issued September 1996

ECONOMICS

AND STATISTICS

ADMINISTRATION

Economics and StatisticsAdministrationEverett M. Ehrlich, Under Secretaryfor Economics Affairs

BUREAU OF THE CENSUS

Martha Farnsworth Riche, DirectorO. Bryant Benton, Deputy Director

Frederick T. Alt, Principal AssociateDirector and Chief Financial OfficerPaula J. Schneider, Principal AssociateDirector for ProgramsFrederick T. Knickerbocker, Associate Directorfor Economic ProgramsThomas L. Mesenbourg, Assistant Directorfor Economic ProgramsMichael S. McKay, Acting Associate Directorfor Planning and Organization Development

For sale by Superintendent of Documents, U.S. Government Printing Office, Washington, DC 20402.

POLICY OFFICEM. Catherine Miller, Chief

AGRICULTURE AND FINANCIALSTATISTICS DIVISIONEwen M. Wilson, Chief

CHAPTERS

1. Introduction-------------------------------------------------------------------------------- 12. Planning and Preliminary Operations---------------------------------------------------- 93. Preparatory Operations------------------------------------------------------------------- 234. Public Awareness Program -------------------------------------------------------------- 375. Data Collection---------------------------------------------------------------------------- 456. Data Processing -------------------------------------------------------------------------- 557. The 1992 Puerto Rico Census of Agriculture ------------------------------------------ 778. 1992 Census of Agriculture for Guam and the U.S. Virgin Islands -------------------- 919. 1994 Farm and Ranch Irrigation Survey ------------------------------------------------ 10110. Coverage Evaluation and Research----------------------------------------------------- 10911. Publication Program ---------------------------------------------------------------------- 119

APPENDIXES

A. Provisions of Title 13, United States Code Relating to the 1992 Census ofAgriculture ------------------------------------------------------------------------------ A-1

B. Historical Notes --------------------------------------------------------------------------- B-1C. Report Forms, Letters, and Envelopes Printed for the 1992 Census of Agriculture -- C-1D. 1992 Census of Agriculture Mailout and Followup Mailings---------------------------- D-1E. Publications in Selected Series ---------------------------------------------------------- E-1F. Chronology of Major Activities ----------------------------------------------------------- F-1G. Report Forms ----------------------------------------------------------------------------- G-1

Publication Program--------------------------------------------------------------- Inside back cover

Contents

Page

CONTENTS IIIHISTORY—1992 CENSUS OF AGRICULTURE

Chapter 1.

Background Information-------------------------------------------------------------------------- 2Legal Authority------------------------------------------------------------------------------------ 2Uses of Agriculture Census Data---------------------------------------------------------------- 3Farm Definition ----------------------------------------------------------------------------------- 3

1992 Census of Agriculture---------------------------------------------------------------------- 4Scope and Reference Dates ------------------------------------------------------------------ 4Scope ---------------------------------------------------------------------------------------- 4Reference periods and dates--------------------------------------------------------------- 4

An Overview of the Census Operation ------------------------------------------------------- 4Data collection------------------------------------------------------------------------------- 4Data processing and publication ----------------------------------------------------------- 4

Geographic Area Detail------------------------------------------------------------------------ 5The 1994 Farm and Ranch Irrigation Survey ------------------------------------------------ 5Expenditures ----------------------------------------------------------------------------------- 5

Organization of the Census Bureau------------------------------------------------------------- 6

Contents

Page

CHAPTER 1 1HISTORY—1992 CENSUS OF AGRICULTURE

Introduction

BACKGROUND INFORMATION

Article 1, Section 2, of the United States Constitutionrequires that a census of population be carried out every 10years to apportion representation of each State in theHouse of Representatives. Even as the delegates to theconvention that produced the Constitution discussed itsvarious provisions, James Madison, its principal author,urged that the census be used for something more thanjust counting heads. Nothing came of his recommenda-tions until 1810, after he had become President Madison.In that year, the census tried to collect information onmanufacturing establishments as well as population, andincluded a single item asking whether the person enumer-ated was engaged in agriculture (approximately 80 percentwere). Another 30 years passed before the census pro-gram included an enumeration of agricultural activities. The1840 census attempted to collect more detailed informationon manufacturing, mining, and agriculture, with limitedsuccess. Despite this, the value of agriculture data (and theother detailed statistics) was so obvious that the censusprogram was permanently expanded to cover economicand agricultural activities.

The agriculture census remained part of the decennialcensus program from 1850 through 1920, while the othereconomic areas were covered first as part of the decennialprogram, and later every 5 years. In 1915, the Congressauthorized the collection of agriculture data quinquennially,but it was not added to the mid-decade enumerationcovering the economic areas until 1925.

Through 1940 the Census Bureau carried out the agri-culture census with the other economic censuses, butchanges in their respective schedules, intended to use theagency’s resources more efficiently and to distribute theworkload over the periods between the decennial cen-suses, caused the reference years to diverge. By the1950’s, the agriculture census was collecting and publish-ing information for years ending in ‘‘4’’ and ‘‘9’’, while theeconomic censuses had years ending in ‘‘2’’ and ‘‘7’’ asreference periods. In 1976, Public Law (P.L.) 94-229 short-ened the intercensal periods after the 1974 and the follow-ing agriculture census to 4 years each, restoring theagriculture census to a concurrent schedule with the 1982and later economic censuses.

The agriculture census is the only source of statistics onAmerican agriculture showing comparable figures, countyby county, and classifying farms by size, tenure, type of

organization, principal occupation and age of operator,market value of agricultural products sold, combined gov-ernment payments and market value of agricultural prod-ucts sold, and Standard Industrial Classification (SIC)code. The 1992 Census of Agriculture covered all agricul-tural operations in the 50 States, Puerto Rico, Guam, andthe U.S. Virgin Islands.

LEGAL AUTHORITY

The Bureau of the Census, an agency of the U.S.Department of Commerce, carried out the 1992 agriculturecensus, under authority granted by Title 13, United StatesCode—Census. Title 13 governs the agency’s operations,establishes what censuses shall be taken and the intervalsbetween them, specifies certain administrative procedures,and describes the duties of particular officials. (See appen-dix A for excerpts of Title 13 applicable to the agriculturecensus.)

Chapter 1 (Administration) of the title covers collectingand handling census and survey data and the qualificationsand duties of census supervisors and other employees.Section 5 of the chapter assigns to the Secretary ofCommerce responsibility for preparing census report formsand determining the number and type of inquiries included.Section 9 sets confidentiality requirements for census dataand forbids the ‘‘use of information furnished... for anypurpose other than the statistical purposes for which it wassupplied....’’ The section also restricts access to censusreport forms to sworn Department of Commerce or Censusemployees and forbids publication of any information fromthe census or any survey that could be used to identify aspecific person or establishment.

Chapter 5 (Censuses), section 142, describes the type,frequency, and geographic scope of the agriculture census,and section 195 authorizes sampling to be used to carryout the provisions of Title 13 (except—applied to thecensuses of population—for determining population forapportionment of congressional representation).

Chapter 7 (Offenses and Penalties), section 214, setsout the penalties (a fine of not more than $5,000, impris-onment for not more than 5 years, or both) for anyemployee or Census Bureau or Department staff memberpublishing or otherwise communicating to any person otherthan a sworn Census Bureau or Department of Commerceemployee any data covered by the confidentiality provi-sions of the law. Sections 221 and 224 require response to

2 CHAPTER 1 HISTORY—1992 CENSUS OF AGRICULTURE

the censuses and establish penalties for noncompliance (afine of $100 for individuals and $500 for organizations).

USES OF AGRICULTURE CENSUS DATA

Agriculture census data are routinely used by the Admin-istration; Congress; Federal, State, and local governmentorganizations; the business community; scientific and edu-cational institutions; and farmer organizations.

x The private sector —businesses, farm cooperatives,commodity and trade associations, utility companies—relieson agriculture census data in developing plans forlocating new plants, service outlets, and sales anddistribution facilities, as well as for allocating researchresources, forecastingmarkets, selectingmarketing areas,and for other activities that provide better services to thefarm community. A major farm organization uses censusdata to develop promotional materials on various seg-ments of American agriculture.

x Farm and agriculture-oriented magazines and newsmedia use census results as technical background forstories and feature articles to determine their share ofthe market and to identify the types of farms they reach.A regional television station, for example, used agricul-ture census data to determine farm operator character-istics and agricultural production levels in each of its 211television markets, enabling its advertisers to targetspecific media markets for particular services and prod-ucts.

x Administrative and legislative bodies at all levels ofgovernment use the census data in planning farm andrural programs and analyzing the results such programs.The Congressional Budget Office uses agriculture cen-sus data to evaluate the farm income-support program;State and county agencies employ census statistics forland planning and zoning, to aid in evaluating environ-mental policy, profiling the States’ labor force, economicplanning, and so on.

x The U.S. Department of Agriculture’s National Agri-cultural Statistics Service (NASS) employs agriculturecensus statistics to develop benchmarks and compari-sons for many of its current estimates, and in evaluatingparticular problems or situations. The USDA’s EconomicResearch Service (ERS) uses census of agriculture datain developing the farm accounts, evaluating the currenteconomic situation, and in monitoring and measuringstructural changes and adjustments in the farm sector.The Animal and Plant Health Service prepares diseaseand pest damage assessments, when needed, usingcensus information.

x Federal Government agencies use production, sales,and size and type of farm data from the census incalculating such economic measures as the gross domes-tic product (GDP), farm income estimates, and indexes

of productivity and price levels. Census data also areused to calculate Federal disaster compensation andenvironmental assessments, and for special projects.

FARM DEFINITION

The reporting unit for the agriculture census has alwaysbeen the individual agricultural operation—the farm (orranch—for census purposes, ‘‘farm’’ and ‘‘ranch’’ are inter-changeable terms). The word ‘‘farm’’ might be applied toany place on which anything called a ‘‘crop’’ is grown or anyanimals are cared for by humans. However, any attempt tocollect data on a specified subject, such as agriculturalproduction, requires that both the subject of inquiry and thesource of the information desired be defined. Conse-quently, the Census Bureau has had to develop an officialdefinition of a farm for census purposes.

The ‘‘farm’’ first was defined for the 1850 agriculturalcensus as any place with annual sales of agriculturalproducts amounting to $100 or more. The definition hasbeen changed nine times since (see appendix B) with newsales or acreage criteria, but the land involved must beused for or connected with agricultural operations, andmust be operated under the day-to-day control of oneindividual or management (e.g., partnership, corporation).Agricultural operations means producing livestock, poultry,or other animal specialties and their products, and/orproducing crops, including fruits and greenhouse or nurs-ery products.

The census results are based on data obtained fromindividual ‘‘farms.’’ The land comprising the farm need notbe a single contiguous tract; that is, several separatepieces of land may be treated as a single farming operationwhen all are operated as a single unit. Since the county isthe smallest geographic unit for which the agriculturecensus tabulated complete data, specific rules cover farmswith land in more than one county. Where land operated asa single farm is located in two or more counties, the dataare tabulated in the county containing the largest value ofagricultural products raised or produced.

The 1992 Census of Agriculture used the same farmdefinition first employed in the 1974 census. A farm wasany place that met the above requirements for ‘‘agriculturaloperations’’ and individual control, and that had, or nor-mally would have had, $1,000 or more in total value ofsales of agricultural products during the reference year.The farm definition for Puerto Rico and the other outlyingareas was different, but also involved minimum levels ofsales: For Puerto Rico, a farm was any place that had $500in annual sales of agricultural products, or a place with atleast 10 cuerdas1 of land and a minimum of $100 in sales.In Guam and the Virgin Islands of the United States, anyplace with $100 in annual sales of agricultural productsqualified as a farm.

1A cuerda is approximately .97 acre.


1992 CENSUS OF AGRICULTURE

Scope and Reference Dates

Scope. The 1992 Census of Agriculture program collectedand published statistical data for all agricultural operationsmeeting the farm definition in the 50 States, Puerto Rico,Guam, and the U.S. Virgin Islands. The Bureau requestedcertain basic data from all farms, while selected informationwas asked of a sample of about 25 percent of all farms. Allfarm operators in the 50 States were asked for informationon:

x Acreage in the place.

x Crops, including fruits and nuts, vegetables, andnursery and greenhouse products.

x Gross value of sales.

x Land use.

x Irrigation.

x Land in government programs.

x Livestock and poultry.

x Animal specialties

x Amount received from Government Commodity CreditCorporation (CCC) loans.

x Payments received for participation in Federal farmprograms.

x Direct sales of commodities to consumers.

x Type of organization, including corporate structure.

x Number of hired workers.

x Injuries and deaths.

x Characteristics and occupation of operator.

The Bureau asked an approximate 25-percent sample ofoperators (i.e., all those with expected annual sales ofagricultural products above a specified value, together witha random sample of all other farms) for additional data. Thesample was small enough to limit respondent burden toreasonable levels, while providing reliable county-levelestimates for:

x Use of fertilizers and chemicals.

x Total production expenses, including interest expensefor secured and unsecured loans.

x Machinery and equipment (inventory and value)

x Market value of land and buildings.

x Income from farm-related sources.

For inquiries in Puerto Rico and the outlying areas seechapters 7 and 8.

Reference periods and dates. The 1992 agriculture cen-sus requested inventory data (e.g., number of livestock) asof December 31, 1992, while production, sales, and otherstatistics (except a few crops, such as citrus, for which datawere collected for the production year) were collected forthe calendar year 1992. In Puerto Rico, Guam, and theU.S. Virgin Islands, inventories were requested as of thedate of the enumeration, while production and sales datawere asked for the 12 months preceding that date.

An Overview of the Census Operation

Data collection. The 1992 census, like its predecessorsfrom 1969 onward, was a mailout/mailback enumeration.The Census Bureau assembled an initial mail list of morethan 14 million addresses from various sources, includingthe 1987 census. This preliminary list was reduced bydeleting duplicates and nonagricultural operations to a finalcensus mail list of 3.55 million (about 550,000 less than inthe 1987 census). In December 1992 the Census Bureaumailed report forms to the names and addresses on thecensus mail list, with a cover letter asking recipients tocomplete the report form(s) and mail them back to theCensus Bureau. Mail and telephone followup (the lattermaking extensive use of computer-assisted telephone inter-viewing (CATI) equipment and techniques) to nonrespon-dents continued over a period of 6 months after the initialmailing.

The farm and ranch irrigation survey, originally sched-uled for 1993, was postponed to 1994 for budgetaryreasons. This sample survey covered the major irrigationStates (see chapter 9, for details), employing a mail listdrawn from the 1992 census respondent list of farmsreporting irrigation.

As the Census Bureau ‘‘closed out’’ collection in eachState or area (i.e., once an acceptable overall responselevel had been achieved), it edited, reviewed, and tabu-lated the data, developed estimates for farms that did notrespond, and prepared the data publication for that State orarea. The 1992 census enumerated a total of 1,925,300farms, with an average farm size of 491 acres, and averageannual sales of over $84,000. By contrast, the 1987 censusshowed 2,087,759 farms, with an average of 462 acres andaverage annual sales of approximately $65,000. Totalacreage in farms for 1992 was 945,531,506 compared to964,470,625 for 1987. Total annual sales of agriculturalproducts for 1992 were approximately $162.6 billion, whilethe 1987 census showed total sales of $136 billion.

Data processing and publication. The Data PreparationDivision (DPD) at Jeffersonville, IN, received mail returns,entered individual report data into the computer file, andresolved edit failures, using interactive minicomputer sys-tems. The Economic Statistical Methods and ProgrammingDivision (ESMPD) then tabulated the data on the CensusBureau’s minicomputer system at the Charlotte, NC, facil-ity. (Report forms from Puerto Rico were processed by theDPD, while those from Guam and the U.S. Virgin Islands


were handled by the AGR staff at Census Bureau head-quarters.) The Census Bureau employed a new tabulationand disclosure system (TADS) to aid analysts in thereview/validation of county and State tabulations beforepublication. The TADS allowed analysts to review thetables on computer terminal screens, calling up individualfarm report files from the computerized data base andresolving any problems that showed up in the tabulations.This eliminated the thousands of pages of computer print-outs required in previous censuses for review and correc-tion.

Budgetary constraints required the Bureau to ‘‘stretchout’’ the 1992 agriculture census processing, and the firstVolume 1, Geographic Area Series printed report wasissued in January 1994. The agency prepared public-usecomputer tapes for each State and released them as theprinted reports were published. All the agriculture censusdata also were published on a set of two compact disc-readonly memory (CD-ROM) discs, and highlights were avail-able online.2

Geographic Area Detail

The 1992 Census of Agriculture provided data for theUnited States, each of the 50 States, Guam, Puerto Rico,the U.S. Virgin Islands, and for over 3,000 counties orcounty equivalents. In addition, selected data were tabu-lated and published at the five-digit ZIP Code level. TheUnited States is the 50 States and the District of Columbia(the 1992 agriculture census did not publish a separatereport for the latter), while ‘‘county equivalents’’ include theparishes in Louisiana and the‘‘census areas’’ in Alaska.State totals are aggregates of the county or county-equivalent totals, while the national ones are aggregates ofthe State data. The 1992 census data publication programalso released selected data items at the five-digit ZIP Codearea level on CD-ROM, and a separate CD-ROM contain-ing selected agriculture data for congressional districtsfrom the 103rd Congress and metropolitan areas (MA’s).3

The 1994 Farm and Ranch Irrigation Survey

This survey was the only follow-on operation providedfor in the 1992 agriculture census program. It supple-mented the basic irrigation data collected from all farm

operators in the 1992 census. The survey used a sample ofoperations from the 1992 census that reported usingirrigation during the reference year to obtain detailed dataabout irrigation practices without increasing the responseburden on all farmers. The survey requested informationon:

x Irrigated acreage.

x Crops—acres, yield, and quantity of water applied.

x Methods of on-farm water distribution.

x Source of water.

x Number of wells.

x Energy use.

x Irrigation maintenance costs.

x Irrigation practices.

Data from the survey was published in 1995 for the 48conterminous States, for the 27 States that account for 98percent of all irrigation including the 17 Western States,and for the 18 water resources areas (WRA’s).

Expenditures

The 1992 Census of Agriculture and its associatedactivities cost $85.6 million. Census operations invariablyoverlap one another in periods—one agriculture censusprogram is still being worked on when planning for the nextbegins. Preliminary work and plans for the 1992 enumera-tion began almost before the actual data collection for the1987 census had been completed. Expenditures by majorcensus activities for each fiscal year for the 1992 censusare shown in table 1-1.

2The agriculture census data were released by State on a flow basis,and the first CD-ROM disc contained the data files for the first 25 Statesreleased for publication. When data for all 50 States were released, theCensus Bureau published a complete file on two CD-Rom’s.

3A metropolian area (MA) is an integrated economic and social unitwith a recognized large population nucleus. Usually, an MA consists ofone or more counties (except in New England, where one or more city ortown may be designated an MA) or their equivalents, including a city witha population of at least 50,000, or an ‘‘urbanized area’’ with a populationof at least 50,000 and a total metropolitan population of at least 100,000(75,000 in New England). The MAmay contain one or more predominantlyagricultural counties.


ORGANIZATION OF THE CENSUS BUREAU

During most of the 1992 Census of Agriculture periodthe Census Bureau was organized under a director, adeputy director, and several associate directors respon-sible for specific operational or administrative areas—demographic fields, decennial censuses, economic fields,information technology, administrative services, field opera-tions, and statistical design, methodology and standards.Subordinate to the associate directors were three assistantdirectors with specific responsibilities for economic pro-grams, decennial censuses, and administration. A fourthassistant directorate, for communications, directly subordi-nate to the director and deputy director, was established inJanuary 1991. The assistant directorate for automated dataprocessing technology (ADP) was abolished in January1993, and the associate directorate for information tech-nology was created.

In June 1994, the Census Bureau underwent a majorreorganization intended to improve control and implemen-tation of Census Bureau policy priorities, establish a plan-ning and organization staff to be responsible for strategicplanning, and exercise better controls over the agency’sfinancial operations. The major features of the reorganiza-tion were the creation of (1) a Principal Associate Directorand Chief Financial Officer, (2) a Principal Associate Direc-tor for Programs, (3) a new Associate Director for Planningand Organizational Development, and (4) an Office of theComptroller at the same organizational level as the asso-ciate directors. The associate and assistant directors over-saw the various divisions and offices that make up theCensus Bureau. These individual units were grouped bybroad functional categories: (1) subject-matter (e.g., Agri-culture, Population, Industry, etc.), (2) data collection andprocessing (Data Preparation, Economic Programming,Systems Support, etc.), (3) statistical research and ser-vices (Statistical Research, Center for Survey Methods

Research), (4) administration and services (Budget, Tech-nical Services), and (5) communications and public ser-vices (Data User Services, Public Information, and theCongressional Affairs Office).

Agriculture census operations remained the responsibil-ity of the Associate Director for Economic Programs, butthe operational structure of the economic programs direc-torate was reorganized as well, consolidating planning staffand resources in a new Economic Planning and Coordina-tion Division. The Agriculture Division’s (AGR’s) own plan-ning staff was transferred to the new division, while thedivision itself was reorganized and acquired responsibilityand staff for collecting economic financial statistics. Table1-2 shows the AGR organization before the reorganization,while Table 1-3 displays the new Agriculture and FinancialStatistics Division (AGFS) organization. (Since the reorga-nization and renaming of the division took place only inJune 1994, this History will use the old title throughout.)

The Agriculture Division (Agriculture and Financial Sta-tistics Division from June 26, 1994) carried out the 1992Census of Agriculture under the overall supervision of theAssociate Director for Economic Programs and the Assis-tant Director for Economic Programs. Other divisions mademajor contributions to the enumeration, especially thefollowing:

x The Economic Programming Division prepared many ofthe computer programs for mail list development anddata processing.

x The Data Preparation Division in Jeffersonville, IN, wasresponsible for handling mailings, part of the telephonefollowup, and most precomputer processing of the reportforms.

x Other divisions handled automated data processing,user services, and publications services.

Table 1.1. Expenditures for Major Census Operations by Fiscal Year(In thousands of dollars)

TotalFiscal year

1990 1991 1992 1993 1994 1995 Total

Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . $2,670 $11,238 $13,664 $30,293 $17,271 $10,469 $85,605Direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 716 2,690 2,231 3,072 3,869 2,811 15,389Content determination/design . . . . . . . . . . . . . 97 2,811 3,493 646 357 376 7,780Mail list development/mailing . . . . . . . . . . . . . 155 1,822 1,889 6,111 834 606 11,417Data collection/processing . . . . . . . . . . . . . . . . 865 3,377 5,348 18,825 8,496 3,392 40,303Publication/dissemination . . . . . . . . . . . . . . . . 837 538 703 1,639 3,715 3,284 10,716


QFR AccountingOperations Branch

Table 1-3. Agriculture Division Organization Prior to June 26, 1994.

AGRICULTURE DIVISION

Procedures Branch

Data Requirementsand Outreach Branch

Publications and Information Office

Farm EconomicBranch

Crop StatisticsBranch

Livestock StatisticsBranch

Special Surveys Branch

Outlying Area Statistics Branch

Program Research andDevelopment Branch

Research andMethods Branch

AGRICULTURE AND FINANCIAL ST ATISTICS DIVISION

Assistant Chief,Financial Statistics

Assistant Chief, Agricultural Statistics

Assistant Chief, Researchand Methodology

Company StatisticsBranch

BusinessInvestment Branch

Frame Research and Development Branch Research and

Methods Branch

Table 1-3. Agriculture and Financial Statistics Division Organization From June 26, 1994.

QFR CorporateAnalysis Branch

Economics andResources Branch

Outlying AreasBranch

CommoditiesBranch

Special Projects Branch

Assistant Chief,Planning

Assistant Chief, Agriculture Statistics

Assistan t Chief , OutlyingAreas and Specia l Surveys

Assistant Chief, Researchand Methodology


Chapter 2.

Planning ------------------------------------------------------------------------------------------ 10Considerations --------------------------------------------------------------------------------- 10

Preliminary Planning--------------------------------------------------------------------------- 10Review of 1987 processing----------------------------------------------------------------- 10Agriculture census steering committee planning activities-------------------------------- 11Changes in computer hardware ------------------------------------------------------------ 11The Structured Program for Economic Editing and Referral (SPEER) system study--- 12

Creation of Computer Assisted Telephone Interviewing (CATI) ----------------------------- 12Census Quality Management (CQM) Process Action Team (PAT) ------------------------- 12

Consultation on the Census --------------------------------------------------------------------- 13General Information --------------------------------------------------------------------------- 13The Census Advisory Committee on Agriculture Statistics---------------------------------- 13Consultation With Governors, State Departments of Agriculture, and Land-GrantUniversities ----------------------------------------------------------------------------------- 14Federal Agency Working Group -------------------------------------------------------------- 15

The 1990 Test ------------------------------------------------------------------------------------ 15General Information --------------------------------------------------------------------------- 15Objectives of Testing -------------------------------------------------------------------------- 15Report Forms ---------------------------------------------------------------------------------- 16Sample Selection ------------------------------------------------------------------------------ 17

Mailout and Followup-------------------------------------------------------------------------- 18Mailing packages---------------------------------------------------------------------------- 18Mailout and mail followup------------------------------------------------------------------- 18Telephone interviews------------------------------------------------------------------------ 18Response rates------------------------------------------------------------------------------ 18

Processing ------------------------------------------------------------------------------------- 18Analysis and Results -------------------------------------------------------------------------- 19

Report Form Design ----------------------------------------------------------------------------- 19General Information --------------------------------------------------------------------------- 19The Regionalized Report Forms-------------------------------------------------------------- 20Map—Report Form Regions for the 1992 Census of Agriculture --------------------------- 21The Screener Section ------------------------------------------------------------------------- 22

Contents

Page


Planning andPreliminary Operations

PLANNING

Considerations

The periodic agriculture census collects and publishesdata on agriculture in the United States. As is the case withthe more familiar decennial census of population andhousing, planning an enumeration of agriculture is anexercise in balancing conflicting and often contradictoryrequirements. The core of the agricultural economy inAmerica is farm and ranch commodity production (theterms ‘‘farm’’ and ‘‘ranch’’ are interchangeable for censuspurposes; both identify a single agricultural establishment),and most farms and ranches in this country still arecontrolled by individual owner-operators (although partner-ships, and corporate organizations play an increasing rolein agricultural production). Farmers, and their counter-partsin agricultural-related industries, are required by law toprovide the census information, though most of them wouldprefer to spend their time doing other things than filling outGovernment questionnaires. The first and greatest com-promise in any agriculture census is made between thewants and needs of data users, and the response burdenthat can be imposed on data suppliers without provokingoutright refusal or other forms of noncooperation.

Cost is a major consideration as well. The CensusBureau developed themailout/mailback, or self-enumerationprocedure for the 1969 census in large part because itwould cut costs in half, but even mail enumeration facescontinually increasing expenses.1 Indeed, mailing costscomprise a very large portion of the cost of each census, sothere is a permanent interest in saving money by reducingthe size of the census mail list. Early response in thedata-collection effort also reduces costs by reducing theneed to followup to nonrespondent addresses that theCensus Bureau, for want of contradictory information, hasto assume are farms.

Once respondents have completed and returned theirreport forms the data must be extracted, processed, andtabulated. The more detailed the tabulations and crosstabu-lations, the more useful the data are to users, but tabulationand cross-tabulation consume both time and money and

increases suppression for disclosure avoidance. The fund-ing available necessarily restricts the volume and detail ofthe tabulations, as does the requirement that the census bepublished on a timely basis. With unlimited resources oftime and money, much more detailed and complex statis-tics could be published, but timeliness in releasing the dataalso is critical, so further compromises have to be made toensure valid data are available to users within a reasonableperiod following the enumeration.

The balancing act is further complicated by the confi-dentiality requirements of the law governing the operationof the Census Bureau. The census data-release programmay not release information that might be used to identifyan individual establishment or operator. This restrictionmeans that all the tabulations and cross-tabulations mustbe checked to ensure that individually identifiable dataitems are suppressed (or are not released).

Preliminary Planning

Review of 1987 processing. The Census Bureau made amajor effort to streamline data processing for the 1987agriculture census and to incorporate improvements insucceeding censuses. Themost significant of these improve-ments was the extensive use of interactive systems formany of the most labor-intensive and time consumingtasks, such as keying data and clerical editing. Up to 70individual work stations at the Data Preparation Division(DPD) office in Jeffersonville, IN, were linked to one ofseveral minicomputers at the Census Bureau’s Suitland,MD, headquarters during processing operations. The mini-computers had substantial data storage and processingcapacity; they dispensed with many of the paper printoutspreviously used by displaying data on the Jeffersonvilleterminals’ screens during name and address research,report form check-in, edit and analytical review, disclosurereview, and problem resolution and correction operations.

While the changes for 1987 led to significant improve-ments in overall processing efficiency, data quality, andtimeliness of release and publication, the relatively lateacquisition of much of the equipment used for the process-ing meant that the benefits realized in the 1987 censuswere only the beginnings of the advantages the agencymight obtain from this technology. This situation was rec-ognized by the agriculture census planning staff, and theinitial planning for the 1992 enumeration included a sys-tematic study of the 1987 processing, tabulation, and

1The 1964 agriculture census was the last carried out as a door-to-door canvass, and cost $25.1 million; the 1969 census, using mailenumeration, cost just $26 million, an actual decline in cost when inflationis considered.


disclosure systems. The general processing strategies forthe 1992 enumeration were developed at a series ofweekly meetings and plans for procuring the requiredcomputer hardware drawn up.

Agriculture census steering committee planning activi-ties. In December 1988, the Agriculture Division (AGR)organized anAgriculture Census Steering Committee, whosemembership included the Chief and Assistant Chiefs of theAGR, and the Assistant Chief of the Economic Program-ming Division for Agricultural Programs. This steeringcommittee oversaw the activities of 13 project committees,and several subcommittees, composed of subject-matterspecialists from the AGR and other Census Bureau divi-sions involved in the agriculture census (primarily theStatistical Research Division (SRD) and the EconomicProgramming Division (EPD)). Each committee and sub-committee was concerned with reviewing and makingrecommendations on specific census activities, as follows:

x Data content/forms design, mail follow-up/mailing.

x Census test subcommittee

x Mail list development

x Coverage evaluation

x Census processing

x Imaging subcommittee

x Edit review

x Analytical review

x Disclosure system

x Nonresponse survey

x Statistical weighting (nonresponse and samples)

x Data release program

x Electronic data release subcommittee

x Print/tabulations subcommittee

x Follow-on program

x Outlying areas program

x Variance estimation

The individual committees were directed to developoverall time schedules, goals, and objectives for planningfor their particular subject areas. The planners worked onthe general assumption that the 1992 census processeswould be similar to those used for the 1987 census, but thatimprovements would be made that would significantlychange the activities required, their sequence, and timing.While the committees addressed virtually every aspect ofthe census operation, the most significant changes as aresult of their recommendations were:

1. Expanding the use of interactive processing and editsystems in agriculture census operations, including

correspondence processing, data entry, and correctionedit operations. (See Chapter 6, ‘‘Data Processing,’’ formore information on these activities.)

2. The design and implementation (including the acquisi-tion of the required computer hardware) of the Tabula-tion and Disclosure System (TADS) using interactivetable review to eliminate paper-intensive table reviewactivies. (See below, and Chapter 6, ‘‘Data Process-ing,’’ for more information on the TADS).

3. The development of improved disclosure analysis meth-ods to reduce the volume of complimentary suppres-sion required in the census tabulations while maintain-ing confidentiality.

4. The adoption of the Census Bureau’s computer assistedtelephone interviewing (CATI) system for the 1992agriculture census followup operations.

Changes in computer hardware. By November 1990,Census Bureau staff had completed their analysis of the1987 processing operation, and began implementing manyof the recommendations that resulted from that analysis. Amajor objective that had emerged from the analysis andplanning activities was expanded automation of censusactivities, particularly the processing and tabulation ofcensus data. The AGR staff had begun plans for a new,automated table design system even before the processinganalysis had been completed, and had finished plans for anew Tabulation and Disclosure System (TADS) require-ment. The TADS requirement called for a standardizedmethod for table development and review that would makemaximum use of available technology and reduce bothprogramming resource requirements and tabulation errors.The system would employ a number of computer worksta-tions at Census Bureau headquarters, using advancedsoftware to provide status tracking, interactive table review,and data flow information to the staff at any time during thedata processing cycle. Further automation of the process-ing also would eliminate the huge masses of paper requiredin previous tabulation review procedures and provide stillbetter online review and research capabilities.

The TADS required the procurement of new computerhardware for the AGR to take maximum advantage of thenew system design. The Census Bureau staff decided thatany new systems should be compatible with the minicom-puter systems employed by the Economic ProgrammingDivision (EPD) staff, since EPD would be heavily involvedin processing planning and program implementation.2

To take best advantage of the TADS, the AGR staffwould have the easiest possible access to the systemcompatible with data file security. To enable the staff to

2The equipment and operating systems selected for employment withthe TADS were not compatible with the majority of equipment then in usein theAGR, and, consequently, the staff concerned had to be trained in theuse of the proposed new systems. This training was conducted shortlyafter the decision to procure the new equipment was approved, so thatonce the hardware was in place, work could be started immediately.


make maximum productive use of the system, the finalhardware configuration called for 20 new computer termi-nal stations, all of which could be used for table review andcorrection. Fourteen of these terminals were ‘‘full capabil-ity’’ work stations, that is, they had substantial internalmemory as well as data and image retrieval capacity, andcould transfer corrected tables and other material to themain-frame computer memory. The remaining six workstationsacquiredaspart of theplanwere ‘‘dumb’’ terminals—thatis, they lacked substantial memory capacity, but could beused for table retrieval and correction, and transfer to thesatellite or boot server stations. Two of the ‘‘full capability’’stations were ‘‘boot servers’’ that could be used to ‘‘bootup’’ the entire system for operations. Twelve ‘‘satellite’’terminals shared the capabilities of the boot servers,except they could not be used for starting up the wholesystem. Ten work stations were assigned to specific mem-bers of theAGR staff. The remaining work stations (satelliteand dumb) were available to anyone working on thesystem.

Procurement of the new TADS stations began with twoterminals and their associated equipment in January 1991.The entire system was in place, and AGR personneltrained to use the hardware and operate the system byAugust 1992.

The Structured Program for Economic Editing andReferral (SPEER) system study. In an attempt to savetime and resources, AGR and the SRD began a project,early in 1989, to adapt the SPEER system, developed bySRD, for use in processing the agriculture census. Theresulting adaptation was designated the ‘‘Ag-SPEER’’ pro-totype, and AGR drew up a list of specific questionnairesections for a comparative test of the Ag-SPEER andconventional complex edit system (Ag-Complex).

The test prototype covered 107 keycodes (i.e., data cellsfrom the report form) dealing with land and crops, since anyedit system would have to be able to work with thesesections if it was to be used in the agriculture census. Therecords passing through both edit systems then were usedas a clean data set, and ‘‘perturbations’’ were introduced tothe data, and the data set then was run through both editsystems once again and the results compared to evaluatethe comparative performance of the systems.

The perturbations added to the data destroyed many ofthe interrelationships among the keycodes, and neither theAg-SPEER or the Ag-Complex systems were able torestore perturbed values to the originals. SRD staff sug-gested that subject-matter analysts also examine the resultsof the edits and produce a more detailed analysis of theresults. This involved a followup study involving addingperturbations to one keycode at a time instead of to all thekeycodes, while the performance measure was changed toinclude a mean statistical error (MSE—an average squared‘‘distance’’ of the output value from the clean value for agiven datum). Three subject-matter analysts were involved;two identified the AgComplex as the more successful editsystem, while one favored the Ag-SPEER.

The AGR staff involved concluded that the Ag-Complexsystem performed the edit as well as, or better than, theAg-SPEER system and recommended that the former beused in the 1992 census. In December 1990, time andbudget constraints prompted the Census Bureau not to tryto develop the Ag-SPEER system for the 1992 census, butthe agency continued the research to determine whetherthe system might be used in 1997.

Creation of Computer Assisted TelephoneInterviewing (CATI)

The Census Bureau had made extensive use of com-puter assisted telephone interviewing (CATI) in its variousdemographic survey and census operations for some timebefore deciding to use it in the agriculture census. TheCensus Bureau first used a CATI system for the agriculturecensus in 1982, when the procedures and equipment weretested using approximately 10,000 delinquent large farmcases. (The followup for the remaining cases referred fortelephone followup used clerks who called respondentsand then wrote in the data on conventional report forms,before sending the forms through the regular processingsystem.)

Planning for the inclusion of CATI operations as part ofthe 1992 Census of Agriculture actually began relativelylate in the census planning operation, when the AGRestablished a CATI Committee (composed of CensusBureau staff knowledgeable about the CATI systems usedin other Census Bureau activities) in August 1991. TheCommittee consisted of members from the Agriculture,Economic Programming, Field, Demographic Surveys, andSystems Support Divisions. For the first 6 months of theplanning phase, this committee met every 2 weeks, there-after dividing into working groups to develop specific CATIprocedures, reconvening once every month to reviewplanning and coordinate activities.

The Census Bureau carried out three small-scale tests(conducted on September 28 and October 28, 1992, andJanuary 14, 1993) of CATI procedures and materials priorto its use for agriculture census followup. Experiencedtelephone interviewers from the Census Bureau’s Hager-stown, MD, CATI facility telephoned AGR personnel for‘‘enumeration.’’3 The tests examined the wording of thespecific questions on the CATI form; routing of interviewingscreens that should be used from question to question;output coding for case types; and nonresponse, transferand installation of files, input file content, and so on. Theresults of the tests were used to refine CATI enumerationprocedures and staff training.

Census Quality Management (CQM) ProcessAction Team (PAT)

The Total Quality Management (TQM) system was firstimplemented by Federal Government agencies in the

3Assigned ‘‘interviewees’’ used scripts designed to produce variedresults, i.e., as if the specific respondent was not an agricultural operator,or a refusal, or an in scope case requiring complete enumeration bytelephone.


mid-1980’s and the Census Bureau moved to adopt acustomized version of this system called the CensusQuality Management (CQM) in 1991.4 The Census Bureauestablished a steering committee of senior officials tosupervise the implementation of the program bureau-widein 1991—as the planning cycle for the 1992 agriculturecensus was approaching completion and the subject-matter divisions were implementing those plans—and overthe following year, all employees of the agency receivedsome degree of training in the CQM program.

The Census Bureau routinely works to improve itsquality of service and its products, so the adoption of theCQM program had its greatest impact on how the plans forimprovement were made and implemented. During thelater planning and the operational stages of the 1992census the AGR established four process action teams(PAT’s) to evaluate and recommend improvements tocensus data collection and processing; the specific teamswere concerned with (1) the citrus caretaker enumeration,(2) census data entry, (3) recording and reporting param-eter sources, and (4) computer-assisted telephone inter-viewing (CATI). Each PAT conducted studies of the specificsubject areas assigned and developed general and spe-cific recommendations for improvements to planning teamsresponsible for specified phases of the census (e.g., thecitrus caretakers enumeration team, the edit team). Therespective teams evaluated the recommendations for incor-poration into the census plan for the 1992 enumeration.

CONSULTATION ON THE CENSUS

General Information

The Census Bureau’s function is to collect and publishstatistical data for use by public institutions, and privateorganizations or individuals. It must, therefore, try to deter-mine which specific statistical information are most needed.Since the data compiled in the statistical tabulations mustbe supplied by individuals and/or organizations outside theagency itself, the Census Bureau must know whether therespondents to its censuses and surveys will be able tosupply the information requested.

The Census Bureau maintains regular contact with datausers and suppliers, and receives their advice and sugges-tions on census content. The agency’s standing CensusAdvisory Committee on Agriculture Statistics is the primarysource of this advice, but other contacts include directconsultation with the Governors’ offices and departmentsof agriculture of all 50 States, the land-grant universities, a

Federal interagency working group (established to advisethe Census Bureau on specific Federal agency data needs),and the Census Bureau’s own extensive outreach to datausers.

The Census Advisory Committee on AgricultureStatistics

Prior to 1940, any advice or recommendations to theCensus Bureau about the agriculture census was given bythe agency’s general statistical advisory committee. Forthe 1940 census, however, the Census Bureau establishedan advisory committee specifically concerned with agricul-ture statistics. From 1940 through 1959 the Census Bureauassembled an agriculture advisory committee as part of theplanning program for each census, and disbanded thecommittee once data collection was completed. In 1962,the agency requested that the Department of Commercecharter a permanent committee on agriculture statistics,and, upon approval of this request, the committee becameone of the Census Bureau’s ongoing advisory bodies.Farmers’ organizations and agriculture-oriented businessand professional associations were selected and invited toparticipate in the census program in an advisory capacity.Eachmemberorganizationnominateda representative—subjectto the approval of the Director of the Bureau of the Censusand the Secretary of Commerce—to participate in theCommittee’s activities.

The member organizations and their representatives forthe 1992 census period (October 1, 1990, through Septem-ber 30, 1995) were:

Organization RepresentativeAgricultural PublishersAssociation

Mr. James D. Rieck

American AgriculturalEconomics Association

Dr. Jerald J. Fletcher

American Association ofNurserymen (memberorganization from 1995)

Ms. Ashby P. Ruden

American Crop ProtectionAssociation (previously theNational AgriculturalChemicals Association)

Mr. Larry L. Harris(from 1991)Ms. Jarrad L. Blank(to 1991)

American Farm CensusBureau Federation

Mr. Mark Jenner(from 1995)Mr. Terry L. Francl(to 1995)

American Feed IndustryAssociation

Mr. David M. Tugend(from 1993)Mr. Norman Coats(to 1993)

American Meat Institute(dropped from committeein 1995)

Mr. Jens Knutson

4The essential characteristics of the ‘‘new’’ management plan wereimproved leadership and organization of the agency involved, improvedplanning, expanded training for all levels of employees, ‘‘implementationfor teamwork’’, and better recognition of contributions to programs byemployees. The objectives of the CQM program were to (1) improve thequality of service to the agency’s customers, (2) improve product quality,(3) involve data suppliers in Census Bureau discussions and plans toimprove their cooperation with the Census Bureau’s data-collectionefforts, and (4) involve and ‘‘empower’’‘‘ Census Bureau employees.


Organization RepresentativeAssociation of ResearchDirectors, Inc.

Dr. Harold Benson(from 1992)Dr. Sidney Evans (to 1992)

Conference of ConsumerOrganizations

Dr. William Fasse

Equipment ManufacturersInstitute

Mr. Douglas E. Petterson(from 1991)Mr. David W. Maaske(to 1991)

The Irrigation Association Mr. Robert C. Sears

National Agri-MarketingAssociation

Mr. Allan J. Hietala

National Association of StateDepartments of Agriculture

Mr. Arthur R. Brown(from 1995)Mr. Thomas W. Ballow(to 1995)

National Association of StateUniversities and Land-Grant Colleges

Dr. B. F. Stanton

National Cattlemen’s BeefAssociation

Mr. John Ross

National Council of FarmerCooperatives

Dr. Joseph D. Coffey

National FarmersOrganization

Mr. Calvin Shockman(from 1991)Mr. Willis Rowell (to 1991)

National Farmers Union Mr. Ivan W. Wyatt

National Food ProcessorsAssociation (dropped fromthe Committee in 1995)

Ms. Regina Hildwine

The National Grange Mr. Leroy WatsonRural Sociological Society Dr. Ronald C. Wimberley

National Agricultural Statis-tics Service, U.S. Depart-ment of Agriculture

Mr. Donald M. Bay(from 1993)Mr. Charles E. Caudill(to 1993)

The Committee’s meetings were open to the public.Outside agencies—e.g., the Office of Management andBudget (OMB), the U.S. Department of Agriculture (USDA),other Government agencies, Statistics Canada, privateorganizations, etc.—and the public could send observersand offer comments, questions, and recommendations tothe Census Bureau and the Committee during periods ofeach meeting reserved for public comment. Census Bureaustaff prepared and published minutes of each meeting,including any Committee recommendations together withthe Census Bureau’s responses.

The Committee met six times during the 1992 censusperiod, as follows:

Date LocationJune 6, 1991 Suitland, MD

May 14, 1992 Suitland, MD

May 26, 1993 Louisville, KY

December 7-8, 1993 Suitland, MD

June 7, 1994 Suitland, MD

May 18, 1995 Fresno, CAThe Census Advisory Committee on Agriculture Statis-

tics served as the Census Bureau’s principal contact withdata users outside the Federal Government. It advised theagency on current and future data needs, the ability (andsometimes, the willingness) of respondents to supply theinformation wanted, general data collection methods, con-tent and format of agriculture census and survey reportforms, and publicity for the census.

Consultation With Governors, StateDepartments of Agriculture, and Land-GrantUniversities

Agriculture is the most important industry in a number ofStates and is a significant industry in all 50, as well as inPuerto Rico and the outlying areas, and the CensusBureau routinely asks the State governments for assis-tance in publicizing the census. Every State in the Unionhas a department of agriculture, and both the Governors,and the State departments of agriculture have a consider-able interest in the content of the census questionnaires,as well as in the completeness and accuracy of theenumeration. In February 1989 the Census Bureau mailedletters to the State Governors and departments of agricul-ture, as well as to their land-grant universities, asking fortheir requests and recommendations on data content forthe 1992 census.

By May 1989, the Census Bureau had received nearly300 recommendations for changes in the content of thereport forms, ranging from requests for additional informa-tion on relatively routine subjects, such as asking forindividual sales data on rice (in Section 9, Gross Value ofCrops Sold), to somewhat more exotic ones, includingadditional data on llamas and ostriches (in Section 16,Horses, Bees, Fish, Goats, Other Livestock or AnimalSpecialties), and for retention or deletion of the specificsections (one respondent wanted to retain all of Section 23,Production Expenses, while another suggested it be deletedaltogether).

The proposed census report form content, reflecting theCensus Bureau’s evaluation of the recommendations and


data requests made by the various offices and organiza-tions consulted was finalized in October 1989, and AGRstaff and the Forms Design Branch of the Administrativeand Publication Services Division (APSD) began workingon test designs for a content test planned for late 1989 orearly 1990 began.

Federal Agency Working Group

The Federal Government is the principal user of theagriculture census data, and early in 1989 the CensusBureau contacted some 25 Federal offices and agenciesthat used statistics from the census and asked them toappoint representatives to an interagency working group todiscuss data needs and make recommendations on con-tent for the 1992 census. The working group as a wholemet twice, on February 23 and on August 16, 1989, to drawup recommendations. The Census Bureau staff also metseparately with representatives of individual agencies todiscuss detailed questions. The following offices and agen-cies participated in the working group:

x Department of Agriculture

x National Agricultural Statistics Service

x Economic Research Service

x Department of Commerce

x Bureau of Economic Analysis

x Economic Development Administration

x National Oceanic and Atmospheric Administration

x Commodity Futures Trading Commission

x Congressional Budget Office

x Congressional Research Service

x Department of Energy

x Energy Information Administration

x Environmental Protection Agency

x Federal Reserve System

x Farm Credit Administration

x Department of Health and Human Services

x National Center for Health Services Research

x Department of the Interior

x Bureau of Indian Affairs

x Bureau of Land Management

x Bureau of Reclamation

x U.S. Geological Survey

x International Trade Administration

x Department of Labor

x Bureau of Labor Statistics

x Occupational Safety and Health Administration

x Library of Congress

x National Science Foundation

x Office of Technological Assessment

x Small Business Administration

x Tennessee Valley Authority

The Census Bureau met with officials from severalDepartment of Agriculture agencies on March 22, 1989, fordetailed discussions of census data content, and theCensus Bureau’s staff also met with individual representa-tives of these offices to review special problems or discussparticular questions throughout the planning phases.

THE 1990 TEST

General Information

Prior to most censuses, the Census Bureau engages indetailed studies and planning aimed at obtaining the mostcomplete and efficient enumeration. Typically, this planningprocess will include one or more field tests of materialsand/or data-collection methodologies, providing an oppor-tunity to evaluate suggested changes in data content,forms design, changes in instructions to respondents, andother factors that might affect the accuracy and complete-ness of the enumeration. Preliminary planning for the 1990Census of Agriculture Test began early in 1989, at whichtime the Census Bureau contacted data users for requestsand recommendations on the data content for the 1992census. The agency contacted State governors and Statedepartments of agriculture, land-grant universities, nationalfarm organizations, and Federal agencies that made exten-sive use of agriculture census data for recommendations toimprove the agriculture census. The Census Bureau origi-nally planned to finalize the proposed data content for thereport forms by the fall of 1989 and to conduct a generaltest late in the year, followed by a second large-scale testto check enumeration procedures and processing in Novem-ber 1990. Budget considerations required cutting back onthis plan, and in November 1989, all work on the censustests was suspended. For a time, the Census Bureau hadno authorization for any test program prior to the 1992agriculture census, but in April, it received permission toconduct a single test, which was rescheduled for Novem-ber 1990.

Objectives of Testing

The Census Bureau had four specific objectives inconducting the 1990 Census of Agriculture Test:


1. To test wording and format of new content items todetermine whether respondents could understand thequestions and supply the information requested.

2. To test varying report form designs (e.g., bookletcompared to foldout) to measure differences in responseobtained by each.

3. To test various screening questions to try to improveand simplify the identification of nonagricultural opera-tions on the census mail list.

4. To test new instructions to determine whether they helprespondents to identify and correctly report for dupli-cate forms.

Report Forms

The 1990 test involved 17 different report form designsin all. The individual report forms themselves were col-lected into seven groups by test objectives (see table 2-1).

The form 90-A1 was used for a ‘‘control panel’’ in thetest. The A1 report form had the same content and design(a 10″ x 14″ foldout with 4 whole and 2 half pages)as the 1987 sample report form. The form 90-A2 was a12-page, booklet measuring 8″ x 11″, but withidentical content to theA1.5 The remainder of the test formsused the foldout design, and all used black ink on whitepaper stock, with yellow shading.

The Census Bureau used a separate, short screenerform (the form 87-A0400) in the 1987 census in an attemptto identify nonfarm operations as early as possible in thedata-collection operation. But the shorter form causedprocessing, and edit and imputation problems not previ-ously experienced with responses from such (generally)very small agricultural operations using the regular reportforms, while not yielding the more detailed informationobtained using the latter. (Ultimately, only about a third ofthe addresses sent these screener forms in the 1987census mailout were identified as farms under the censusdefinition and were included in the census tabulations.)

The 1990 test included a reexamination of the 1987screener idea; a control panel received a screener formidentical in content to that employed in 1987, while four testpanels were sent variations.

1. The control panel received the 90-A17 form, which wasidentical in content to the 1987 form 87-A0400 screenerform, but had a differently worded screener question.

2. The second panel was sent the 90-A18, with the samecontent as the 90-A17, but no screener question.

3. The third panel was sent the form 90-A3, which wasidentical in content to the 1987 nonsample report form,with the new screener question.

4. The fourth test panel got the form 90-A4, which was ashortened version of the 90-A3 with a screener ques-tion.

Five panels were used for testing various ways toidentify duplicate addresses in the mail list. Three of thesepanels were composed of known pairs of duplicate addressesidentified in the 1987 census, while the remaining two werenonduplicate addresses. The duplicate addresses panelsreceived the forms 90-A13, -A14, or -A20. All three formswere identical to the 1987 sample form, but the 90-A13package included a special insert with instructions forreporting duplication. The 90-A14 used a new ‘‘instruc-tions’’ box just below the census logo and address box onthe front page of the form for the duplication instructions,while the 90-A20 included only the 1987 instructions, whichwere part of the overall instruction sheet. The two nondu-plicate address panels used forms 90-A15 and -A16, whichhad the same format, content, and instructions arrange-ments as the 90-A13 and -A14 respectively. These panelstested the new duplicate instructions’ effect on responserates for the general (nonduplicate) universe.

Four panels tested specific data items and data con-cepts new to the census. All four panels employed formdesigns generally similar in layout and content to the 1987sample form, but with specific data item variations. Theobjective of this specific part of the test was to determine ifspecific new data items were understood and accuratelyanswered by the respondents, decide which of severalvariations of an item obtained the best overall response,and determine whether a particular item was reportedfrequently enough to include it in the census. All four reportforms used—the forms 90-A6, -A7, -A21, and -A22—testedlayout and wording for the following data sections/items:

x Section 1—Number of Landlords: Number of land-lords from which the operator rented land.

x Section 2—Types of Wheat: Acres of three differentkinds of wheat (winter, durum, and spring).

x Section 20—Farming Practices: Farming practicesthat reduced the use of chemicals and manufacturedfertilizers, such as low-impact sustainable agriculture(LISA).

x Section 21—Direct Sales: Direct sales of agriculturalproducts for human consumption.

x Section 22—Production Contracts: Commodities pro-duced under production contracts.

x Section 26—Injuries and Deaths: Farm-related inju-ries or deaths.

x Section 28—Farm Labor: Total farm labor, includingthe operator and unpaid labor.

x Section 29—Fertilizers and Chemicals: Use of com-mercial fertilizers and chemicals.

5The main reason for testing a booklet format was the possible use ofelectronic imaging of the 1992 census report forms in processing the 1992census. Funding restrictions precluded purchasing the needed hardwarefor such a system, and the idea was abandoned for the 1992 census.


Sample Selection

The 1987 census mail file served as the base file for thetest sample selection. The file contained approximately 4.1million records. Prior to sampling, records for operationsthat were to be excluded from the test were deleted fromthe sample file, these were—

x All farms with TVP’s of $500,000 or more.

x Multiunits and abnormal farms.6

x 1987 census follow-on survey cases.

x 1987 Classification Error Survey and June AgriculturalSurvey cases.

x Addresses in Alaska and Hawaii.

x Postmaster return cases (i.e., cases undeliverable asaddressed), correspondence records, and any othercase not classified as in-scope, out-of-scope, or nonre-spondent in the 1987 census.

x Selected nonrespondent cases.

The resulting sample file was sorted into an in-scope file,an out-of-scope file, and a nonrespondent file. The in-scoperecords then were sorted again, by ascending census filenumber (CFN), and the out-of-scope and nonrespondentrecords were sorted by ascending State/county codes fromthe CFN, then by ascending 1987 mail size code.

The test sample was organized into 17 test panels,which could be grouped into four general categories (therewas considerable overlap among the panels, as severaltest panels were used to evaluate various form character-istics or instructions, see table 2-1):

1. Long form panels: Selected from the entire eligiblemail file and made up of 1987 in scope (70 percent),out-of-scope (20 percent), and nonrespondent (10percent) cases. The long form panels were numbers 1,2, 5, 6, 7, 8, 10, 11, 12, 13, 14, and 15 (see table 2-1for report used for each).

2. Short form panels: Selected from 1987 mail list caseswith TVPs of less than $25,000 (in-scope, out-of-scope, and nonrespondent proportions were the sameas for the long form panels). The four short-form panelswere numbers 3, 9, 16, and 17 (see table 2 1 for reportforms used for each).

6Multiunits were companies or organizations with substantial agricul-tural operations at more than one location; abnormal farms were farmsoperated by institutions, such as Indian reservations, State agriculturalresearch stations, prison farms, etc.

Table 2-1. Forms Tested

Panelnumber

Formnumber Content description Sample drawn from

Testobjective

1 90-A1 1987SampleReport Form (Control panel) Mail list (with restrictions) Foldout vs. booklet response2 90-A2 1987 Sample Report Form (Booklet)

3 90-A3 1987 Short Form Mail list farms with Screener questions4 90-A4 1987 Booklet Short Form estimated TVP1 less and short form vs.

than $25,000 with Nonsample formrestriction

5 90-A6 1987 Sample Form (Version 1) Inscope farms only Content6 90-A7 1987 Sample Form (Version 2) Cluster and noncluster sample

7 90-A21 1987 Sample Form (Version 3)8 90-A22 1987 Sample Form (Version 4)

9 90-A8 1987 Short Form (Version 1) and Mail list farms with Content and screenerduplication instruction estimated TVP less question

than $25,000 withrestrictions

10 90-A11 1987 Sample Form (Version 2) and Mail list with Content and foldoutduplication instructions restrictions vs. booklet response

11 90-A13 1987 Sample Form with duplicationinstruction insert

Out-of-scope check-in with linkages toinscope farm

Duplication instructions

12 90-A14 1987 Sample Form with new duplica-tion Instruction on form

13 90-A20 1987 Sample Form with no insert ornew duplication instructions

14 90-A15 1987 Sample Form with duplication Mail list withInstruction insert restrictions

15 90-A16 1987 Sample Form with newduplication Instructions on form

16 90-A17 1987 Short Form with new screenerquestion

Mail list farms with estimated TVP of$25,000 with restrictions

Screener questions

17 90-A18 1987 Short Form with no screener ques-tion

1Total value of agricultural products sold.


3. Content change panels: Selected from 1987 in-scopecases only, with part of this sample drawn from tar-geted counties and States. The content change panelsall received variations of the long report forms. Thepanels were numbers 5, 6, 7, and 8 (see table 2-1 forthe report forms used for each).

4. Duplication panels: Selected from 1987 cases origi-nally checked in as inscope, but later changed to out ofscope because of duplication (linkage) to anotherinscope record. The duplication panels were numbers11, 12, 13 (see table 2-1 for report forms used foreach).

The AGR staff provided the applicable specifications,and the EPD staff used the Census Bureau’s mainframecomputers to select the specific mail lists for each samplepanel from the mail list file prepared for use for thoseparticular panels. The computer selected cases based onthe established criteria at a specified sampling rate from arandom start in each file. For example, ‘‘panel X’’ required2,600 cases chosen from a total file of approximately 4million addresses that met the sampling requirement crite-ria; the computer’s counter was set to start with the 56thcensus file number (CFN) in the file, and selected every1,500th CFN thereafter. (The actual selection process wasmore complicated than this for most of the panels, sincethe samples were also split into in-scope (70 percent—fora panel of 2,600 cases, 1,820 were to be drawn from the1987 in-scope list), out-of-scope (20 percent—520 cases),and nonrespondent (10 percent—260 cases) portions, butthe general procedure was identical for each individualcomponent of the sample panel.)

The samples for panels 5, 6, 7, and 8 (those receivingreport forms 90-A6, -A7, -21 and -A22) included ‘‘cluster’’samples drawn for selected counties, as well as a nationalsample. The remainder of the samples generally were notdesigned to meet any rigorous geographic dispersion orfarm type requirements, but to provide a general cross-sectional sample of addresses within prescribed param-eters from each file.

Mailout and Followup

Mailing packages. The report forms, envelopes, informa-tion sheets, and cover letters were printed by privatecontractors and the materials sent to the Census Bureau’sData Preparation Division’s (DPD’s) facility in Jefferson-ville, IN. DPD staff printed the mailing labels using high-speed printers and the computerized address file preparedby headquarters, then assembled the mailing packages.Each package consisted of a cover letter explaining thetest, the appropriate report form for each panel, anyinstruction sheet required, and a return envelope, all foldedand inserted so that the blank address box showed throughthe cut-out ‘‘window’’ of the envelope. The labeling machinesapplied the adhesive address labels through the window ofthe outgoing envelope.

Mailout and mail followup. The Jeffersonville office mailedtest forms to 44,292 addresses in 17 test panels onNovember 30—about 1 month earlier than the typicalagriculture census mailout—with a response-due date ofJanuary 1, 1991. (Panels 1-4, and 9-17 consisted of 2,600addresses each, while panels 5-8 had 2,623 addresseseach.) In mid-December, the Census Bureau sent areminder/thank you card to all addresses on the samplemailing list. On January 14, 1991, followup packagescontaining the appropriate test report form, instructionsheet, return envelope, and a cover letter requestingresponse were mailed from Jeffersonville to 19,275 nonre-spondent addresses. By the end of January, overall responsehad reached 66.4 percent. The mail data-collection phaseof the test was closed down on February 8, by which timeoverall mail response was over 69 percent.

Telephone interviews. The Census Bureau planned tocarry out telephone interviews of a one-in-four randomsample of each of the four new-content panels (5, 6, 7, and8) for telephone contact, or 2,364 farm operators in all. TheAGR designed and printed test questionnaires, arrangedfor telephone interviewing, and the staff began callingsample subjects in mid-December. Farm operators weregenerally cooperative with the telephone interviewers andthe operation proceeded well ahead of schedule; the first800 interviews had been completed by the end of Decem-ber, and all 2,364 telephone-survey sample addressesfrom the content panels had been contacted by February11, 1991—more than 2 weeks ahead of schedule.

The telephone interview program contacted test respon-dents. During later processing and analysis, the answersgiven by each respondent on the report form were com-pared to those provided by the same respondent to thetelephone interviewer to identify problem areas on theproposed questionnaires.

Response rates. The test achieved a relatively highresponse rate, with over 55 percent of all addresseesreplying by mid-January 1991, i.e., before the report formfollowup mailing was done. Final overall response was justover 69 percent, while responses within individual panelsvaried from a low of 62 percent to a high of over 75 percent.Nine of the 17 panels had response rates over 70 percent,which is considered excellent for a test.

Processing

The Census Bureau carried out only limited processingof the returned report forms, since no tabulations as suchwere needed. Report forms received from respondents onselected panels (those designed to test response to speci-fied content items and to form design) were reviewed andkeyed to a data file at the Jeffersonville office to enable theCensus Bureau to measure the effect of new content and


form design on data item and section responses. Approxi-mately 20,000 report forms were keyed to produce therequired analysis and tabulations.

Analysis and Results

The census test included a wide variety of ‘‘new’’ contentmaterials. TheAgriculture Division staff compiled responsesreceived to all the test report forms and to particularsections and individual items. This information was ana-lyzed to identify any patterns of response and the percent-ages of ‘‘correct’’ and ‘‘incorrect’’ responses to particularitems or sections. Statistical analyses of the telephonesample cases and of all formatted records for the new-content panels allowed comparisons between these panelsfor comparable items. Finally, two groups of farmers wereassembled and asked their opinions and observationsabout selected sections of the report forms.

The three principal general areas of interest were (1)possible variations in response for foldout compared tobooklet questionnaires, (2) the screener questions, and (3)the effectiveness of the duplication instructions. The stafffound there was no significant difference in the responseobtained between the foldout and booklet versions of thereport forms. Similarly, there seemed to be little or nodifference in the effectiveness of the various screenerquestions tested, although the 1987 version (used on form90-A17) proved more reliable in identifying nonfarms, whilethe screener questions on the nonsample forms providedsimilar results with reduced content. Consequently, thestaff suggested the Census Bureau use the 1987 screenerwording on the nonsample report forms, eliminating the1987 short form. With regard to the duplication instructions,the test revealed no improvement in reporting of duplica-tion from using either the duplication insert or the additionalnew instructions on the form. Telephone interviews ofrespondents who received the alternate instructions showedthat only about one-third actually read them. The staffrecommended that the 1987 duplication instructions beused again for the 1992 enumeration, although simplifiedlanguage be employed if possible.

The test also looked at a number of specific data contentitems to try to determine the practicality of collectinginformation on them. The staff submitted the followingrecommendations on these specific items:

1. Section 1—Number of Landlords: Add a question onthe number of landlords from which the operator rentedland and delete request for the name(s) of landlord andtenants.7

2. Section 2—Types of Wheat: Include the followingtypes of wheat on regionalized report forms for thefollowing States:

A. Winter wheat, durum, and other spring wheat:Arizona, California, Michigan, Minnesota, Mon-tana, North Dakota, South Dakota, Wisconsin, andWyoming.

B. Winter wheat and spring wheat: Colorado, Idaho,Nevada, New Mexico, Oregon, Utah, and Wash-ington.

3. Section 20—Farming Practices: Do not include aFarming Practices section on the 1992 census reportforms. (The results of all versions of the test, includingmajor items, showed results to be very unreliable, anddata published from the responses to these itemswould be highly suspect.)

4. Section 21—Direct Sales: Include the section ondirect sales, using the version used on form 90-A21(panel 7), adding the phrase ‘‘door to door’’ in the itemitself, with strengthened purchase and resale instruc-tions on the information sheet.

5. Section 22—Production Contracts: Drop any item onproduction contracts from the 1992 census reportforms. The relatively small number of farms operatingunder production contracts, together with the highlevels of uncertainty about what a production contractis, resulted in unacceptably high error levels in the test.

6. Section 26—Injuries and Deaths: Include a sectionon injuries and deaths, using the version tested on form90-A21 (panel 7).

7. Section 28—Farm Labor: Include paid labor with twocategories of workers—those working less than 150days per year, and those working 150 days or more. Donot include items on operator labor or unpaid labor.

8. Section 29—Fertilizers and Chemicals: Use the1987 version of the fertilizer and chemical sections,since lack of any additional information on farmingpractices would seem to make more detailed data onchemical use necessary.

The specific content recommendations all were incorpo-rated into the final report form designs. The Census Bureaudecided to adopt the booklet format for the 1992 question-naires to facilitate a possible test of electronic imagingequipment and procedures in census processing.

REPORT FORM DESIGN

General Information

The Census Bureau consults a variety of public agen-cies and private groups to determine the content andoverall design of the report forms it uses in its censuses

7The landlord item test yielded an extra practical lesson in report formdesign and its effect on response when the printer accidentally failed toprovide an unshaded box over the response line for the ‘‘number oflandlords’’ item on the panel 7 questionnaires. Less than half therespondents from the panel realized they were supposed to respond tothis item, whereas approximately 85 percent of respondents in the otherthree test panels concerned did so.


and surveys (see above for agencies and associationsconsulted specifically for the agriculture census). Theformat and overall design of the report forms are the resultof the work of the Census Bureau’s Agriculture (AGR) andthe Forms Design Branch of the Administrative and Publi-cations Services (APSD) Divisions, and reflect the resultsof tests, studies carried out between the censuses, tech-nology requirements of processing or other systems, andthe experience of the staff. The content of the report formsis more particularly the product of a variety of legislativeand administrative mandates, and data needs expressedby public and private users.

In determining the content of the 1992 agriculture cen-sus report forms, the Census Bureau considered thejustification for items based on the following criteria:

x Specific Congressional mandates, or strong Congres-sional interest or support.

x Requests from Federal agencies for data to meet legis-lative requirements to provide information to Congress.

x Evaluation requirements for existing Federal programs.

x Data items, which, if omitted, would result in addedrespondent burden and costs for a separate survey forother agencies or users.

x Historical farm classification requirements.

x Coverage improvement needs.

x Data required on a current problem of particular interestto the Federal Government or the general public.

The Regionalized Report Forms

Title 13, United States Code—Census, authorizes theSecretary of Commerce to determine the content of allcensus report forms. The Secretary normally delegates theactual task of designing the questionnaires to the CensusBureau of the Census, and for 1992, this work was carriedout by the AGR, assisted by the Forms Design Branch ofthe APSD.

Design of the 1992 forms began with a review of therecommendations of various data users regarding content,and of the 1987 report forms. The Census Bureau retainedthe overall report form plan employed in the 1992 enumera-tion, including using ‘‘regionalized’’ report form design. Forthe 1992 census, the 50 States were organized into 9multi-State regions, plus separate ‘‘regions’’ for Alaska,Florida, Hawaii, and Texas, each with a customized reportform. The regionalization of the questionnaires actuallyinvolved only limited changes, primarily to the specifictypes of crops listed in the appropriate sections. All of thereport forms had identical formats, consisting of a core ofstandardized inquiries asked of all agricultural operators(e.g., acreage, total value of sales, operator characteris-tics), and a set of production, inventory, and sales items(e.g., field crops, fruit trees and nuts) that applied specifi-cally to agricultural operations within each region. Usingregionalized report forms enabled the census to collect

more specific information on locally important agriculturalproducts, while avoiding the apparent inappropriateness oflisting items obviously not applicable to a given area (e.g.,listing sugarcane on a report form received by a farmoperator in Maine). The ‘‘nonsample’’ and ‘‘sample’’ reportform concept was retained as well, with nonsample formscontaining questions asked of all operators, while samplequestionnaires were used to collected specified additionalinformation (e.g., production expenses, machinery andequipment, farm related income) from a 25-percent sampleof the all farms.

Four report forms were used for each region—twononsample versions (one with the screener question andone without), and two sample versions (one for ‘‘must’’ andone for ‘‘certainty’’ cases8). The questionnaires were num-bered for identification, using the prefix ‘‘92-A’’ followed bya numeric designator identifying form type (‘‘01’’ for non-sample, ‘‘02’’ for sample-nonMust, ‘‘03’’ for Must, and ‘‘04’’for nonsample with screener question), and the regionnumber (‘‘01 through ’’13‘‘ without screener). For example,a small dairy farmer in Wisconsin might be sent report formnumber 92-A0103; the nonsample questionnaire for region3 (Wisconsin and Michigan).

The States composing each ’’region‘‘ for report formdesign purposes were as follows:

Region States1. Connecticut, Delaware, Maine, Maryland, Mas-

sachusetts, New Hampshire, New Jersey, NewYork, Pennsylvania, Rhode Island, Vermont,West Virginia.

2. Illinois, Indiana, Iowa, Kansas, Nebraska, Ohio.3. Michigan, Wisconsin.4. Alabama, Georgia, Kentucky, North Carolina,

South Carolina, Tennessee, Virginia.5. Florida.6. Arkansas, Louisiana, Mississippi, Missouri, Okla-

homa.7. Texas.8. Minnesota, Montana, North Dakota, South

Dakota.9. Colorado, Nevada, New Mexico, Utah, Wyo-

ming.10. Idaho, Oregon, Washington.11. Arizona, California.12. Hawaii.13. Alaska.

The Census Bureau added a general purpose reportform (92-A0214), with nonregion specific crop, livestock,and animal specialty listings, for use in correspondenceand for general information.

8‘‘Must’’ cases were agricultural operations so large that failure toinclude their data in the census tabulations would significantly distort thestatistics, or which required special handling (e.g., multiunits, abnormals,and so on). ‘‘Certainty’’ cases represented operations that were not largeenough to qualify as ‘‘must’’ cases, but were considered sufficiently largeto justify intensive followup. The precise sales levels and/or acreagesrequired to qualify as ‘‘must’’ and ‘‘certainty’’ cases varied by State.


TIPS [UPF] BATCH_146 [ACEN,C_ARLEDGE] 10/25/94 12:24 PM MACHINE: EPCV22 DATA:VOL1_TIPS_APXB_01.TIPS;1 * 10/12/94 09:48:00 TAPE: NOreel FRAME: 3TSF:TIPS92-09482232.DAT;1 10/12/94 09:48:35 UTF:TIPS93-09482232.DAT;1 10/12/94 09:48:36 META:VOL1_TIPS96_APXB_01.DAT;6 10/12/94 09:49:15

Map not available.

In addition to the standard nonsample and sample reportforms, the Census Bureau designed an agricultural ques-tionnaire for use with its film optical sensing device for inputto computer (FOSDIC) equipment (the form 92-AO202(F)),and tested it in the census mailing itself, with a sample offarms in region 2.9

The agency tested a variety of report form formats in anational test involving a mailing of test questionnaires to44,292 addresses in November 1990 (see above). TheCensus Bureau used the responses from this test to refinethe report form design prior to finalization of the content.

The Screener Section

The Census Bureau has been under continual pressureto restrain or reduce costs, and reduce respondent burden,in all of its operations. The mail operations of the censusare a very expensive budget item, and significant savingsin overall cost could be attained by reducing the volume ofmailing required by the census. There are several ways tomake savings in the mail operations; for example, reducingthe size of the initial mailout, obtaining earlier and morecomplete response, and quickly identifying nonfarm opera-tions retained on the mail list to avoid followup costs. All ofthese methods also reduce overall respondent burden.Prior to the 1987 agriculture census the Office of Manage-ment and Budget (OMB) issued a directive to the CensusBureau that it reduce the size of the census mail list to 4.2million addresses overall, and that no more than 3.2 millionregular report forms be sent in the initial mailing, toaddresses estimated to be most likely to represent farms.The remaining 1 million addresses were those for whichthe Census Bureau lacked the information needed toreliably identify them as farms, or nonfarms. Since addresses

in this group probably included a large percentage ofnonfarms, they were to be sent a short form (form 87-A400)asking for basic data required to determine (1) whether theaddress met the census farm definition, and (2) if it did, toenable the agency to impute for any information notactually collected.

The Census Bureau made strenuous efforts to furtherrefine its mail list and eliminate nonfarms for the 1992enumeration. Improved mail-list compilation procedures,particularly with regard to identifying and deleting duplicateaddresses and nonfarm agricultural operations, enabledthe agency to reduce the size of the initial census mail listto 3.8 million addresses. Budget constraints compelled theCensus Bureau to reduce the list to 3.55 million addressesfor the census mailout. The short form used in 1987 hadproven highly successful in identifying out-of-scopeaddresses early in the data-collection phase of the census(the 1987 short form was a 2-page questionnaire sent toaddresses considered least likely to represent farms). Forthe 1992 census, the Census Bureau used ’’screener‘‘forms that were virtually identical to the nonsample ques-tionnaires, but with an added ’’screener‘‘ question askingwhether (1) any crops were grown, or (2) any livestockraised, kept, or sold during 1992. Respondents answering’’yes‘‘ to either or both these inquiries were asked tocomplete the rest of the form; those who said ’’no‘‘ to both,were directed to section 26 (Person Completing ThisForm), and asked to return the form as quickly as possible.

The Census Bureau tested variations of the screenerform and the addition of a screening question to thenonsample census report form in the November 1990 test(see above). The results of the test led to a recommenda-tion from theAGR that a short, two-page form with screenerbe dropped from the 1992 enumeration, in favor of using ascreener question on the regular nonsample report forms.For 1992, the Census Bureau produced and used 11regionalized versions—forms 92-A0401 through -A0411(as well as a general-reference screener form, the -A0414)for the 48 conterminous States. The data from the screenerforms enabled the Census Bureau to (1) determine whetherthe address involved represented a farm, and (2) if so, toimpute census data not specifically requested on theforms.

Facsimiles of a representative standard report form, theFOSDIC experimental form, and a screener form, arereproduced in appendix F.

9The FOSDIC equipment is widely used in processing populationcensus report forms. For use with FOSDIC, population census reportforms were designed to have the respondent select one of a number ofpossible statistical ranges, and fill in a circle specifying that particularresponse. The forms were microfilmed during processing, and the FOS-DIC equipment ‘‘read’’ the responses from the microfilm by identifying andautomatically tabulating the specified response. The agriculture censusreport forms were a little different, using fill-in squares for respondents toidentify items for which they provided specific responses; the responsesthemselves still had to be keyed to the data file. The principal advantageto using a FOSDIC form for the agriculture census was that it streamlinedprocessing of the report forms.


Chapter 3.

General Information ------------------------------------------------------------------------------ 25

Address List Development----------------------------------------------------------------------- 25Introduction------------------------------------------------------------------------------------- 25General Procedures --------------------------------------------------------------------------- 26Sources----------------------------------------------------------------------------------------- 26General information ------------------------------------------------------------------------- 26Preliminary list ------------------------------------------------------------------------------- 26Final list -------------------------------------------------------------------------------------- 26Source priority codes ----------------------------------------------------------------------- 27

Format and Standardization------------------------------------------------------------------- 27General information ------------------------------------------------------------------------- 27Source file numbers (SFN’s) --------------------------------------------------------------- 27Source record edit -------------------------------------------------------------------------- 27Name control -------------------------------------------------------------------------------- 28Surname locator----------------------------------------------------------------------------- 28Size coding ---------------------------------------------------------------------------------- 28Possible partnership or corporation (PPC) flags ------------------------------------------ 28

EIN/SSN Record Linkage --------------------------------------------------------------------- 28General information ------------------------------------------------------------------------- 28EIN linkage ---------------------------------------------------------------------------------- 29SSN linkage --------------------------------------------------------------------------------- 29

Geographic Coding ---------------------------------------------------------------------------- 29General information ------------------------------------------------------------------------- 29Geographic reference file ------------------------------------------------------------------- 29Mail-file processing-------------------------------------------------------------------------- 30

Name and Address Linkage------------------------------------------------------------------- 30General information ------------------------------------------------------------------------- 30Identification of name parts----------------------------------------------------------------- 30Record linkage ------------------------------------------------------------------------------ 31

Statistical Modeling ---------------------------------------------------------------------------- 31General information ------------------------------------------------------------------------- 31Classification tree methodology ------------------------------------------------------------ 31Source-list record linkage------------------------------------------------------------------- 32

Clerical Resolution of Possible Duplicates --------------------------------------------------- 32Controls ---------------------------------------------------------------------------------------- 32General information ------------------------------------------------------------------------- 32Trace sample -------------------------------------------------------------------------------- 32Control counts ------------------------------------------------------------------------------- 33

Final Mail List------------------------------------------------------------------------------------- 33General Information --------------------------------------------------------------------------- 33Census File Numbers (CFN’s) ---------------------------------------------------------------- 33Must and Certainty Cases--------------------------------------------------------------------- 33Mail List Sampling ----------------------------------------------------------------------------- 34

Contents

Page


Printing and Addressing Report Forms --------------------------------------------------------- 34General Information --------------------------------------------------------------------------- 34Address Labels -------------------------------------------------------------------------------- 35Printing--------------------------------------------------------------------------------------- 35Quality control ------------------------------------------------------------------------------- 35

Printing, Assembling, and Addressing Mailing Packages ----------------------------------- 35Quantities ------------------------------------------------------------------------------------ 35Quality control ------------------------------------------------------------------------------- 35Labeling-------------------------------------------------------------------------------------- 36


Page

Preparatory Operations

GENERAL INFORMATION

By early 1991, the Census Bureau had determined theprincipal characteristics of the 1992 Census of Agricultureand began final preparations for the data-collection effortitself. These preparatory operations encompassed fourmajor activities—

x Determination of census report form content. (See chap-ter 2 for details of the final report form design, andappendix H for a facsimile of a representative question-naire.)

x Compiling the census mail list.

x Promoting the census to encourage cooperation byagricultural operators. (See chapter 4 for information onthe promotion campaign.)

x Printing the report forms and other enumeration materi-als and preparing census mailing packages.

The census mail list is the core of the modern agriculturecensus; the Census Bureau collects the bulk of all thestatistics published from the census through responses toreport forms mailed to farmers and ranchers. Compiling thecensus mail list involved acquiring, compiling, and match-ing (or ‘‘linking’’) records from various source lists ofaddresses believed to represent agricultural operations.For the 1992 census, the mail list compilation was carriedout in two phases—the first completed in the spring of 1992and a second beginning in the early summer using recordsthat became available later in the year. In both phases,records from a variety of sources were compiled andlinked—that is, source and address information analyzedto identify duplicate records. When two or more recordsappeared to be duplicates, the record, or records, withlower priority (that is, those drawn from sources consideredless reliable in providing current and complete mailingaddresses) were deleted from the file after transferring theidentification data to the higher priority record. During bothmail-list compilation phases, the Census Bureau tried toclassify by size addresses believed to represent agricul-tural operations meeting the census farm definition. Afterthe second major linkage operation, the Census Bureauapplied a classification model (see pp. 29-30) to theresulting address file, grouping addresses within the file bytheir likelihood of representing a farm. Addresses believedleast likely to be farms were identified and either deletedfrom the file, or were sent a ‘‘screener’’ form in the initialcensus mailout.

After completing the final mail list, the Census Bureauassigned individual identifying numbers—the census filenumbers (CFNs)—to each address in the file, and drew thesample file from the general census mail file (see p. 34 fordetails of the mail list sampling).

Once the report form content and design were finalized,the Census Bureau contracted with private printers to printthe report forms and associated materials (envelopes,instruction sheets, cover letters, and so on), assemble themailing packages according to specifications supplied bythe agency, and deliver the packages to the Data Prepa-ration Division (DPD) in Jeffersonville, IN. The DPD staffprinted adhesive address labels, using source files sup-plied by Census Bureau headquarters, and applied thelabels to the individual mailing packages just before mail-out.

ADDRESS LIST DEVELOPMENT

Introduction

The Census Bureau first used the mailout/mailbackdata-collection method in the 1969 Census of Agriculture.This self-enumeration procedure reduced personnel andother costs compared to the personal-interview methodol-ogy, but required a complete and accurate address list foragricultural operations that met the census farm definition.Duplicate and nonfarm records on the census mail list notonly added to the overall cost of the census, but increasedresponse burden and general respondent irritation—undermining respondent cooperation—with the enumera-tion, so duplicates and identifiable nonfarms had to beeliminated from the list as well.

All this meant that compiling the census mail list was acritical part of the census operation. Continuing emphasison reducing overall costs for the census meant the CensusBureau had to take particular care with the 1992 list.Budget constraints on mailing costs compelled the agencyto restrict the first census mailing to no more than 3.55million census packages, of which not more than 3.1 millioncould be mailed standard report forms. The mail-list com-pilation operation itself produced a preliminary census mailfile of approximately 3.78 million names and addresses. Toreach the required figure, the Census Bureau employedstatistical modeling techniques (see pp. 31-32) to identifyand delete from the file records believed least likely torepresent farms. For the 1992 census, the Census Bureau


identified some 411,000 addresses to be sent ‘‘screener’’report forms. The screener form included clear instructionsthat enabled out-of-scope recipients to return the formwithout having to fill out all the sections. This identifiednonfarms early in the collection cycle, saving followupcosts as well as response burden for the individualsinvolved.

General Procedures

The Census Bureau compiled the 1992 census mail listfrom the records of the previous census and from currentadministrative records from a variety of Federal agenciesand private associations. Names and addresses frequentlyappeared on more than one of the source lists used for thecompilation, so the various lists had to be matched to oneanother and duplicate records identified and deleted.

The Census Bureau conducted the 1992 mail list com-pilation in two phases—the first done over the period July1991 to April 1992, and the second and final phase fromJune to November 1992. The agency used essentiallyidentical procedures in both phases of the compilationprocess, the principal difference being the addition to thelist of updated source records. Each list construction phaseinvolved seven major operations:

1. Source list format and standardization.

2. Employer identification number (EIN) and social secu-rity number (SSN) record linkage.

3. Geographic coding.

4. Name and address coding and record linkage.

5. Resolution of possible duplicates.

6. Statistical modeling (that is, identifying groups of recordsby expected (or estimated) proportion of census farmsin each).

7. Establishing controls and assigning identification num-bers and other processing codes.

Sources

General information. The Census Bureau began devel-oping the 1992 mail list in December 1990, when Agricul-ture Division staff and representatives of other CensusBureau divisions met in the first of a series of conferencesto study the acquisition of records for the list. The CensusBureau made a determined effort to include all importantsources of agricultural information in the mail-list compila-tion, including the Internal Revenue Service (IRS), the U.S.Department of Agriculture’s (USDA’s) various offices, pre-vious census records, and private records from trade andagricultural associations. The Census Bureau used itsmain computer facility at Suitland, MD, to compile andprocess the source-list records in two phases, a prelimi-nary (phase 1) linkage operation completed in the spring of

1992, and a final (phase 2) compilation and linkageoperation—including the results of the preliminary linkageand records not available earlier in the year—completed inthe fall of 1992.

Preliminary list. The Phase 1 (spring 1992) linkage opera-tion involved approximately 9.1 million records drawn fromthe following sources:

Source Records

Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9,158,514

National Agricultural Statistics Service (NASS):Farms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,594,125Nonfarms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631,274Special list* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69,627

Special list (other) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107,603

1987 Census of Agriculture:Inscope. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,826,042Out of scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,534,398Nonresponse. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585,810

1990 Internal Revenue Service (IRS) files:1040F (Schedule for Farm Income and Expenses(attached to form 1040 Individual Tax Returns)) . . 2,242,3561120 (Corporation Income Tax Return (equivalentto standard industrial classification (SIC) codes01 and 02). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21,1521065 (Partnership Return of Income (equivalent toSIC codes 01 and 02) . . . . . . . . . . . . . . . . . . . . . . . . 67,710941/943 (Employers’ Quarterly Tax Returns)(941 coded 01 and 02 (Agriculture) for nonagri-cultural workers, and 943 for agricultural work-ers) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406,772

Business Master File (BMF—IRS 1120/1065 and941/943 combinations) . . . . . . . . . . . . . . . . . . . . . . . . . . 71,645

*The Census Bureau began collecting the ‘‘special lists’’ in April 1991.This involved contacting various Federal and State agencies, as well asbusiness associations and corporations to request lists of addresses ofindividuals and companies conducting agricultural operations.

The first phase of the mail-list compilation and linkageoperation (see below for details of the phase 1 and 2linkage) was completed in April 1992, resulting in a prelimi-nary mail file of 4,704,331 addresses.

Final list. In June 1992, the Census Bureau began thesecond and final compilation and linkage operation, usingthe preliminary mail file, and adding new source recordsfrom the IRS 1991 tax-year files, supplemental (NASS)National Agricultural Statistics Service files, updated mul-tiunit and abnormal lists from the 1987 agriculture census,the USDA’s Agricultural Stabilization and ConservationService’s (ASCS’s) Conservation Reserve Program (CRP)files, and the NASS’s JuneAgricultural Survey (used by theCensus Bureau in its coverage evaluation program)—atotal of approximately 3.28 million additional records. Thesecond compilation phase of the operation involved almost8 million records, as follows:


Source Records

Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7,990,944

Preliminary mail file . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4,704,3311987 Census of Agriculture, multiunits and abnormalfarms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10,670NASS farm adds (active) . . . . . . . . . . . . . . . . . . . . . . . . . 156,336NASS nonfarm adds (inactive) . . . . . . . . . . . . . . . . . . . . 207,561NASS evaluation file (June Agricultural Survey) . . . . . 64,136ASCS Conservation Reserve Program . . . . . . . . . . . . . 237,443

1991 Internal Revenue Service (IRS) files:1040F (Schedule for Farm Income and Expenses(attached to form 1040 Individual Tax Returns)) . . 2,056,9661120 (Corporation Income Tax Return (for SICcodes 01 and 02)) . . . . . . . . . . . . . . . . . . . . . . . . . . . 17,1161065 (Partnership Return of Income (for SICcodes 01 and 02)) . . . . . . . . . . . . . . . . . . . . . . . . . . . 63,912941/943 (Employers’ Quarterly Tax Returns) (941coded 01 and 02 (Agriculture) for nonagriculturalworkers, and 943 for agricultural workers) . . . . . . . 404,424

1991 BMF (1120/1065 and 941/943 combinations). . . 68,049

The second compilation and linkage operation created amail list containing 3,783,302 names and addresses. TheCensus Bureau reduced this preliminary mail file to approxi-mately 3.55 million by statistical modeling (see below fordetails).

Source priority codes. The source priority code identifiedthe specific source from which a name and address recordhad been drawn, and established its selection priorityrelative to suspected duplicate records from other sources.That is, when two records had been identified as suspectedduplicates, the one with the higher source priority wasselected for inclusion in the file, and the record with thelower source priority was deleted. Thus, a record drawnfrom the IRS 1040F file (priority code ‘‘2’’) would beselected in favor of an apparent duplicate record from theNASS (USDA list frame) file (priority code ‘‘7’’). The sourcepriority codes used in both phases of the mail list compila-tion, in priority order, were as follows:

PriorityCode Source

1 1987 Census of Agriculture Multiunits andAbnormal Farms.

2 IRS 1040F (Individual tax return).3 IRS 941 and 943 (Agriculture employers tax

returns).4 IRS 1065 (Partnership tax returns).5 IRS 1120 and 1120S (Corporate tax returns).6 1987 Census of Agriculture Farms.7 NASS (USDA List Frame).8 Special lists.9 1987 Census ofAgriculture Nonrespondents.10 1987 Census of Agriculture Nonfarms.11 NASS Nonfarms.12* NASS Evaluation File (June Agricultural

Survey).

*Used in the phase 2 compilation operation only.

Format and Standardization

General information. The Census Bureau’s two-phasemail-list compilation effort for the 1992 agriculture censusinvolved over 12.4 million individual address records from avariety of sources. Before these records could be matchedand the duplicates removed from the mail file, the agencyhad to establish a computer-record format compatible withits processing programs, and then apply that standardizedformat to the variety of computerized records assembledfrom the source lists. This format standardization placedeach source record into a standard format for name andaddress information and for generating processing codefields. The program functions included:

1. Assignment of unique identification number (source filenumbers (SFN’s)).

2. Source record edit.

3. Assignment of name control.

4. Assignment of processing codes/flags.

5. Size coding.

Source file numbers (SFN’s). The format program assigneda unique identification number—the source file number(SFN)—to each computerized record to enable specificrecords to be located and identified, together with thesource from which it had been drawn. Ranges of eight-digitnumbers were reserved for each source (for example,SFN’s from 15,000,000 to 19,999,999 were assigned to1991 IRS 1040F source records; 30,000,000 to 30,999,999to NASS farm adds, and so on) used for the compilation ofthe census mailing list. The computer program assignednumbers from the reserved set assigned for each source tothe appropriate records during the initial processing run.

Source record edit. The source record edit placed allsource records (i.e., names and addresses from the vari-ous lists used in compiling the mail list) into a commonformat for computer processing. The edit established twoname fields—a primary field that would always be filledfirst, and a secondary field used (if needed) for additionalnames (such as farm names, business names, or addi-tional individual names). The secondary field remainedblank in most records. Separate fields also were set up foraddress, place (city, State, and ZIP Code), and for process-ing codes.

For source lists that used the ‘‘last-name-first’’ format, anedit subroutine switched the order of names. The editprogram also deleted commas, periods, selected specialsymbols from the name and address fields, inserted aspace between adjacent alphabetic and numeric charac-ters so that each word could be classified as numeric ornonnumeric, substituted uppercase for lowercase alpha-betic characters, and replaced standard two-digit Stateabbreviations for State names or old-style abbreviations.For example:


Doe, Mr. John J., Jr. MR JOHN J DOE JR530 Euclid #48 became 530 EUCLID 48Chessie, Mary. XXXXX CHESSIEMD XXXXX

Name control. The ‘‘name control’’ for a record was thefirst four alphabetic characters of either an addressee’ssurname, or of a corporation’s main name, used to deter-mine possible duplicate status when linking records basedon EIN’s or SSN’s.

The formatting program read the name field of eachrecord from right to left until it identified an alphabetic wordwith three or more characters, then matched that word to a‘‘skip list’’ dictionary containing words or abbreviations(e.g., ‘‘Farm’’ or ‘‘Sons’’) that might appear in the name fieldbut were unlikely to be a surname. The first alphabetic wordwith three or more characters—a ‘‘3+’’ word—on the recordthat was not on the skip list was used to establish the namecontrol for that record. The first four (or first three if athree-character word) alphabetic characters—from left toright—were inserted by the program into the name controlfield.

For example, for a record with a name field reading ‘‘MrJames Smith & Sons Dairy,’’ the formatting program wouldread from right to left, identifying ‘‘Dairy’’ and ‘‘Sons’’ aswords to be skipped. The ‘‘&’’ was a single character andalso would be skipped. ‘‘Smith,’’ however, was a legitimatepossible surname, and the program extracted the first fourcharacters from left to right—‘‘Smit’’—and placed them inthe name control field. If the computer program could notidentify any usable word after scanning the entire primaryname field the name control field remained blank.

Surname locator. The formatting program inserted anindicator—a surname locator—in each individual record toidentify the field position of the first character of the namecontrol. If the name-control field was blank, the recordcould not be recoded (for details of the name recode, seebelow) for name and address linkage.

Size coding. Each source record was assigned a measureof estimated size derived from size indicators present in therecord. The size code was an estimate of the total value ofsales (TVP) of agricultural products by the agriculturaloperation represented by each record. The computer insertedthe size code for each record in a specific data field,depending on the source of the individual record. All thesize codes for an individual record were retained duringrecord linkage—that is, as records were linked and dupli-cates deleted from the file, the size codes from the deletedrecords were transferred to the appropriate field in theretained record. After all identified duplicates had beendeleted, the computer scanned all of the size codes foreach remaining record. If multiple codes were present, thespecific code retained depended on the priority assigned tothe size codes for particular sources (i.e., the CensusBureau’s estimate of the reliability of size information froma given source). The sources, and the size code indicatorswere used as follows:

Source Size indicatorIRS form 1040F Gross incomeIRS 1065 and 1120 Net receipts1987 Census of Agricul-ture farm records

Total value of products sold,from 1987 census records

1987 Census of Agricul-ture nonfarm records

Designated size code (17)

IRS form 941 and 943 Annual payroll1987 Census of Agricul-ture nonrespondents

1987 mail-list size code

Multiunits Designated size code (15) *Abnormals Designated size code (16)*Special lists Varied by list (usually based on

commodity inventory)NASS farm list USDA calculated farm value of

salesNASS nonfarms Designated size code (17)Conservation ReserveProgram

Total dollars paid*

*Used only in the phase 2 compilation effort.

Possible partnership or corporation (PPC) flags. Theformat program identified certain cases during the listbuilding and matching processing and ‘‘flagged’’ their com-puterized records as possible partnership or corporation(PPC) cases. This prevented the automated deletion ofpartnership or corporation records that had been matchedto individual records. For example, John Doe might operatean individual farm as a sole proprietor, while also having apartnership operation with Joseph Roe. In this case, thecomputer would compare the Doe/Roe partnership recordto Doe’s individual record on the basis of his name andEIN, and delete one or the other as a duplicate. The PPCflag on the Doe/Roe record changed the match status from‘‘duplicate’’ to ‘‘possible duplicate’’ and the record wasassigned to clerical review to determine the final disposi-tion.

EIN/SSN Record Linkage

General information. EIN’s and SSN’s provided the easi-est methods of linking duplicate records from the varioussource lists. Ninety percent or more of the records from thedifferent sources used in compiling the mail list includedeither an EIN or an SSN, or frequently, both. The CensusBureau’s computer programs compared the numbers oneach record to those on all other records in the files—aswell as checked name controls and PPC flags—to identifypositive duplicate and possible duplicate records. Thecomputer flagged and displayed possible duplicate recordsfor clerical review. Records identified as positive duplicates—those with matching EIN’s or SSN’s, and name controls,but without PPC flags—were reviewed by computer forsource priority codes; the record with higher numericalpriority code was deleted from the file. (Source priority


codes were assigned in reverse numeric order; that is, arecord with a priority code of ‘‘3’’ had a lower priority than arecord with a code of ‘‘1.’’)

The linkage operation for the 1992 census was essen-tially identical to the highly successful procedures used forthe 1987 enumeration. The computer programs linkedrecords by matching EIN’s to EIN’s and SSN’s to SSN’s.Records with both an EIN and an SSN were linked in twoseparate cycles. The number of records deleted duringeach phase of the EIN/SSN linkage operations was asfollows:

Recordsdeleted

Total fileafterlinkage

Phase 1 1992 linkageoperation:

EIN linkage 412,332 8,738,958SSN linkage 2,438,303 6,300,655

Phase 2 1992 linkageoperation:

EIN linkage 593,850 7,397,094SSN linkage 2,010,671 5,386,423

EIN linkage. Every record with an EIN was subject to theEIN linkage process. The computer sorted the record filesby EIN, then by PPC flag, name control, and address-priority code, and transmitted the sorted files to the match-ing cycle in code-priority order. (Records entered the cyclein descending priority order, so that records that would bedeleted always entered the cycle after the record thatserved as the ‘‘original’’ (called the ‘‘deleting record’’).) Thematching cycle moved the records from the sorted input fileinto temporary storage for the linkage operation. Thecomputer ‘‘wrote’’ the processed records to one or the otherof two output files, one for records with EIN’s only (no SSN)and all records for deletion, and a second for records withboth EIN’s and SSN’s. (The latter file would be subject to asecond linkage operation using SSN’s.)

The computer established two record locations, record1, containing the first (that is, the record with the highestsource-priority codes) of a suspected pair or group ofsuspected duplicates, while record 2 contained successivesuspected matching records.

When EIN’s matched, the computer compared namecontrols and checked for a PPC flag; if the name controlmatched and there was no PPC flag, the records wereidentified as a positive match. The sorting done prior to thelinkage operation ensured that record 2 had lower prioritysource-address codes than record 1, so the computerflagged record 2 for deletion and transferred all of record2’s source, size, and geographic codes to record 1, thenwrote it into the appropriate output file (EIN’s only andrecords for deletion, or EIN’s with SSN’s), while a newrecord moved into the record 2 location. When EIN’smatched, but the name controls did not, or when one or

both records contained a PPC flag, the records weredeclared possible duplicates. No codes were transferred,but a ‘‘possible-duplicate pair’’ number was inserted in bothrecords, linking them so they could be displayed togetherfor review. If record 1 already had a pair number, the samenumber was inserted into record 2; record 1 then waswritten to the appropriate output file and record 2 movedinto the record-1 location. This cycle continued until theinput file was exhausted, all duplicates had been flagged,and all possible duplicates had been assigned pair num-bers.

SSN linkage. After each EIN linkage operation, the Cen-sus Bureau merged the ‘‘EIN with SSN’’ output file with the‘‘SSN only’’ file to create the input file for the SSN-linkageprocess. The computer sorted the combined file by SSN,PPC flag, name control, and address priority in the samefashion used for the EIN-linkage input file, and used similarlinking procedures, except that ‘‘dummy’’ file records werecreated and different methods were used in assigning pairnumbers.

The SSN linkage operation created extra records forinput records containing more than one SSN. The creationof these dummy records allowed the original and thedummy to be linked, because IRS 1040F files sometimescontained two SSN’s (usually those of spouses) and had tobe linked to both SSN’s. The computer linked only one datafield for each record, hence dummy records were createdfor 1040F records with two SSN’s. After linkage, thecomputer matched the dummy records to their masterrecords, transferred any codes picked up during process-ing to the master, and deleted the dummy record.

Geographic Coding

General information. The 1992 census mail-list compila-tion processing operation included a geographic codingoperation designed to ensure that all records entering therecord-linkage system contained the appropriate standardgeographic codes. The agriculture census did not requirethe very detailed geographic coding required for the eco-nomic censuses, since agriculture census data generallywere compiled for only three primary geographic units—States,counties, and five-digit ZIP Code areas (data for othergeographic entities, such as census divisions, regions, andthe United States, were aggregations of State totals).Nevertheless, accurate geographic coding was vital to boththe mail-compilation operation (the name and addresslinkage operation used ZIP Codes as a blocking param-eter) and the census mailout itself.

The agriculture census geographic-coding operationused census State and county codes, alphabetic countyabbreviations, and ZIP Codes. The geographic informationused was drawn from the master geographic reference fileassembled by the Census Bureau’s Geography Division.

Geographic reference file. The source for the 1992agriculture census geographic codes was the 1992 Eco-nomic Geographic Reference File (EGRF), compiled by the


Geography Division. The 1992 EGRF was a concise andrelatively easily maintained computerized geographic filecontaining approximately 50,000 place codes covering the50 States, the District of Columbia, Puerto Rico, Guam, theVirgin Islands of the United States, and the Commonwealthof the Northern Mariana Islands. It contained the geo-graphic code structure—including two-digit State, three-digit county, two-digit congressional district, and five-digitZIP Codes—required for tabulating economic (and agricul-tural) data for specified geographic entities, as well as forediting other geographic files, and for producing recodefiles, geographic stub file, and other geography-relatedreference materials and products. Geography Divisioncompiled the 1992 EGRF by updating the 1987 economicgeographic information reference tape (EGIRT) file, usingindependent sources of geographic information.1

After updating, the Geography Division used the CensusBureau’s mainframe computer to edit the EGRF to checkfor consistency in the file. Thereafter, the EGRF containedthe following records used in the agriculture census:

1. One record containing the name and current andhistoric codes for specified census geographic entities(for agriculture census purposes, these entities were,regions, divisions, States and equivalents, countiesand equivalents, congressional districts, and five-digitZIP Codes with proper and variant spellings of mostpost office names, as well as the most likely countylocation for each ZIP Code).

2. One record representing the United States as a whole.

The Geography Division created several products fromthe EGRF used in the 1992 Census of Agriculture. Theseincluded the county/place code change file, used to convert1987 geographic codes to the 1992 code structure; thecounty alpha recode file, which converted the FIPS countycode into a six-character alphabetic abbreviation of thecounty name (used to prepare the questionnaire mailinglabel); the duplicate names file identifying each place orequivalent name that occurred more than once in a givenState; and the publication geographic stub file, used toinsert geographic entity names in the stubs for publicationtables.2

Mail-file processing. The Census Bureau used the 1992EGRF to edit the agriculture census mail records in aseries of computer operations that (1) checked the validityof the ZIP Code/post office name match on each record; (2)inserted ZIP Codes, post office names, and State andcounty alpha codes into records lacking those items; (3)standardized spellings of post office names; and (4) assigned(mailed and reported) county and State numeric codes.

Geographic coding was carried out as part of both thephase 1 and 2 mail-list processing operations (followingEIN and SSN linkage and the deletion of duplicatedrecords identified during those operations). In phase 1processing, 6,040,354 records underwent geocoding and1,366 records were rejected by the coding program.3

Records geocoded during phase 1 processing (over 4.7million in all) were not recoded in phase 2. A total of632,662 records underwent geocoding during the phase 2operation, and 2,026 were rejected as uncodeable.

After geographic coding, the mail file was ready forname and address linkage.

Name and Address Linkage

General information. After EIN/SSN linkage and deletionof duplicates, and geographic coding, the remaining mail-file records underwent a third linkage operation that matchednames and addresses. The name and address linkageprocess (1) identified name parts and other variables touse for the matching program, (2) recoded the names andaddresses to create short records for linking, and (3)matched the names and addresses in the file and deletedduplicate records.

Identification of name parts. The contents of the first andsecond name fields for each record had to be identifiedbefore the names themselves could be passed for linkage.The computer did this by comparing all the words in eachname field to the ‘‘skip list’’. Words matched to words onthe skip list were ignored, and the computer scanned andclassified all the remaining characters and/or characterstrings as a surname, single letter, conjunction (‘‘&,’’ ‘‘and,’’and so on), or ‘‘other.’’ The surname was identified usingthe surname locator assigned in the initial format program(see above). The computer identified conjunctions by com-paring each word to another dictionary, and classifying thewords accordingly. Each word was coded with a numericdesignator (for example, surname = ‘‘3,’’ conjunction = ‘‘4’’).

After classifying each character and character string ineach field, the computer retained the assigned codes, insequence, as the name pattern. This pattern identifiedeach word (character or character string) in the field, andthe computer compared the name pattern to a file ofacceptable name patterns that identified each word as afirst name or initial, middle name or initial, or surname.

1The sources used included (1) lists of new boundary changes thataffected governmental units through January 1, 1992 (based on theCensus Bureau’s annual boundary and annexation survey); (2) the Officeof Management and Budget’s (OMB’s) changes to the metropolitan area(MA) list inventory and components; (3) a list of towns in New England,New York, and Wisconsin, and of townships in Michigan, Minnesota, NewJersey, and Pennsylvania that qualified as special economic urban areas(SEUA’s); (4) the appropriate Federal Information Processing Standards(FIPS) publications (the FIPS codes were the source of the State andcounty geographic identification codes used in various agriculture censusprocessing, tabulation, and publication operations); and (5) selected 1990census population reports.

2The 1992 EGRF also was used in publishing the 1992 EconomicCensus Geographic Reference Manual, a printed report containing thecodes assigned to geographic areas for which the Census Bureautabulated data for the 1992 economic and agriculture censuses.

3The usual reason for rejecting a record was that it was uncodeable;that is, the information in the record was so incomplete, or so obviouslywrong, that no reasonable identification of its geographic location (at therequired level) could be made.


The computer created dummy records for any recordwith a multiple name pattern, so that each possible namehad an individual record. Each dummy record carried allthe identification codes of the original record so that it couldbe matched back to the original after linkage. Dummyrecords also were created for spouse names (except thosefrom the IRS 1040F files), names in the second name field,and partnerships.

Record linkage. Prior to 1992, census record linkage wasdone within each ‘‘block’’ of records—a block consisting ofall records from a single ZIP Code (or ZIP Code group4)that had the same recoded surname. The linkage processused a limited number of variables (surname, first name,middle initial, box/house number, rural route number) asmatch keys. While generally effective, this match proce-dure failed to detect a high proportion of duplicate namesand addresses, identifying many address as ‘‘possibleduplicates,’’ which required clerical review.

In an effort to increase the numer of duplicates detectedand to reduce the need for costly and time-consumingreview, for the 1992 census the Census Bureau used anew name-and-address linkage method based on a statis-tical model developed by the Census Bureau’s StatisticalResearchDivision (SRD).The newprocedure used frequency-based probabilities and statistical-match weighting pro-duced by an expectation-maximization (EM) algorithm,along with a match weight adjustment based on expertjudgment.5 More information was extracted from the indi-vidual records—including street name, telephone number,and SSN—for comparisons, which significantly improveddetection of duplicate records. Blocking for linkage in thenew system was based on ZIP Code or ZIP Code groupand the first character of the surname, which meant thatmore comparisons were made within blocks.

The procedure also used ‘‘string comparators’’ to com-pare names, telephone numbers, and identification num-bers. The name and address linkage operation in previouscensuses classified match keys as agreeing, disagreeing,or missing. The string comparators used for the 1992census identified degrees of similarity between two stringsof letters or numbers. For example, if three of the lastfour-digits of two telephone numbers checked agreed, thecomparison was not given the full disagreement weight,and thus was more likely to be classified as a possibleduplicate rather than a nonduplicate record.

The new procedures allowed the Census Bureau to sethigh and low limits, or cutoffs, for assigned duplication‘‘weights’’ for each group of records processed. Pairs ofrecords having weights above the high cutoff were identi-fied as duplicates, those with weights below the low cutoff

were nonduplicates. Pairs with weights falling between thecutoffs were identified as possible duplicates for clericalreview.

Statistical Modeling

General information. The various matching operationsused to compile the 1992 census mail list produced a‘‘final’’ mail file of approximately 3.78 million records. TheCensus Census Bureau used statistical modeling to iden-tify those records remaining in the file that were least likelyto represent farms for deletion to reduce the file to 3.55million addresses. The Agriculture Division staff selected a‘‘classification tree’’ methodology as the means to separatemail records into probable-farm, and probable-nonfarm,operations.

The Census Bureau used a classification tree model inthe 1987 census as well. This form of statistical modelingused the known characteristics of farms from the previouscensus to determine which were most useful in predictingfarm/nonfarm status. The 1987 census classification modelhad been developed by the Census Bureau’s AgricultureDivision and Economic Programming Division. For the1992 census, the Census Bureau used classification andregression trees (CART) software purchased from a privatevendor.

Classification tree methodology. The 1987 census mail-list records served as the source for the classification treedefinitions. The CART software partitioned the 1987 maillist into model groups (MG’s) defined by the informationknown prior to the mailout and common to both the 1987and 1992 mail lists (for example, geography, record source,previous census status, and expected value of sales). Thecomputer calculated the proportion of 1987 farm records ineach group, and used that as the expected proportion offarms in the same group for the 1992 mail file. Afterupdating the definitions (for example, using tax records for1990 and 1991 instead of 1985 and 1986) the staff appliedthese model groups to the 1992 preliminary mail file.

The CART software defined the model groups for the1992 census by the values of the classification variablesdetermined most likely to identify farm/nonfarm status forall the records under consideration. Fourteen of the clas-sification variables were defined according to whether agiven record had the following characteristics:

1. Was this record a 1987 Census of Agriculture nonre-spondent?

2. Was this record a NASS farm?

3. Was this record a NASS nonfarm?

4. Was this record a 1987 Census of Agriculture farm?

5. Was this record a 1987 Census of Agriculture non-farm?

6. Did this record submit an IRS 941 or IRS 943 form intax years 1990 or 1991?

4A ZIP Code group combined all records for a multi-ZIP Code city intoa single block. In the agricultural census, subject addresses in cities wererelatively few in number, and typically were horticultural or other specialtyoperations.

5The weighting adjustment was necessary because the EM algorithmlacked acceptable accuracy for finding probabilities associated with rareevents, such as agreement on SSN’s from two separate records.


7. Did this record submit an IRS 1065 form in tax year1990 or 1991?

8. Did this record submit an IRS 1120 form in tax year1990 or 1991?

9. Did this record submit an IRS 1040F form in tax year1990 or 1991 and was it a NASS farm and/or a censusin-scope record in the 1987 Census of Agriculture?

10. Did this record submit an IRS 1040F form in tax year1990?

11. Did this record submit an IRS 1040F form in tax year1991?

12. Did this record submit an IRS 1040F form in tax years1990 or 1991?

13. Did this record submit any IRS forms, except a 1040Fof either tax year 1990 or 1991?

14. Was this record on a special list?

The Census Bureau also used expected mail size (thatis, expected total annual value of sales of agriculturalproducts) as a classification variable. The expected mailsize had values ranging from 1 (estimated annual totalvalue of products sold (TVP) greater than or equal to$1,000,000) to 14 (estimated TVP less than $1,000). TheCensus Bureau derived this variable from informationreceived from the source lists used to compile the initialmail list. The following categories or records were includedin the mail file as certainty records, and therefore excludedfrom the modeling operation:

x Multiunits and abnormals (see p. 31).

x Most special list records and other source recordsmatched to special list records.

x NASS June Agricultural Survey (JAS) records and othersource records matched to JAS records.

Source-list record linkage. The phase 2 linkage opera-tion began in June 1992, matching 4.7 million records inthe preliminary list to approximately 3.29 million from newsource lists. After linkage, source and size codes requiredfor modeling and sample selection were assigned to theindividual records, and a file of 3,783,302 records wascreated. Applying the statistical model to this preliminaryfile created 787 MG’s. Records in the MG’s with the lowestexpected proportion of farms were identified as least likelyto be farms for the 1992 census, and were flagged fordeletion from the 1992 census mail file. Agriculture Divisionstaff reviewed the results of the initial modeling and adjustedthe model based on past experience, to shift specifiedgroups of records from the mail list to the delete file, andfrom the delete file to the mail list. This adjustment added145,026 cases to the delete file, but shifted 134,445 casesfrom the delete file to the mail file, so that the total deletionsfrom statistical modeling totaled approximately 229,000,with the final mail file reduced to 3,553,639 records.

Once the final mail file was defined, the Census Bureaualso used the statistical model to identify records to receivethe census screener report forms. This file was created byselecting records with (1) an estimated TVP of $25,000 orless and (2) the lowest farm probability among records notselected to receive the long (sample) report form. Themodel identified 411,640 records in the census mail file toreceive the screener forms.

Clerical Resolution of Possible Duplicates

After completing each EIN/SSN and name and addresslinkage operation the computer sorted the possible dupli-cate cases file by pair number for clerical resolution. Inprevious censuses, the clerical review operation involvedprinting thousands of pages of computer printouts andclerical processing of possible duplicate records and datakeying to incorporate clerical actions for computer process-ing. For the 1992 census, the Census Bureau developedan interactive computer system for reviewing and process-ing possible duplicate records. The new system displayedcomputerized records of possible duplicate sets on acomputer terminal screen for clerical review. Using specificprocedures and guidelines, the clerks reviewed the recordsto determine whether the records in each set were dupli-cates. Records to be retained were identified by comparingsource-priority codes (when duplicate records had identicalsource priority codes, the clerks retained the one with themost complete addresses). The clerks identified duplicaterecords electronically for deletion by keying the appropriateaction code into the computer. (A small sample of possibleduplicate sets were resolved by telephone calls to therespondents involved.)

The interactive system eliminated the need for printingand control of the paper listings previously used, as well asthe need for a separate data-entry operation to capture theclerical actions. The 1992 census clerical resolution opera-tion processed 769,267 record sets containing 1,979,936individual records, compared to 1,100,900 sets and 2,430,019individual records for the 1987 census, but required aclerical staff of only 24 (compared to 59 employed in theprevious census), and cost only about half as much perrecord processed.

Controls

General information. The Census Bureau creates a set ofchecks and controls on the mail list compilation operationin each census to enable it to track the actual source recordprocessing and collect materials to test specific phases ofthe operation. The 1992 census used two major controlprocedures—a ‘‘trace sample’’ of individual source recordsundergoing compilation and individual process controlcounts. The agency used the trace sample for qualitycontrol review of the overall operation, while the controlcounts recorded the number of individual records retainedor deleted from the file at each processing step.

Trace sample. The trace sample was a file sample ofrecords selected from the various source lists used to


monitor the mail list development processing. The trace-sample records were selected as part of the format andstandardization operation for each source list. The com-puter selected the first record, and every 1,000th recordthereafter from each file, setting a ‘‘trace flag’’ in each toidentify it as a trace sample record. Each trace-samplerecord was copied to a file for storage and display (whenspecified) at each stage of processing for review bystatistical analysts. This produced a file for each samplerecord showing it as it entered the compiling operation andthe changes made to it at every point during processing.The Census Bureau’s staff used the trace sample as aquality control tool, and for research projects concernedwith address file processing.

Control counts. The Census Bureau’s computer pro-grams generated processing-control counts of input files,deletions (and the point in processing at which recordswere deleted), and output files at each stage of the mail fileprocessing cycle. These counts served as checkpoints ateach phase of the processing. The principal matchingoperations, and the control counts generated in the phase1 and 2 mail list compilation were as follows:

Count RecordsPhase 1 Phase 2

Total input file 9,151,290 7,990,944EIN linkage deletes 412,332 593,850SSN linkage deletes 2,438,303 2,010,671Exact name matches 260,301 49,430Name and address link-age deletes 814,691 171,019Possible duplicatedeletes* 457,779 128,629Geocoding rejects 1,366 2,026Final processing rejects 62,187 1,106,130Out-of-scope ‘‘2+’’ deletes 140,783Duplicate 1987 CFNs 5,104Statistical model drops 231,895Output file (preliminary(Phase 1) and final(Phase 2) mail files 4,704,331 3,551,407

*Includes deletions resulting from clerical and telephone review.

FINAL MAIL LIST

General Information

The final mail-list preparations involved assigning cen-sus file numbers (CFN’s) and other processing codes toeach record, identifying ‘‘must’’ and ‘‘certainty’’ cases,selecting records to receive sample report forms, andidentifying cases to receive the screener form.

Census File Numbers (CFN’s)Census processing required a unique identification for

each data record—the census file number (CFN). TheCensus Bureau assigned a CFN to each address on the

final mail list. Each CFN consisted of 11 digits arranged inthree groups. The first five digits were the State and countycodes for the expected location of the agricultural opera-tion, the second five a serial number identifying the specificoperation within its county, and the last digit was a checkdigit to provide for a quality control check during process-ing.

The CFN was printed in both numerics and as a barcode on the address label affixed to each report formmailed.

Must and Certainty Cases

‘‘Must’’ cases were agriculture operations that (1) wereso large that failure to include their data would significantlydistort the census statistics, or (2) required special han-dling, such as multiunits. ‘‘Certainty’’ cases were agricul-tural operations expected to have large acreages or vol-umes of sale, but did not qualify as must cases in terms ofsize or type of farm, but were important enough to justifyautomatic selection for the long (sample) report form.

Must cases were identified by computer review of thecensus mail file after record linkage had been completed.The selection program used size codes for individualoperations, lists of multiunits from the 1987 census inscopelist, and other size indicators from the mail files, to identifyrecords for (1) farms so large that some data had to becollected, rather than imputed, in cases of nonresponse;(2) operations for which some explanation was needed ofwhy they were not engaged in agricultural operations (i.e.,an address that had large-scale agricultural activities recordedin the 1987 census); and (3) addresses with indicationsthat census response would require special analyst’s review.These general categories included—

x Multiunits. Multiunits were companies or organizationswith significant agricultural operations at more than onelocation. Multiunits typically required a separate reportform for each agricultural establishment, since each wasconsidered a separate farm for census purposes. Indi-vidual files were maintained for each ‘‘master’’ (i.e.,company or organization) record and each associatedfarm. Multiunits identified prior to the census mailoutwere assigned multiunit identification numbers in thealpha/plant field6 of the address label indicating whetherthe report form was for the master company or for anassociated farm.

x Abnormal farms. Abnormal farms were those operatedby institutions, such as Federal or State agriculturalresearch facilities, prison farms, Indian reservations, etc.

6The alpha/plant field identified the company with a six-digit number inthe alpha field of each record. The ‘‘plant’’ code was a four-digit estab-lishment identifier. The master record for a multiunit would have thecompany identifier in the alpha field, and four zeroes in the plant field,while each associated farm had the company identifier in the alpha field,and an individual identification number in the plant field. Each report formfor a master company or associated farm had a serial number, the farmsreceiving numbers in sequence following the master establishment.


x Other farms. Other must farms included addresses theCensus Bureau believed represented large individualfarms. The size criteria (expected total value of productssold (TVP) or total acreage) used to determine muststatus varied from State to State (e.g., in Texas, a mustcase had to have an expected TVP of $500,000, or2,000 acres of land; while in West Virginia, $100,000 inTVP or 1,000 acres qualified).

Other large cases also were selected for telephonefollowup on the basis of acreage and TVP. The minimumacreage requirement generally was the same as for themust cases—i.e., 1,000 to 10,000 acres, depending on thespecific State, while minimum TVP varied from $40,000 to$500,000. Must and telephone followup cases receivedintensive telephone followup during the census processing.When addresses could not be reached by telephone, or thefarm operators refused to respond, secondary sources,such as the USDA’s Extension Service (ES) and/or Agri-cultural Stabilization and Conservation Service (ASCS)offices were asked to provide information on any agricul-tural operations of nonrespondent addressees. Data fromprevious censuses, together with information from othersources, were used to impute responses for nonrespon-dents.

Mail List Sampling

The Bureau of the Census introduced large-scale sam-pling for agriculture data in the 1945 Census of Agriculture.The Census Bureau used post-census sample surveys tosupplement the basic data collected in the 1959 and 1964agriculture censuses, and employed a 50-percent samplein the 1969 census to collect data for farms with annualsales of less than $2,500. In the 1978 and followingcensuses, the Census Bureau sampled its census mail listto collect specified additional data from selected agricul-tural operations. The censuses asked all farms for basicdata, while a sample of approximately 25 percent ofaddresses on the census mail list received a sample reportform requesting additional information on such items asvalue of machinery and equipment, production expenses,and use of fertilizers and insecticides. To further reduceoverall response burden in the 1987 census, the CensusBureau introduced a ‘‘short’’ form (one sheet, front andback) with abbreviated versions of the standard items.Addresses on the mail list, but believed least likely to meetthe census farm definition, received these short forms. Forthe 1992 census, a screener section was added to the frontof the standard nonsample questionnaires to produce theForm 92-A401 to -A411, and -A414, Screener Forms. Thescreener section enabled recipients who were out of scopeto determine that fact, and skip the rest of the reportingsections of the form.

The sampling method used for the 1992 census wasessentially the same as that used for the 1982 and 1987enumerations. The mail-list compilation operation identi-fied addresses as ‘‘certainty’’ (including multiunits, abnor-mal farms, and all farms in counties reporting fewer than

100 farms in the 1987 census) or ‘‘noncertainty’’ based onexpected value of sales of agricultural products or acreage(the exact requirements for designation as a certainty farmvaried by State). After linkage and deletion of duplicaterecords, and the statistical modeling of the final mail list,the regular census mail file was sorted by CFN for sampleselection.

The sample file included all mail-list records in Alaska,Hawaii, and Rhode Island, and a sample of records in allother States. Records selected for the sample included all‘‘certainty’’ records, a systematic sample of 1 in 2 of allnoncertainty records in counties reporting 100 to 199 farmsin the 1987 census, and a systematic sample of 1 in 6 of allnoncertainty records in counties reporting 200 or morefarms for 1987. This differential sampling scheme providedreliable data for the sample items at the county level. Whena nonsample large farm was identified during processing,the Census Bureau mailed it a supplemental form contain-ing the additional sample data inquiries.

The Census Bureau identified mail-list records to receivethe screener form by statistical modeling (for details, seeabove). All records not designated for the sample weresorted by model-group farm probability as specified by themail-list group. The records in the groups with the lowestprobability of being farms, and with an expected total valueof agricultural product sales below $25,000 were added tothe screener form file. The remaining mail list recordsreceived the nonsample form.

The final mail list file was as follows:

Report form type Records

Total 3,551,407Sample/certainty (including must) 1,008,068Nonsample 2,131,699Screener 411,640

PRINTING AND ADDRESSING REPORT FORMS

General Information

The Census Bureau contracted with commercial printersto print the report forms, information sheets, mailout andreturn envelopes, and other enumeration materials.7 Thecontractors printed the various forms, and assembledspecified numbers of mailout packages for the initial andfollowup mailings, using written specifications provided by

7Other materials included special instruction inserts for multiunits,abnormal farms, cattle feedlots, nursery and greenhouse operations,poultry contractors, bee and honey producers, fish and aquacultureoperations, and laboratory animal producers. The number of operations ineach of these categories varied widely; there were fewer than 200laboratory animal producers on the census mail list, while there werenearly 57,000 nurseries and greenhouses.


the agency, and under quality control supervision of Cen-sus Bureau personnel.8 The contractors sent completedpackages and other printed materials to the DPD office inJeffersonville, IN, for final preparation (essentially affixingmailing labels and sealing packages) and mailout.

Address Labels

Printing. The 1992 agriculture census mail list comprisedover 3.55 million names and addresses. The CensusBureau created a computerized mailing list at its maincomputer facility in Suitland, MD, then transmitted the list tothe DPD in Jeffersonville, IN, by telephone datalink. TheDPD staff in Jeffersonville used the address list files to printthe adhesive address labels using six high-speed printers.The check-in operation updated the response list daily, andthe Census Bureau created an address file of nonrespon-dents following each response cutoff date. Computer-generated adhesive address labels (the addresses werefrom the nonrespondent lists produced by the CensusBureau after each mail response cutoff date) were used forall the mailouts except the first followup, which used the92-A01(L2) Reminder Card (the equipment printed addressesdirectly onto the face of the about 1 million cards; theremainder were addressed using the adhesive labels).

Quality control. As the labels were printed (for the initialmailout and after each mail closeout), DPD quality control(QC) clerks monitored the printing to ensure that the labelswere in the proper format, legible, aligned so that when cutthe address and identification information would be visible,and that the bar codes were readable and in the correctformat. The QC clerks checked the entire first file for eachfarm type from each printer, for each printing. They alsospot checked pages of labels at specified intervals in eachprinting run. QC problems with any file resulted in partial orcomplete reprinting, as needed.

Printing, Assembling, and Addressing MailingPackages

Quantities. The quantities of report forms, letters, andenvelopes printed for the 1992 agriculture census aresummarized in table 3-1 (for more detailed information onprinted quantities of materials, see appendix C).

Table 3-1. Quantities of Materials Printed

Form number Description Quantity

Information sheets andform letters:

92-A01(I) and -A02(I) Instruction sheets 9,289,00092-A01(L1) and (L1A) Transmittal letters (initial mail-

out (L1) and UAA’s (L1A) 4,563,00092-A01(L2) Reminder card 4,200,00092-A01(L3) through (L6) Followup letters 4,656,000

Envelopes

92-A7.1 through -A7.5; -A7B,-A7C, and -A7 Outgoing envelopes 9,862,00092-A8A(SC), (N), (S), and(M)

Return envelopes (screener(SC), nonsample (N), sample(S), and must (M)) 9,800,000

Report Forms

92-A0401 through -A0411,and -A0414 Screener report forms 1,816,0002-A0101 through -A0111 Nonsample report forms 5,137,50092-A0201 through -A0214 Sample report forms 2,667,90092-A0201(F) Sample report form (FOSDIC

test) 129,00092-A0301 through -A0311 Must report forms 466,000

Facsimiles of the instructions sheets, reminder card, andprincipal followup letters are included in appendix G.

The mailing package contents for the initial mailout inDecember 1992 were as follows:

Type Report formInforma-

tion sheetReturn

envelopeCoverletter

Screener . . 92-A0401 through-A0411*

92-A01(I) 92-A8A(SC) 92-A01(L1)

Nonsample . 92-A0101 through-A0111*

92-A01(I) 92-A8A(N) 92-A01(L1)

Sample . . . . 92-A0201 through-A0213*

92-A02(I) 92-A8A(S) 92-A01(L1)

Must . . . . . . . 92-A0301 through-A0311*

92-A02(I) 92-A8A(M) 92-A01(L1)

*As appropriate, the mail packages include Hawaii (-A0212) andAlaska (- A0213), and special instructional inserts.

Quality control. Private contractors printed and assembledthe 1992 agriculture census mailing packages to specifica-tions supplied by the Census Bureau. Teams of two orthree DPD quality control (QC) personnel visited eachcontractor’s printing facility when the forms and packageswere being printed and assembled to oversee the printingand to inspect the contractor’s printedmaterials andassembledpackages. Report forms and envelopes were subject to avisual review to make certain the printing was of accept-able quality, the proper colors and shading were used, andso on. The QC staff also checked a random sample ofassembled mailing packages to ensure that they werecomplete and the materials had been inserted in the properorder.

Each contractor boxed each day’s production of assembledpackages for QC review, which involved the Census Bureau’sQC staff selecting a specified number of boxes (the exact

8In addition to the complete packages of each form type, a certainnumber of each type of form was printed and reserved for remailing to‘‘undeliverable as addressed’’ (UAA) cases, for correspondence, or forinformational uses. For example, the nonsample printing run included notonly the regular mailing packages for regions 01 through 11, but also atotal of 81,000 UAA packages for those regions, as well as 22,000correspondence packages (2,000 per region, the UAApackages preparedfor each region varied proportionally to the initial mailout for a particularregion), and 90,000 ′information’ copies of the report forms (5,000 to11,000 per region).


number depended on the total number of boxes of pack-ages produced by that days’ printing run) for the qualitycheck. The QC staff then pulled three packages at randomfrom each box for inspection. When an error was identified,the rest of the packages in the box involved were checkedas well. If similar, or other, errors were found, the surround-ing packages also were inspected. All detected errors hadto be corrected before the packages were shipped to theDPD office at Jeffersonville for labeling and mailing.

Labeling. The adhesive address labels for the initial mail-out (and for all but the first (thank you/reminder card) mailfollowups, were printed by form number in ZIP Codesequence. Labeling machines at the DPD office in Jeffer-sonville, IN, applied the labels through the open windows of

the outgoing envelopes. The machines labeled mailingpackages at the rate of up to 10,000 per hour. QC staffinspected the labeling machines prior to each productionrun and checked at random intervals during each run toensure that the labels were applied to the correct forms andthat torn or mutilated labels were removed and the pack-ages recycled for correction.

The bulk of the labeling was completed by the end ofNovember, and the Census Bureau released the mailingpackages for abnormal and multiunit operations to the U.S.Postal Service for mailing on December 8. The remainderof the approximately 3.55 million census mailing packageswere mailed during the following week.


Chapter 4.

Introduction --------------------------------------------------------------------------------------- 38General Information --------------------------------------------------------------------------- 38Objectives------------------------------------------------------------------------------------ 38Consultation --------------------------------------------------------------------------------- 38Program activities --------------------------------------------------------------------------- 38

Theme and Logo------------------------------------------------------------------------------- 38

Census Publicity Campaign --------------------------------------------------------------------- 39Broadcast Materials --------------------------------------------------------------------------- 39Printed Materials------------------------------------------------------------------------------- 39General -------------------------------------------------------------------------------------- 39Posters--------------------------------------------------------------------------------------- 39Brochures, standardized speeches and statements, and order forms ------------------- 41Guide to the 1992 Census of Agriculture and Related Statistics------------------------- 41

Newspapers and magazines --------------------------------------------------------------- 41Agriculture report form guide --------------------------------------------------------------- 41Precensus lesson plans--------------------------------------------------------------------- 42Special materials ---------------------------------------------------------------------------- 42

Agribusiness and Agricultural Organizations ------------------------------------------------- 42

Postcensus Publicity and Assistance to Data Users ------------------------------------------- 43News Releases -------------------------------------------------------------------------------- 43Professional Meetings ------------------------------------------------------------------------- 43State Farmer Meetings ------------------------------------------------------------------------ 43

Contents

Page


Public Awareness Program

INTRODUCTION

General Information

Objectives. The public awareness program for the agricul-ture census has two major parts, the first addressing datacollection and the second involving promoting data dissemi-nation and use. The data-collection outreach phase has asits principal goal to persuade farm and ranch operators tocomplete and return their census questionnaires. Theprogram’s objectives were to—

x Encourage farmers and ranchers to respond to theagriculture census by February 1, 1993.

x Create public awareness of the agriculture census.

x Inform farmers and ranchers of the benefits of thecensus data to their own operations. .

x Emphasize the confidentiality of the census data.

x Defuse negative attitudes towards the census.

The data-dissemination, or post-census, phase of theawareness program was intended to—

x Increase public awareness of the agriculture census andits data products.

x Increase public access to, and use of, agriculture censusdata products.

Consultation. The Census Bureau’s Agriculture Division1

(AGR) formed a census awareness program working groupin the summer of 1990 to prepare a promotion and mar-keting campaign. Early in 1991, the agency formed an adhoc committee with representatives of various agenciesfrom the U.S. Department of Agriculture (USDA) to coordi-nate cooperation between the Census Bureau and theUSDA in promoting the 1992 census.2 Originally, the

Census Bureau considered contracting with a private com-pany to develop and implement television and radio activi-ties for the 1992 census promotional campaign; in Decem-ber 1992, however, the AGR was directed to use agencyresources for its public awareness effort. The division’sstaff, in cooperation with the agency’s Public InformationOffice (PIO), Data User Services Division (DUSD), andCongressional Affairs Office (CAO), designed a multimediapublicity program for the 1992 census.

Program activities. Publicity and promotional activitiesinvolved precensus mailings to inform respondents anddata users about the agriculture census, efforts to promoteearly response to the census, and post-census mailoutnews releases about the enumeration and the data col-lected. Census Bureau staff briefed Members of Congress,farm and agribusiness organization representatives, andagriculture-related media editors and broadcasters. Theagency distributed information materials—brochures, les-son plans, news releases, special stories, and the like—toschools, businesses, private associations, and individualsthroughout the country.

Before the census mailout in December 1992, promo-tional activities concentrated on raising general awarenessof the census and encouraging early and complete response.After the bulk of the data had been collected, the focus ofthe program shifted to informing the public, and particularlypotential data users, about census product data content,format, media, and availability.

Theme and Logo

The Census Bureau used two logos for the 1992 agri-culture census. The first, used in the title boxes of reportforms and on the spines of printed publications, duplicatedthe 1987 logo depicting a farm and silo, with ‘‘AG CENSUSUSA’’ below it.

1On September 18, 1994, the economic directorate was reorganizedand the Agriculture Division was renamed the Agriculture and FinancialStatistics Division.

2The participating USDA offices were the Radio and Television Divi-sion, Office of Public Affairs; Extension Service (ES); Agricultural Stabili-zation and Conservation Service (ASCS); Soil Conservation Service(SCS); Farmers Home Administration (FmHA); National Agricultural Sta-tistics Service (NASS); and the Animal Plant Health Inspection Service(APHIS).


A second logo appeared on the front covers of printedreports; this was a stylized representation of a barn andsilo, and plowed fields, with stylized cows peeking out atthe viewer from the lower left side of the circular field. Thelogo was a circle, with green ink on an white background.

A third frequently employed image was a map of the 50States with symbols of various agricultural products super-imposed over the approximate areas of the country inwhich they were grown. This image was used on all threeposters, information kits, covers, and video tapes, as wellas in the ‘‘drop-in’’ ads provided to magazines and news-papers (see page 40).

The theme for the 1992 promotional campaign dupli-cated that used for the 1987 effort—‘‘America Counts onAgriculture,’’ and this motto, usually preceded by theadmonition ‘‘Make It Known,’’ was used on the posters,press releases, and other publicity materials.

CENSUS PUBLICITY CAMPAIGN

Broadcast Materials

The AGR staff drafted initial work proposals for a seriesof radio and television promotional spots in August 1990,and some preliminary video-taping of promotional materi-als began in October 1990. However, substantial work onthe radio and television components of the public aware-ness campaign did not begin until early 1992, when theCensus Bureau began recording a series of radio andvideo promotional spots for distribution to Census RegionalOffices (RO’s), State/Business and Industry data centers(S/BIDC’s), and cooperating broadcast outlets. However,the Census Bureau decided to concentrate its broadcastoutreach activities in radio, in part because radio broad-casters proved more cooperative about using promotionalmaterials, and because the agency determined that farm-ers spent more time listening to radio than watchingtelevision.

The Census Bureau prepared two series of publicservice announcements (PSA’s) as the primary radio broad-cast publicity vehicles for the agriculture census. The firstwas a set of four PSA’s (one 10-second and three 30-secondspots) in which members of Congress recorded scriptedstatements supporting the agriculture census for distribu-tion and broadcast by radio stations in their districts. Thesecond was a series of 11 recorded statements of supportfor the census from agribusiness leaders, government

officials, and representatives of farmer and agriculturalorganizations, such as the Pennsylvania Farmers’ Asso-ciation and the National Cattlemens’ Association. TheCensus Bureau began distributing copies of the PSA’s inNovember 1992, and continued sending materials out untilMay 1993, supplying copies to every member of theNational Association of Farm Broadcasters (some 280full-time members in all, each of which could represent anetwork of 2 to 200 broadcasting outlets).

During and immediately following the enumeration, theAGR and the PIO cooperated in preparing agriculture-census related interviews for radio broadcast. Each month,beginning in May 1993, PIO staff conducted a series of fourrecorded interviews with Agriculture Division staff coveringa variety of census-data related subjects—e.g., Federalpayments to farmers, catfish farming, young farmers, rank-ing agriculture States and counties, etc. The interviews,edited to one minute each, then were broadcast as the‘‘Just a Minute’’ segment of the PIO’s own ‘‘Windows onAmerica’’ radio program, which was distributed to cooper-ating broadcasters. At its inception, this program was usedby only a few radio stations, but within 18 months as manyas 200 broadcast outlets were involved.

Printed MaterialsGeneral. While broadcast and other electronic media areincreasingly influential in reaching the public, printedmaterials—newspapers, magazines, posters, informationalbrochures, and the like—remained an important source ofinformation. The 1992 public awareness program contin-ued to make use of these materials, providing posters andbrochures to offices and agriculture-related organizationsall over the country for display; distributing wallet cardscontaining basic agriculture information as well as anintroduction to the 1992 census at conferences and meet-ings; providing articles, press releases, and drop-in adver-tisements to magazines and newspapers; and writing anddistributing standardized speeches, agriculture census guides,and lesson plans.

Posters. The Census Bureau printed 64,750 copies eachof a 8-1/2″ x 11″ and an 11″ x 14″ 1992 agriculture censusposter, using green and black ink, with gray shading, onheavy white glossy stock. The Census Bureau mailed16,000 of the smaller posters to rural post offices (i.e.,those whose areas included rural delivery routes) fordisplay, and distributed a further 42,500 copies of each tovocational agriculture teachers and programs (the teachersalso received a cover letter requesting that they ask theirstudents to place the posters in the windows of localbusinesses). Approximately 10,000 copies of the largerposter were sent to USDA agencies (National ResourcesConservation Service (NRCS—3,150 copies), the Exten-sion Service (ES—4,800 copies), and the Farmers HomeAdministration (FmHA—50 copies)), while the CensusBureau distributed 200 more to Land Grant Colleges.

The Census Bureau also printed 250 copies of an 11″ x24″ version of the poster for use in displays at meetingsand conferences.


1992 Census of Agriculture

Make It Known—America Counts

on Agricultur e

U.S. Department of CommerceEconomics and Statistics AdministrationBUREAU OF THE CENSUS

Farmers andRanchers!

Please return yourcensus form by

February 1, 1993


Brochures, standardized speeches and statements,and order forms. The Census Bureau prepared pre- andpost-census promotional brochures as part of the aware-ness campaign. The pre-mailout brochure, Form A92-R1,The 1992 Census of Agriculture, was an eight-page bro-chure, outlining the history, legal authority, and scope of theagriculture census. It briefly described the need for thecensus and the principal uses of the information collectedand published, and alerted operators about when theywere likely to receive the census questionnaire.

Agriculture Division and the PIO also wrote a shortstandardized speech promoting the census for CensusBureau employees and interested outside persons to giveat local meetings and presentations. Approximately 10,000copies were printed and distributed to ES and ASCScounty offices, as well as to land-grant colleges anduniversities, various agricultural, trade and public service(e.g., Junior Chambers of Commerce, FFA, and Rotary)associations, and State departments of agriculture.

The post-mailout items included the 1992 Census ofAgriculture Data Products Order Form. This covered muchof the same background material given in the Preview, butalso provided basic descriptions of the data-release pro-gram and included ordering and pricing information for bothprinted reports and electronic media and telephone num-bers for accessing online services and for obtaining addi-tional information.

Guide to the 1992 Census of Agriculture and RelatedStatistics. While most agriculture census data users areprimarily interested in the agriculture data themselves,there are a number of other Census Bureau programs thatpublish statistical information that is useful in any study ofAmerican agriculture or the people concerned with agricul-ture. Moreover, some specialized data users are not famil-iar with the overall character of the agriculture censusstatistics program. In order to provide a compact generaloverview of agriculture-related statistics available from theCensus Bureau, the DUSD, in cooperation with the AGRfirst produced a general guide to the agriculture censusand related statistics for the 1982 enumeration, and addedthis publication to the general program for succeedingcensuses. The Guide to the 1992 Census of Agricultureand Related Statistics was a 58-page booklet that outlinedthe background and procedures of the agriculture census,and described the agriculture census data series andmedia used for each. The publication also reviewed otherCensus Bureau data sets from census and current dataprograms that included agriculture-related information orthat might be of interest to agriculture census data users.The guide covered publication plans for the 1992 agricul-ture and economic censuses, and listed reference sources.Appendixes included a list of tables in the Volume 1 datareports, and a facsimile of a representative 1992 agricul-ture census report form.

The Census Bureau distributed copies of the Guide to itsregional offices and public advisory committee members,State data centers, clearinghouses for Census data ser-vices, and as part of its census curriculum support project.

The Guide was available on request at Census Bureauexhibit booths or presentations at conferences and con-ventions around the country and from Agriculture Divisionand DUSD.

Newspapers and magazines. The Census Bureau begandistributing printed promotional materials in November1992. Special information kits were assembled, each con-taining some or all of the following—

x A cover letter.

x Agriculture census fact sheet.

x Form 92-A15, ‘‘Questions Frequently Asked About theCensus of Agriculture.’’

x The appropriate sample report form.

x The 1992 Census of Agriculture Report Form Guide.

x Telephone contacts for the Agriculture Division.

x Scheduled release dates for 1992 Census of Agriculturepublications.

x Data release program.

x An insert suitable for use in Congressional newsletters.

x The script for four radio PSA’s.3

Two thousand eight hundred similar information kits(without the PSA scripts and with a different cover letter)were distributed to newspaper and magazines news edi-tors, and to some radio news directors as well.

The Census Bureau prepared 11 data-collection seriesnews releases (forms 92-A24(A) through -A24(K)) aboutthe census for release between November 1992 and May1993. The first three releases announced that the censuswas about to begin and told why agricultural operatorsshould cooperate. The remaining eight concentrated onreminding farmers to respond, and described the impor-tance of agriculture and its changing characteristics in theUnited States.

Agriculture report form guide. The AGR prepared formAC92-R-7, 1992 Census of Agriculture Report Form Guide,as a reference manual for county agents, vocational agri-culture teachers, USDA agencies (e.g., the National Agri-cultural Statistics Service (NASS), Farmers Home Admin-istration (FmHA), ASCS, etc.), and others to use in helpingfarmers and ranchers, or other respondents, complete theirreport forms. The Guide was 64 pages long, includingappendixes and index, and contained explanations anddetailed instructions for completing each item on the sample

3The PSA scripts were included only in the kits sent to the adminis-trative aides of Members of Congress. The Census Bureau requested thatMembers of Congress support the census by participating in these radioPSA’s by making a generic statement of support. The four PSA’s hadscheduled release periods; the first was to be used before December 15,1992, the second between December 15, 1992 and February 1, 1993; thethird from February 1 to June 1; and the final one after June 1, 1993.


and nonsample questionnaires. Moreover, each item instruc-tion included what should not be included in their response.For example, the instructions for section 2, ‘‘CroplandHarvested,’’ item C, ‘‘Cropland used for cover crops,’’asked for land used in 1992 only to grow cover crops forcontrolling erosion or to be plowed under for improving thesoil. Respondents were specifically instructed not to includeacreage from which crops were harvested or land used forpasture or grazing.

The Census Bureau printed 75,000 copies of the 1992Census of Agriculture Report Form Guide, together withcover letters and/or additional materials (such as theprecensus lesson plan), were distributed as follows:

Organization or agency Copies

National Resource ConservationService (NRCS)

3,150

Farmers Home Administration(FmHA)

50

Extension Service (ES) 4,800National Agricultural Statistics Service(NASS) State office (5 each)

250

High school vocational agricultureteachers

8,500

Land grant colleges and universityagriculture departments (2 each) 74State data centers (SDC’s), and leadagencies (2 each)

100

Census Bureau regional offices(10 each)

120

Precensus lesson plans. The AGR staff developed theprecensus agriculture census lesson plan for use by voca-tional agriculture classes; ‘‘agriculture in the classroom’’teachers; student teachers at land-grant universities; math-ematics and geography teachers; and elementary, middle,and high school teachers. The plan—form AC92-R-3(A)Census of Agriculture Lesson Plans—provided backgroundinformation on the census and six lesson plans:

x Plan 1. Kindergarten through grade 2.

x Plan 2. Grades 3 through 6.



x Plans 5 and 6. Grades 7 through 12.

Lesson plans 1 through 4 were designed to introducestudents to the agriculture census; each was geared to aspecific age group. Lesson plan 5 was designed to be usedwith each schools’ language arts program, while lessonplan 6 was intended to be employed with the schools’mathematics program.

The Census Bureau distributed the precensus lessonplan to some 8,000 vocational agriculture instructors and to

over 3,000 USDA ‘‘Agriculture in the Classroom’’ instruc-tors. Several thousand copies also were distributed toland-grant universities, teacher education programs, com-munity colleges that had agriculture programs, and to farmand trade organizations.

Special materials. In addition to the usual press releases,advertisements, public service announcements, and otherconventional publicity materials, the Census Bureau usedseveral special promotional items to try to increase publicawareness about the agriculture enumeration. During the1982 agriculture census, the Census Bureau had pur-chased a supply of baseball-style hats, bearing the censuslogo, for distribution by NAFB members. The hats provedso popular that the Census Bureau used them again in the1987 census, and by 1992 they had become almost astandard part of the promotional program. For the 1992census, the Census Bureau ordered 4,000 hats with thecensus logo; 20 of the hats were supplied to cooperatingNAFB member stations, while the rest were given away atconferences and meetings, or on request.

The Census Bureau also ordered 20,000 corn starchbook bags, imprinted with the product map design andtheme used on the census posters, for distribution atmeetings and conferences as well, together with 10,000rolodex cards showing the agriculture census logo andtelephone information contact numbers at the CensusBureau, and 30,000 pencils stamped with slogan ‘‘AmericaCounts on Agriculture.’’

Agribusiness and Agricultural Organizations

The largest single user of agriculture census data, afterthe Federal Government, is the agribusiness sector of theeconomy. About 16 percent of the Nation’s gross domesticproduct (GDP) is involved with agriculture, which meansthat a very large number of business enterprises arenecessarily very interested in what happens down on thefarm. Consequently, the Census Bureau and the agribusi-ness community have a mutual interest in achieving themost complete and accurate census possible and theCensus Bureau has always enlisted the help of concernedbusinesses in publicizing the census and encouragingcooperation. To help in this effort, the Census Bureau madeall of its publicity materials (posters, brochures, censusguides, drop-in advertisements, and so on) available tointerested agribusiness companies and associations.

In May 1991, the Census Bureau sent a letter to majoragricultural organizations in each State, informing themthat Census Bureau personnel were available to makepresentations on the agriculture census, or staff booths ordisplays at business conferences, meetings, and conven-tions. The agency also provided copies of news releases,information kits, and feature articles to farm organizations’publications, and made staff and materials available forinformation booths at conventions and conferences held byvarious agriculture-oriented organizations. AGR staff regu-larly attended the annual conferences of several major


agribusiness and agriculture organizations, including theNational Association of Farm Broadcasters, Future Farm-ers ofAmerica,American Farm Bureau Federation, NationalCattlemen’s Association, National Agri-Marketing Associa-tion, Rural Sociological Society, National Farmers Union,and so on. During 1991-92, the staff also participated inapproximately 20 State meetings, concentrating on thoseStates (primarily in the southeast) that had the lowestresponse rates to the 1987 census.

The member organizations of the Census AdvisoryCommittee on Agriculture Statistics (see chapter 2 formember organizations and their representatives) cooper-ated in the publicity campaign as well. The organizations’representatives on the Committee recorded and/or filmedinterviews and supporting statements, and the groupsthemselves publicized the census in their own advertisingand promotional materials.

POSTCENSUS PUBLICITY AND ASSISTANCE TODATA USERS

News Releases

Following the completion of data collection for thecensus, the Census Bureau conducted a publicity cam-paign designed to inform potential users, and the public atlarge, about the kinds and availability of the data to bepublished.TheAGRprepared a series of 50 news stories—onefor each State—using the 1992 census data, and releasedthe specific story for each State just prior to the publicationof the printed Volume 1 report for that State. Copies of thestories were mailed to the national wire services, nationaland local newspapers (with copies sent to all the majornews outlets in the State for which the census data wereabout to be published), and agriculture-oriented maga-zines. Each article contained summary State-level datafrom the 1992 census and comparative 1987 statistics ontotal number of farms, land in farms, total value of sales,and so on, and graphs of selected items. Copies of thestories also were released on the Census Bureau’s onlineinformation service—CENDATATM—and on the AgriDataNetwork’s online system as well.

A summary story, using the same format as the Statestories but displaying data for the United States, wasprepared and released when the volume 1, United StatesSummary, was published in October 1994.

Professional Meetings

The Census Bureau sent representatives to a variety oftrade shows, professional conferences, agricultural newsmedia conferences, and commodity producers associationmeetings to help publicize the census. Agriculture Divisionand other Census Bureau staff made special efforts toattend meetings of such organizations as the AmericanAgricultural Editor’s Association and the National Associa-tion of Farm Broadcasters to request their help, and thehelp of their associations’ members, in promoting coopera-tion with the census. These meetings enabled CensusBureau officials to hear suggestions for improving thecensus and their own particular efforts and supplementedthe advice and assistance provided through the CensusAdvisory Committee on Agriculture Statistics (see chapter2). AGR personnel staffed information booths, made pre-sentations and speeches, or participated in discussionsand workshops at an average of four or five meetings everymonth for most of the census period. Division representa-tives also regularly participated in meetings of prominentorganizations, such as the National Agri-Marketing Asso-ciation and National Association of State Departments ofAgriculture, and attempted to reach more specialized audi-ences by participating in the 1993 Industry Trade Fair inPuerto Rico and the 1993 annual conference of the Ameri-can Society of Agricultural Engineers.

State Farmer Meetings

Census Bureau staff routinely attend the annual meet-ings of a number of State- and national-level farm organiza-tions—the American Farm Bureau Federation, Future-Farmers of America, National Farmers Union, NationalCattlemen’s Association, and others—making presenta-tions, staffing information booths and exhibits, and so on, topromote the census to the leaders and members of thevarious organizations. The Census Bureau also tried totarget specific organizations and meetings in agriculturallyimportant States (e.g., the California Farm EquipmentExpo). In an average year between the agricultural cen-suses, the Census Bureau participated in 20 to 25 nationalconferences and meetings. During the run up to the 1992census, staff also took part in about 20 meetings ofState-level organizations, concentrating in States (particu-larly in the South) with low response rates to the 1987census.


Chapter 5.

General Information ------------------------------------------------------------------------------ 46Census Mail Operations ------------------------------------------------------------------------- 46General Information --------------------------------------------------------------------------- 46

Initial Mailout ----------------------------------------------------------------------------------- 47General information ------------------------------------------------------------------------- 47The census mailout ------------------------------------------------------------------------- 47

Followup Mailings------------------------------------------------------------------------------ 47General information ------------------------------------------------------------------------- 47First followup—remainder/thank you cards ------------------------------------------------ 47Second followup----------------------------------------------------------------------------- 48Third followup ------------------------------------------------------------------------------- 48Fourth followup ------------------------------------------------------------------------------ 48Fifth followup -------------------------------------------------------------------------------- 48

Undeliverable As Addressed (UAA) ---------------------------------------------------------- 48

Telephone Operations---------------------------------------------------------------------------- 48General Information --------------------------------------------------------------------------- 48

Computer Assisted Telephone Interviewing (CATI) ------------------------------------------ 49Introduction ---------------------------------------------------------------------------------- 49CATI staff training --------------------------------------------------------------------------- 49Field organization and assignments ------------------------------------------------------- 49

Telephone Followup Operations -------------------------------------------------------------- 49Telephone numbers search----------------------------------------------------------------- 49Telephone interviewing procedures -------------------------------------------------------- 50Output files ---------------------------------------------------------------------------------- 50Results --------------------------------------------------------------------------------------- 50

Jeffersonville Telephone Operations ---------------------------------------------------------- 51General information ------------------------------------------------------------------------- 51Secondary source operations -------------------------------------------------------------- 51

Model Drop Survey------------------------------------------------------------------------------- 51

Nonresponsive Survey --------------------------------------------------------------------------- 52General Information --------------------------------------------------------------------------- 52Data Collection--------------------------------------------------------------------------------- 52

Citrus Caretakers--------------------------------------------------------------------------------- 53Background Information ----------------------------------------------------------------------- 53The 1992 Enumeration --------------------------------------------------------------------- 53

Results -------------------------------------------------------------------------------------------- 54

Contents

Page


Data Collection

GENERAL INFORMATION

The Census Bureau employed mail self-enumeration asthe principal data-collection methodology for the 1992agriculture census in the 50 States (Puerto Rico, Guam,and the Virgin Islands of the United States were enumer-ated in separate operations, see chapters 7 and 8 fordetails). The Data Preparation Division (DPD) at the Cen-sus Bureau’s Jeffersonville, IN, facility carried out the bulkof the mail operations involved in the enumeration, with theclose supervision and assistance of theAgriculture Division(AGR). The DPD mailed some 3.55 million agriculturecensus report forms in December 1992, and carried out upto five followup mailings to nonrespondents—the first areminder card requesting early response, mailed in the firstweek of January to all addresses on the census list—betweenthe beginning of January and the end of June 1993.

The vast majority of agricultural operations were enu-merated by mail, but the Census Bureau supplemented themail enumeration with a telephone followup program inwhich telephone interviewers tried to contact selectednonrespondent cases with large estimated annual totalvalue of sales (TVP) of agricultural products and, in coun-ties with unacceptably low overall levels of response (i.e.,less than 75 percent). For the 1992 census, the agencyintroduced computer-assisted telephone interviewing (CATI)as the principal technique employed in the telephonefollowup operations for the agriculture census. The tele-phone followup to large nonrespondent cases began inMarch 1993, while the low-response county telephoneoperation began in May.

The AGR conducted a separate enumeration of citruscaretaker operations1 in the summer and early fall of 1992,in order to obtain data on citrus production in Florida,Texas, and Arizona at the close of the growing season inthose States.

In addition, the data-collection operations for the 1992census included mailouts to addresses for the 1992 Non-response Survey from April through July 1993 (while mailfollowup to the census was still underway), and the 1993Model Drop Survey (used to evaluate the effectiveness ofthe Census Bureau’s classification tree methodology inaccurately classifying addresses on the mail list).

CENSUS MAIL OPERATIONS

General Information

The DPD office in Jeffersonville, IN, had the primaryresponsibility for handling the agriculture census mailings(as it did for the bulk of all the Census Bureau’s maildata-collectionoperations).TheDPDoffice receivedassembledmail packages from the private contractors, prepared adhe-sive address labels using computerized files provided bythe Census Bureau’s headquarters computer facility, appliedthe labels (or, in the case of the reminder cards, printed theaddresses directly onto the cards using DPD equipment) tothe packages, and conducted the mailings for the censusmailout and the followup mailings. The data-collectionmailing for the 1992 Census of Agriculture involved over 8million separate cards or packages in the initial and fivefollowup mailings.

The 1992 agriculture census report-formpackages includeda cover letter asking addressees to respond by February 1,1993. The first followup mailing, carried out in the first weekof January 1993, consisted of a reminder/thank you cardthat was sent to all addresses on the initial census list. Thefour regular followup mailings that followed were con-ducted at 4 to 5 week intervals, beginning in the secondweek of February, and continuing into the first week of June1993. The second, third, and fifth followup mailings allinvolved complete report form packages, each includingthe report form, instruction sheet, cover letter, and returnenvelope. The fourth followup consisted of only a letter(form 92-A01(L5)) asking for a response.

The Census Bureau organized the census mail list bytype of case (i.e., ‘‘must’’ or ‘‘nonmust’’), and by State, innine geographic segments, to distribute the processingworkload. Segment 1 comprised all ‘‘must’’ cases for allStates. The remaining eight segments each contained‘‘nonmust’’ cases for specified States, as follows:

Segment States1 Must cases for all States.2 Arizona, California, Colorado, Idaho,

Nevada, New Mexico, Oregon, Utah,Washington, Wyoming.

3 Texas, Michigan, Minnesota, Montana,North Dakota, South Dakota, Wisconsin.

1Citrus producers, whose groves were covered in the caretakerenumeration, also received census report forms by mail in the Decembermailout so they could report any other agricultural activities.


Segment States5 Connecticut, Delaware, Maine, Maryland,

Massachusetts, New Hampshire, NewJersey, New York, North Carolina,Pennsylvania, Rhode Island, Vermont,Virginia, West Virginia.

6 Alabama, Georgia, Kentucky, SouthCarolina, Tennessee.

7 Arkansas, Louisiana, Mississippi, Missouri,Oklahoma.

8* Florida, Kansas, Indiana, Iowa.9 Alaska, Hawaii, Illinois, Nebraska, Ohio.

*FOSDIC test cases for Iowa were mailed as part of the segment 9mailout.

These segments did not correspond to the agricultureregions established for the design of the census reportforms, or to the official census geographic regions ordivisions used for tabulating data. Rather, they were aprocessing and administrative convenience, with stag-gered response closeout and mailing dates.

Initial Mailout

General information. The 1992 Census of Agriculturedata-collection effort began on December 8, 1992, whentheCensusBureaubeganmailing questionnaires to addresseson the mail list. Between December 8 and 18, question-naires and associated materials were mailed to 3,551,407addresses throughout the 50 States. The DPD office inJeffersonville, IN, conducted the mailing operations for this,and the followup mailouts. The census employed third-class bulk rate postage for most of the mailing packages,using first-class postage only for multiunits, abnormals,‘‘births’’ (i.e., newly identified agricultural operations), Alaskaand Hawaii addresses and, in followup operations, forundeliverableasaddressed (UAA)casesand the reminder/thankyou card sent in January mailing. First class postage alsowas used on the return envelope included in each reportform package.

The census mailout. The initial mailing packages eachcontained the form 92-A01(L1) cover letter requestingresponse, the appropriate report form, the form 92-A01(I)information sheet with instructions for completing the report

form (the form 92-A02(I) was used for addresses in Hawaii),any special instruction sheet (required for known feedlots,nurseries, certain animal specialties, and so on), and areturn envelope.

The quantities of each general type of form sent in theinitial mailing were as follows:

Table 5.1. 1992 Census of Agriculture Mailout:December 8-18, 1992

Form type Quantity

Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3,551,407Nonsample (forms 92-A0101 to -A0111) . . . . . . . . . . . . 2,131,699Sample (total (excluding multiunits and abnormals)) . 1,000,442General sample (forms 92-A0201 to -A0213). . . . . . 865,226Must cases (forms 92-A0301 to -A0311). . . . . . . . . . 135,216

Multiunit (form 92-A0301 to -A0311, form 92-A0201and -A0212) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5,727Abnormal (forms 92-A0301 to -A0311, 92-A0201,and 92-A0213) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,899Screener (forms 92-A0401 to -A0411) . . . . . . . . . . . . . . 411,640

Followup Mailings

General information. Private contractors printed the reportforms and other census mailout materials, assembled themailing packages for the initial and followup mailingssubject to quality control inspection by Census Bureaupersonnel on the premises, and delivered them to the DPDfacility at Jeffersonville, IN, where the individual packageswere labeled and mailed on a flow basis. (See chapter 3 forthe details of the label printing and assembly operations.)The DPD staff used similar procedures for each mailout.Table 5.2 summarizes followup mailings for the 1992census:

First followup—reminder/thank you cards. The DPDmailed form 92-A01(L2) reminder/thank you cards to alladdresses (except abnormals) on the census mailing list.Originally, all the address labels for the reminder/thank youcards were to be printed using the new inkjet printer atJeffersonville, but only 1 million labels could be preparedon this equipment before the scheduled mailing date.Accordingly, the DPD staff brought the older Printronixprinters online to complete the label printing before thedeadline. Address labels were applied to over 3.5 million

Table 5.2. Summary of 1992 Census of Agriculture Mail Followup

Followup Form content Dates of mailing Total NonsampleGeneralsample Must Screener

First . . . . . . . Reminder/thank you card Jan.6-8, 1993 3,543,781Second . . . . Report form Feb. 11-24, 1993 1,521,702 858,117 398,505 66,957 198,123Third. . . . . . . Report form Mar. 18-29, 1993 1,102,924 610,123 295,772 48,782 148,247Fourth . . . . . Letter Apr. 23-May 7, 1993 856,191Fifth . . . . . . . Report form May 27-June 3,1993 722,874 401,852 191,766 25,841 103,415


cards using the mechanical labeling equipment used forlabeling the census mailing packages. The requestedresponse date for the census was February 1; the cardswere mailed on a flow basis as they were labeled, fromJanuary 6 through January 8, 1993.

Second followup. By early February 1993, response tothe census was over 50 percent, and the Census Bureauwas preparing for the first followup to nonrespondentaddresses. Cut-off dates for response for each geographicsegment were established, ranging from February 8 to the21st. As each date was reached, updated address listsconsisting of all addresses within a particular segment forwhich no response had been checked in at Jeffersonville,were generated and address labels printed. Mail followuppackages for each segment had been prepared in advanceby the printing contractors, each consisting of the appro-priate report form (i.e., nonsample, certainty, and must, bysegment), the cover letter form 92A01(L3), requestingprompt response, any special instruction sheets, and areturn envelope. The address labels were mechanicallyapplied through the open ‘‘windows’’ of the outgoing enve-lopes and the 1,521,702 packages were mailed on a flowbasis, by geographic segment, between February 11 andthe 24th. (See appendix D for details of the followupmailings.)

Third followup. The Census Bureau’s study of previousmail-enumeration operations, particularly response to fol-lowup mailings, revealed that report form mail-followupachieved better and earlier response than letters, so the1992 mail-followup operation emphasized mailing com-plete enumeration packages to nonrespondent operations.The third mail followup used a complete census report-form package, with the original cover letter replaced by theform letter 92-A01(L4) requesting prompt response, remind-ing the addressee that response was required by law, thatinformation provided would be kept confidential, and givingthe toll-free telephone assistance number. Closeout datesfor response varied by segment from March 15 throughMarch 24. By the third week of March, overall responsehad reached 68 percent (approximately 2 percent belowthe 1987 rate at the same date). The address labels wereprinted and the preassembled report-form packages mailedon a flow basis from March 18 through March 29. The thirdfollowup mailings totaled 1,102,924 packages.

Fourth followup. Closeout dates for the fourth followupranged from April 21 through May 4, with mailout on a flowbasis fromApril 23 through May 7. The fourth mail-followupwas the only ‘‘letter’’ followup, and used the form 92-A01(L5)letter to ask addressees to respond to the census. The L5letter repeated the information given in the L4 used in thethird followup, with the applicable sections of Title 13, U. S.Code printed on the reverse side of the letter. A total of856,191 letters were mailed to nonrespondent addresses.The final closeout dates for response to the fourth followupranged from May 24 to May 28, with two segments closing

out each day. By May 29, the response rate to the 1992census had reached 78.1 percent. This was about 2percent lower than at the comparable point in the 1987census.

Fifth followup. The fifth mail-followup was a report formmail operation, with each mailing package consisting of theappropriate report form and information sheet, a coverletter, form 92-A01(L6), requesting response to the censusand noted that this was a final notice. This letter assuredconfidentiality of the information supplied and included (onthe back of the letter) excerpts from Title 13, U.S. Code,relating to the authority for collecting the data, the manda-tory response provisions of the law, and the guarantee ofconfidentiality.

The DPD staff in Jeffersonville prepared adhesive addresslabels for the report form packages using the censuscheck-in file for the segments as each was closed out. Thelabels were applied by machine to the mailing packages,and mailout, by segment, began on May 27 and finished onJune 3. A total of 722,874 census packages were mailed toaddresses still on the nonrespondent list.

Undeliverable As Addressed (UAA)

Census packages for which no physical place matchingthe address could be found were returned by the PostalService as UAA. The DPD staff at Jeffersonville identifiedUAA cases during the receipt and check-in phase of theprocessing (see chapter 6 for details) and prepared specialpackages for remailing. These packages contained theappropriate report form, instruction sheet and return enve-lope, together with a form letter specifically written for UAAcases. The DPD remailed—

x First-time UAA’s believed to represent larger agriculturaloperations.

x UAA cases with address changes.

x ‘‘Deceased’’ UAA’s (the report forms were mailed to the‘‘Estate of ...’’).

The UAA mail followup used only two closeout dates,February 4 and March 1, 1993. The Jeffersonville officereceived 278,424 UAA cases from the Postal Service, ofwhich 33,393 were remailed (UAA cases were remailedfirst class).

TELEPHONE OPERATIONS

General Information

Telephone operations for the 1992 Census of Agricultureencompassed an ‘‘incoming call’’ activity to assist respon-dents and answer general inquiries about the census, anda data-collection and followup operation. The DPD officetelephone operations staff in Jeffersonville, IN, handledincoming calls, as well as calls to secondary sources for


selected nonrespondent cases. The data-collection andfollowup operation introduced computer-assisted telephoneinterviewing (CATI) as the principal telephone followupactivity to the agriculture census.

Computer Assisted Telephone Interviewing(CATI)

Introduction. CATI collected data through respondenttelephone interviews, and the interviewers keyed the datasupplied directly to the data file. The data then wereprocessed electronically, eliminating paper report and pro-cessing forms. Prior to the 1992 agriculture census, TheCensus Bureau had used CATI extensively in its demo-graphic surveys program, particularly the Current Popula-tion Survey (CPS). For the 1992 Census of Agriculture,delinquent large farms once againwere referred for telephone-followup, but this time CATI was used for all the referrals,as well as for followup for low response counties (i.e., thosewith response rates below 75 percent at specified dates),and for the 1992 Nonresponse Survey, and for casesrelated to the June Agricultural Survey (JAS).

CATI staff training. Field Division personnel trained CATIenumeration staff for the agriculture census CATI followupin January and February 1993. The training included anintroduction to the agriculture census, using agriculturecensus reference materials, farm operations, concepts andprocedures, ‘‘walk-through’’ training interviews, and a finalreview. Each operator was given a binder containingrelevant reference materials (e.g., a glossary of agriculturecensus terms, the 1992 report form guide, a table of unitsof measures with a conversion chart, crop yield and priceguidelines, and so on), together with written procedures toconsult once operations began. The CATI staff beganinterviewing large-farm delinquent cases on February 22,1993, with referrals for the nonresponse survey added tothe workload beginning in April, and those for low-responsecounties in May.

Field organization and assignments. The CATI opera-tion used two dedicated offices, one in Hagerstown, MD,and the second in Tucson, AZ. Cases selected for tele-phone followup to the nonrespondent survey, and thelow-response county followup were assigned to the CATIoffices on a flow basis, with Hagerstown handling casesfrom States in the eastern and central time zones, while theTucson office was responsible for those in mountain andpacific time zones. Each office operated three shifts—morning,afternoon, and evening—each approximately 5 hours long.Each CATI interviewer used a work station consisting of amonitor and a keyboard, both networked to the individualoffice’s computer with the active CATI files, and telephonesystems with headphones. When operators began work onindividual records, the CATI program displayed file infor-mation and the interview questions on the monitor, and theinterviewer keyed responses directly to the respondent’srecord in the CATI file.

The Hagerstown facility employed 45 operators andworkstations on the first shift, 50 on the second, and 20 onthe third shift for agriculture-followup operations during thefirst 5 days of each month, reducing this to 30 operators forthe first and second shifts, and eliminating the third shiftduring Current Population Survey (CPS) operations (thefirst four full work days of the week of the 19th of eachcalendar month), then using 55 operators and stations foreach shift during the remainder of each month. The Tucsonoffice used 85 operators and stations for each shift exceptduring CPS week, when the first and second shifts werereduced to 20 operators each, and the third shift to 10.

The cases referred to the CATI followup operationincluded—

x Large nonrespondent farms (i.e., farms with an esti-mated annual value of sales of agricultural products of$100,000 or more or with 1,000 acres or more (depend-ing of the specific State)).

x Nonrespondent cases that matched to the JAS farm list.

x A sample of the general nonrespondent list (used for the1992 Nonrespondent Survey).

x After May 1993, a sample of nonrespondent addressesin those counties with response rates below 75 percent.

Initial planning for the telephone-followup projected atotal workload of up to about 167,000 delinquent large-farmcases. The actual number of these cases submitted to theCATI-followup operation between February and Septem-ber 1993 was approximately 152,000. This total included—

x Low-response county referrals.Approximately 9,700 cases,in 179 low-response counties scattered across 31 Stateswere referred for CATI followup in eight ‘‘waves,’’ begin-ning in May and ending in August 1993.

x Any nonrespondent case with expected sales of $250,000or more, or with expected size of 3,000 acres or more.

x 1992 Nonrespondent Survey. A total of 14,271 caseswere referred for CATI enumeration.

x Any case with an address that matched to or originatedin the JAS farm list (important for estimating farms not onthe mail list for the coverage evaluation).

Telephone Followup Operations

Telephone numbers search. The Census Bureau’s Eco-nomic Programming Division (EPD—now the EconomicStatistical Methods and Programming Division (ESMPD))prepared 49 computerized State files of delinquent large-farm cases selected for telephone followup for installationon the systems serving the Hagerstown and Tucson facili-ties (the DPD staff in Jeffersonville handled cases in Hawaiibecause of the unique nature of much of agriculturalproduction there). The Census Bureau’s records originallycreated for many of these large operations included tele-phone numbers whenever possible (i.e., whenever the


administrative records used to compile the lists included atelephone number for an address), but many did not, soField Division staff conducted a computerized review ofeach State file to identify specific records that did not havea usable telephone number. This process created separateState files of ‘‘bad’’ telephone number cases, which wereinstalled on the computer system at the responsible CATIfacility to enable the Directory Assistance subunit (DAsubunit—a number of CATI operators on each shift wereassigned to this subunit) to call directory assistance fortelephone numbers. As numbers were identified, the infor-mation was added to the State files and subsequentlysubmitted for CATI enumeration.2

The EPD assembled each State file on a Thursday, andDA subunit searches began on the following Monday,lasting for approximately one work week. Telephone num-bers research for the 1992 Nonresponse Survey andlow-response county referrals (beginning in April and Mayrespectively) was handled in the same way, although casesfor several States were organized into a composite, ratherthan an individual file, for installation on the CATI system.

Telephone interviewing procedures. When each shiftbegan work at the CATI facilities, interviewers ‘‘logged on’’the CATI system—that is, turned on their individual workstations and keyed ‘‘P’’ (for ‘‘proceed), which allowedaccess to the CATI file. The CATI system automaticallyassigned records for followup to available stations thatwere not already actively working on a file.

The same basic procedures were followed for all casesreferred for CATI-telephone followup. When a case wasassigned to an interviewer, the work-station monitor dis-played identification information for the specific case, includ-ing a ‘‘label line’’ that might include the name and addressof the operator, the name of the operation, and/or thecensus farm number (CFN) for the operation. The pagealso showed available background information, including aspecified contact person when available, any previouscontacts made with the particular operation, and informa-tion on the operation from previous censuses. After review-ing the historical and background information, the inter-viewer entered the identification information and name ofthe contact person from the label line, and the telephonenumber, then attempted to call the number.

Once the interviewer contacted an operation; he or shecould interview any knowledgeable family member whowas at least 14 years old, an employee of the farmoperator, or an accountant if the operator permitted. Ideally,the interviewer spoke directly to the operator. The inter-viewer asked the person if he or she was willing to respond

to the census by telephone. If the answer was yes, theinterviewer confirmed the identity of the respondent andthe agricultural operation involved and checked whetherthe respondent had received any census report formsunder a different name or CFN. (If the respondent hadreceived forms under another name or CFN, the inter-viewer entered these into the record at this time formatching to the nonrespondent file.) The interviewer enteredthe respondent’s name and began the interview proper.The CATI system displayed each question and the inter-viewer entered the appropriate codes or responses (e.g.,crops and livestock/animal specialties were assigned numericcodes (1 = field corn for grain, 13 = soybeans, and so on;quantities reported, such as ‘‘770 acres’’ were entered asnumerics in response to specified questions, i.e., ‘‘770’’)).

After completing the interview, the interviewer keyed ‘‘P’’once again, displaying a section in the record (the ‘‘i-notes’’) for any comments or notes about anything thatmight be considered atypical for a farm operation. Forexample, for a small operation that grows only roses, theinterviewer might add a note—‘‘grows roses only.’’ Theinterviewer then entered the work ‘‘COMPLETE’’, his or herown interviewer identification code, and the date of theinterview.

Output files. The results of the CATI operations at theHagerstown and Tucson facilities were transmitted to theEPD at the Suitland, MD, headquarters by telephonedatalink, in the form of four output files for each State:

1. Answer file: Interviewer coding including respondents’data and interviewer remarks for resolved cases.

2. F7 file: Interviewer remarks made during the interview-ing process for resolved cases.

3. History of access file: ‘‘Snapshot’’ of installed casesshowing each time accessed.

4. Case master file: System management informationfor each installed case.

The ‘‘Answer’’ and ‘‘F7’’ files for each State were trans-mitted nightly after the last shift for each facility closeddown telephone operations. At the closeout for each State(i.e., when all telephone referral cases for a specific Statehad been resolved) a cumulative version of all four fileswas sent to the EPD. The EPD assured the receipt of thesefiles and subsequent processing for merging them into the1992 Census of Agriculture data file.

Results. Altogether, some 175,900 cases were referred tothe CATI units for telephone followup, the bulk of which(over 152,000) represented delinquent large farms. (Anadditional 7,897 cases composing the 1993 Model DropSurvey also were referred to the CATI staff for telephoneinterviewing in September 1993.) The CATI units contactedand enumerated 57,708 large farm cases by telephone.(Approximately 29,300 cases originally referred for tele-phone followup responded by mail during the followup

2During the early stages of the CATI large-farm followup operation,evencases that lacked usuable telephone numbers after DA subunitsearch were merged with the general State file so that they would beidentified in the output files. The Census Bureau changed this procedurefollowing the first‘‘wave’’ of large-farm referrals by assigning appropriatecoding identifying them as lacking telephone numbers; this dropped thecases from the ‘‘calling queue’’ while still identifying them in the outputfiles.


operation.) Some 15,000 large-farm cases referred to theCATI operation were found to be out of scope while about5,000 were duplicates and were dropped from the file.

The low-response county operation ultimately involved atotal of over 9,700 individual cases, but the objective ofCATI enumeration in this operation was limited to raisingoverall county response to 75 percent; when this wasachieved, further telephone-followup activities ended. Finalresults for the low-response county file showed 4,358cases resolved, either by enumeration or identification asout of scope (excluding mail returns received while theCATI operation was underway).

The 1992 Nonresponse Survey CATI followup achieveda 94.5 percent response rate (13,486 cases in all).

Jeffersonville Telephone Operations

General information. The Data Preparation Division (DPD)office at Jeffersonville, IN, included a telephone interview-ing staff and facilities that had handled most of the CensusBureau’s telephone enumeration and survey activities priorto the adoption of the CATI systems and the establishmentof the dedicated CATI offices in Hagerstown, MD, andTucson, AZ. With the advent of the CATI offices, the bulk ofthese telephone canvassing and survey operations weretransferred to those offices, and the Jeffersonville staff wasassigned to handle calls for assistance (see chapter 6 fordetails), consult secondary sources for problem cases andrefusals, and followup correspondence and other specialcases that required access to the census data file andprocessing operation.

Secondary sourceoperations. Thepurposeof thesecondary-source operation was to obtain information to determinethe farm status of all cases that could not be completed byrespondent contact, including refusals and ‘‘no telephonenumber listed’’ nonrespondents, as well as other noncon-tact cases. All followup cases the CATI telephone unitscould not complete as either inscope or out of scope weretransferred to the EPD for assignment of secondary-sourceflags based on the presence in the individual records ofvarious types of administrative data. For unsatisfied CATIrecords having associated NASS data, and with an indica-tor that the NASS data had been updated in 1992, thein-scope census record was generated by computer with-out any telephone contact with a secondary source. Tele-phone calls to secondary sources were made for all otherunsatisfied CATI records to verify that they were agricul-tural operations as of December 31, 1992. After confirmingfarming operations for these cases the Census Bureaucreated an in-scope report from available administrativedata. The administrative records used included NASSrecords that lacked a 1992 update indicator, and recordswithout NASS data but with an historic record (e.g., 1987census) that had not been created from a 1982 censusrecord.

The telephone control unit sorted the incoming files byState and county groups, using the individual records’CFN’s and assigned records by county groups to tele-phone operators, who contacted the appropriate U.S.Department of Agriculture (USDA) Agricultural Stabilizationand Conservation Service (ASCS) county office to try toobtain the necessary information.3

Six telephone clerks worked each shift, calling ASCSoffices in this operation. When the ASCS offices were ableto provide information, cases were classified as inscope orout of scope. The latter were deleted from the file. Forin-scope cases with NASS data with no 1992 update dataand nonreplicated historic data cases, the clerks enteredthe appropriate flag code in the ‘‘Census Use Only’’ box onthe front of a labeled blank form 92-A0214 (general)questionnaire, and routed the case to the data keying unit.For cases with no replicated historical data the clerks triedto obtain additional data from the ASCS office contacted,summarizing the data on Form 92-A417, General SourceWorksheets, then transcribing the totals to a form 92-A0214report form.

A total of 25,966 cases were processed by the Jeffer-sonville telephone staff for secondary source contact;13,083 of these cases proved to be inscope, and theimputed data were incorporated into the census data file.The remaining 12,883 cases were dropped from the file asout of scope.

MODEL DROP SURVEY

In September and October 1993, the Census Bureaucarried out a Model Drop Evaluation Survey to evaluate theefficiency of the agency’s Classification and RegressionTree (CART) methodology used to identify and deleteaddresses least likely to represent farms from the 1992agriculture census mail list. (See chapter 10 for moreinformation on the actual evaluation.) The national sampleframe for the Model Drop Survey comprised the 229,180addresses deleted from the census mail list by the CARTprocedures and by analysts’ adjustments, organized in fivestrata (1 through 5) based on calculated probability ofmeeting the farm definition. (See chapter 3 for moreinformation on the statistical modeling used in preparingthe census mail list.)

3The Census Bureau and the ASCS tested a different method ofcollecting secondary-source data—mailing lists of names to ASCS offices.This was tried for secondary-source cases in Oregon, but proved unsuc-cessful for a variety of reasons, including slow delivery and nondelivery ofthe mailed lists, and slow response by some offices. In general, theagency found that telephoning the individual offices should be the firstapproach, with facsimile transmission or mailing of lists resorted to only ifthe ASCS office(s) could not or did not respond to telephone inquiries.


The Suitland headquarters staff drew systematic samplesfrom each of the defined strata, as follows:

Stratum Total addresses Sample selected

Total . . . . . . . . . . . . . . 229,180 7,8971 . . . . . . . . . . . . . . . . . . . 48,082 1,6572 . . . . . . . . . . . . . . . . . . . 60,261 2,0773 . . . . . . . . . . . . . . . . . . . 57,668 1,9874 . . . . . . . . . . . . . . . . . . . 33,558 1,1565 . . . . . . . . . . . . . . . . . . . 29,611 1,020

The EPD prepared a computerized file of the totalstratified sample of 7,897 records for referral to the CATIfacilities for telephone numbers research and interviewing.Some 5,421 sample cases were contacted and resolved byCATI interview during September 1993; 2,476 cases forwhich telephone numbers could not be found or that theCATI staff could not contact, were mailed survey packagescontaining a 92-A414, Screener Form, a cover letter explain-ing the survey and asking for prompt response, and areturn envelope, and were processed in DPD’s coverageevaluation unit. Mail data collection continued to the end ofthe 1993.

A total of 5,892 responses (82.2 percent (excluding UAAcases)) were obtained by the survey, of which 5,526 wereclassified as farm or nonfarm addresses. An additional 729(mailed) cases were UAA, and 1,276 addresses werenonrespondent. Overall, the Model Drop Survey achieveda 33.1 percent mail response rate; the CATI followupattained 96.8 percent response from cases submitted fortelephone followup.

NONRESPONSE SURVEY

General Information

The Census Bureau surveyed a sample of agriculturecensus nonrespondents to estimate the proportion of non-respondents to the census in each State that met thecensus farm definition. These estimates, and the finalnumber of nonrespondents for each State, were used toestimate the number of census nonrespondents that actu-ally were farm operations in each county. The CensusBureau used a whole-farm imputation procedure to ‘‘inflate’’the data from respondent farms to represent ‘‘all farms,’’including nonrespondents, in the statistical publications.(See volume 1, Geographic Areas Series, appendix C, fordetails of the statistical estimation methodology and thereliability and coverage estimates for each State.) Thesurvey sample excluded all must, abnormal, and Alaskaaddresses. The specific maximum sales and acreage limitsvaried from State to State—the total value of agriculturalproducts sold (TVP) from $100,000 to $200,000, and theacreage from 1,000 to 3,000 acres.

The Census Bureau selected the 1992 NonresponseSurvey from the agriculture census check-in file. Theagency stratified the eligible address file based on expected

value of sales, information from previous censuses, andreport form type. The strata codes assigned are shownbelow.

The Census Bureau staff used a single-stage, system-atic sample of eligible records for each State to select atotal of 18,569 addresses for the survey. Individual State’snonrespondent lists were used as the sample frame, withselection intervals calculated to produce samples largeenough to produce reliable estimates for each State.Samples were selected at five points during processing,depending on the census data-collection closeout dates forthe States involved. The selection dates by States were asfollows:

Group* Selection Date States

01 Apr. 12, 1993 Delaware, Indiana12 Apr. 26, 1993 Iowa, Missouri, Oregon, Wash-

ington, Wisconsin03 May 10, 1993 Connecticut, Maine, Maryland,

Massachusetts,Michigan,NewHampshire,Ohio,Rhode Island,Vermont, Virginia, WestVirginia

13 May 10, 1993 Illinois, Kansas, Wyoming04 June 28, 1993 Kentucky,NewJersey,NewYork,

Pennsylvania, Tennessee14 June 28, 1993 California, Colorado, Idaho,

Minnesota,Montana,Nebraska,Nevada, Utah

05 July 12, 1993 Florida, Georgia, Mississippi,North Carolina, SouthCarolina, Texas

15 July 12, 1993 Alabama, Arkansas, Arizona,Louisiana, NewMexico, NorthDakota,Oklahoma,SouthDakota

26 July 12, 1993 Hawaii

*The first digit of the group number identified the survey center forthose States, and the second digit the selection ‘‘wave’’. The surveycenters were: ‘‘0’’ - Hagerstown, MD; ‘‘1’’ Tucson, AZ; and ‘‘2’’ - Jefferson-ville, IN.

Data Collection

The 1992 Nonresponse Survey was conducted usingthe CATI system. As each wave of sample selection was

Stratum Description

0 All cases ineligible for nonresponse weighting and all 100-percent followup records for each State.

1 All eligible screener cases for each State.2 All eligible nonscreener cases with an estimated 1992 TVP of

less than $2,500.3 All eligible nonscreener cases with an estimated 1992 TVP

between $2,500 and $9,999.*4 All eligible nonscreener cases with a 1987 census in-scope

source combination code and an estimated 1992 TVP of$10,000 or more.*

5 All eligible nonscreener cases with no 1987 inscope sourcecombination code, but with an estimated 1992 TVP of $10,000or more.

*Since the telephone cutoff levels varied by State, the mail-size codesfor cases included in strata 4 and 5 also varied by State.


completed, the sample names and addresses were referredto the telephone offices in Hagerstown, MD, and Tucson,AZ, for identification of telephone numbers. The telephonesearch staffs found current telephone numbers for 14,271cases in the sample, and these were referred for CATIenumeration. As the telephone numbers research pro-ceeded, the staffs inserted computer flags for cases forwhich no telephone numbers could be located, and twicegenerated output files (one during telephone operations(containing 4,563 records) and a second (with 1,015 records)after calling was completed) of names andmailing addressesfor which directory assistance could find no telephonenumbers or the Census Bureau was otherwise unable tocontact. These files were sent to the EPD at headquarters,which used it to generate mailing address labels. TheCensus Bureau used certified mail to send the 92-A0414report forms (a nonregion-specific sample form used forboth the Nonresponse Survey and for mailing to certaincorrespondence cases) to addresses for which no tele-phone numbers could be found.

Telephone operations (see above) began in April andcontinued through August 1993. The telephone followupachieved a 95.6 percent response rate for cases with validtelephone numbers, contacting 11,032 survey sample cases.Mail followup was considerably less successful, with a finalresponse rate achieved of just 47.1 percent.

CITRUS CARETAKERS

Background Information

A citrus caretaker is an organization or individual caringfor, supervising, or managing citrus groves for the groveowners. Individual caretakers’ activities varied consider-ably in scope, from doing only selected grove work tohandling the entire care and management of the groves(although many did not do harvesting). The Census Bureauintroduced mail enumeration in the 50 States in the 1969agriculture census, but continued using field enumerationfor citrus caretakers in several States—initially Florida, butlater including Texas and Arizona. The field interviewprocedure eliminated the difficulty of identifying and enu-merating absentee owners who, in any event, frequentlyemployed caretakers to manage their groves and did nothave the information needed to complete the report form.

The first separate field operation to collect data fromcitrus caretakers was undertaken in the 1964 agriculturecensus in Florida, where caretakers received special atten-tion in an effort to improve coverage of the citrus industry.The field interview staff visited caretakers and completed areport form for each that listed the names, addresses, andacres owned by each grove owner employing the care-taker. The Census Bureau staff then matched the owners’names and addresses to the census respondent file toeliminate duplicate reports. Direct canvassing of caretak-ers continued in the censuses that follows, and expandedto cover caretakers in Texas in the 1974 and later enumera-tions, and in Arizona from 1978.

The 1992 citrus caretakers’ enumeration covered allthree States included in previous censuses—Florida andTexas by field interview and telephone interview inArizona—where their employment by grove owners continued to bewidespread, and they remained the most reliable source ofinformation.

The 1992 Enumeration. The Census Bureau designedand printed a special questionnaire (i.e., form 92-A0215)for enumerating the caretaker operations. The A0215 wasan 8-1/2″ x 13″ 8-page booklet, printed in black ink on whitestock, with salmon shading. Content was considerablyabbreviated (compared to the regular report forms) andcontained only those items applicable to citrus caretakeroperations. There was one section each on citrus produc-tion and nursery and greenhouse crops, and a singlesection on ‘‘other crops, livestock, or poultry,’’ while theremainder of the form incorporated most of the noncropinquiries from the sample report form (i.e., sections onirrigation, farm labor, production expenses, use of chemi-cals and fertilizers, machinery and equipment, currentvalue of land and buildings, income from farm-relatedsources, products sold directly for human consumption,injuries or deaths, organization, corporate structure, andcharacteristics of the operator (of the caretaker operation)).

The AGR compiled a list of citrus caretakers fromvarious administrative records and mailed the censusreport forms, together with a cover letter, and an instructionsheet, to identified citrus caretaker operations in Texas inMay 1992, in Florida in early September 1992, and inArizona in October. The caretakers were asked to look overthe report form and complete it if possible, and then to holdit until an enumerator visited or telephoned. Caretakers inTexas and Florida were visited by Census Bureau inter-viewers, while those in Arizona were telephoned and thedata transcribed to a report form(s) by the telephoneinterviewer. The staggered schedule of mailings and enu-merations were intended to contact the caretakers whentheir workloads were lightest and information from the1991-92 harvest season would be available.

A 92-A0215 report form was completed for every care-taker that had any citrus operations in 1991-92, and eachcaretaker enumerated was assigned a unique ‘‘caretakernumber.’’ In cases where a caretaker was responsible forcitrus operations in more than one county, the countycontaining the most citrus acreage was designated the‘‘principal’’ county of operations. When caretakers hadsignificant citrus operations in more than one county, theCensus Bureau asked that a report form be completed foreach county with 500 acres or more of citrus.

The interviewers not only obtained at least one com-pleted report form for each caretaker, but also obtainedlists of the names and addresses of grove owners, acres ingrove and county, and grove location, so that the CensusBureau could ensure that duplicate reports from the groveowners were not incorporated into the census data file. Thecaretakers were asked to inform their grove owners thatthey had provided citrus production data to the Census


Bureau, and supplied the owners with their caretaker’snumbers. (The Census Bureau matched the names andaddresses of grove owners listed in the caretaker enumera-tion against the census mail lists during data processing.)Caretakers also could request the Census Bureau to mailthem the special instruction sheet Form 92-A31(L), for useby grove owners who also received a 1992 agriculturecensus report form. The A31(L) instructed owners to writein their caretaker’s ‘‘caretaker number’’ on the report form,and to supply any additional information requested onnoncitrus operations.

The number of citrus caretakers enumerated, number ofgrove owners they served, and the approximate acreage ofcitrus production in their operations, by State for 1992 and1987, were as follows:

State CaretakersGroveowners

Citrusacreage

1992 1987 1992 1987 1992 1987

Total . . 61 92 2,465 3,975 171,300 196,500Arizona . 5 7 65 175 7,300 12,000Florida . . 44 65 2,300 3,000 150,00 0 170,000Texas . . . 12 20 100 800 14,000 14,500

Citrus operations not associated with caretakers, both inthe States specifically covered by the caretaker enumera-tion, and in other States (e.g., California, Hawaii) wereenumerated in the regular census data-collection effort thatbegan in December 1992.

RESULTS

The 1992 Census of Agriculture achieved an overallresponse rate of 84.5 percent—about 1.3 percent belowthe final response rate for the 1987 census—obtainingresponses from approximately 3,000,940 addresses, froma total census mail file of 3,551,407. The UAA casesaccounted for another 33,983 cases. The CATI and sec-ondary source telephone data-collection operations col-lected data for 76,079 additional farms.

The Census Bureau published statistics for 1,925,300agricultural operations that met the census definition of afarm. Major summary results of the census data-collectionoperation, with comparisons to selected previous cen-suses, are given in table 5.3.

Table 5.3. Summary Census Counts

1992 1982 1974

Total number of farms. . . . 1,925,300 2,240,976 2,314,013Land in farms (acres) . . . 945,531,506 986,796,579 1,017,030,357Estimated value of landand buildings per farm . $357,056 $345,869 $147,838Total value of sales ofagricultural products . . . $162,608,334 $131,900,223 $81,526,126


Chapter 6.

Introduction --------------------------------------------------------------------------------------- 57

Preparatory Processing -------------------------------------------------------------------------- 57General Information --------------------------------------------------------------------------- 57

Receipt and Check-In ------------------------------------------------------------------------- 59Receipt and mechanical sort --------------------------------------------------------------- 59Remove contents and sort------------------------------------------------------------------ 60Quality control ------------------------------------------------------------------------------- 60

Tracking System ------------------------------------------------------------------------------- 60

Correspondence ------------------------------------------------------------------------------- 61General information ------------------------------------------------------------------------- 61Interactive processing subunit (readers/keyers) ------------------------------------------- 61Quality control ------------------------------------------------------------------------------- 62Suspense file -------------------------------------------------------------------------------- 62

Telephone Assistance Unit -------------------------------------------------------------------- 62Telephone operations ----------------------------------------------------------------------- 62Quality control ------------------------------------------------------------------------------- 63

Special Case and ‘‘2+’’ Case Processing ---------------------------------------------------- 63Special cases-------------------------------------------------------------------------------- 63‘‘2+’’ cases ----------------------------------------------------------------------------------- 64Quality control ------------------------------------------------------------------------------- 65

Large Farms Coverage Unit ------------------------------------------------------------------ 65General information ------------------------------------------------------------------------- 65Multiunits------------------------------------------------------------------------------------- 65Abnormals ----------------------------------------------------------------------------------- 66

Data Entry-------------------------------------------------------------------------------------- 66General information ------------------------------------------------------------------------- 66Batch for data keying ----------------------------------------------------------------------- 66Data keying operations --------------------------------------------------------------------- 67Quality control ------------------------------------------------------------------------------- 68FOSDIC data keying unit ------------------------------------------------------------------- 69

Computer Processing ---------------------------------------------------------------------------- 69General Information --------------------------------------------------------------------------- 69Format------------------------------------------------------------------------------------------ 70

Computer Edit---------------------------------------------------------------------------------- 70General information ------------------------------------------------------------------------- 70Computer edit and imputation -------------------------------------------------------------- 70

Failed Edit Review----------------------------------------------------------------------------- 71General information ------------------------------------------------------------------------- 71Interactive edit ------------------------------------------------------------------------------- 71Computer out-of-scope processing--------------------------------------------------------- 71Format reject processing ------------------------------------------------------------------- 72

Contents

Page


Statistical Estimation -------------------------------------------------------------------------- 72General Information ------------------------------------------------------------------------- 72Nonresponse estimation -------------------------------------------------------------------- 72Sample estimation -------------------------------------------------------------------------- 72

Post-Edit Correction Processing -------------------------------------------------------------- 73General Information ------------------------------------------------------------------------- 73Duplication review--------------------------------------------------------------------------- 73Analytical review and data correction ------------------------------------------------------ 74Tabulation for counties, States, and the United States ----------------------------------- 74Disclosure analysis-------------------------------------------------------------------------- 75Table review and preparation--------------------------------------------------------------- 75


Page

Data Processing

INTRODUCTION

The data processing operations for the 1992 Census ofAgriculture can be very approximately divided into a pre-paratory phase and computer processing. The formeractually made extensive use of automated equipment andcomputers, but was primarily concerned with the receipt,sorting, and handling of census report forms, and the entryof the data from those forms into the census data file. Thecomputer processing consisted of editing the data file,imputation of data for nonresponse, and tabulation.

The bulk of the preparation processing was done at theCensus Bureau’s Data Preparation Division (DPD) office inJeffersonville, IN, while the computer processing of thedata was done interactively, using minicomputers at Suit-land and the Census Bureau’s Charlotte, NC, facility.1 Thevarious activities at Jeffersonville made extensive use ofinteractive systems linked electronically to the Charlotteoffice. The minicomputers had substantial data storage andprocessing capacity, and their use enabled the DPD staff todispense with paper printouts, using data displayed directlyon the individual work stations’ monitor screens for reviewand edit purposes.

The Census Bureau’s computer facility at Charlotte, NC,used minicomputer systems to format, edit, and tabulatethe data received from Jeffersonville. Individual censusrecords that failed the computer edit were electronicallyreferred to the Jeffersonville unit and displayed on terminalscreens there for review and correction. The DPD staffcould enter any corrections required directly to the fileusing the interactive systems.

The Census Bureau conducted 1992 economic andagricultural censuses concurrently and integrated many ofthe processing activities for the censuses (e.g., check-in,correspondence), but the data from the economic andagriculture census report forms were keyed separately andeach of the census operations employed specialized com-puterized edit and tabulation programs.

PREPARATORY PROCESSING

General Information

The DPD mailed approximately 3.55 million agriculturecensus report forms in December 1992 and conducted an

extensive mail and telephone followup over the succeeding7 months. Responses to the census began arriving at theJeffersonville facility almost immediately after the mailout.By the end of January 1993, response had reached over34.3 percent (about 1.2 million responses). As the reportforms arrived at the Jeffersonville office, they were pro-cessed to update the census respondent lists and toprepare for data tabulation. This involved—

x Receiving and checking in the report forms.

x Sorting the report forms and removing the contents fromthe envelopes.

x Evaluating and responding to census-related correspon-dence.

x Reviewing nonagricultural, ‘‘2+,’’ multiunit, and ‘‘Special’’cases.

x Microfilming evaluation forms and film optical sensingdevice for input to computer (FOSDIC) test report forms.

x Keying the data from the report forms to the data file.

x Edit review of keyed work units containing at least onereport form requiring correction.

x Maintaining central files containing all the report formsreceived by DPD during the data-collection operation.Out-of-scope forms were referred to the central filesimmediately upon being identified; in-scope report formsgenerally arrived at central files after edit review.

x The DPD staff carried out several other functions, includ-ing evaluation studies of census data (e.g., comparingcensus data to that obtained from the National Agricul-tural Statistics Service’s (NASS’s) June Agricultural Sur-vey (JAS)). The ‘‘flow’’ of each kind of report formthrough the DPD processing operation is illustrated infigure 6-1.1The Census Bureau introduced interactive computer systems to the

agriculture census processing in the 1987 enumeration.


�

Large FarmCoverage Unit

(Multiunits, Must“Special Cases,”and “Abnormals”)

Figure 6.1. 1992 Census of Agriculture Report Form Flow

Remove Contentsand Sort

Microfilm Unit

Evaluations Unit

Outlying Areas

Correspondence

Pre-Key Unit

(InScope)

(Out ofScope)

DataKeying Unit

EditReview

(Cases for telephone resolution)

TableReview

Check-In

(Puerto Rico, Guam,and the Virgin Islands)

(Receives film)

(Receives, JAS, CES, and FOSDIC materials

( “2+” “ Special” cases)

�

�

�

�

�

CentralFiles

�


Receipt and Check-In

Receipt and mechanical sort. The U.S. Postal Servicepresorted incoming agriculture census mail packages byusing different mail box numbers for must, sample, non-sample, screener, multiunit/abnormals, and Alaska/Hawaiireceipts.2 The packages were placed in trays by type ofreceipt and then sent to check-in clerks, who fannedthrough the receipts in each tray, checked to make certainthat each tray contained one type of receipt, and thenfurther sorted to identify agriculture receipts as follows:

x Materials addressed to a specific analyst.

x Agriculture nonresponse survey receipts (form 92-A46).

x Classification error survey receipts (form 92-A90).

x Undeliverable as addressed (UAA).

x Single-unit agriculture receipts (exceptAlaska andHawaii).

x Single-unit agriculture receipts for Alaska and Hawaii.

x Multiunits and abnormals.

x No barcode visible.

x PS form 3811 (certified mailout return receipts from theU.S. Postal Service; these were referred directly to thecoverage evaluation unit).

x Other receipts.

The unit also received materials that were not checkedin because the packages included correspondence. Theclerks scanned the correspondence to determine whetherit was a ‘‘congressional’’—i.e., the return envelope or theletterhead was from a Senator or Member of the House ofRepresentatives, or any representative of the legislative orexecutive branch of the Federal Government, or if the letterwas from a respondent and indicated that a copy had beensent to a Senator or Member of the House of Representa-tives (the threat to write to any of these was not considereda congressional). Congressional cases were referred to theunit supervisor, while for all other cases the clerks tran-scribed the census file number (CFN) of the case on theupper right hand corner of the letter, stapled the correspon-dence to the back of the report form, and placed it in a mailtray to batch for check-in.

Correspondence was grouped into batches of approxi-mately 100 cases at least once each day for referral to thecorrespondence unit. Single-unit agriculture census receipts(in envelope) and UAA batch sizes were based on thecapacity of the rolling bins used to transport materials—abatch consisted of a full bin, or 10,000 documents, which-ever was less (smaller batches could be used to clear theunit). The unit control clerk used the unit’s interactive workstation to enter the information needed to register each

batch on the Data Entry Control System, keying thenecessary code, user name, and password, then indicatingthe specific operation involved (in this case, check-in batchregistration), survey code (agriculture (except Puerto Rico)),and the correct document type (selecting from reportforms, 2+ report forms, UAA, or respondent originatedcorrespondence (ROC)). The computer generated aCheck-inBatch Cover Sheet with a sort number, batch number andcheck-in action code for each batch of work requiringlaser/wand/keyboard check-in. The clerk placed the coversheet on top of the appropriate batch and sent the materi-als for check-in by laser sorter or wand/keyboard check-in.

The rolling bins of single-unit receipts were sent to thecheck-in/lasersorterunit,where the56-pocket laser reader/sorterwas used to sort the packages. The sorter operator createda header record for each batch, keying the sort number,batch number, and batch special code from the BatchCover Sheet. The operator jogged the receipts (to makecertain they did not stick together and that the addressbarcode was visible through the envelope window) andplaced them upside down facing the laser, and then startedthe sort. The laser ‘‘read’’ the barcodes showing throughthe address windows on the return envelopes, and sortedthe packages as follows:

x By State (except Alaska and Hawaii).

x Multiunits and abnormals.

x Other returns.

x Classification error survey receipts.

x June agriculture survey receipts.

x Undeliverable as addressed (UAA).

x Machine failures (machine failures were not resubmittedfor each batch, but were held and rebatched as ‘‘machinefailures’’ and resubmitted later).

x Machine rejects (rejects were resubmitted three times; ifstill unreadable they were returned to the opening stafffor opening).

Materials for multiunits, abnormals, and Alaska andHawaii were sorted separately. After a batch had beensorted, the operator keyed the relevant identification datato the tracking system and transmitted the check-in infor-mation to the mail-update file.

Materials requiring laser wand/key check-in includedreport forms and UAA’s with unreadable barcodes, 2+cases, multiunit report forms and UAA’s, respondent origi-nated correspondence, agriculture census out-of-scopecases, and secondary source referrals from the telephoneunit. Wand/key operators also used the interactive systemsto keep track of their work, keying batch numbers and otheridentification information as work batches arrived for check-in. The operators used a hand-held laser wand to check in

2Must (P.O. box 5105), sample (nonmust: 5115), nonsample (5125),agriculture screener (5135), multiunit/abnormals (5145), and Alaska andHawaii (5165).


those materials with a visible barcode. If the barcode couldnot be read by the wand, or no label was present, theoperator used a keyboard station to key the CFN directly tothe file.

After completing check-in, the materials were sent on tothe remove contents and sort unit.

Remove contents and sort. The remove contents andsort unit received envelopes from the initial sort grouped bytype of report form (nonsample; must; sample; screener(A400); and multiunit/abnormal, Alaska/Hawaii, and PuertoRico). Clerks examined the contents of each envelope andsorted the contents into specific groups for further process-ing, placing the report forms in trays on rolling bins used totransport the report forms between units. The clerks sortedall the report forms first byState, thenby typeof form—updatingthe check-in records of report forms identified as ‘‘2+’’cases, and 0/7 screeners to show that status—as follows:

Destination trays(by State andform type) Forms

Must special cases Must special cases

Multiunits/abnormals Agriculture multiunit ‘‘2+ Cases,’’Agriculture multiunit ‘‘2+’’abnormals

‘‘2+’’ processing Agriculture single-unit ‘‘2+’’ cases,Agriculture single unit ‘‘2+’’coverage cases

Microfilming Special cases—June AgricultureSurvey (JAS), JAS—other,Special cases—Classification ErrorSurvey (CES),CES—other

Special cases Special cases (all other)

0/7 Screeners Screeners 5-7

Good receiptsdata entry)

FOSDIC cases,Alaska and Hawaii cases,all other cases

Quality control. The lead clerk for the unit conductedquality control checks on the staff’s work each day andreported the results to the unit supervisor. The checks weremade by the lead clerk selecting one case from eachdestination tray twice each day (morning and afternoon),and reviewing the selected cases to determine if they hadbeen sorted correctly. If an error was identified, the clerkthen verified the four preceding and four succeeding casesin that tray. If there were no additional errors, the clerkreturned the cases to the tray. If additional errors wereidentified, then the cases in that tray were subjected to100-percent verification and correction. After verificationwas completed for a particular tray, that tray was releasedfor further processing. The clerk entered the number of

forms verified for each group, the number of forms in error,and descriptions of the errors for each group on the DailyVerification Record, and sent it to the unit supervisor.

Tracking System

About 3.5 million report forms were processed by theDPD during processing operations for the 1992 agriculturecensus. Each of these documents represented a response(or nonresponse) record, and the DPD had to maintaincontrol of how these millions of documents and recordswere handled during the processing operation. The systemdeveloped to do this employed elements of the CensusAutomated Tracking System (CATS) originally used for the1990 Census of Population and Housing, and adapted forthe agriculture census.

The agriculture census CATS contained a file of all thecensus file numbers (CFN’s) mailed out in the census. Asreport forms arrived at the Jeffersonville office, the CATStracked the CFN’s by integrating available information fromthe various automated data-capture systems used in thecensus processing—(1) the laser sorter, (2) wand/keycheck-in, (3) batching for data entry, (4) computer-assistedtelephone interviewing (CATI) check-in, (5)data transmis-sions, and (6) computer editing. At each of these stages inprocessing, the individual report form/record CFN wasentered into the tracking system, together with the processpoint identification.

The system produced five basic data capture resolu-tions:

1. Undeliverable as addressed (UAA): The censusform was returned by the U.S. Postal Service asundeliverable.

2. Out of scope (O/S): The case has been identified atsome point in processing not to represent an operationmeeting the census farm definition.

3. Keyed and transmitted: Data have been keyed fromthe report form and transmitted to Charlotte and editedby the Economic Programming Division (EPD), and thecase has not been identified as O/S.

4. CATI resolution: The data for the case have beencaptured by computer assisted telephone interviewing(CATI) and has not been identified as O/S.

5. Not satisfied: The case does not fit any of the fourcategories above, no data have been captured orscope determination made, and it is not UAA.

At the time of the mailout, every CFN was ‘‘Not satis-fied.’’ As report forms were returned to Jeffersonville, theywere checked in by laser sorter (including UAA’s) andproceeded through the processing system, with additionalcheck in and out of specific units as they were processed.The CATS generated daily reports showing the cumulativereceipts, backlog of forms to be processed, cumulativeprocessed, and number of forms processed each day for


seven categories or processing units: (1) not satisfied, (2)open and sort unit, (3) microfilm processing, (4) batch fordata entry, (5) data entry, (6) correspondence unit, and (7)CATI followup. These daily reports were available for theUnited States and for individual States. Each CFN in themail file had a CATS record that showed the currentlocation of the CFN, its destination and the location fromwhich it was being sent, and the previous location, as wellas the form type (must, sample, nonsample, screener).When a Volume 1 (i.e., for a State, area, or the U.S.summary) report was produced, the CATS generated areport for that State or area with the tallies of CFN’s—

x Transmitted for tabulation.

x Resolved by CATI.

x O/S.

x UAA.

x Not satisfied.

Correspondence

General information. The correspondence unit of theprocessing staff was responsible for reading and process-ing incoming correspondence (respondent-originated cor-respondences (ROC)—letters or notes), as well as forcases referred by other units of the processing staff. Theunit handled correspondence cases for both the economicand agriculture censuses, and for the 1992 agriculturecensus, the unit also printed labels for remails to respon-dents who telephoned the Census Bureau and for UAAname and address change cases. The following informa-tion covers agriculture census correspondence operationsonly.

Interactive processing subunit (readers/keyers). Theinteractive processing subunit read correspondence referredto it from the check-in operation and the telephone assis-tance unit, and used interactive processing system toupdate the computerized census mail list. Telephone assis-tance cases arrived at the correspondence unit with formBC-435, Record of Telephone Call, documents attached.(Incoming correspondence was read on a first in/first outbasis, giving priority to BC-435 materials.) Keyers checkedincoming correspondence to determine whether it was

agriculture or economic census related, or referred toanother Census Bureau operation (such as the CurrentIndustrial Reports survey). Agriculture and economic cen-sus correspondence was retained for processing, whilecases involving other surveys were referred to the unitsupervisor for disposition. Keyers identified agriculture-related correspondence by the 11-digit CFN; an ‘‘A’’ pre-ceded the first digit of an agriculture CFN. When a caselacked a CFN, the keyer determined the status of thecorrespondence by checking for a business letterhead, orfor any mention of nonagricultural economic activity in theletter itself. (If a keyer had any doubt about the status of acase, the case was referred to a supervisor.) Agriculturecensus correspondence without CFN’s were referred forCFN search.

Keyers read the correspondence and used problemdescription tables to select a description that most closelymatched the subject of the correspondence, and tookwhatever action was indicated by the table. For example, inthe case of a respondent that did not receive a report form,but wanted to report agricultural operations, the keyerannotated the letter in the top right corner of the first pageof the correspondence with the appropriate two-digit unitnumber (in this case ‘‘12’’) and three-digit correspondencecategory (CORCAT) number (‘‘708’’—identifying this caseas a ‘‘Name (address add)’’ to the census file). For casesrequesting time extensions, the unit and CORCAT num-bers were followed by a six-digit time extension date. Afterall correspondence in a batch had been processed by areader, the reader separated the completed batch into 10categories:

1. Name/address (corrections).

2. Name (address adds).

3. Name/address search.

4. CFN research.

5. Check-in updates.

6. CORCAT keying.

7. Mailout.

8. Other.

9. Referrals (separated by type (i.e., large farm coverage,supervisor, correspondence analyst).

10. N/A changes HOLD.


Materials in categories 7-10 were referred to the appro-priate units. Keyers processed materials in categories 1-6using interactive programs to update the census mail file.The programs and the general actions were as follows:

Program Action

Name and addressupdates

Select menu option ‘‘name/address updates’’and enter necessary changes. If another COR-CAT code has been entered on the correspon-dence, hold the case and source materials forup to 5 days. If corrections had not beenmade, the keyer referred the case to the unitsupervisor; if corrections had been made, thecase was held for 3 days and rechecked foradditional CORCAT codes and recycled ifnecessary. After completing, the case wasreferred to central files.

Name and address adds The keyers used the ‘‘name and addressadds’’ routine to add new cases to the data-base and held them for 3 days to checkagainst CFN data display (recycling if neces-sary).

Name and addressresearch

The keyers performed name searches of thedatabase as needed (e.g., additional namesin correspondence, no CFN present).

CFN If more than one CFN is given in the corre-spondence, the keyers checked the statuscode of the additional CFN(s), and wrote itafter the CFN; then referred the case to ananalyst.

Check-in updates When a missing CFN was found during nameresearch, the keyer selected the ‘‘check-inupdate’’ routine and entered themissingCFN(s)with ‘‘3-0’’ status and continued processingusing the action tables for correspondencewith a CFN.

CORCAT keying Follow CORCAT keying instructions.

As many as 30 CFN’s and their associated codes couldbe keyed at a time.

After completing all interactive keying or updates for abatch, the keyer sent the materials to the unit control clerksfor disposition (except for name and address update casesbeing held for verification and CORCAT keying being heldfor label generation). The control clerks sorted the com-pleted work each day for referral to the appropriate units—

x Referrals from correspondence readers were directed tothe—

x Large farm coverage unit.

x An analyst.

x The unit supervisor.

x Central files.

x To the special case unit.

x Verified completed work went—

x For mailout.

x To central files.

x To batch and control for data entry.

x To the special case unit.

The control clerks prepared a control sheet for eachbatch of completed work indicating the referring unit (‘‘From09 Reading’’), the receiving unit, and the count sent. Thebatches were held until CORCAT labels had been printedfor them (each morning the unit supervisor checked theprinter for labels run up for the previous day’s work). Theclerks reviewed the labels and referred the completed workbatches and labels to the appropriate processing unit.

Quality control. Correspondence mailout materials weresubject to quality control procedures requiring 100-percentverification of each batch. Correspondence mailout pack-ages were batched into work units containing one day’swork of a particular type (e.g., complete remail, specialrequest, Census Bureau-originated correspondence, etc.),and referred to quality control review with a correspon-dence inspection record with identification information, andspaces to record the number of packages in the batch,number of errors detected, number of packages found tobe defective, and the date. Verification clerks checkedeach mailout package in each batch for mailout to ensurethat (1) the address label was correct, (2) each requireditem was included in the package, (3) and that all requiredcopies had been made and reports or copies stamped anditems circled as needed. Any errors were recorded on thecorrespondence mailout inspection record; an error rate of3 percent or more (i.e., 3 percent of the packages inspectedwere defective) resulted in a rejection of the batch involved.The clerks corrected any errors, added any materialsomitted, and released the batch for further processing.

Suspense file. The suspense file held all respondent-originated correspondence (ROC) cases that required areply by the agency, regardless of the unit of origin, as wellas all Census Bureau-originated correspondence (BOC)cases. Cases remained in the suspense file for a maximumof 35 days. If no additional response had been received atthe end of that period, the cases involved were referreddirectly to an analyst who determined what further action, ifany, should be undertaken.

Telephone Assistance Unit

Telephone operations. The DPD office at Jeffersonville,IN, included a staff and facilities for conducting telephoneenumerations and surveys. Prior to the introduction of theCATI system and the establishment of the dedicated CATIoffices at Hagerstown, MD, and Tucson, AZ, this staff hadbeen the principal telephone data-collection operation withinthe Census Bureau. However, the adoption of the CATIsystem resulted in the transfer of this function largely to the


two CATI offices, while the Jeffersonville telephone staffwas given the task of providing assistance to census andsurvey respondents who called with questions or problemsabout the censuses or surveys to which they were beingasked to respond, and also handled secondary sourceinquiries as part of the data-collection operation (seeChapter 5, Data Collection for details).

The Jeffersonville telephone unit was organized to handleincoming calls from agricultural census respondents (thestaff also handled incoming calls from economic censusrespondents, for details see the History of the 1992 Eco-nomic Census). Each census mail package included atoll-free telephone number at the DPD office in Jefferson-ville, IN, for respondents to use if they needed assistancecompleting their report forms, or if they had any otherquestions. The telephone staff handled 74,862 telephoneinquiries (many involving multiple contacts with a givenrespondent (approximately 114,000 individual telephonecontacts were made)), the bulk of which were from respon-dents claiming they had already filed a completed reportform, or who believed they were out of scope of theagriculture census.

The DPD staff had 30 individual work stations, eachequipped with a telephone and headset, and an interactivecomputer station. In December 1992, the Census Bureautrained telephone clerks in telephone techniques, subjectmatter, basic telephone procedures, and using the interac-tive computer system. (New staff assigned later weretrained as they joined the telephone operation.) The staffactually assigned to the operation varied from a maximumof 44 clerks in May, to only a single clerk by December1993.

Quality control. The quality control regime for the tele-phone unit involved verification monitoring by the unit’slead clerks or supervisors of incoming and outgoing calls.The verifiers used interactive computer stations and tele-phone instruments similar to those employed by the tele-phone interviewers, together with a monitoring packagethat enabled the verifiers to listen to individual telephonecalls and monitor the keystrokes of the telephone clerkhandling the monitored call. To begin monitoring, theverifier logged onto computer, entered his or her ownsurname and a password that activated the InteractiveTelephone Calls (ITC) Monitoring Menu. Selecting theappropriate commands, the verifier identified the specifictelephone instrument to be monitored. The verifier’s screenremained blank until the interviewer made contact during acall, and thereafter displayed each of the interviewer’skeystrokes as they were made.

During the first week that an interviewer began handlingcalls, the lead clerk monitored one call per day per inter-viewer. If performance was satisfactory, individual interview-ers then were monitored at the rate of three calls per weekper interviewer. The lead clerk checked five intervieweractions—

1. Question asking. The interviewer’s performance inasking questions correctly.

2. Probing. Whether the interviewer attempted to obtainmore data or a clearer answer when necessary.

3. Responsiveness. Whether the interviewer answeredrespondents’ questions and requests for any kind ofinformation.

4. Accuracy. Whether the interviewer provided correctresponses and information to respondents.

5. Correct entries. Monitored data entry by interviewersand other keystroke actions.

The lead clerk maintained a record of errors assigned toindividual interviewers and reported these to the unitsupervisor. The unit supervisor was responsible for makingcertain that interviewers were aware of any errors madeand how to correct mistakes.

Special Case and ‘‘2+’’ Case Processing

Special cases. Special cases were nonmust report formsreceived from the open and remove contents operation,with attached correspondence, remarks entered on thefront or back page, a blank front page with no positive data,or with acres reported in section 1, but with no crops orlivestock shown on the form. The special cases staffreviewed the report forms and attached materials usingthree condition/action tables to determine what action, ifany, should be taken with each case. The three tablesaddressed particular kinds of cases: Table I covered cor-respondence and remarks, a congressional test (i.e., wasthe case a congressional case), blank forms, and reportswith land, but no crops or livestock; Table II addressedcorrespondence, remarks, and reported data indicating achange in status; and Table III indicated actions for corre-spondence requiring reply. Reviewers began work on eachcase with Table I and continued through Tables II and IIIuntil the case met a condition in a table that resulted in theassignment of an out-of-scope (O/S) or referral code.

If the case met none of the conditions in the tables, itwas considered inscope, and was sent to the data keyingunit. The reviewing clerks entered O/S codes (O/2 fordeceased addressee, and O/7 for all other O/S cases) inthe upper right hand corner of the front of the report form.Selected referral codes (S for successor, P for partnership,and CF for claims filed) were entered in the middle of thetop margin of the report form, while all other codes werewritten to the right of the label area near the right margin.After reviewing each case, the clerk initialed the report formin the lower right corner of the front page of the report form,and after completing an entire work unit (up to 100 cases),the clerk wrote the data on a yellow post-it note and appliedit to the top report of the work unit.

The control clerks for the special cases unit separatedcompleted work units into groups by priority and disposi-tion, and routed them to the appropriate units for furtherprocessing. The groups and disposition were as follows:


Groups by priority Disposition

2+ cases 2+ processing unitREM, R-AG, or R-LL* Large farm coverage unitForm letter assigned Correspondence readingCorrespondence analyst Correspondence agriculture

analystSuccessor, partnership,or claims filed

Research clerk, special caseunit

Out of scope O/S wanding withinunit/forward checked formsto central files

Conservation reserveprogram (CRP)

Batch and hold in unit

Inscope Batch for data keying

*A case was coded ‘‘REM’’ when attached correspondence conflictedwith data reported on the form; code R-AG indicated doubt about farmstatus, or that the place was a partnership, but the name of the seniorpartner was not provided; code R-LL indicated that some land was rentedout, but that crops were reported.

‘‘2+’’ cases. ‘‘2+’’ cases resulted when; (1) two or morereport forms were mailed to the same individual (whomight, or might not, operate more than one farm or ranch),(2) two or more report forms were mailed to differentindividuals involved in the same operation (e.g., a husbandand wife, or two partners), or (3) two or more unrelatedreport forms were mailed to an accountant or a bank trustmanager who returned multiple report forms together in asingle envelope. All 2+ cases had to be reviewed todetermine whether they involved a single or multiple farms,and to ensure that all related report forms were checked inand the records and farms were properly linked within thecensus data file.

Materials arrived at the 2+ unit on a flow basis, routedfrom the remove contents and sort unit after check-in, andfrom the special cases unit. The control clerks at theoriginating units placed the report forms involved, and anyrelated correspondence, in folders marked ‘‘AG SU 2+’’ or‘‘AG SU 2+ COVERAGE’’ and batched them into work unitsof approximately 100 each for referral to the 2+ unit. Clerksin the 2+ unit reviewed the report forms and all separatepieces of correspondence in each folder to determinewhether congressional remarks were present, or if therespondent indicated correspondence with other Federaloffices, or if there was any mention of the Freedom ofInformation Act, and referred any cases with any suchmaterial present to the unit supervisor for disposition. Allother cases were reviewed to determine if the report formsrepresented a single farm, or multiple farms, and, if pos-sible, the scope of each report form.

The clerks also had to determine whether all the CFN’spresent for a specific case had to be linked to preventduplication of data. The CFN’s had to be linked if—

x The materials in a folder included several CFN’s, all ofwhich related to a single farm.

x A single report form was returned with multiple CFN’sreported on the front page or in attached correspon-dence.

x When the owner or operator respondent was involved inmultiple farm operations.

The CFN’s were not linked if—

x The 2+ identification was in error.

x Unrelated report forms had been returned in a singleenvelope.

x Multiple report forms had been returned with the sameCFN attached or written in.

x The case included one or more preidentified abnormal ormultiunit farm, or Alaska or Hawaii report forms (all thematerials for these cases were kept in their folders, andthe individual folders annotated ‘‘2+ Abnormal,’’ ‘‘2+Multi,’’ ‘‘2+ Alaska,’’ or ‘‘2+ Hawaii’’ as appropriate).

The clerks assigned linkage codes to each CFN in eachcase determined to require linkage. A primary-linkage codewas assigned in each case; for cases involving a single-report form with multiple CFN’s, the clerk wrote the primarycode ‘‘1’’ in the upper right-hand corner of the report formaddress label and circled it, then wrote a secondary code‘‘5’’ to the right of any additional CFN’s that had been addedin the write-in space. When multiple (but duplicate) formswere in a folder, with only one in-scope CFN, the clerkassigned the primary code to the in-scope CFN, and thesecondary code to the out-of-scope CFN(s), circling bothprimary and secondary linkage codes.

If more than one in-scope report was involved in a case,the clerk checked each report to determine whether theywere duplicates. If two or more of the report forms involvedwere duplicates, the clerk checked which contained themost information, and assigned the primary code ‘‘1’’ to theCFN for that report and the secondary code to the others.If there were no duplicate reports, but there was a commonownership relation, the clerk assigned a primary code ‘‘1’’to one of the reports, and a secondary code of ‘‘9’’ to theothers. The clerks assigned out-of-scope reports second-ary linkage codes of ‘‘5.’’ When all the CFN’s for a casewere out of scope, the reviewing clerk assigned a primarycode of ‘‘2’’ to one of them and a secondary code of ‘‘6’’ tothe remaining report(s), writing the codes in the upper rightcorner of the address label and circling them. When aprimary or unlinked report was determined to be out ofscope, the reviewing clerk annotated report ‘‘O/2’’ (in casesinvolving a deceased owner or operator) or ‘‘O/7’’ (all otherout-of-scope cases).

After coding, 2+ case CFN’s were linked using theinteractive system. The clerks entered the primary CFN foreach folder and the linkage code assigned to it, then thesecondary CFN(s) and linkage code(s). After all the CFN’sand linkage codes for a folder had been keyed, the keyerpressed the DO key and the system carried out the linkageand cleared the screen.


After linkage, the folders and the materials contained inthem were disposed as follows:

Groups Disposition

Abnormals, multiunits,Alaska, and Hawaii

Large farm coverage unit

AG 2+ SU COVERAGEfolders

Microfilm unit

Referrals 2+ Agricultural Analyst in unitForm letter assigned Correspondence unitSuccessor, partnership,or claims filed

Research clerk in unit

O/2 and O/7 codedreports

Out-of-scope interactivecheck-in unit/boxed forburning

Out-of-scope linked sec-ondary

Boxed for burning

Void duplicates Boxed for burningInscope reports Batch for data entry

Quality control. The work of both the special case and 2+case processing units was subject to quality-control proce-dures before being released from the units. For the specialcase clerks, the first 100 cases processed were verified100 percent, while for the 2+ unit, 100-percent verificationcontinued until 25 consecutive error-free cases had beenverified. The special case clerks qualified for sampleverification if they had achieved an error rate of 5.0 percentor less. When qualifed for sample verification, the work ofspecial cases processing clerks was checked at a 1-in-10rate, while 2+ case processing clerks’ cases were reviewedat a 1-in-8 rate. To remain qualified for sample verification,special cases clerks had to have at least 7 ‘‘accept’’decisions in each sequence of 10 decisions made; receiv-ing a fourth reject decision meant the clerk was returned to100-percent verification until qualifying for sample verifica-tion again. For 2+ case clerks, any error identified duringsample verification meant returning to 100-percent verifi-cation status until requalifying.

In their quality-control verification of special and 2+cases, the verification clerks checked for specified errorsand coded the records with identified problems. Casescould be rejected for any of the following errors:

Code Description

Special Cases

1. Error in scope classification of report forms.

2. Error in coding report for research (code CF (claimsfiled), S (succesor), or P (partnership)).

3. Referral error.

4. Error in transfer of data from remarks to report form.

5. Error in form letter designation.

2+ Cases

1. Failure to refer Congressional case to supervisor.

2. No linkage made as required (failure to assign linkagecodes or enter related CFN’s).

3. Linkage made when not required.

4. Incorrect linkage codes assigned.

5. Other error in coding/annotation of report form.

6. Error in performing interactive linkage.

Verifiers corrected all errors identified before referringthe individual cases reviewed for further processing. Thequality control staff maintained individual weekly verifica-tion records for each processing clerk and submitted aweekly summary verification report to the AGR staff atSuitland.

Large Farms Coverage Unit

General information. The large farms coverage unit (LFCU)reviewed multiunit (MU) and abnormal cases, resolved anyproblems and made any corrections necessary, and con-ducted any mail or telephone followup required to completethe enumeration of delinquent cases.

Multiunits. The Agriculture Division established companyfolders for multiunit (MU) farms identified prior to the 1992census mailout, and report forms were sent to all of theaddresses on file for MU operations as part of the censusmailing. Agriculture Division analysts accumulated theresponses from MU operations in the company folders untilall the addresses listed for a MU had responded or beenotherwise accounted for, then reviewed all the materialsbefore they were referred for data entry. The reviewensured all the agricultural operations of the MU companyhad been enumerated, satisfied, and corrected, and thateach in-scope report was ready for keying to the data file.

Analysts reviewed all the report forms for a singlecompany together, checking for duplication between indi-vidual establishments in a MU, correcting name and addresserrors, telephoning nonrespondents when necessary toobtain data, and, when a report form was missing, or analpha-plant number or employer-identification (EI) numberneeded to be corrected, used the interactive name andaddress update routine to make any changes required. Theanalysts conducted a section by section review of eachreport form in each folder, decided what action should betaken to make any corrections needed (e.g., allocatedbracketed entries to the appropriate cell, calculated correctunits for individual crops and regions, converted fractionsto tenths, and so on), and compared data between sectionsto ensure consistency (e.g., the land in farms reported insection 1 should be the same as that reported in section


10). The analysts also made historical comparisons, match-ing selected 1992 responses to 1987 data and verifyingany changes that exceeded specified limits.3

The staff made telephone calls or prepared explanatoryletters and mailed report form packages to respondentswho had returned incomplete report forms or to resolvespecific problems, and held the company folders out of theprocessing cycle until followup was completed. ‘‘New’’farms could be added to a MU if (1) a farm was part of theoperation, but did not duplicate the plant listed in inventoryfor that MU, (2) data from a report had to be split betweenacross county lines, or (3) telephone or correspondencefollowup determined that the farm in question had not beencovered in the report form mailout.

After completing the review of all the cases in eachfolder, resolving any problems or referring special prob-lems to the appropriate analyst, the staff updated thecensus mail file using the interactive systems to identifyout-of-scope cases in the file. The report forms and othermaterials for out-of-scope cases were retained in thecompany 1992 folders (including the MU inventory sheetsand any additional notes or attached materials), which thenwere filed by alphabetic name. The staff removed in-scopereports from the company folders and batched them intowork units for referral to the Data Services Branch for dataentry.

Abnormals. Analysts in the LFCU reviewed report formsand attached materials for all abnormal farms—farmsoperated by grazing associations, Indian reservations,government agencies (including Federal agencies, such asresearch stations), church held farms, and selected pri-vately held operations) responding to the census beforethe report forms were sent for data keying. The review wassimilar to that done for multiunits; the analysts checked—

x Name and address corrections and, if necessary, updatedthe mail file interactively, then carried out a section bysection review of the report form, using a set of printedguidelines to resolve identified problems, and readingany attached correspondence or other materials.

x Carried out historical comparisons, using the samegeneral guidelines used for MU’s (e.g., reviewing formsreporting changes of 1,000 acres or more in land for1992 compared to 1987, and so on).

x For abnormal cases involving two or more farms (i.e., itwas part of a 2+ case) the mailed abnormal report formswere used as the primary case and its CFN as theprimary numeric identifier.

x Conducted any telephone or mail followup required tocomplete report forms.

After completing the review of each case, making anycorrections required, and obtaining any missing data, theanalysts determined whether the individual case was inscopeor out of scope. Out-of-scope cases were checked out ofthe mail file interactively, and the report forms and attachedmaterials were retained in the LFCU file. For in-scopecases, the analysts checked in the CFN using a laser wandand the interactive systems, batched the report forms, andsent them for data keying.

Data Entry

General information. The DPD’s Data Services Branch(DSB) received data keying work units (DKWU’s) of reportforms batched by State and type (must, nonmust, non-sample, and screener (with and without geographic codechanges) and FOSDIC4). The DKWU’s arrived in plasticenvelopes with a Data Entry Batch Cover Sheet attached.The cover sheet included the DKWU number assigned bythe CATS, the number of documents (report forms, linkagedocuments, and mail file update materials) in the unit, theState code, the date the DKWU was prepared, and theoriginating unit.

Data entry (or keying) involved transcribing data fromthe census report forms to a machine-readable data file foredit and tabulation. The DPD staff used a key-to-diskinteractive system that combined the clerical review of theindividual census questionnaires with the data entry opera-tion. Each key station had a keyboard and monitor thatallowed the keyer to display and edit keyed data, as well asreceive messages or questions from the input program.Quality control procedures included reviewing samples ofeach keyer’s work and, when necessary, correcting keyererrors and retraining keyers.

As data were keyed and verified, DSB lead operatorstransferred the data electronically to the Charlotte, NC,facility for further processing.

Batch for data keying. After check-in and/or other prekey-ing processing, in-scope agriculture census report formswere referred to the batching control unit where the controlclerks batched them into data keying work units (DKWU’s)by form type and State, using the CATS system. The CATSsystem was menu-driven, and the clerk(s) successivelyselected appropriate items from census and State menusto begin batching. After selecting ‘‘Agriculture’’ from theAg/Econ data entry batching menu, the agriculture sub-menu was displayed and the clerk chose the correct formtype (i.e., must, sample, nonsample, or screener) and theappropriate State. With form type and State identified, theclerk used a laser-wand to read the barcoded labels on theforms to be batched. The computer rejected barcodes withinappropriate State codes in the CFN, or with wrong check

3The processing staff carried out historical comparisons for selectedfarms for land in farms, total value of agricultural products sold, andspecified commodities. For example, if either 1992 or 1987 acres reportedexceeded 1,000, the 1992 acres should be within 500 acres (or 150percent) of the 1987 acreage. Changes of 1,000 acres or more in easternStates, or 5,000 acres in other States, also required review and resolution(checking changes in acreage or land rented in or out, omission of leasedland, grazing permits, and so on). Similarly, the staff compared 1987 datato the 1992 report for any farm reporting $500,000 or more sales ofagricultural products in the 1992 census. 4Film optical sensing device for input to computer.


digit, alerting the clerk with an audible ‘‘beep’’ and display-ing an error message on the station’s screen. Report formswhose barcodes were rejected by the CATS were pulledfrom batching and referred to analysts for problem reviewand resolution. The clerk continued to wand barcodedlabels until 95 report forms had been accepted, thenpressed a function key to indicate that the batch wascomplete. The CATS system emitted a tone, and thenprinted a Data Entry Batch Cover Sheet for that batch. Atthe same time the system accepted the batch, it automati-cally updated the CFN tracking record to show that thereport forms in the batch were now going to ‘‘data entry.’’The clerk placed the batch, with the cover sheet on top, inplastic envelopes and placed the envelopes in a rolling binfor transfer to the data keying unit.

Data keying operations. The data entry operation for the1992 census represented a refinement of the proceduresused for the 1987 enumeration. As in the previous census,data entry combined clerical screening and data entry in asingle operation. Keyers identified problems on the reportforms and used guidance and instructions imbedded in thekeying programs to decide whether a given problem shouldbe keyed, flagged, ignored, or handled in some othermanner.

As DKWU’s were distributed to keyers by the unitsupervisors, each keyer opened the plastic envelopescontaining the report forms and other documents, wrote akeyer/verifier identification number on the cover sheet, andreviewed each report form for problems as data wereentered. Keyers rejected report forms for data entry andassigned ‘‘reject reason codes’’ for any of the followingreasons:

Code Reject reason

01 Geographic Area Code (GAC) check digit failure.The check digit failed (i.e., the check digit enteredwas not within acceptable ranges) on the GACchange located in the Census Use Only (CUO) Box036 on the front page of the report form.

02 Report form was not keyable. The majority of datavalues and/or their location on the report form wasillegible.

03 Blank report form. Sections 1-26 of the nonsample/screener forms, or Sections 1-32 of the must/sampleforms, were blank.

04 Maximum value failure. A data field entry exceedsthe maximum value.

05 Remark(s) requiring a reply. The respondent hasentered a remark or remarks on the report form thatrequired a response by the Census Bureau.

06 Report form nonmatch. The State code (the firsttwo digits in the CFN) in the CFN is not a valid one forthe geographic region (the last two digits of the reportform number); and/or more than one type of reportform is included in the WU.

Code Reject reason

07 GAC validation State change. The State reportedfor the principal county (containing agriculture opera-tions) does not match the ‘‘alpha’’ State entered onthe Data Entry Batch Entry Sheet.

08 Batch size exceeds established limit. The batchcontains more than 99 report forms with valid data(excluding any rejected report forms). After 99 ‘‘good’’report forms were keyed, all others in the batch wererejected.

09 Duplicate CFN. The CFN for a report form hadalready been keyed, transmitted, and formatted, bythe computer; and ‘‘1’’ is not entered in cell K039.

10 State nonmatch. The State code (the third andfourth digits in the WU batch number) does notmatch the first two digits of the CUO Box 036 datafield.

11 Secondary source. The first two digits of the batchnumber are ‘‘58’’ and CUO Box 037 is equal to:

x ‘‘1, 3, 7, or 8’’ with no data reported on the form.

x ‘‘5 or 6’’ with data reported on the form.

x ‘‘2, 4, 0, or 9.’’

12 Invalid GAC State code. The first two digits of theGAC in CUO Box 036 are not a valid numeric Statecode.

The keyers rejected, but did not key a reject reason codefor, report forms with invalid CFN check digits and/or invalidCFN State codes.

Four typesof report formscouldbeaccepted forkeying—must,sample, nonsample, and screener forms. The interactivedata entry system program assigned ‘‘screens’’ to specificsections of each type of report form for keyers to use inentering the data from the individual forms. For must andsample forms, screens 1-35 were ‘‘keyed,’’ while for non-sample and screener forms, screens 1-23 and 30-35 wereused. Keyers used screen 1 to key the CFN, and extraCFN’s and linkage codes, and data reported in the CUOboxes below the label area, and screen 2 for entering nameand address corrections and GAC information. Screens3-35 were used for entering data from the various sectionsof the report forms.

After entering the identification and address data fromthe address label area and section 1 of each report form,keyers began keying the data sections. The keyer enteredthe two-digit section identifier for each section, then thethree-digit keycodes for items within each section contain-ing data, using a nine-digit data field for the information ineach item. For example, for section 2 (‘‘Were any of thefollowing CROPS harvested on ’THIS PLACE’ in 1992’’),the respondent might report that 5,000 bushels of field corn(for seed) was produced on 50 acres. The keyer wouldenter the section identifier (‘‘S2’’) followed by the keycodes


for any specified crops for which the respondent reportedcrops harvested (e.g., ‘‘067’’ for acres and ‘‘068’’ forquantity (in bushels) of field corn harvested for grain orseed), and used a nine-digit data field to enter the acres(50) and quantity in bushels (5,000) harvested followingeach keycode. After keying each data field, the keyer usedthe field release code, or pressed the ‘‘F6’’ function key tomove on to the next code of data field.

For sections requesting write-in responses e.g. section 8(‘‘Were there a combined total of 20 or more FRUITTREES, including GRAPEVINES, CITRUS, and NUT trees,‘‘THIS PLACE’’ in 1992?’’) the keyers did not key thepreprinted keycodes unless data were reported in the cells.Instead, the keyers entered the section identifier (‘‘S8’’) the‘‘main’’ keycode located under the heading ‘‘code’’ only,followed by each keycode (with reported data) in the orderin which they appeared in the section. (Sections requestingwrite-in responses for fruits, crops, or other products notspecifically listed in the response boxes included listings ofprobable crops/products and their respective codes in thesection below the response boxes. Respondents wrote inthe name and keycode for each crop/product not prelisted,together with the production and inventory data requestedon the appropriate line.)

The keyer continued on through each report form,entering the various codes as needed. They were expectedto key, reject, or flag any problem item, using detailedinstructions as shown in table 6-1.

Quality control. The Census Bureau used formal qualitycontrol (QC) procedures for the data keying operation toensure that the information on the report forms was accu-rately recorded in the data file for editing and tabulation.Verifiers independently keyed data for report forms forselected records and matched the original keyer’s data setwith the verifier’s set. Any differences were displayed forevaluation and assignment of error. All mistakes identifiedduring this review were corrected and reverified before thedata were released for computer processing. The qualitycontrol procedures defined errors as either keyer or non-keyer errors. Keyer errors were generally miskeying thatcaused errors in data fields, field or document omission orduplication, unnecessarily keyed fields, etc. Nonkeyer errorswere those involving mechanical problems, or supervisoror verifier errors.

Data keyers progressed through a two-stage verificationregime after training, which included instruction on thespecific agriculture data entry procedures (the ‘‘T’’ stage;during training, all keyers’ work was verified 100 percent) aprequalification (‘‘P’’) stage, and a qualified (‘‘Q’’) stage.After completing training, all keyers were rated in the Pstage, and work units keyed by them were verified at the100 percent rate (i.e., all report forms keyed were checkedfor chargeable (keyer) errors). During the P stage, all thedata keying work units (DKWU’s) completed by a keyerwith a batch error rate higher than 2.0 percent had to berepaired by the original keyer and resubmitted for verifica-tion. To move on to the Q stage, a keyer had to key oneDKWU with a keyer error rate of 1.5 percent or less.

Table 6-1. Keyer Problem Instructions

Problem Description Keyer action

Alpha entries Respondent used an alphabetic equivalent for a numericvalue (i.e., ‘‘ten’’ acres instead of ‘‘10’’ acres)

Keyer interpreted the value and keyed it in numeric

Dollars/cents Respondent reported dollars and cents instead of dollarvalue only

Keyer entered only dollar value

Altered stub A change or addition to the preprinted items by therespondent

Keyer flagged the problem by entering ‘‘-’’ for the datacell(s) involved.

Bracketed entries A single entry reported for multiple data cells. Keyer keyed the data in the cell in which it appears, or, ifthe data is outside any single cell, key the data in the firstof the bracketed data cells, and flag the entry by keying‘‘&’’

Data reported outside of a datacell

Data reported written outside corresponding data cell Keyer keyed data outside a data cell

Double entries More than one entry (not totaled) reported in a single datacell

Keyer entered the last entry and flagged it by entering ‘‘(’’

Fractions and decimals Fractions and/or decimals reported when not requested,or when ‘‘tenths’’ requested

Keyer decided how to key based on rules provided in keyerinstructions

Illegible entries - Referred to supervisor for resolution; if unable to resolveimmediately, keyer keyed a remark flag

Negative entries Negative values reported for acres or dollars Keyer keyed a ‘‘+’’ flag for the sectionRange entries Data reported with an upper and lower limit instead of a

specific numberKeyer keyed upper limit only

Reference to other data Symbols such a arrows, ditto marks (‘‘), or remarks (‘‘all’’)used to indicate ‘‘the same as’’ referring to other data

Keyer decided to key data or flag (‘‘+’’) based on instruc-tions and examples in keying instructions

Wrong units Units reported are inconsistent with those listed in the datacell or preprinted to the right of crop names below thewrite-in section

Keyer compared the reported units to the preprinted unitsand keyed the data if the units were the same, or flaggedthe data by keying the reported unit (the first unit if severalwere used) followed by a ‘‘+’’

Remarks Comments or reporting errors that (1) required a change tothe reported data, (2) contained data, (3) related to themanner in which data were reported, or (4) required areply

Keyer rejected form with remarks requiring a reply. Forothers, decided to key data and/or flag (’’+‘‘) based oninstructions and examples in keying instructions


The work for Q stage keyers was verified on a samplebasis, with the number of report forms checked based onthe size of the DKWU being verified, as follows:

Work unit size Verification rate(percentage)

Samplinginterval

Less than 24 100 All24-44 12.5 845-74 6.67 1574-99 4.0 25

Keyers at the Q stage had to pass two quality standardson a weekly basis; if a Q keyer had three or more DKWU’srejected during the week, or if a keyer has one or twoDKWU’s rejected and the overall error rate exceeds 2.0percent for the week, the keyer was reclassified to the Pstage, and continued on 100-percent verification until requali-fied.

The interactive systems used for keying designated asystematic random sample of report forms from eachDKWU for verification based on the sample rates pro-grammed. The first report selected was one whose sequencewithin the DKWU corresponded to a random integer betweenthe first report in the DKWU and the sampling interval (e.g.,for DKWU’s with 75-99 report forms, the selection wouldbegin with by randomly selecting one of the first throughthe twenty-fifth reports in the unit). Keyed DKWU’s wereaccepted if the number of fields with keying errors did notexceed a specified number based on the total number offields sampled. For example, for relatively small DKWU,with 139-199 fields verified, 4 errors or less would beacceptable, while more than 4 fields in error led to rejectingthe DKWU and its resubmission for keying and verification.For a comparatively large DKWU, with 1,588 to 1,668 fieldsverified, 23 or fewer fields with errors was acceptable,while more than 23 errors meant rejection and rekeying.

After data keying and verification, the accepted DKWU’swere moved to a holding area and held until dispositionlistings were generated showing which records had failedand which had passed the computer edit. The processingstaff pulled the keyed DKWU’s for the interactive editreview and correction process. Thereafter, the report formswere routed to central files for sorting, boxing, and storage.

FOSDIC data keying unit. The 1992 agriculture censusincluded testing a report form designed for use with theCensus Bureau’s film optical sensing device for input tocomputers (FOSDIC) equipment.5 The test employed cen-sus sample (nonmust) questionnaires from selected Statesin questionnaire region 2 (i.e., Illinois, Indiana, Iowa, Kan-sas, Nebraska, and Ohio); respondents in Indiana and

Kansas, and half the respondents in Iowa received thestandard form 92-A0202 questionnaire, while the other halfof the respondents in Iowa, as well as those in Ohio,Illinois, and Nebraska were sent the form 92-AO202(F)(FOSDIC) test report form. (The Census Bureau used thereport forms for Indiana and Kansas to monitor responserates for FOSDIC, compared to nonFOSDIC, question-naires.)

The DPD established a special keying unit to handledata entry for the FOSDIC test cases, organized into twosubunits, one to key data from the paper questionnaire (theFOSDIC/PAPER subunit), and the second to key datausing the FOSDIC/microfilm access device (MAD, hencethe FOSDIC/MAD subunit)). The FOSDIC/PAPER keyersentered data directly from the paper report forms, while theFOSDIC/MAD keyers used the MAD and its associatedsoftware to key data to the file from microfilm of the reportforms. Using the MAD equipment increased the informa-tion available to the keyer before the keying operation,identifying all keycodes that the FOSDIC reader equipmenthad identified and setting the necessary flags for yes/no/blankresponse or check marks. Furthermore, the reader infor-mation was used to run the microfilm readers, which wereprogrammed to skip pages with no markings, to go directlyto specified batch numbers for work, and even to skipentire questionnaires.

Each keying subunit consisted of 5-7 keyers with therequisite work stations. The FOSDIC/PAPER keyers receivedwork units of approximately 100 report forms for keying,while the FOSDIC/MAD subunit used microfilm reels con-taining the images of report forms. Originally, the CensusBureau planned to begin keying using the paper forms inFebruary 1993, and start work from microfilm in May.However, delays in selecting and training the staffs (DPDdeliberately chose keyers from the economic area with noexperience in the agriculture census operation for theFOSDIC keying unit) led to the paper keying operationbeginning in March, and the microfilm keyers starting workin August. All FOSDIC unit keying was completed by earlyOctober 1993. Total workload for the combined unit wasapproximately 72,000 cases.

COMPUTER PROCESSING

General Information

After being keyed to a computer file, the data from eachreport form were formatted, edited, and tabulated usingminicomputer systems at the Census Bureau’s Charlotte,NC, facility. The data from each report form were edited,item-by-item, in a comprehensive check for consistencyand reasonableness. During the edit, the computer cor-rected erroneous or inconsistent items, supplied missingdata based on similar farms in the same county, andassigned any classification codes required. (Agriculturesubject-matter specialists reviewed major changes to thefile by the computer edit to ensure no computer-generated

5The report forms used circles or boxes filled in by respondents forreporting selected data, which could then be read directly to the data fileby the FOSDIC equipment. Report forms using this design technique hadbeen previously used to collect selected information in the population andhousing censuses. The agriculture census report form used the fill-inboxes to indicate that data were reported in a specific cell, that is, as anenhancement to data keying rather than for reporting information directly.


errors were retained in the census data file.) The statisticalestimation and imputation for nonresponse also were doneby computer following the data file edit. The CensusBureau carried out an interactive pre-analytic review of theindividual data records, matching all the records in the finalfile to identify any duplicate responses. This final matchingeffort found 7,500 duplicate records and deleted them fromthe final files before tabulation.

The computer work cycle began almost as soon as thefirst report forms were received and the data keyed andtransmitted to the Charlotte, NC, facility in January 1993and continued until the tabulation and disclosure reviewwas completed in August 1994. The principle componentsof the computer processing cycle were—

x Format.

x Computer edit and edit correction.

x Statistical estimation.

x Data tabulation and disclosure review.

Format

The data entry operation created an individual record foreach agriculture questionnaire containing a series of ‘‘fixed’’record layouts that, for computerized editing and tabula-tion, had to be converted to ‘‘variable’’ output records withbinary codes for numeric values. The data files thenunderwent a computerized format and simple edit process.

The formatting program converted the data records intoa series of fixed and variable portions. The fixed portionscontained standardized identification information—Stateand county codes, CFN, standard industrial classification(SIC) code, and so on—and the variable portions heldfields for each data item reported, imputed, or changed.The computer recognized the individual data items fromthe keycodes at the beginning of each data segment andignored blank segments. Historical data for individual itemswere added at this time and were compared to the reporteddata for completeness and reasonableness. The programalso carried the flags set during data entry to the formattedrecords and set new flags for any problems identifiedduring the formatting cycle, as follows:

x Illegal geographic or report-form codes.

x Cases with no reported sales or livestock inventory.

x Cases with individual items flagged (i.e., illegal key-codes, invalid crop codes, etc.).

x Cells or records that, compared to historical data for thatitem(s), exceeding established limits, or as incomplete.

Flagged cases were displayed for analyst’s review andresolution before being submitted to full computer edit.

Computer Edit

General information. Computer editing is the mechanicalprocess of checking and reviewing reported data andcomparing it to established parameters. For the 1992agriculture census, all reported data were keyed and thenedited by computer. The format and simple edit procedureconverted the raw data records produced by the data-keying operation into binary codes and flagged selectedproblem cases.

Computer edit and imputation. The complex edit andimputation programs were designed to carry out severalhundred individual editing operations in all (850 keycodescould be reviewed for each nonsample form and 900 foreach sample form) although generally only a part of thepossible total was required for any individual record. Agri-culture subject-matter specialists wrote and transmitted thecomputer edit specifications to the computer programmersin the EPD using decision logic tables (DLT’s). Each DLTwas a tabular display of the elements comprising a specificedit operation from inception to resolution.

Prior to submission for the complex edit, the formatteddata files were sorted by State and by county and CFNwithin each State. The data from each farm record weresubjected to a detailed, item-by-item, computer edit. Thiscomplex edit—

x Determined whether each record represented an agri-cultural operation meeting the census farm definitionand deleted out-of-scope operations from the file.

x Assigned farm classification codes needed for tabulatingthe data, including acreage, tenure, product sales, orga-nization, and SIC code.

x Identified nonsample farms representing farms that metthe ‘‘certainty’’ criteria for each State, and convertedthose records to sample records.

x Checked consistency between and within sections ofeach record.

x Checked for reasonable relationships between and amongdata items, values for various sizes of farms, andcombinations of commodities.

x Checked that geographic, legal, and physical constraintswere met.

The complex edit operation also imputed missing datafor farms in the census files. Whenever possible, editimputations, deletions, and changes were based on otherdata in the same record, or for some items (such asoperator characteristics), on historical information from theprevious census. Other missing items were calculatedbased on reported quantities and average commodityprices in the same State. When these methods could notbe employed, the imputation program used informationreported by other, similar farm operations in nearby geo-graphic areas (e.g., the same county). For example, a


record that contained acres of corn harvested, but notquantity, would be assigned the bushels of corn per acreharvested reported for a nearby farm with otherwise similarcharacteristics.

Data records that failed to meet the census farm defini-tion, or that had undergone substantial computer-generatedchanges to the data (see below), were reviewed to ensurethat the data had been keyed correctly and/or that thechanges were justified. Edit referral cases (i.e., cases thatfailed edit and were flagged for review by the computer)were reviewed for keying accuracy to ensure that the editresults were correct. Any cases for which the computer editresults were found to be unacceptable were corrected asrequired and reedited.

Failed Edit Review

General information. Data records rejected by the editprograms—i.e., ‘‘failed edit review’’ cases—were referredto the DPD staff at Jeffersonville, which carried out thefailed edit review. This procedure determined, (1) whetherthe information was keyed from the report forms to the datafile and, (2) that any changes made to the individualrecords in the computer edits were correct and acceptable.The staff also reviewed cases identified as format rejects,or determined to be out of scope by the computer (COS’s).For general failed edit cases, flags set by the computeridentified specific problems, and/or data items that hadbeen changed by the computer edit programs. The failededit review operation began in the second week of Marchand continued through December 1993. At peak levels inJune and July, the edit review staff processed as many as8,000 records per week.

Interactive edit. Failed edit cases were referred to the editunit in their original data keying work units (DKWU), so thatthe edit clerks worked through batches of reports orga-nized by State and form type (must, sample, nonsample,and screener). The edit clerks used interactive minicom-puter systems to review and correct the data recordsflagged during data keying or computer edit. The ‘‘fail editscreen’’ (FES) was the principal tool used for the editreview. When in use, the FES was divided into four areas;the first, at the top of the screen, contained identificationinformation for the individual record being reviewed (e.g.,CFN, sequence number, DKWU number, etc.); a secondarea (taking up the bulk of the screen) contained keycodesand their current data; the third area (on the right side of thescreen) contained the keycodes and values entered ascorrections to the record; and the last area consisted ofreference screens that could be accessed from the FES(e.g., displays of original keyed data, historical data, NASSdata, etc.).

To begin edit review, each clerk ‘‘signed on’’ the system,called up the production screen, and entered the seven-digit DKWU number of the first work unit to be edited. Thesystem then displayed a CFN menu showing the CFN’s,

sequence numbers, disposition codes, and referral dispo-sition (RD) codes for each record in the work unit. Thelistings for reports that had passed the initial edit weredisplayed in reverse video (i.e., with the background colorchanged from the standard used on the screen). Casesflagged by the computer as out of scope (‘‘computerout-of-scope’’ records) were processed separately (seebelow) and records that were rejected because of format-ting problems were referred to the large farm coverage unit(LFCU) for processing.

The clerks called up the first accessible record insequence that had failed the initial edit to begin theirreview. There were four types of failures addressed by theedit review:

1. Simple edit failure. Bracketed entries, double entries,other crops or livestock reported, altered stub, wrongunits, remarks that required resolution, and geographicchange not valid in section 1.

2. Historical failures. Large farm records with substan-tial differences between 1992 and 1987 reported data.

3. Inconsistency failures. Entries for keycodes are notconsistent with item to be reported or measures used.

4. Complex edit failures. The complex computer editchanges in the reported data was inconsistent betweenor within sections, or reported data exceeded estab-lished limits.

The edit clerks used detailed instructions describing thespecific edit failures and corrective actions to be taken toreview each flagged keycode. After entering corrections forall the flagged keycodes, the clerks pressed the key andentered one of the following RD codes as appropriate:

RD code Action1 Re-edit record.2 Re-edit record—by-pass specified sections

based on edit skip code.4 No changes made—accept as is.5 Delete from file (out of scope).7 Million dollar + and/or 30,000+ acres—refer

to the LFCU.9 All other referrals.

If a record referred for re-edit failed a second time, it wasimmediately redisplayed with the new flagged keycodes,and the clerk reworked the case and resubmitted it forediting. Only when a record had passed the edit did theinteractive system display the next available CFN forreview. Edit review processing for a DKWU was completedwhen every accessible CFN in the work unit had an RDcode applied.

Computer out-of-scope processing. The computerizedcomplex edit of the 1992 agriculture census data filesidentified certain records as out of scope, either due to data


keying errors (a respondent entry was missed, or mis-keyed), or remarks on the form contained data or clarifieda situation on the form the computer did not recognize. Thecomputer out-of-scope review determined whether theserecords actually were out of scope by an interactive clericalreview of each record, using a ‘‘point’’ system for agricul-tural operations reported by computer out-of-scope cases.The edit unit conducted this review; the interactive systemsdisplayed individual records, identifying out-of-scope caseswith a ‘‘reverse video’’ (i.e., the background color of themonitor screen is different than for in-scope cases) fordetailed review. The clerks deleted cases confirmed as outof scope from the data file, and pulled the report forms fromthe batches involved for referral and storage with otherout-of-scope cases.

For cases that appeared to be inscope, the clerksreviewed reported crops or other production, and used atable of point awards for acreage and production of speci-fied crops or other activity. Any record accumulating morethan 1,000 points was considered inscope. For example, arespondent might have reported having 3 acres of corn forsilage, and half an acre of Irish potatoes. Corn for silagewas worth 220 points per acre, or 660 points for 3 acres,while Irish potatoes was worth 1,400 points per acre, or700 points for half an acre. This particular farm then had anaccumulation of 1,360 points and could be consideredinscope for the census. Cases identified as inscope werere-edited and incorporated into the data file.

Format reject processing. Format rejects were createdwhen keyed data failed specified checks prior to thecomplex edit. The records and the census report formsinvolved then had to be clerically reviewed and any prob-lems resolved before being rekeyed. Format reject reviewand processing was handled by the LFCU and werereferred to the unit from the edit review unit. As the reportswere received, the LFCU staff checked coded reasons forrejection and used detailed written instructions to resolvespecified format problems. For example, for a case referredbecause of ‘‘item rejects’’ (i.e., 10 or more keycodes and/ortheir associated data were rejected), the clerk checked theinstructions for specific actions to be taken for identifiedproblems, and reviewed the entire report form to correct allpossible item rejects. Corrected forms were routed to thebatching unit for rebatching and rekeying. Report formsidentified as out of scope were stored, in CFN sequence, inthe LFCU.

Statistical Estimation

General information. The final response rate for the 1992Census of Agriculture was 84.5 percent. The publishedstatistics from the 1992 census represent all farms in theUnited States because the Census Bureau used statisticalestimation to inflate the data supplied by respondents tocompensate for nonresponse and the use of sampling.

Nonresponse estimation. The statistical estimation pro-cedure used by the Census Bureau excluded large, abnor-mal, and unique farm operations, since they were sub-jected to intensive followup during census processing. Theagency contacted a stratified sample of the remainingnonrespondents, using the computer-assisted telephoneinterviewing (CATI—see chapter 5 for details of the CATIoperations) system.

For nonresponse estimation, a five-strata sample—basedon expected value of sales, previous farm census status,and whether the record was identified by the mail list modelto receive the screener report form (the forms 92-A0401through 92-A0411)—was defined and drawn from thenonrespondent file for telephone contact. When a nonre-sponse survey case could not be contacted, or no tele-phone number was found, the appropriate screener formwas sent to the address by certified mail.

The Census Bureau estimated the proportion of censusnonrespondents that operated farms for all five strata at theState level by applying the nonresponse survey results tothe total number of nonrespondents in each stratum. Theagency assumed that the distribution of farms and nonre-spondent farms in each stratum at the county level was thesame. County-level estimates for nonrespondent farmswere synthetically estimated from estimated totals calcu-lated for each stratum within each county.

Within each stratum in a county the Census Bureaucalculated the ‘‘nonresponse weight’’ and assigned it toeach eligible respondent farm record. This ‘‘weight’’ wasthe ratio of the sum of the estimated number of nonrespon-dent farms and the number of eligible respondent farms tothe number of eligible respondent farms. The weight wasnever allowed to be greater than 2.0 using controlledcollapsing procedures. The nonresponse weight was ran-domly rounded to 1.0 or 2.0 for each record for tabulatingthe complete count items for publication.6

Sample estimation. The 1992 census collected ‘‘sample’’data (for items 21 through 26 on the sample form7) fromabout 20-percent of the agriculture census mail universe.The following types of addresses on the census mail listreceived sample report forms:

x Special insert cases.

x All addresses in Alaska, Hawaii, and Rhode Island.

x All must cases.

x All addresses expected to represent ‘‘large’’ farms (thedefinition of a large farm varied by State, from a mini-mum of 1,000 acres or $100,000 in sales, to as high as10,000 acres and $200,000 in sales; these were allconsidered certainty cases).

6Certain respondent farms that produced ‘‘rare’’ commodities wereidentified as ineligible to represent the typical nonrespondent farms andwere excluded from the nonresponse weighting operation.

7These items requested data on the following: (21) commercialfertilizers, (22) use of specified chemicals, (23) production expenses, (24)machinery and equipment, (25) estimated current market value of landand buildings, and (26) income from farm-related sources.


x All farms in counties with less than 100 farms identifiedin the 1987 census.

x A systematic sample of 1-in-2 addresses in counties thatcontained 100 to 199 farms in the 1987 census.

x A systematic sample of 1-in-6 addresses of counties thatcontained 200 or more farms in the 1987 census.

The sample estimates were designed to develop thetotals that would have resulted had all census respondentsbeen asked for the data requested for items 21-26. Theestimates were calculated using a controlled collapsingprocedure that assigned a ‘‘weight’’ to each respondentrecord containing a sample item or items. For any givencounty, the sample item total was estimated by multiplyingthe sample data items for each farm in the county by therespective sample weight assigned, and summing over allthe sample records for that county.

The Census Bureau classified respondent sample recordsas ‘‘certainty’’ sample farms (i.e., the first five types ofoperations listed above, and hence, mostly large opera-tions) or sample farms (mostly smaller operations). Cer-tainty sample farms were assigned a weight of ‘‘1’’, since allfarms of these characteristics were subject to intensivefollowup. The sample farms within each county wereassigned weights based on specified characteristics. Tocalculate these weights, the Census Bureau used a three-step process, employing three variables. The first variablecontained eight 1992 total value of sales (TVP) groups; thesecond contained two SIC code groups; and the thirdcontained two acreage groups. The variables were asfollows:

TVP SIC CodeMajor Group Acreage

$1 to $999 01 All crops 1 to 69$1,000 to $2,499 02 All livestock 70 or more$2,500 to $4,999$5,000 to $9,999$10,000 to $24,999$25,000 to 49,999$50,000 to $99,999$100,000 or more

In the first step of the estimation procedure, all farmswere classified into one of 32 mutually exclusive stratausing the three variables groups. The total and samplefarm counts were expanded to account for nonresponse.Each stratum then was assigned an initial sample weightequal to the ratio of the total farm count to the sample farmcount. This weight was supposed to be approximatelyequal to the inverse of the probability of selecting a givenfarm for the census sample.

Once the farms had been classified and initial weightsassigned, any stratum containing less than 10 samplefarms (after nonresponse adjustment), or that had a weightmore than twice the mail sample rate, was collapsed withanother stratum. (The mail sample rate was either 2 or 6,depending on whether the county involved had a 1-in-2 or

a 1-in-6 sampling rate.) After the collapsing process wascompleted, new total and sample farms counts were com-puted from each of the final strata and were used tocalculate the final sample weights.

In the final step of the estimation process (actuallyperformed after the analytical review), final weights wereassigned to the sample farm records in each stratum. Thisweight was the ratio of total farm count to sample farmcount in each stratum expanded to account for nonre-sponse. The noninteger weights were randomly rounded toan integer weight for tabulation. (E.g., the final weightassigned to farms in a particular stratum was 7.2, thenone-fifth of the sample farms in the stratum were randomlyassigned a weight of 8.0, while the remaining four-fifthsreceived a weight of 7.0.)

Post-Edit Correction Processing

General information. After the computer edit and editcorrection processing were completed, the Census Bureauconducted a final review of the data files to removeduplicate records that had not been identified before aggre-gating the individual records for analytical review. Theindividual records were tabulated by computer into a matrixcalled the analytical table. The main analytical table con-sisted of 2,200 rows of data. Accessory matrices—‘‘calltables’’ (CT’s)—also were built. There were 10 CT’s thatcollectively comprised approximately 2,000 additional rowsof data. The Census Bureau used the matrices to extractdata for analysis and correction of the county-level tabula-tions.

After analytical review, the data underwent final weight-ing. A small, State-level, table of selected basic data (calledthe ‘‘C99’’ or ‘‘hand-off’’ table) then was created andreviewed to ensure that the final weighting had not causedsignificant shifts in the data. Once the handoff table wasapproved, the data were aggregated into a single file—themaster matrix (MM)—for each State and for the UnitedStates. The master matrix also underwent disclosure pro-cessing, and then was used to build all of the Volume 1tables except the cross tabulations. The cross tabulationsfor the Volume 1 reports were generated separately fromthe master matrix and underwent independent disclosureanalysis. All of the Volume 1 tables were downloadedelectronically to a local area computer network for tablereview using the Census Bureau’s Tabulation and Disclo-sure System (TADS).

Duplication review. Prior to conducting the analyticalreview of the census data, AGR staff carried out a finalduplication review of each State file. During mail-list devel-opment, many names and addresses from different sourcesthat, while similar, were not exactly alike, were retained bythe matching operation done before mailout. Even exactname matches could be mailed two report forms when the


associated information (e.g., employee identification num-ber (EIN), social security number (SSN), telephone num-ber, etc.) differed. This final duplication review was designedto identify and delete these duplicate records from the datafile.

In a further computer matching operation, the computerprogram matched records in each county in the State fileunder review. Records were sorted and matched in threephases, and any suspected duplicates identified on onephase of the review was removed from possible consider-ation in any other phase. The three phases successivelyreviewed and matched records by CFN, telephone number,and three-digit SIC code, then by selected keycode itemswithin the matching records. If five keycode items matched,the records involved were displayed at an interactiveworkstation for analysts to review.

In their on-screen review,AGR staff checked the matchedrecords to determine whether any should be deleted fromthe file. Reviewers assigned alpha codes to the CFN’sinvolved for records they reviewed. CFN’s requiring noaction (i.e., that were not duplicates) were assigned a ‘‘K’’code, and retained in the file. Those CFN’s identified aspositive duplicates were assigned ‘‘D’’ and were deletedand made out of scope by assigning them a referraldisposition (‘‘RD’’) code of 5. CFN’s about which somequestion remained were designated ‘‘U’’ and were referredto analytical review for final disposition.

The duplication review operation began as soon as theafter-edit failure review was finished, and continued until all50 States had been reviewed. Approximately 70,000 dupli-cate records were identified and deleted from the final filesin this review.

Analytical review and data correction. Analytical reviewis the review of all census of agriculture data items, values,and selected data relationships in the data files for eachState and county. The review staff used unpublishedanalytical tables (the computer generated a single analyti-cal table for each State, and for each county within eachState—these were the first tabulations of census datacreated after duplication review) displaying all census dataitems with positive data values in the 1992 or 1987censuses—and the analysts used these tables to interac-tively review the data on the minicomputer system. Agri-culture Division analysts performed the review and correc-tion operation in several phases.

In an initial summary table review, senior subject-matteranalysts checked the data for their specific areas ofresponsibility (farm economics, livestock, and crops). Theylooked for completeness and comparability with historicalcensus data within expected limits. The interactive processallowed the analysts to examine detailed county data forany specific item that apparently had problems at the Statelevel. This summary review was followed by an analyticreview involving a systematic review of the county data bysubject-matter specialists. Any problems identified early inthe review process were checked for validity, and thecurrent census data were compared to NASS and historical

census data to check for disappearances, duplications, orchanges involving larger agricultural operations in thecounty. During summary and analytical review, analystswrote any specific criticisms to a ‘‘criticism shell’’ for eachcounty.

After AGR analysts completed review of all the counties’data for a State, the tables were ready for review by NASSrepresentatives. Visiting statisticians from each of theNASS’s State offices examined State and county data andreviewed the problems identified in the Census Bureau’sanalytical review. (The NASS reviewers used the sameanalytical tables used by the Census Bureau analysts andwere sworn in as temporary Census Bureau employees tomaintain confidentiality protections, and to enable them tohave the same access to the detailed databases forindividual records that the Census Bureau’s own analystshad.) The NASS reviewers could write criticisms from thetables, just as Census Bureau analysts did in the summaryand analytical table review.

Once the NASS statisticians completed their review of aState file, supervisors in the analytical review staff carriedout a final criticism review and edit to make certainidentified problems were valid and merging NASS withCensus Bureau criticisms. This final review included anautomated outlier check to ensure that no large, uncriti-cized, differences between 1992 and historical census dataon specific items escaped identification. After completingthis review for a given State, EPD created a file containingthe CFN’s of records criticized in the table review, andprinted out the list in the DPD office in Jeffersonville, IN.The DPD staff used the CFN listings to pull the report formsreviewed in the criticism-resolution process and estab-lished county folders containing the report forms requiringreview. The DPD staff worked through each county filecontaining criticisms in criticism order and by CFN withincriticism, using the interactive minicomputer systems tocheck each individual record for identifiable problems.

When all of the required corrections for all counties in aState were completed they were reviewed by a supervisorystatistician in Jeffersonville. Once all corrections had beenconfirmed and the entire State was approved, the datawere ready for data review and release for Volume 1tabulations.

Tabulation for counties, States, and the United States.As data are keyed, edited, and reviewed, they wereincorporated into the master data matrix for the census.There was a master matrix for each State and for theUnited States, each containing the number of farms anddata values for every item defined in the master matrixdictionary. The data were stored in the master matrix in two‘‘universes’’—all farms, and farms with annual value ofsales of $10,000 or more.

The census tables, that is, the aggregation of data inrows and columns, were populated using the data in themaster matrix. County and State table data were takenfrom the master matrix, while State cross-tabulation data


were taken from the detail data file. The tables for geo-graphic areas above the State level—i.e., for the UnitedStates—were compiled by summing data from the indi-vidual States. Selected tables included historical data fromprevious censuses. The 1992 publication program did notinclude any advance reports; the tables were released inthe Volume 1, Geographic Area Series reports.

Disclosure analysis. The Census Bureau is prohibited bylaw from publishing information that could be used toidentify individual respondents. To ensure that this confi-dentiality is maintained, all of the data were checked priorto publication in a procedure called disclosure analysis.This involved analysts’ reviewing the data tables that haditems suppressed that, if published, would; (1) result indirect disclosure of data reported by a respondent, or (2)reveal information about a respondent by derivation—thatis, by a data user adding or subtracting a publishedsubtotal from a published total to reveal individual data.

The disclosure guidelines set lower limits on the numberof farms that were required to have reported an item beforeit was published. Since some tables included identicalinformation arranged under several different classifica-tions, the suppression of data in one table required thesuppression of the same data in other tables. Publishingthe number of farms in a particular size or other categorywas not considered a disclosure.

The Census Bureau’s automated equipment and pro-grams performed the bulk of the analysis and suppression,but Agriculture Division staff carried out interactive tablereview using the Tabulation and Disclosure System (TADS).

Table review and preparation. For the 1992 agriculturecensus, the Census Bureau introduced a new table reviewsystem—TADS—to simplify the table review process byminimizing programming resources, reducing table adjust-ment errors, and eliminating nearly all of the paper involvedin previous table reviews.

The TADS operated on DEC workstations. (For informa-tion on the procurement of the hardware for the TADS seeChapter 2, ‘‘Planning and Preliminary Operations.’’) TheTADS procedures incorporated status tracking, interactivetable review, and data flow. The status tracking washandled through a combination of a code managementsystem (CMS) and relational database procedures. TheCMS tracked the status of each State file and eachtable—whether pending review, approved, rejected, etc.

Table review was conducted using a spreadsheet pro-gram. Each table had a template—a ‘‘shell’’ containingvarious reference files, as follows:

x Table sourcing document. A grid consisting of rowsand columns giving the master matrix location or key-code sourcing for each data cell in a publication table.

x Data relationships. The relationships of master matrixitems written in equation form, used to accomplishcomplementary disclosure on data in a table.

x Master matrix dictionary (relevant to specific projects,such as the Volume 1).

x Table cross reference. Cross references to other pub-lication tables where particular item is located.

x Variable crop reference. Information on which cropsare published for a given State.

x Acronyms. A file of the acronyms used in the mastermatrix and associated with the published tables.

The data files (published and unpublished) provided byEPD were merged with the templates to create the tablesthat the analysts reviewed. The TADS could interactivelydisplay for any data cell in a table the corresponding itemscontained in the template. Three types of tables werereviewed—

1. State tables with master matrix sourcing in each cell.

2. Cross-tabulation tables with a combination of detailand master matrix sourcing in the stub and header ofthe table, and cells tabulated by the intersection of thestub and header sourcing.

3. County tables with master matrix sourcing for the firstcolumn only.

Analysts, using their workstations, clicked on buttons toperform ‘‘lookups’’ of the sourcing of an individual cell, thetable cross reference(s) of cell, etc. The analysts checkedthe tables primarily for accuracy and completeness of thesuppression patterns, and could, with supervisory approval,make changes to the data. Any changes were written to aState-level transaction file, which then was electronicallytransferred to the EPD to update the data contained in themaster matrix for the State involved.

TADS processing began in October 1993 and continuedthrough August 1994. After the tables for a State had beenprocessed through the TADS, the EPD generated thetables using the Table Image Processing System (TIPSII—a processing system employed for large, centralizedcomputer-based tables), and transmitted these tabulationsfor AGR review. Upon completing the TADS processingand review, the tables for each State were released forpublication preparation.


Chapter 7.

Map—Puerto Rico-Regions and Municipios ---------------------------------------------------- 79

Introduction --------------------------------------------------------------------------------------- 80Historical Background ------------------------------------------------------------------------- 80Uses of Agriculture Census Data ------------------------------------------------------------- 80Legal Authority and Special Agreement ------------------------------------------------------ 80Farm Definition--------------------------------------------------------------------------------- 81

1992 Census of Agriculture ------------------------------------------------------------------- 81Census methodology------------------------------------------------------------------------ 81Scope and content of the census ---------------------------------------------------------- 81Reference periods and dates--------------------------------------------------------------- 82Data collection------------------------------------------------------------------------------- 82Data processing and publication ----------------------------------------------------------- 82

Planning ------------------------------------------------------------------------------------------ 82General Plans---------------------------------------------------------------------------------- 82Interagency Planning Committee ------------------------------------------------------------- 82

Preparatory Operations -------------------------------------------------------------------------- 82Report Form Design --------------------------------------------------------------------------- 82Mail-List Development ------------------------------------------------------------------------- 83Sample Design and Selection ---------------------------------------------------------------- 83

Printing and Addressing Report Forms------------------------------------------------------- 84Printing report forms and assembling mailout packages --------------------------------- 84Address labels------------------------------------------------------------------------------- 84Maps ----------------------------------------------------------------------------------------- 84

Field Organization ----------------------------------------------------------------------------- 84Field office organization--------------------------------------------------------------------- 84Areas of responsibility ---------------------------------------------------------------------- 85Recruiting and training ---------------------------------------------------------------------- 85Training and reference materials ----------------------------------------------------------- 85

Agricultural Extension Office Support--------------------------------------------------------- 85General activities ---------------------------------------------------------------------------- 85Farmer assistance activities ---------------------------------------------------------------- 85

Publicity ------------------------------------------------------------------------------------------- 86

Data Collection ----------------------------------------------------------------------------------- 86General Information --------------------------------------------------------------------------- 86Mailout and Mail Followup -------------------------------------------------------------------- 86Telephone Followup Operations -------------------------------------------------------------- 87

Area Sample Procedures --------------------------------------------------------------------- 87Field canvassing ---------------------------------------------------------------------------- 87Quality control ------------------------------------------------------------------------------- 87

Data Processing---------------------------------------------------------------------------------- 88General information---------------------------------------------------------------------------- 88

Contents

Page


Data Processing—Continued

Precomputer Processing ---------------------------------------------------------------------- 88Receipt and check-in------------------------------------------------------------------------ 88Consistency and coverage review --------------------------------------------------------- 88Data entry------------------------------------------------------------------------------------ 88

Computer Processing-------------------------------------------------------------------------- 89General information ------------------------------------------------------------------------- 89Computer edit and imputation -------------------------------------------------------------- 89Sample estimation -------------------------------------------------------------------------- 89

Tabulation and Data Review ------------------------------------------------------------------ 90Tabulations----------------------------------------------------------------------------------- 90Table review --------------------------------------------------------------------------------- 90Disclosure analysis-------------------------------------------------------------------------- 90

Publication Program------------------------------------------------------------------------------ 90


Page


Map not available.

The 1992 Puerto Rico Census of Agriculture

INTRODUCTION

Historical BackgroundThe Census Bureau carried out the first agricultural

census of Puerto Rico as part of the 1910 decennialcensus program, and the Commonwealth continued to becovered in the decennial agricultural censuses from 1910through 1950. The U.S. Congress authorized quinquennialcensuses of agriculture of the United States in 1915(although the first mid-decade agricultural enumerationwas not done until 1925), but Puerto Rico was not includedin this program until 1959. (The Puerto Rico Reconstruc-tion Administration conducted a special census of agricul-ture in the Commonwealth in 1935, but this was a localeffort.)

Congress modified the schedule of the agriculture cen-suses in 1952, requiring that they be conducted every 5years for years ending in ‘‘4’’ and ‘‘9.’’ The legislationchanging the dates of the censuses did not include PuertoRico in the program, but the 1959 agricultural census in theCommonwealth collected data for the 1958-1959 crop yearand the data were published as part of the 1959 censuspublication program. Thereafter, Puerto Rico remained partof the regular quinquennial agricultural enumeration.

In 1972 Congress changed the census schedule onceagain, directing that the agricultural and economic cen-suses be conducted for the same reference periods, andauthorizing the Census Bureau to shorten the intercensalintervals between the 1974 and the two subsequent agri-cultural censuses by 1 year so that the census schedulesconverged by 1982 with a minimum disruption of censuswork. Following the 1977 Economic Censuses, however,the Census Bureau advanced the date of the agriculturalenumeration in Puerto Rico by a full year to take advantageof the offices and office staff organized for the economicprogram. The agency repeated this arrangement for thefollowing agricultural census, although the alteration of theschedule for the census in the 50 States meant that thePuerto Rico enumeration began just as the statesidecensus completed data collection.

For the 1987 and later censuses, the Census Bureauassigned the Agriculture Division (AGR1) responsibility for

both the agricultural and the economic censuses in PuertoRico. This enabled the agency to take advantage of someeconomies of scope created by combining control of thecensuses in one area. The field office supervised censusoperations within the Commonwealth, including the fieldenumeration of the area sample farms, and field andtelephone followup of nonresponse cases.

Uses of Agriculture Census Data

The census of agriculture is the principal source ofagricultural production data for Puerto Rico and is the onlysource of consistent, comparable data at the municipiolevel. Census data are used by—

x The Federal Government to administer programs, includ-ing such things as relief efforts after hurricanes.

x Local governments to develop and change farm pro-grams, measure the effects of these programs, bench-mark their own data-collection activities, and administera variety of other programs.

x Private industry in planning production and distributionof its products, as well as in designing and implementingmarketing programs aimed at the agricultural commu-nity.

Legal Authority and Special Agreement

Title 13, United States Code, sections 142(a) and 191require that the quinquennial censuses of agriculture foryears ending in ‘‘2’’ and ‘‘7’’ cover the Commonwealth ofPuerto Rico. The specific features of the census aregoverned by the provisions of Title 13 and the Common-wealth of Puerto Rico, Law No. 11—to the extent theprovisions of such legislative acts are not in conflict withFederal law. The Director of the Bureau of the Census andrepresentatives of the Commonwealth Government drewup an agreement concerning the conduct of censuseswithin the Commonwealth in 1958, and prior to eachsucceeding census officials of the Census Bureau and theGovernment of the Puerto Rico met to conclude similaragreements establishing their respective responsibilitiesand functions in each enumeration. The successive agree-ments have been added to the original 1958 accord asamendments. In October 1991, the Census Bureau and theCommonwealth Government signed the ‘‘Memorandum ofAgreement for Conducting the 1992 Agriculture and Eco-nomic Censuses in Puerto Rico,’’ which became Amend-ment XIV to the 1958 document.

1The Census Bureau reorganized its Economic Directorate in October1994, and the Agriculture Division (AGR) became the Agriculture andFinancial Statistics Division (AGFS). Since the bulk of the work on the1992 Census of Agriculture Census for Puerto Rico was carried out underthe old organization, the History will use the AGR designation whenreferring to activities of the division throughout.


The Commonwealth Government agreed to—

1. Secure adequate space for testing and training enu-merator personnel.

2. Encourage a wide variety of organizations and agen-cies to participate in recruiting qualified candidates forcensus jobs.

3. Assist the Census Bureau in publicizing the census.

4. Assist the Census Bureau in obtaining planning infor-mation from sources within the Commonwealth and incertifying boundary information and names for geo-graphic areas to be used for data tabulation andpublication.

The Census Bureau had final responsibility for planningand carrying out the enumeration, including appointing thecensus supervisor; assigning Census Bureau staff as cen-sus advisors; interviewing, testing, selecting, appointing,and training persons, and determining pay rates for thecensus staff; and maintaining confidentiality of the censusdata. The agency also had final authority to determinereport form content and design, enumeration procedures,tabulations to be made, and data to be published, butagreed to consult advisory committees and interestedagencies of the Government of Puerto Rico on thesematters.

The Census Bureau bore the entire cost of the basiccensus program, with the Commonwealth Governmentresponsible for the cost of any expansion of the basiccensus program (there was none). The agency also agreedto make available to the Puerto Rico Planning Board, on acost-reimbursable basis, special statistical tabulations andtapes not included in the publication program. (The Com-monwealth Government had the authority to publish theseadditional tabulations if they complied with census require-ments.)

Farm Definition

The farm definition for the 1992 agriculture census inPuerto Rico was based on value of sales of agriculturalproducts during the reference period. A place qualified as afarm, for census purposes, if it had, or normally could beexpected to have, $500 or more in sales of agriculturalproducts in the 12 months preceding June 30, 1993. Thisdiffered from the definition used for the 1987 census in thatit no longer included operations with 10 or more cuerdas2

of land and sales between $100 and $499. The newdefinition represented the Census Bureau’s attempt tofocus more on commercial, rather than subsistence, farm-ing.


Census methodology. The 1992 Census of Agriculture forPuerto Rico differed significantly from earlier censuses inthe Commonwealth in that it used a mailout/mailback

strategy as the principal means of data collection. Inaddition, an area sample, conducted by personal interview,supplemented the mailout and provided coverage of smallfarms. In previous censuses, a field enumeration staffcanvassed the island searching for places that met thefarm definition, but this method was increasingly expen-sive, burdensome to nonfarmers, and yielded low qualitydata. The Census Bureau decided to redesign its method-ology to reduce nonfarm response burden, contain costs,and improve the quality of the data.

The Census Bureau mailed report forms to farm opera-tors with an expected total value of production of $2,500 ormore and/or at least 20 cuerdas of land in the 1987 census,and newly identified farms that began operation after 1987.Farm operators were requested to complete and returntheir forms within 21 days—telephone and personal inter-viewers followed up nonrespondents. The area samplewas designed to provide estimates for small farm opera-tions (i.e., those with sales of less than $2,500). TheCensus Bureau adjusted the data from this sample torepresent all farms that were not mailed report forms. (Fordetails on sample design, see below.)

Scope and content of the census. The basis of theagriculture census was the individual operating unit—usuallythe individual farm. The 1992 agriculture census in PuertoRico covered all farms that met the census definition, butmade extensive use of sampling.

The census requested data on—

x Land (cuerdas) and land use in the last 12 months.

x Crops (cuerdas harvested and production).

x Irrigation (cuerdas irrigated and major source of water).

x Livestock, poultry, aquaculture, and other animalspecialties.

x Total value of sales (crops, livestock, and aquaculture).

x Farm-related income.

x Type of organization.

x Operator characteristics.

x Fertilizers and agricultural chemicals used.

x Production expenses.

x Machinery and equipment.

x Hired workers, agregados3, and sharecroppers.

2A cuerda is approximately .97 acre.

3An agregado is a member of a family living on a farm not operated byany member of the family. An agregado might or might not be anemployee of the farm operator and might or might not produce anyagricultural products. An agregado might own, rent, or use rent-free thehouse he/she lives in. The land operated by, livestock belonging to, andthe products marketed by an agregado are included in the totals of thefarm operator in charge of the place on which the agregado lives.


Reference periods and dates. The census requestedland, land use, production, expenditure, farm labor, andsales data for the 12 months between July 1, 1992 andJune 30, 1993. Data on inventory (livestock, poultry, andhogs), machinery and equipment, buildings and facilities,and number of sharecropper and agregado families, wererequested as of July 1, 1993.

Data collection. The bulk of the data for the 1992 agricul-ture census in Puerto Rico was collected by mail. TheCensus Bureau assembled a mailing list of farms that hadreported sales of $2,500 or more, and/or 20 cuerdas ormore of land, in the 1987 census in the Commonwealth,and mailed report forms to approximately 14,500 addressesin June 1993. The initial mailout was followed by a remindercard sent to all addresses on the initial list, and bytelephone followup of nonresponse cases by the censusfield office’s staff. In addition, an area sample of smallerfarms in selected barrios was carried out by enumerators,who visited and interviewed agricultural operators notincluded on the census mail list.

Data processing and publication. The Puerto Rico fieldoffice handled the data collection phase of censusoperations—other than mailout and mail followup—including interviewing sample farms and telephone andfield followup. The completed report forms wereprocessed and the data transcribed for editing andtabulation at the Census Bureau’s Data PreparationDivision’s (DPD’s) office in Jeffersonville, IN. The DPD staffused interactive systems employed in processing thecensus for the 50 States to check-in, review, and keyresponses to the census data file. The Outlying AreasBranch of the AGR handled the majority of work involvingcoverage review, correspondence, data review, disclosureanalysis, table preparation, and table review. DPD staffassisted with pre-key processing and performed edit cor-rections.

The data from the 1992 Census of Agriculture in PuertoRico were published in a printed report—Volume 1, Geo-graphic Area Series, Part 52, Puerto Rico.

PLANNING

General Plans

Planning for the 1992 Census of Agriculture in PuertoRico began early in 1991, when the Census Bureau beganpreliminary planning for data content and enumerationmethodology, and the Commonwealth Government estab-lished an interagency planning committee (see below) toconsult with the Census Bureau regarding the census. Theobjective of the preliminary planning was to reduce respon-dent burden and improve data quality without significantlyincreasing costs. By July 1991, the Census Bureau hadproposed the final content for the Puerto Rico report formand had developed plans for using mail enumeration in the

agriculture census. During the early fall of 1991 the CensusBureau and the Commonwealth Government drew up thespecial agreement covering the two parties’ responsibilitiesin the enumeration (see above for details of the agree-ment), and began preparations for the census.

By the end of 1991, the general plan for the census inPuerto Rico called for a mail enumeration of large farms(those with sales of $2,500 or more, or 20 cuerdas of landor more), and an area sample to collect data for smalleroperations.

Interagency Planning Committee

The Government of Puerto Rico organized an informalcommittee composed of representatives of various agen-cies of the Commonwealth Government concerned withthe agriculture census. The following offices or agencieswere represented on the committee:

x Bureau of the Census.

x Puerto Rico Planning Board.

x Puerto Rico Department of Agriculture.

x Puerto Rico Farm Bureau.

x Puerto Rico Agriculture Extension Service.

x University of Puerto Rico

x Agriculture and Economics Department.

x Puerto Rico Rural Development Corporation.

x Credit and Rural Development Corporation.

x Sugar Corporation.

x Agricultural Development Administration.

x Bank of Puerto Rico.

Census Bureau officials met representatives of themember agencies and offices periodically, and communi-cated with them on a continuing basis, beginning in March1991, to discuss plans for report form content and enu-meration methodology.

PREPARATORY OPERATIONS

Report Form Design

The Outlying Areas Branch of the AGR, with the coop-eration of the Puerto Rico Planning Board, designed asingle report form for the agriculture census in Puerto Rico,the form 92-A1(PR)(SP). The report form was an 8-1/2″ x11″ 8-page booklet, of white stock with printing and shadingin brown ink. The standard version was in the Spanishlanguage (hence the ‘‘SP’’ suffix); an English-languagereference version also was produced with black ink ongreen paper.


During review of data users’ recommendations, thereport form design team decided to delete the section onproduction for home consumption. In addition, tobacco,molasses grass, milo, and elephant grass were no longerprelisted on the form, although the respondent could writethem in under ‘‘other crops.’’ The value of sales for indi-vidual vegetable crops was dropped and replaced by aquestion asking for value of all vegetables and melons.

Two new questions—asking the number of cuerdasplanted for individual fruit trees, and aquacultureproduction—were added to the form.

The final version of the 1992 report form consisted of 23numbered sections and asked for data on the following—

x Cuerdas owned, rented, or used in the last 12 months.

x Agricultural products (including sugarcane, coffee, fruits,grains or farinaceous crops, vegetables or melons, lawngrass or ornamental plants, grasses or seedlings) har-vested and products sold.

x Cultivated and/or improved pastures.


x Hired farm workers.

x Irrigation.

x Land use.

x Cattle or calves, hogs, poultry, and their products sold.

x Aquaculture products for sale.

x Livestock inventory and sales.

x Farm related income.

x Farm organization.

x Characteristics and occupation of operator.

x Agricultural chemicals used.

x Production expenses.

x Machinery equipment, buildings, and facilities on farms.

x Number of agregado and sharecropper families on farms.

Mail-List Development

The mailing list for the 1992 Census of Agriculture inPuerto Rico was compiled from four principal sources—(1)the 1987 agriculture census database, (2) the Puerto RicoAgriculture Department’s (PRDA’s) general farm list, (3)the University of Puerto Rico’s Extension Service’s (UPR-ES’s) farm list, and (4) the Puerto Rico Poultry Industry list.The AGR sorted the lists from the PRDA by region,municipio, and last name, and submitted the resulting file tothe Economic Programming Division (EPD) for processingfor mail list production. The EPD edited the mail list files to

eliminate blank lines and add sequence numbers to recordsto provide a unique identification during manual review andas a reference number for matching and deleting duplicateaddresses from the lists.

The edited files containing the PRDA address lists werematched against the 1987 census database for PuertoRico to identify duplicate addresses. The EPD deletedpositive duplicates from the final file, then reviewed the fileonce again to select all records with a total value ofproducts sold (TVP) of $2,500 or more, or with a TVP of atleast $500 and a minimum of 20 cuerdas of land (theserecords were considered to represent ‘‘certainty’’ farms,i.e., farms that would receive the census report form in themail). To develop estimates on agricultural operations thatdid not meet the minimum criteria established for themailing list—as well as any new operations that were noton the mailing list—the Census Bureau designed a statis-tical area sample to account for smaller farms (see below).

After editing, matching, and deleting all identified dupli-cate records, the Census Bureau merged all four majorsource files to create the 1992 Puerto Rico mail list file.After completing mail list compilation from available sources,the Census Bureau employed a private contractor in PuertoRico, who had been authorized by the U.S. Postal Serviceto process USPS address files, to update and standardizethe census mail list. After completing this procedure, thecontractor returned the standardized mail file to the CensusBureau for the census mailing.

Sample Design and Selection

The agriculture census in Puerto Rico used sampling tocollect data from selected municipios and developed sta-tistical estimates of agricultural operations at the Common-wealth and municipio levels. The census used a mailsegment that included all certainty farms and all noncer-tainty farms in specified enumeration districts (ED’s), andan area sample comprising all farms in selected ED’s ineach municipio. The sample ‘‘universe’’ for the Puerto Rico,that is, the group or area from which the sample wasselected, consisted of farms in all the ED’s in Puerto Rico.4

The Census Bureau used records from the 1987 PuertoRico Census of Agriculture file to identify ED’s that includedcertainty or noncertainty farms, based on the 1992 certainty/noncertainty definition. Certainty farms were those withreported sales of $2,500 or more, and 20 cuerdas or moreof land. All other farms were noncertainty operations.Certainty/noncertainty status was coded on each farmrecord. The Census Bureau classified individual ED’s ascertainty or noncertainty based on the total number of

4The single exception to this was the single ED comprising the entiremunicipio of Catano, just west of San Juan. Catano has never reported afarm in the agriculture census. The Census Bureau contacted the PuertoRico Government prior to the mail-list compilation to determine whether ithad any indication that a farm or farms existed in Catano, and upon beinginformed that there were none, excluded the municipio from the areasample.


noncertainty farms identified in each; an ED with four orfewer noncertainty farms became a certainty ED. Thecensus involved four categories of records

1. Certainty farms. TheCensus Bureau identified 10,659certainty farms for the 1992 census in Puerto Rico.Additional farms were added from special lists suppliedby Puerto Rico Government agencies (see mail listdevelopment above). To ensure coverage of this seg-ment of agricultural activity (which accounted for 97.6percent of the value of agricultural products sold inPuerto Rico in 1987); certainty farms were part of themail enumeration.

2. Noncertainty farms in certainty ED’s. The CensusBureau identified all ED’s with fewer than four noncer-tainty farms from the 1987 Census of Agriculture nameand address list and designated them as certainty mailED’s. The census included 293 certainty ED’s that, in1987, had contained 308 noncertainty farms.

3. Noncertainty farms from a sample of noncertaintyED’s. The Census Bureau used an area sample tocollect data from noncertainty farms in noncertaintyfarms in noncertainty ED’s. In each municipio, theagency grouped noncertainty ED’s into clusters withapproximately equal numbers of noncertainty farms.The area sample randomly selected one cluster fromeach municipio and the field interview staff canvassedall farms in the cluster. The sample included a total of108 ED’s.

4. ‘‘New’’ (i.e., newly identified) noncertainty farmsfrom a coverage sample of certainty mail ED’s. TheCensus Bureau selected a sample of certainty ED’s—3for each agricultural region—for canvassing to identifyany farms missed by the mail list compilation opera-tion.

Printing and Addressing Report Forms

Printing report forms and assembling mailout pack-ages. Private contractors printed the report forms, enve-lopes, instructions sheets and other enumeration materialsand assembled the mailing packages before deliveringthem to the DPD office in Jeffersonville, IN. The quantitiesof report forms and principal associated materials printedare shown in table 7-1.

Table 7-1. Report Forms, Envelopes, Letters andOther Enumeration Materials Printed

Form Description Quantity

92-A1(PR)SP . . . . . Report form (Spanish) 35,00092-A1(PR) . . . . . . . . Report form (English) 1,00092-A1(PR)L1(SP). . Cover letter (Spanish) 35,00092-A1(PR)L1 . . . . . Cover letter (English) 1,00092-A1(PR)SP(I) . . . Information sheet (Spanish) 35,00092-A1(PR)L2. . . . . . Follow-up postcard (Spanish) 20,00092-A7A(PR) . . . . . . Outgoing envelope 35,00092-A8(PR) . . . . . . . . Return envelope 35,000

Address labels. The Census Bureau prepared an addresslabel for each address on the mail list. Each label containedthe printed address and a machine-readable barcodecontaining the address as well as size and farm-type codesfor the addressee. The EPD provided the mail-address fileto the DPD in the second week of June and the DPD usedthe high-speed Printronix printers to produce the addresslabels for the mailing packages. Clerks affixed the labels tothe report forms through the open windows of the outgoingenvelopes. Labeled mailing packages were packed incartons (each containing approximately 225 mail pack-ages) according to postal requirements for presorted first-class mailings (i.e., by 3- and 5-digit ZIP Code), and sentfor mailout.

Maps

The area census office and the field enumeration staffneeded maps to carry out the area sample. Four officemaps, showing municipio boundaries and names, wereprepared for use in the Hato Rey field office, together withfour sets of 123 enumeration district (ED) maps for theED’s selected for the agriculture census area sample. The1992 ED’s corresponded generally with those of the 1990census—that is, they generally were drawn within therecognized barrio boundaries, although some boundarieswere altered in built-up areas to allow for changes in landuse.

The ED maps were scaled to fit on a single page (8-1/2″x 11″) and had the same level of detail as the 1990decennial census maps, showing roads, trails, water fea-tures, geographical boundaries, landmarks, and powerlines. Office maps were wall maps.

The census maps for Puerto Rico were produced by theCensus Bureau’s Geography Division, using it’s automatedgeographic database, the Topologically Integrated Geo-graphic Encoding and Referencing (TIGER) system—developed in a cooperative effort with the U.S. GeologicalSurvey (USGS). The TIGER system contained informationfor each geographic feature (e.g., the various individualsegments that make up roads, railroads, rivers, etc.) suchas geographic area codes, latitude/longitude coordinatesof boundaries and features, and the name and type of anyfeatures. Geography Division delivered the various mapsrequired for the agriculture census in Puerto Rico in earlyApril 1993.

Field Organization

Field office organization. The Census Bureau estab-lished a field office in the San Juan suburb of Hato Rey toprovide an administrative headquarters for the 1992 agri-culture and economic censuses in the Commonwealth. Theoffice opened in December 1992, after the Field Divisionrented suitable office space and rented or borrowed thenecessary furniture. The Field Division assigned an employeeto the office as office manager while the remaining officeand field staff were recruited locally and employed for theduration of census operations in Puerto Rico.


The field office was organized under an area censussupervisor, with an office manager responsible for admin-istrative activities in support of the census (personnel,payroll, clerical cost and progress reporting, supply requi-sitioning, etc.) as well as the telephone followup staff, anda field operations supervisor who recruited, selected, trained,and supervised the field interview staff. The office managerhad 3 lead clerks, 2 office clerks, and 10-15 telephoneclerks (the staff size varied based on the level of activity inthe field office; the highest staffing levels were attainedduring periods (May-early July) when substantial activitieswere underway for both the economic and agriculturecensuses) to assist in office operations, while the fieldoperations supervisor oversaw the field enumerator staff.

Areas of responsibility. The Hato Rey office and the fieldstaff carried out the data-collection phase of the censuswhile the Census Bureau’s headquarters staff providedsupervision and technical assistance. The Census Bureau’sField Division had primary operational responsibility for thefield office and the field enumeration staff, while the Outly-ingAreas Branch of AGR, and the DPD in Jeffersonville, IN,prepared mail lists and mailed out report forms. (The mailportion of the agriculture census began in June. Thearea-sample data-collection effort began in June and lastedthrough mid-September.)

Respondents to the mail census returned their com-pleted report forms to the Jeffersonville office. The reportforms completed in the area sample canvass, and by thetelephone followup operation at the Hato Rey office, wereshipped to Jeffersonville. The report forms and their datawere processed by the staff at Jeffersonville, and the AGRstaff at Census Bureau headquarters, using interactivesystems linking both offices.

Recruiting and training. The Hato Rey field office openedon December 7, 1992, but data collection for the agriculturecensus did not begin until the following June. Prior to theagriculture census the office worked on the economiccensuses in the Commonwealth. Most of the office stafforiginally recruited for the economic censuses stayed onfor the agriculture enumeration as well. The agriculturecensus field staff—12 crew leaders, who would be theimmediate supervisors of the field enumeration staff, and80 enumerators—was recruited in April and early May1992, with selection based on scores on standardizedtests. The office and field staffs were salaried, temporaryemployees of the Census Bureau and were paid at stan-dard U.S. Government general schedule pay scales for theappropriate grades.

The field staff underwent training during the week of May24. Census Bureau staff conducted a 3-day training ses-sion for the crew leaders, covering enumeration proce-dures, coverage and quality control procedures, adminis-trative requirements, and so on. The crew leaders theneach trained the enumerators they would be supervising.

Training and reference materials. The AGR and FieldDivisions prepared training and reference guides for use in

the agriculture census in Puerto Rico. The principal admin-istrative reference used in the field office was the OfficeProcedures Manual. The manual covered basic adminis-trative procedures for the field office, including local tele-phone followup operations and processing activities. Thefield operations assistant was responsible for training thecrew leaders of the field enumeration staff, and used theForm A53(PR)SP, Guide for Training Crew Leaders as theprimary training manual. The individual crew leaders eachreceived a copy of the Form A7(PR)SP, Crew Leader’sManual, for their own reference, and used the FormA52(PR)SP, Guide for Training Enumerators for training theenumeration staff.

Each enumerator received a copy of the FormA10(PR)SP,Enumerator’s Manual as the primary reference for the fieldenumeration. The enumerators also were provided withenumerator kits that included Form A5, Record Book, a listof large or special farms in the assigned ED, and a supplyof form 92-A1(PR)SP report forms, and a map of theassigned ED. The list of large and special farms includedthe names and addresses of all farm operators that hadbeen mailed a report form in the June mailout. Enumera-tors had to verify that the farmer had returned a completedcensus form, or, if no report form had been received, to visitthe nonrespondent address and complete a report form bypersonal interview. The Record Book contained a supply ofform A2(PR), Listing Sheets, which served as a record ofthe canvass of the ED. Each A3 sheet included a listing ofthe screening questions enumerators were to use to iden-tify households that qualified as farms under the censusdefinition and space for listing each household canvassed.

Agricultural Extension Office Support

General activities. The UPR-ES functions in the samefashion as the U.S. Department of Agriculture’s ExtensionService, that is, local offices assist farmers with informationand advice on agricultural programs, problems, legal ques-tions, and the like. The local offices have considerableknowledge of farming and farmers within their areas. TheUPR-ES assisted the Census Bureau by providing its ownfarm list for the Census Bureau’s census mail list compila-tion effort, in distributing publicity materials provided by theCensus Bureau and promoting the enumeration amongfarmers in personal contacts, and by providing help tofarmers in completing the census report forms.

Farmer assistance activities. The UPR-ES’s network oflocal offices offered an obvious source of assistance tofarmers who were being asked by the 1992 agriculturecensus to complete report forms distributed by mail. TheCensus Bureau conducted a series of four 3-hour trainingpresentations for ES agents during the last week of May1993 (held at the ES regional offices in Arecibo andCaguas, at the Tropical Research Station in Mayagez, andat the Agricultural Research Station in Rio Piedras) tofamiliarize the agents with the census program and toprepare them to answer questions from farmers. Approxi-mately 100 ES agents received the training, which covered


an overview of the census, data collection methodology,the ES role in the census, and the 1992 report form, andconcluded with practice exercises to familiarize the agentswith the task of filling out the report form.

Publicity

The Census Bureau’sAGR and Public Information Office(PIO) cooperated in developing the publicity plan for the1992 agriculture census in Puerto Rico. The program hadthree major objectives—

x To aid in recruiting staff for the census operation.

x To encourage cooperation and prompt response byfarmers to the census mail enumeration and area samplesurvey operation.

x To provide information to the public about the release ofcensus data products.

The Census Bureau staff developed several items spe-cifically for the publicity effort in Puerto Rico. The printedmaterials included two posters, an agriculture censusinformation packet, a newsletter article with general infor-mation about the census (including timing, data collected,uses of the data, and so on), two information brochures(one containing general agriculture information and onespecifically describing the agriculture census data), and aseries of three press releases (‘‘America Counts on PuertoRico’s Agriculture,’’ ‘‘Puerto Rico Counts on CompleteAgriculture Census Data,’’ and ‘‘Puerto Rico’s 13th CensusofAgricultureGetsUnderway’’); while public service announce-ments (PSA’s) for use by local radio stations were recordedand distributed as well. The posters came in a large (11″ x14″) and small (8-1/2″ x 11″) size, and announced thecensus and that census staff job openings were available.In April 1993, 1,300 copies of the posters were distributedthrough local government offices and businesses for dis-play in windows and on bulletin boards. The informationpacket contained—

x A transmittal letter.

x A sheet of frequently asked questions about the census,with answers.

x Copies of the Puerto Rico report form (92-A1(PR) andinstruction (form 92 A1(PR)(I) sheet.

x The agriculture census information brochure.

x A census telephone contacts list.

x A copy of the small poster.

x The newsletter article.

The agency assembled and shipped the information kitsto the Commonwealth for distribution to (and through) thePuerto Rico Planning Board, Commerce Department, andDepartment of Agriculture; local newspapers; the Small

Business Administration’s Hato Rey field office, the UPR-ES, and the U.S. Postal Service; and local colleges andagriculture-oriented organizations.

The PIO produced a series of four 30-second radiospots and distributed copies to radio stations throughouttheCommonwealth for broadcast as public service announce-ments (PSA’s) during the census. In addition, the agencyproduced a 30-second video spot, featuring the PuertoRican Resident Commissioner, Carlos Romero Barcelo,and distributed copies to television stations in the Com-monwealth for broadcast in July 1993.

The Census Bureau also asked the Governor of PuertoRico to issue an official proclamation about the census. OnJune 28, 1993, the Governor signed a proclamation des-ignating July 1993 ‘‘Agriculture Census Month’’ in theCommonwealth.

DATA COLLECTION

General Information

Early in 1993, the Census Bureau finalized its mail list ofall agricultural operations in the Commonwealth that reportedtotal value of agricultural production of $2,500 or more,and/or at least 20 cuerdas of land, in the 1987 census, andany known farms with these characteristics that beganoperation after 1987. On June 16, 1993, the CensusBureau mailed report forms to the approximately 14,500addresses on its mail list, asking operators to complete theforms and return them within 21 days. The agency maileda reminder/thank you card to all addresses on the mail liston June 30. Nonrespondents were followed up by tele-phone and personal interviews whenever possible. Themail enumeration achieved a 78.2 percent final mail responserate.

In addition to the mail enumeration, the Census Bureaudeveloped and carried out an area sample to collect dataon small farm operations. Selected ED’s in each municipio(except Catano, which had no agricultural operations meet-ing the farm definition) were canvassed by a field staff, andany farms that had not received a report form wereenumerated by personal interview. Approximately 2,500farms were enumerated in the area sample. Their datawere weighted based on the number of farms in the samplearea compared to the number of farms in the municipiofrom the 1987 census.

Mailout and Mail Followup

The DPD mailed 14,468 census report form packagesby first-class mail to farms in Puerto Rico on June 23, 1993.This was the only mailing involving report forms for theagriculture census in the Commonwealth. The CensusBureau carried out a single mail followup to all addresseson the census mail list for Puerto Rico, using Form


2-A1(PR)L2, Reminder/Thank You Postcards, by third-class mail on June 30. The DPD printed the addressesdirectly onto the cards using the ink-jet printers at theJeffersonville, IN, office, then shipped the cards to the HatoRey field office for mailout at the end of the month.

Telephone Followup Operations

By July 21 the Census Bureau had received 2,650completed agriculture report forms in response to themailout, together with some 3,741 forms returned asundeliverable as addressed (UAA). That left approximately8,000 nonrespondents for followup operations. The DPD inJeffersonville, IN, generated adhesive address labels fornonrespondent addresses in descending order of TVP andshipped the labels to the field office in Hato Rey for use ina telephone followup to nonrespondents. Clerks at theHato Rey office researched telephone numbers for thenonrespondent addresses and began telephoning nonre-spondents to try to collect the census data by telephoneinterview, or to encourage the respondent to agree tocomplete and return the report form by mail. If the farmoperator involved was willing to provide the information theclerk interviewed him or her, and filled out a standard A1report form using the data supplied. If the respondent wasnot willing to cooperate in a complete interview, or tocomplete and return a report form, the clerk tried to obtainenough basic production and inventory information (e.g.,total acreage, number of cattle, etc.) to enable the CensusBureau to impute data for the operation based on itsgeneral characteristics and geographic location.

Clerks applied the appropriate adhesive address label,with the bar code required for check-in, to the completedreport form for each telephone case, and shipped thesereport forms to Jeffersonville, IN, for check-in and process-ing.

At the same time, the DPD generated a duplicate set oflabels, by municipio, attached them to one-page question-naires, and shipped them to the UPR-ES. UPR-ES agentsreviewed the cases for their respective municipios andprovided the Census Bureau with any information they hadon each nonrespondent case. By the end of the followup,all cases were resolved as (1) out of scope, (2) will file, (3)completed by telephone, or (4) completed using secondary-source information.

Area Sample Procedures

Field canvassing. The area sample for the agriculturecensus in Puerto Rico covered all the places that met thecensus definition of a farm, but had not been mailed areport form, in 123 selected ED’s (108 selected for thegeneral area sample, and 15 ‘‘certainty’’ ED’s (3 from eachagricultural region in the Commonwealth) chosen as acoverage sample to identify ‘‘new’’ noncertainty farms)throughout Puerto Rico. Two canvassing procedures wereused in the field enumeration. In predominantly rural ED’s

the enumerators visited every household, while in urban-ized ED’s, and in built-up areas (i.e., a group of 25 or morehouses or other structures, each on less than half a cuerdaof land) of rural ED’s, they were instructed to ‘‘consultknowledgeable people’’ to identify persons within the areathat operated farms.

The enumerators began interviews in both urbanizedand rural areas by asking for the name of the head of thehousehold, and a series of screening questions to deter-mine whether the household qualified as a farm. Thequestions asked whether—

1. The respondent had, in the previous 12 months, raised,produced, or sold any crops, vegetables for sale,ornamental flowering plants, or had 1 or more cattle, 1or more pigs, or 15 or more poultry.

2. The respondent had sales of agricultural products inthe previous 12 months, or expected sales this year, of$500 or more.

3. This place had 10 or more cuerdas of land, or sales ofagricultural products in the previous 12 months, orexpected sales this year, or $100 or more.

Enumerators assigned a farm serial number and inter-viewed the respondent to complete a report form for eachplace that met the census farm definition. The enumeratorstried, whenever possible, to interview the head of eachhousehold visited, but when the head was not available,some other responsible adult member of the householdcould be asked to provide the information needed.

As they canvassed each ED the enumerators plottedeach farm, nonfarm operator visited, built-up area, and soon, on their ED maps, so that they and their crew leaderscould monitor each ED’s coverage as the census contin-ued. The enumerators also completed a Form 92-A4(PR)SP,Nonresident FarmOperator/CloseoutDataCard, andassignedan A4 serial number, for any place that qualified as a farm,but (1) contained no housing unit, (2) had no operator(s)living there, or (3) for which no responsible and knowledge-able person could be located to provide the requiredinformation. The A4 card listed the name and address ofthe nonrespondent or absent operator and whatever basicinformation could be obtained from neighbors or othersources. The enumerators gave copies of the A4 cardscompleted each week to their crew leaders who referredthe cases to either the appropriate enumerator within theirown area, or to another crew leader responsible for thecanvass in the area where the absentee operator lived.

Quality control. Crew leaders supervised the field can-vass quality control operation, observing enumerators dur-ing canvassing to ensure that interviewing and coveragemet requirements, and formally reviewing each enumerator’swork during weekly meetings, and again at the completionof canvassing for each ED. The formal review checked—

x Records on the listing sheets with plotted line numberson the ED maps to make certain the ED was completelycanvassed.


x The A3 listing sheets and A4 Nonresident OperatorCards to ensure they had been completed accuratelyand legibly.

x Completed records against the listing sheets.

x Completed report forms for completeness, legibility, andaccuracy.

x Coverage for prelisted large and special farms, and forbuilt-up areas.

Prior to the census, quality control (QC) enumeratorsvisited each area sample ED and, starting from fourrandomly selected locations on the ED map, proceeded ina specified direction and canvassed the first five housesvisited. The QC enumerator carried out a short interview todetermine whether a report form should be completed foreach place and entered the information on A3 ListingSheets. The crew leaders used these sheets to checkcoverage obtained by the regular enumerators for eachED. In cases where a matching address was identified asa farm in the QC prelisting operation, but not in the areasample canvass, the enumerator was counseled aboutprocedures and instructed to revisit the place to resolve thediscrepancy. Crew leaders carried out interviews and com-pleted report forms as needed to ensure complete andaccurate coverage. Typically, the crew leaders handleddifficult refusal cases that enumerators had not been ableto resolve.

DATA PROCESSING

General Information

The Census Bureau’s DPD in Jeffersonville, IN, pro-cessed the report forms from the Puerto Rico agriculturecensus. After check-in and review of any problem cases orcorrespondence attached to completed forms, the DPDstaff keyed the data from the questionnaires. The resultingcomputerized records were subjected to a detailed com-puter edit for consistency and reasonableness. The editalso corrected obviously erroneous or inconsistent data,supplied missing data based on imputation (using charac-teristics from similar farms to impute information), andassigned farm classification codes needed in tabulating thedata. After editing, records were classified as either passedor failed. The DPD staff corrected all failed records untilthey passed the edit, or were determined to be out ofscope.

Before publication, AGR statisticians reviewed the tabu-lations for inconsistencies and potential coverage prob-lems. The 1992 totals were compared to previous censusdata, as well as other available information and anyproblems was examined. When necessary, the staff madecorrections to the data records and retabulated the affectedtotals.

Precomputer Processing

Receipt and check-in. Returned mail cases were checkedin by optical scanning equipment that identified each caseby the bar code on the mailing label, while report formscompleted by personal interview were checked in usingassigned identification numbers keyed directly to the data-base. The first receipts arrived at the DPD office in June,and continued on a flow basis until the last week ofSeptember 1993, when the last of the completed reportforms were shipped to Jeffersonville from the Hato Reyoffice at the close of the data collection operation.

After check-in the report forms were routed to thebatching control unit where the control clerks batched theminto work units of up to 95 report forms using the CATSsystem (see Chapter 6, ‘‘Data Processing,’’ for a details ofthe CATS system), which printed a Data Entry Batch CoverSheet for that batch. At the same time the system acceptedthe batch, it automatically updated the CFN tracking recordto show that the report forms in the batch were now goingto ‘‘data entry.’’

Consistency and coverage review. The Census Bureaumaintained historical data from the 1987 census for PuertoRico in the database used in processing the 1992 agricul-ture census for the Commonwealth. Large cases prese-lected for review were identified by a specific processingsort code and were automatically sorted for review byanalysts. In their review of the individual cases, analystschecked for internal inconsistencies and matched the 1992data for a specific case against the historical file to evaluatethe reasonableness of any changes.

Data entry. Data entry (or keying) involved transcribingdata from the census report forms to a machine-readabledata file for edit and tabulation. The DPD’s Data ServicesBranch (DSB) used a key-to-disk interactive system thatcombined the clerical review of the individual censusquestionnaires with the data entry operation. Each keystation had a keyboard, and monitor that allowed the keyerto display and edit keyed data, as well as receive mes-sages or questions from the input program. Quality controlprocedures included reviewing samples of each keyer’swork and, when necessary, correcting keyer errors andretraining keyers.

The keying unit supervisors distributed work batches tokeyers, who opened the plastic envelopes containing thereport forms and other documents, wrote a keyer/verifieridentification number on the cover sheet, and reviewedeach report form for problems as data were entered.Keyers rejected report forms for data entry and assigned‘‘reject reason codes’’ for any of the following reasons:

Code Reject reason04 More than 99 CFN’s in a batch06 Blank report form08 Maximum value failure (a data field entry

exceeds the maximum acceptable value)10 Report form not keyable


The keyers rejected, but did not key a reject reason codefor, report forms with invalid CFN check digits.

The interactive data entry system program assigned‘‘screens.’’ Keyers used screen 1 to key the CFN, andscreen 2 for entering name and address corrections andadds, and GAC (municipio) information. Screens 3-26 wereused for entering data from sections 1-24 of the reportforms.

After entering any written in corrections to the identifica-tion and address data from the address label area (keyersalso entered these data manually for the area samplecases) and section 1 of each report form, keyers begankeying the data sections. Keyers did not enter the key-codes for data cells; rather, the keycodes were prepro-grammed and assigned to individual data cells based onthe location of the data cell on the screen. Once the keyerentered the data for a specific field, he or she used the fieldrelease or pressed the ‘‘F6’’ function key to go to the nextfield.

The keyer continued on through each report form,entering the various codes as needed. As data were keyedand verified, DSB lead operators transferred the dataelectronically to the Charlotte, NC, facility for further pro-cessing.

Computer Processing

General information. After keying, the data from eachreport form were subjected to a computerized edit. Ana-lysts reviewed and verified any substantial changes gen-erated by the computer edits to the data file prior totabulation. The data were tabulated by municipio and forthe Commonwealth and AGR statisticians reviewed alltabulated totals to identify inconsistencies and potentialcoverage problems. The statisticians carried any requiredcorrections to the individual data records and the specifictotals involved were retabulated. After disclosure analysis,the data file was ready to be released for publication.

Computer edit and imputation. The data from each farmrecord were subjected to a detailed, item-by-item, com-puter edit. This complex edit—

x Determined whether each record represented an agri-cultural operation meeting the census farm definitionand deleted out-of-scope operations from the file.

x Assigned farm classification codes needed for tabulatingthe data, including acreage, tenure, product sales, andtype of organization (SIC) code.

x Checked consistency between and within sections ofeach record.

x Checked for reasonable relationships between and amongdata items, values for various sizes of farms, andcombinations of commodities.

x Imputed missing or obviously erroneous data for farmsbased on information in the same record, or on responsesof similar farms in the same geographic area.

Data records that failed to meet the census farm defini-tion, or that had undergone substantial computer-generatedchanges to the data, were reviewed to ensure that the datahad been keyed correctly and/or that the changes werejustified. Edit referral cases (i.e., cases that failed edit andwere flagged by the computer for review) were reviewed forkeying accuracy to ensure that the edit results werecorrect. Any cases for which the computer edit results werefound to be unacceptable were corrected as required andreedited.

Whenever possible, edit imputations, deletions, or otherchanges were based on related data from the respondent’sreport form. For some items, such as operator character-istics, data from previous censuses could be used. Valuesfor missing or unacceptable reported data were calculatedbased on reported quantities and known prices, or by usinginformation from other, generally similar farm operations.

Sample estimation. The 1992 agriculture census in PuertoRico collected data from only a sample of noncertaintyfarms, so the Census Bureau had to assign weights to thesample farms to account for data not obtained from thosefarms excluded from the sample. The Census Bureau alsoassigned weights to farms in certainty mail ED’s to accountfor farms not on the mail list, and used weighting tocompensate for nonrespondent farms as well.

The agency calculated weights separated for threegroups of farms—

x Certainty farms. For certainty farms, respondent andnonrespondent farms were classified into three stratabased on value of sales reported in the 1987 census,and the Census Bureau calculated the number of respon-dent and nonrespondent farms in each strata. Thenonresponse weight assigned to each respondent farmin each stratum was equal to the sum of the respondentand nonrespondent farms divided by the number ofrespondents.All certainty farms received a sample weightof one, while the final weight for each certainty farm wasequal to the product of the nonresponse and sampleweights; the final weight assigned to a given farm forestimation purposes was equal to the nonresponseweight assigned to that farm.

x Noncertainty farms in noncertainty ED’s. In eachmunicipio, the Census Bureau assigned a sample weightto each noncertainty farm in each noncertainty ED equalto the number of clusters (see above) in that municipio.In addition, in each municipio the agency also assigneda nonresponse weight equal to the sum of the numbersof respondent and nonrespondent noncertainty farms,divided by the number of respondent noncertainty farmsin that municipio. The Census Bureau multiplied thesample and nonresponse weights for each noncertaintyfarm to yield the final weight.


x Noncertainty farms in certainty mail ED’s. In eachregion, the Census Bureau obtained an estimate offarms in certainty mail ED’s that were not on the mail listfrom the coverage sample, and developed estimates offarms not on the mail list at the municipio level using asynthetic estimation procedure. The agency assignedfinal weights at the municipio level to sample respondentfarms in certainty mail ED’s that were equal to the sumof the number of respondent and nonrespondent farmsmailed report forms, and the estimated number of farmsnot on the mail list, divided by the sum of the numbers ofrespondent farms mailed a report form and respondentfarms enumerated for the coverage sample, but not onthe mail list.

The Census Bureau estimated item totals by roundingthe final weight assigned to each farm to an integer, thenmultiplying each farm’s data values for each item by theinteger weight and summing those weighted values over allfarms in the municipio. Repeating the process at theregional level produced the regional estimates, and for allfarms in Puerto Rico for the island-level estimates.

Tabulation and Data Review

Tabulations. For Puerto Rico, the Census Bureau pre-pared and published data tables all farms, and (for sum-mary tables at the Commonwealth level only) for farmstotal value of sales of agricultural products reported of$2,500 or more. Tables showed data for the Common-wealth, the five agricultural regions defined by the PuertoRico Department of Agriculture, and for municipios.

Table review. The Census Bureau’s automated equipmentand programs performed the bulk of the analysis andsuppression, but AGR staff carried out interactive tablereview using the Tabulation and Disclosure System (TADS)procedures prior to release of the data for publication. TheTADS used interactive workstations (for more informationon the TADS see Chapters 2, ‘‘Planning and PreliminaryOperations,’’ and 6, ‘‘Data Processing’’), and the workprocedures incorporated status tracking, interactive tablereview, and data flow. Table review was done using aspreadsheet program; each table had a template contain-ing the various reference files, and the data files weremerged with the templates to create the tables that ana-lysts reviewed.

Analysts could make changes to the data, and onsupervisory approval, the changes were written to theappropriate file, then the file was electronically transferredto the EPD to update the master matrix.

After all the tables had been reviewed and data andsuppression patterns verified, the EPD produced the tablesusing the Table Image Processing II (TIPS II—a processingsystem used for large, centralized, computer-based tables)and transmitted these tables to the AGR for review.

Disclosure analysis. The Census Bureau is prohibited bylaw from publishing information that could be used toidentify individual respondents to any of its censuses or

surveys. To ensure that this confidentiality is maintained, allof the data tabulations are checked prior to publication in aprocedure called disclosure analysis. This involved a reviewof the data tables that identified and suppressed specificitems that, if published, (1) would result in direct disclosureof data reported by a particular respondent individual orcompany or (2) would reveal information about an indi-vidual by derivation—that is, by a user adding or subtract-ing a published subtotal from a published total to revealindividual data.

PUBLICATION PROGRAM

The Census Bureau published the 1992 agriculturecensus statistics for Puerto Rico in Volume 1, GeographicArea Series, Part 52, Puerto Rico. The report showedestimates for all farms in the Commonwealth, for 5 agricul-tural regions, and for 77 individual municipios. Tables 1-15contained data for all agricultural operations in Puerto Rico;tables 16-68 showed municipio-level data; and tables69-71 presented more detailed tabulations for major dataitems for farms with sales of $2,500 or more.

The basic data shown for all farms included the follow-ing:

x Farms, land in farms, and land use.

x Tenure, characteristics, and main occupation ofoperator.

x Hired workers, agregados, and sharecroppers.

x Selected machinery, equipment, and buildings.

x Agriculture chemicals used, including fertilizers.

x Irrigation.

x Selected farm production expenses.

x Market value of agricultural products sold.


x Livestock and poultry (inventory and sales (includingsales of livestock and poultry products)).

x Crops harvested.

x Horticultural specialties.

x Fish and aquaculture (for the Commonwealth).

The tables showed 1992 and comparable 1987 data.For farms with sales of $2,500 or more, tables 69-74

showed summary statistics (i.e., at the Commonwealthlevel) for 1992 classified by tenure, type of organization,main occupation and age of operator, size of farm (cuer-das), market value of products sold and type of farm.

The Census Bureau released data highlights from theprinted report through its CENDATATM ‘‘online’’ system. Inaddition, tables taken from the TADS were placed into aspreadsheet format and provided on flexible diskette to thePuerto Rico Department of Agriculture and the University ofPuerto Rico Extension Service.


Chapter 8.

Introduction --------------------------------------------------------------------------------------- 92Historical Background ------------------------------------------------------------------------- 92Uses of Agriculture Census Data ------------------------------------------------------------- 92Scope and Legal Authority -------------------------------------------------------------------- 92Reference Periods and Dates ---------------------------------------------------------------- 92Farm Definition--------------------------------------------------------------------------------- 92

Preparatory Operations -------------------------------------------------------------------------- 93Planning ---------------------------------------------------------------------------------------- 93General Information ------------------------------------------------------------------------- 93Special agreements ------------------------------------------------------------------------ 93Report form content ------------------------------------------------------------------------- 93

Map—Virgin Islands of the United States-Islands---------------------------------------------- 94Map—Guam-Election Districts------------------------------------------------------------------- 95

Preparation of Enumerator Materials --------------------------------------------------------- 96Printing report forms and enumeration materials------------------------------------------ 96Maps ----------------------------------------------------------------------------------------- 96Record books and enumeration kit--------------------------------------------------------- 96

Staffing and Training--------------------------------------------------------------------------- 96

Data Collection ----------------------------------------------------------------------------------- 97Enumeration Methodology -------------------------------------------------------------------- 97Callbacks --------------------------------------------------------------------------------------- 97Refusals ---------------------------------------------------------------------------------------- 97Field Review ----------------------------------------------------------------------------------- 97Results ----------------------------------------------------------------------------------------- 98

Data Processing---------------------------------------------------------------------------------- 98Precomputer Processing ---------------------------------------------------------------------- 98

Computer Processing-------------------------------------------------------------------------- 98General information ------------------------------------------------------------------------- 98Computer edit and tabulation--------------------------------------------------------------- 99Disclosure analysis-------------------------------------------------------------------------- 99Table review --------------------------------------------------------------------------------- 99

Publication Program------------------------------------------------------------------------------ 100

Contents

Page


1992 Census of Agriculturefor Guam and the U.S. Virgin Islands

INTRODUCTION

Historical Background

The United States purchased the Virgin Islands fromDenmark in 1917 and carried out a special census thatincluded an agricultural enumeration that same year. Theislands were not included as part of the regular agriculturecensus operation until 1930, when the Federal Govern-ment incorporated an enumeration of the islands into thedecennial census program. The first agricultural census onGuam was carried out in 1920, as part of the decennial-census of that year. Thereafter, an agricultural enumerationof the island continued as part of the decennial censusprogram through 1960. In 1964, Title 13, United StatesCode, Section 191(a) was changed to include both the U.S.Virgin Islands and Guam in the quinquennial censuses ofagriculture. (Two other areas—American Samoa and theNorthern Marianas Islands—have been included in theagricultural censuses every 10th year; the data are col-lected during the decennial population and housing cen-sus, but are published as part of the preceding agriculturecensus program. The Northern Marianas Islands became aCommonwealth in association with the United States in1987. Thereafter, Title 13 was amended to include them inthe quinquennial census, beginning with the 1997 enumera-tion.)

Uses of Agriculture Census Data

The census of agriculture is the principal source ofagricultural production data for Guam and the U.S. VirginIslands and is the only source of consistent, comparabledata at the detailed geographic level. Census data areused by the local governments in (1) developing andchanging farm programs, (2) measuring the effects of theseprograms, (3) benchmarking for designing and evaluatingtheir own data collection activities, and (4) for administeringa variety of other programs. Private industry uses censusstatistics in planning production and distribution of itsproducts, and in designing and implementing marketingprograms aimed at the agricultural community.

Scope and Legal Authority

The conduct of the agriculture censuses is authorized byChapter 5 of Title 13, United States Code—Census. Sec-tion 142(a) of Chapter 5 directs that agriculture censuses

be carried out in 1979, 1983, and every fifth year thereafter,and Section 191(a) specifies that the agricultural enumera-tions may cover the U.S. Virgin Islands, Guam, the Com-monwealth of the Northern Marianas Islands, and otherpossessions and areas over which the United Statesexercises jurisdiction, control, or sovereignty. Section 191(b)authorizes the Secretary of Commerce to use data col-lected by the Governors or other Federal officials (providedthe data are collected using plans approved or prescribedby the Secretary) for censuses in any of these places.

In practice, agriculture censuses have been carried outon Guam and in the U.S. Virgin Islands every five yearssince 1964, while agricultural enumerations in the otheroutlying areas generally have been conducted decennially,as a component of the population and housing censusesprogram. (The 1990 decennial census of the Common-wealth of the Northern Marianas Islands included an agri-cultural enumeration (the data were released as part of the1987 Census of Agriculture publication program), so thearea was not enumerated again for the 1992 census.)

Reference Periods and Dates

Reference periods and dates differed between Guamand the Virgin Islands. On Guam, the agriculture censuscollected inventory data (i.e., acreage, numbers of live-stock and poultry, and so on) as of the day of enumeration,while crop and livestock production, sales, and expensedata were requested for the calendar year 1992. In theVirgin Islands, inventory data were asked as of the day ofenumeration as well, but crop and livestock sales, produc-tion, and expense data were for the 12-month periodpreceding June 30, 1993.

Farm Definition

Prior to the 1987 agriculture census, any place on Guamor in the Virgin Islands from which any crop, vegetable, orfruit was harvested or gathered, or on which there were anylivestock or a specified number of poultry, was identified asa farm. The agriculture census in the 50 States beganusing volume of sales of agricultural products as theprincipal criterion for defining a farm in the late 1970’s, butlocal conditions in the areas led the territorial governmentsand the Census Bureau to retain the broader definition until1987. The 1987 Census of Agriculture on Guam and in theVirgin Islands introduced a new farm definition based on


dollar value of sales; any place that had, or normally wouldexpect to have, a total annual value of sales of agriculturalproducts of $100 or more was considered a farm. TheCensus Bureau continued to use this definition for the 1992census in these areas.

PREPARATORY OPERATIONS

Planning

General information. The 1992 agriculture census inGuam and the U.S. Virgin Islands was a cooperative effortof the Census Bureau and the respective territorial govern-ments. Census Bureau staff began meeting with represen-tatives of the governments of the U.S. Virgin Islands andGuam in May and June 1991, respectively, to negotiate thespecial agreements that would govern the conduct of thecensus in each area, and to develop general plans for theenumeration. The Census Bureau also began preparingthe procedures, edit programs, and tabulation programs forhandling the information to be collected.

Initially, the overall designs for the censuses werealmost identical to those for the 1987 agriculture censuses;field interviewers would canvass the islands and collect theagriculture data, using the same definitions employed in1987, and the report forms and data would be processed atthe headquarters.

The general plans for the censuses were formalized inspecial agreements negotiated by Census Bureau and thearea governments.

Special agreements. On October 23, 1991, the Commis-sioner for the Department of Economic Development andAgriculture for the U.S. Virgin Islands, and on November 6,1991 the Director of Commerce for Guam, signed memo-randa of agreement with the Census Bureau for carryingout agricultural censuses in their jurisdictions. Under theterms of these agreements, the governments of the respec-tive territories assumed responsibility for appointing acensus coordinator for each, and for conducting the fieldenumeration. The local authorities also were responsiblefor—

1. Recruiting qualified personnel for census jobs.

2. Training persons hired for the census following proce-dures established by the Census Bureau.

3. Determining local pay rates, subject to review by theCensus Bureau for consistency and overall fundingavailability.

4. Arranging office space, equipment, and supplies requiredby the census operation within each jurisdiction.

5. Maintaining administrative and financial records for thecensus and providing weekly reports to the CensusBureau.

6. Publicizing the census locally (the Census Bureauprovided promotional materials).

The Census Bureau was responsible for procuring anddistributing maps, manuals, and supplies, and for thedevelopment of any special procedures that might berequired for the enumeration within each territory, togetherwith designing (in consultation with the respective localgovernments) and printing the report forms, instructionmanuals, training materials, and related forms. In addition,the agency provided training for the enumerators and crewleaders, established a calendar of operations, and pro-vided technical advice, as needed, to clarify concepts andprocedures. Finally, the Census Bureau bore the total costof the agricultural censuses in each of the areas.

Report form content. The Census Bureau designed thereport forms for Guam and the U.S. Virgin Islands incooperation with the respective governments. The reportforms—the form 92-A1(G) for Guam, and the form 92-A1(VI)for the Virgin Islands—were of similar design and layout.Each form was a single sheet of white stock, measuring17″ x 14″ folded to 8-1\2″ x 14″, with four numbered pages.The A1(G) was printed in black ink with salmon shading,while the A1(VI) had blue shading. Each form requestedinformation on—

x Land in farms and land use.

x Farm labor.

x Organization.

x Crops harvested (acres and pounds) for sale.

x Vegetables or melons (acres and pounds harvested forsale).

x Fruits, nuts, or nursery crops (inventory and poundsharvested for sale).

x Livestock and poultry (inventory and sales).

x Fish or aquaculture (number and acres of ponds, quan-tity (pounds) and value of sales).

x Total value of agricultural products sold.

x Expenditures.


The A1(G) included an additional section requestingdata on irrigation (acres irrigated, major source of water,whether a private or public system was used, and the typeof rate charged).

The A1(VI) had 13 and the A1(G) 14 sections, includingan enumerator’s record of the individual interview. Theenumerator completed this section with information aboutthe person who supplied the data for the questionnaire, anyremarks about the place, owner or operator, crops orlivestock, involved, as well as the location of the land, andthe enumerator’s signature.



Map not available.


Map not available.

Preparation of Enumerator Materials

Printing report forms and enumeration materials. Thereport forms, materials for the enumerator record books,and administrative forms used in the 1992 agriculturecensuses for Guam and the Virgin Islands were printed byprivate contractors supervised by the Government PrintingOffice (GPO). All printed materials were delivered to theData Preparation Division (DPD) office in Jeffersonville, IN,where the staff assembled 500 enumerator record books(using instructions provided by the Agriculture Division(AGR) staff) for each of the areas, and shipped the recordbooks and agriculture census report forms to the AGR inSuitland, MD, at the end of April 1992. The AGR forwardedthe materials to the respective census managers for distri-bution to the field enumeration staff.

Maps. The agriculture census on Guam and in the VirginIslands was a door-to-door canvass of agricultural opera-tions and accurate maps were crucial to a successfulenumeration. The Census Bureau’s Geography Divisionprepared a set of maps for use by the area offices andstaffs on Guam and in the Virgin Islands, using the 1990Topologically Integrated Geographic Encoding and Refer-ence (TIGER) Line files as the geographic data base forthe maps. For each area, four office maps were required,showing enumeration districts (ED’s), and ED numbers,together with three complete sets of ED maps for use bythe enumerators.1

Record books and enumeration kit. Each enumeratorwas provided with a record book and enumeration kit. Therecord book consisted of a rigid chipboard (for protection ofmaterials and to serve as a writing surface when required),2 copies of the Form 92-A4(OA), Enumerator’s DailyRecord, 2 copies of the Form 92-A3(OA), Enumerator’sWeekly Report, 20 Form 92-A2(OA), Listing Sheets, and 1copy each of the Form 92-A4(OA), Cover Sheet and FormBC-356, Envelope. The chipboard, envelope, individualrecord/report and listing sheets, and the cover sheet wereperforated and bound together when each record book wasassembled (the envelope was bound to the chipboard withthe flap up and to the right, so that it could be openedwithout disassembling the record book). After assembly ofthe individual record books, ED maps were folded andinserted in the envelope with the map identifier facingupward for quick identification.

The A2 Listing Sheets served as the record of thecanvass in each ED, and listed a series of screeningquestions enumerators’ used to determine whether a placequalified as a farm. The A2 also had prelisted names andaddresses of known farms in each ED. Each day, enumera-tors listed the island name and ED number of the area

being canvassed, date, miles traveled, hours worked,number of report forms completed that day, and number ofcases pending on the A4 Daily Record. After each week ofcanvassing, the enumerator completed a Form A3, WeeklyReport using this information to date, and turned it over tothe responsible crew leader.

A total of 250 record books were prepared for eachterritory, with about a dozen available for each ED ifneeded.

Each enumerator also was supplied with a Form A20Enumerator Manual, and a supply of the appropriate reportforms.

Staffing and Training

The staffs for the agriculture censuses in the areas wereorganized into a small office staff and the field enumerationstaff. The census manager for each area functioned asboth the general supervisor for the census and the head ofthe census office. Each area’s staff was as follows:

Staff GuamVirgin

Islands

Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 7Project manager . . . . . . . . . . . . . . . . . . . . . . . 1 1Crew leaders . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1Enumerators . . . . . . . . . . . . . . . . . . . . . . . . . . 16 5

The census staffs received salaries as temporary employ-ees of the local governments, paid on the standard U.S.Government general schedule (GS) pay scale for theappropriate grades.

The local governments appointed the project managerswho were responsible to the Census Bureau for theconduct of the enumeration in their areas. Their broadresponsibilities included precensus preparatory activities,such as securing office space, recruiting, testing, andselecting personnel, and publicizing the census. They alsohad general supervisory responsibility for the enumerationand for keeping Census Bureau headquarters informedabout the progress of the enumeration. Relatively littleclerical work was done at the area offices in the 1992census, but the project managers were responsible forensuring that once the enumeration was complete allrequired materials were secured and forwarded to Bureauheadquarters for processing and tabulation.

The crew leaders assisted the managers and directlyoversaw the enumeration, reviewed enumerators’ work,and made periodic progress reports to the manager. Theyalso carried out any related duties assigned by the man-ager. The enumerators actually conducted the census,interviewing agriculture operators in their assigned dis-tricts.

Census Bureau staff visited Guam in June and the VirginIslands in July 1993 to train the census managers, crewleaders, and enumerators. Some enumerators left the

1The AGR carried out both the 1992 Census of Agriculture and the1992 Economic Census in the areas. The set of four office maps was usedfor both census operations, but separate sets of enumerator maps, onefor the agriculture operation and one for the economic census, weresupplied for both the Virgin Islands and Guam.


census operation prior to completing the census; anyreplacements that had to be hired received training inenumeration procedures from the local managers.

DATA COLLECTION

Enumeration Methodology

The Census Bureau employed personal interviewing bya field staff for the agriculture censuses for Guam and theVirgin Islands. The enumeration began in July and wascompleted by mid-August 1993. On Guam, the island’s 19election districts were used as enumeration districts (ED’s),while the Census Bureau delineated 20 ED’s in the VirginIslands. An enumerator was assigned to canvass each ED,visiting addresses provided to them as probable farms andasking enough questions about tracts of land on which noone lived to determine whether the person in charge of theland should be interviewed to collect agricultural data.Enumerators also looked for other places, not prelisted, butthat evidently had commercial agricultural operations. Theenumerators checked with respondents, local feed stores,service stations, and any other knowledgeable source toobtain information on nonlisted places with agriculturalactivities. Any place that had any crops, livestock, orpoultry was listed in the enumerators’ record books. If theplace(s) appeared likely to have annual sales of agricul-tural products of $100 or more, the enumerators visited theplace and completed a report form for any operation. Incases when no one was at home, the enumerators tried toobtain information from other persons, such as hiredworkers or neighbors, to decide whether a report form hadto be completed for that place. If so, the enumerator madea ‘‘callback’’ later to complete the enumeration of theoperation.

When visiting a place for field enumeration, enumeratorsidentified the ‘‘operator’’ as the person with day-to-daymanagement of the farm and interviewed that person toobtain the necessary information. For partnerships, thepartner in charge of the actual farm operations, or thesenior partner, was listed as the operator. For land ownedby institutions, and used for agricultural purposes, theinstitutional owner was listed as the operator, and the nameof the individual in charge of activities at the place wasentered as the ‘‘manager’’ in the remarks section of thereport form.

For places with two or more tracts of land, or with land inmore than one ED, the enumerator completed a singlereport form covering all the land operated by one person,regardless of location. The enumerator identified the loca-tion of each tract of land included on the form to avoidduplication of the data. Operators with land and agriculturalactivities in more than one ED were enumerated in the EDin which they resided. Once the enumerator identified theperson who operated the farm, and could supply therequested information, he or she assigned the place a

10-digit farm serial number (FSN)2 as a unique identifica-tion and wrote it into the appropriate space on the reportform, then went ahead with the interview.

Callbacks

For a variety of reasons—the operator was not present,necessary records were not available, or some otherreasons not connected with an operator’s outright refusalto respond—enumerators sometimes were unable to com-plete report forms during the first visit to a household. Inthese cases, the enumerator made arrangements for areturn visit—a ‘‘callback’’—at a time convenient to theoperator. Callbacks were to be made as soon as possibleafter the initial visit, but enumerators were not to conductmore than two personal visit callbacks unless their crewleader decided special circumstances warranted additionalattempts.

Refusals

In cases where an operator refused to respond to thecensus, the enumerators were to first try to persuade theoperator to provide the data needed and to explain thelegal requirement for response. When individuals contin-ued to refuse to cooperate, the enumerator identified thecase either as a partial or complete refusal (some refusalsdid provide partial information) in the record book andreported the case to the crew leader (on Guam) or to theoffice supervisor (in the Virgin Islands). The supervisor wasthen responsible for determining the correct course ofaction for obtaining the data.

Field Review

The crew leaders were responsible to the project man-agers for the actual conduct of the enumeration. Theysupervised and reviewed the work of their enumeratorsand made periodic progress reports to the census manag-ers. The crew leaders’ reviews included observing each oftheir enumerators at least once during canvassing, select-ing for early observation those enumerators that seemedmost likely to have difficulty getting started on the job. Thecrew leader spent approximately half a day accompanyingeach enumerator, observing canvassing procedures, includ-ing route planning for covering the assigned ED, interviewprocedures, whether forms and listing sheets are fully andaccurately completed, and so on. When interview or otherproblems were observed during canvassing the crew leaderdiscussed these with the enumerator involved after theinterview. After completing each observation period, thecrew leader and the enumerator observed/discussed thelatter’s performance, and the crew leader completed an

2The FSN was a 10-digit identification number composed of fourparts—the geographic area code, island code, ED number, and linenumber from the listing sheet.


observation report form. The crew leaders arranged sec-ond observations for enumerators that were having difficul-ties and, where necessary, conducted retraining to correctproblems identified during the field observations. (Crewleaders could recommend that enumerators who they rated‘‘unsatisfactory’’ be replaced. In those cases, the finaldecision was made by the census manager, based on thecrew leader’s recommendation.)

Crew leaders also monitored the progress of the enu-meration, receiving periodic progress reports from theirenumerators (the form 92-A3(OA) Enumerator’s WeeklyRecord of Progress). The crew leaders verified both thecumulative figures reported by each enumerator, and com-pleted weekly progress reports for the census manager.

Results

The 1992 agriculture census data collection effort inGuam and the Virgin Islands began in the second week ofJune 1993 and was completed at the end of July. (Theagriculture field enumeration was carried out simultaneouslywith field work associated with the 1992 Economic Censusin the Outlying Areas.) In Guam, the agriculture censusenumerated 199 farms with 1,919 acres, of which 1,373acres were in cropland. The average size of a farm inGuam in the 1992 census was just 9.6 acres. In the VirginIslands, the census counted 202 farms, with 13,666 acres,of which 861 acres were cropland. The average farm in theVirgin Islands had 67.7 acres.

DATA PROCESSING

The AGR staff at Census Bureau headquarters in Suit-land, MD, processed the report forms from the agriculturecensuses in Guam and the Virgin Islands. After check-inand review of any problem cases, the AGR staff keyed theinformation from the questionnaires to a computer data file.The resulting computerized records were subjected to adetailed computer edit for consistency and reasonable-ness. The edit also corrected obviously erroneous orinconsistent data, supplied missing data based on imputa-tion (using characteristics from similar farms to imputeinformation), and assigned farm classification codes neededin tabulating the data. Any significant change in the data inany given report by the computer edit was reviewed andverified by agricultural analysts in the DPD office.

Before publication, AGR statisticians reviewed the tabu-lations for inconsistencies and potential coverage prob-lems. The 1992 totals were compared to previous censusdata, as well as other available information and anyproblems were examined. When necessary, the staff madecorrections to the data records and retabulated the affectedtotals.

Precomputer Processing

While the local area staffs reviewed the report forms toensure they had been filled out correctly, they did not carryout any detailed editing or other processing. After the field

enumeration was complete, the census manager boxedand shipped the report forms directly to theAGR in Suitlandfor data preparation and processing. At Suitland, AGR staffreviewed the individual report forms for—

x A farm serial number.

x An ED number.

x Correct geographic area code.

x The name and address of the operator.

x A positive entry under ‘‘land in agriculture.’’

x Either crop production or livestock/poultry inventory.

The edit identified report forms for operations that didnot meet the farm definition; each such case was verifiedby an analyst. The remaining report forms were reviewedfor accuracy, consistency, and completeness. Reportingerrors in computations, units of measures, data inconsis-tencies, misplaced entries, and so on, were corrected,deriving the missing information from reported data forsimilar type and size farms in nearby areas.

After review, AGR’s Outlying Areas Branch staff keyedthe data from each record using a database program onmicrocomputer equipment at Census Bureau headquar-ters. The program employed was designed to create datafor output in the same format used by the DPD staff atJeffersonville, IN, for transmitting keyed census data to theEconomic Programming Division (EPD) at Suitland forcomputerized processing. The keyers selected the data-base option when beginning data entry, creating a file forkeying. The computer screen displayed the layout forentering data into the file and the keyer then entered therequired identification information for each record, a sequencenumber (i.e., the number of that record in keying sequenceduring each data entry session), and the form number. Thekeyer then moved on to the first data field, entered the fieldkeycode, and keyed the data in that field, and continued onthrough the report form, entering the keycodes and data foreach field containing information. After keying the lastentry, the keyer added the completed record to the file bypressing the ‘‘y’’ key, which saved the record and automati-cally moved on to the first field of a new record.

After data entry was completed, the AGR submitted theresulting files to an edit program that prepared them fortransmission to EDP for processing, loaded the files to theminicomputer system used for file transfer, and the EPDcopied each file to the appropriate data base for processingand tabulation.

Computer Processing

General information. After loading the file to the system,the data from each report form were subjected to anitem-by-item computerized edit. Analysts reviewed andverified any substantial changes generated by the com-puter edits to the data file prior to tabulation. The data weretabulated by ED and for each area as a whole, and AGR


statisticians reviewed all tabulated totals to identify incon-sistencies and potential coverage problems. The statisti-cians carried any required corrections to the individual datarecords and the specific totals involved were retabulated.After disclosure analysis, the data file was ready to bereleased for publication.

Computer edit and tabulation. The computerized datafiles were edited by computer for completeness and con-sistency. Inconsistent entries or other problem items were‘‘flagged’’ by the edit program and were reviewed by AGRanalysts. Inconsistent or other problem data were com-pared to previous census data, as well as to other availableinformation. The interactive computer system enabledAGRanalysts to review up-to-date tallies of selected data itemsfor various criteria or sets of criteria, which could includegeographic levels, farm types, sales levels, or other spe-cific characteristics. Any errors or problems were reviewedand researched by reexamining individual data records.Any corrections required were keyed to the records and thecorrected data file reedited.

Whenever possible, edit imputations, deletions, or otherchanges were based on related data from the respondent’sreport form. For some items, such as operator character-istics, data from previous censuses could be used. Valuesfor missing or unacceptable reported data were calculatedbased on reported quantities and known prices, or by usinginformation from other, generally similar farm operations.The data from the individual records then were tabulated toproduce the statistical table files that would be used forpublication.

Disclosure analysis. The Census Bureau is prohibited bylaw from publishing information that could be used toidentify individual respondents to any of its censuses orsurveys. To ensure that this confidentiality is maintained, allof the data tabulations are checked prior to publication in aprocedure called disclosure analysis. This involved a reviewof the data tables that identified and suppressed specificitems that, if published, (1) would result in direct disclosureof data reported by a particular respondent individual orcompany or (2) would reveal information about an indi-vidual by derivation—that is, by a user adding or subtract-ing a published subtotal from a published total to revealindividual data.

Table review. The Census Bureau’s automated equipmentand programs performed the bulk of the analysis andsuppression of the data tabulations, but the AGR staffcarried out interactive table review using the Tabulation andDisclosure System (TADS) procedures prior to release ofthe data for publication. The TADS used interactive work-stations (for more information on the TADS see Chapters 2,

‘‘Planning and Preliminary Operations,’’ and 6, ‘‘Data Pro-cessing’’), and the work procedures incorporated statustracking, interactive table review, and data flow. Tablereview was done using a spreadsheet program; each tablehad a template containing the various reference files, andthe data files were merged with the templates to create thetables that analysts reviewed.Analysts could make changesto the data, and on supervisory approval, the changes werewritten to the appropriate file, then the file was electroni-cally transferred to the EPD to carry the changes to thesystem files.

After all the tables had been reviewed and data andsuppression patterns verified, the EPD produced the tableimages using the Table Image Processing II (TIPS II—aprocessing system used for large, centralized, computer-based tables) and transmitted these tables to the AGR forreview.

PUBLICATION PROGRAM

The Census Bureau published the statistical data fromthe 1992 Census of Agriculture for Virgin Islands andGuam, in March and June 1995, respectively, in Volume 1,Geographic Area Series, Part 53, Guam, and Part 54,Virgin Islands of the United States.The publication programdid not include any advance reports for the outlying areas.

The report for the Virgin Islands showed statistics for theterritory, for Saint Croix, and for Saint John and SaintThomas (combined). The report for Guam showed data forthe island, and for 19 election districts. The statistical tablesincluded data on—

x Number of farms.

x Farm characteristics.

x Land in farms and land use.


x Selected farm expenses.

x Acres planted, amount harvested, and sales value offruits and nuts, vegetables and fieldcrops.

x Selected machinery and equipment.

x Inventory and sales of livestock and poultry and theirproducts.

The Census Bureau initially released highlights from theprinted reports for Guam and the U.S. Virgin Islands on theagency’s CENDATATM online information service. The datafiles for the areas later were made available in spreadsheetformat on the final 1992 Census of Agriculture CD-ROMrelease.


Chapter 9.

Introduction --------------------------------------------------------------------------------------- 102General Information --------------------------------------------------------------------------- 102Legal Authority --------------------------------------------------------------------------------- 102Scope and Reference Year ------------------------------------------------------------------- 102Estimation -------------------------------------------------------------------------------------- 103

Preparations -------------------------------------------------------------------------------------- 103Planning ---------------------------------------------------------------------------------------- 103Sample Design and Selection ---------------------------------------------------------------- 103Report Form------------------------------------------------------------------------------------ 104

Data Collection ----------------------------------------------------------------------------------- 104General Information --------------------------------------------------------------------------- 104Mailout and Mail Followup -------------------------------------------------------------------- 104Telephone Followup -------------------------------------------------------------------------- 105Results ----------------------------------------------------------------------------------------- 105

Data Processing---------------------------------------------------------------------------------- 105General Information --------------------------------------------------------------------------- 105Receipt and Check-In ------------------------------------------------------------------------- 105Correspondence and Telephone Assistance ------------------------------------------------- 105Prekey Review --------------------------------------------------------------------------------- 106

Data Entry-------------------------------------------------------------------------------------- 106Batch for data entry ------------------------------------------------------------------------- 106Data entry------------------------------------------------------------------------------------ 106Quality control ------------------------------------------------------------------------------- 106

Computer Edit and Final Edit ----------------------------------------------------------------- 107Tabulation -------------------------------------------------------------------------------------- 107

Publication ---------------------------------------------------------------------------------------- 107

Contents

Page


1994 Farm and RanchIrrigation Survey

INTRODUCTION

General Information

The agriculture census began collecting selected dataabout on-farm irrigation in 1890, and in 1900 the Congressauthorized a census of farms using irrigation. The CensusBureau conducted censuses of irrigation (and later drain-age) as part of the decennial censuses through 1950, anda survey of on-farm irrigation in selected States was addedto the 1954 and 1959 agriculture censuses. Surveys ofon-farm irrigation, using samples drawn from agriculturecensus respondent lists have since been carried out fol-lowing the 1978, 1982, 1987, and the 1992 Censuses ofAgriculture.

In 1990, approximately 81.1 percent of all fresh waterconsumed for all purposes in the United States was usedfor agricultural irrigation. The 1992 Census of Agricultureshowed that 14.5 percent (approximately 279,000) of allfarms in the United States were irrigated, and those farmsaccounted for approximately 35 percent of total value ofsales of agricultural products and 51.9 percent of the valueof all crops sold. Information on agricultural irrigation wascrucial to legislators and policymakers, economists andfarmers, and planners and hydrologists concerned aboutthe nation’s supply of both food and fresh water. The 1994Farm and Ranch Irrigation Survey (FRIS) supplementedthe basic irrigation data collected from all farm and ranchoperators in the 1992 agriculture census, asking a sampleof 1992 agriculture census respondents who reportedusing irrigation for information about their irrigation prac-tices.

Legal Authority

Title 13, United States Code—Census, authorizes theCensus Bureau to conduct periodic censuses of agriculturein the United States. Title 13 governs the agency’s opera-tions, establishes what censuses shall be taken, the inter-vals between them, specifies certain administrative proce-dures, and describes the duties of particular officials. (Seeappendix A for excerpts of Title 13 applicable to theagriculture census.) Chapter 5 (Censuses), of the title,section 142, directs that agriculture censuses shall betaken ‘‘in every fifth year beginning after 1983.’’ Section182 of the chapter authorizes the use of surveys, whileSection 195 provides for the use of sampling (except for

the determination of population for purposes of apportion-ment), to collect and furnish annual or other data onsubjects covered by the census. The Census Bureauconducted the 1994 FRIS as a sample survey under theprovisions of this section.

Scope and Reference Year

The Census Bureau has normally carried out a farm andranch irrigation survey, and other follow-on sample sur-veys, in the year immediately following the agriculturecensus. That is, in the usual course of events, census datacollection would be done in the first six or seven months ofthe year following the census reference year and theagency would have drawn the survey sample and doneother preparations for the survey(s) immediately after-wards. For the 1992 census, this would have meant theirrigation survey report forms would have been mailed inJanuary 1994, to request irrigation data for 1993. However,budget restrictions required a general ‘‘stretching out’’ ofcensus operations to save cost. In consequence, theirrigation survey was delayed to the following calendar yearand asked for irrigation data for calendar 1994.

The survey requested relatively detailed data, but limitedoverall response burden by asking only about 1 in every 12irrigators—identified from the 1992 census file—for infor-mation. The sample excluded all farms in Alaska andHawaii, as well as abnormal and horticultural specialtyoperations, and was designed to provide reliable estimatesof irrigation practices for the 18 water resources areas(WRA’s)1 of the 48 conterminous States, and for the 27leading irrigating States.2

The survey asked respondents to supply data on landuse, irrigation, and maintenance expenditures, as well asinventory items for calendar year 1994, while irrigated andnonirrigated crops data were requested for the 1994 grow-ing season.

1Defined by the Water Resources Council, WRA’s are geographicunits, fluvial drainage areas tied to county boundaries at the lowest level.They may intersect State boundaries, but cannot intersect county bound-aries.

2The leading irrigation States were Arizona, Arkansas, California,Colorado, Florida, Idaho, Illinois, Kansas, Louisiana, Michigan, Minne-sota, Mississippi, Montana, Nebraska, Nevada, New Mexico, NorthDakota, Oklahoma, Oregon, South Dakota, Texas, Utah, Washington,Wisconsin, and Wyoming.


Estimation

The survey used two types of statistical estimationprocedures. Statistical estimation was used because not allcensus irrigators were included in the sample and not allirrigators contacted by the survey responded. The surveyestimates were computed by weighting the data for eachrespondent irrigator by an expansion factor that was theproduct of the whole farm nonresponse weight, the sampleweight, and the ratio estimation weight.

The Census Bureau used whole farm nonresponseweight to expand the survey data to account for theirrigators who did not respond to the survey—for whateverreason—and for the survey report forms that could not bedelivered (undeliverable as addressed). For each stratumgroup, a noninteger nonresponse weight was calculated atthe stratum level and assigned to each in-scope respon-dent record. (The noninteger nonresponse weight is theratio of the total number of in-scope sample cases to thetotal number of in-scope responding cases within a stra-tum.) The underlying assumption of this approach was thatsurvey respondents and nonrespondents within a stratumconstitute a homogenous population, allowing respondentsto represent nonrespondents.

The sample weight expanded the survey data to esti-mate universe totals as if a complete census of irrigatorshad been conducted.All respondent survey records receiveda sample weight. The sample weight, calculated at stratumlevel, is the ratio of the universe estimated number ofirrigating farms in a stratum to the count of sample farmswithin the same stratum. A third weight, the ratio estimationweight, was used to obtain agreement between surveyestimates and census estimates of universe in-scope irri-gated acres at publication level within stratum.

The final weight, the product of the sample weight, thenonresponse weight, and the ratio estimation weight, wasrandomly integerized for tabulation. If, for example, thefinal weight for the number of irrigators in a particularstratum was 7.2, then one-fifth of the irrigators in thisstratum were randomly assigned a weight of 8 and theremaining four-fifths received a weight of 7. The surveytotal for a given characteristic was estimated by multiplyingthe data value by the corresponding sample farm finalweight and summing over all sample farms for the respec-tive geographic area.

PREPARATIONS

Planning

Planning for the 1994 FRIS began in 1992, when theCensus Bureau’s Agriculture Division (AGR) conducted areview of previous farm and ranch irrigation surveys. The

Census Bureau mailed letters to selected persons inwater-related government organizations, the irrigation andagriculture industries, and academic positions, asking forcomments and suggestions on report form content. Inaddition, the Census Bureau also mailed letters to selectedrespondents to the 1988 irrigation survey, enclosing a briefdescription of the proposed new survey, and asking forsuggestionson content andany comments on the respondents’experiences in the 1988 survey. Census Bureau staffreviewed responses for use in evaluating data needs andother questions, such as respondent burden.

Comparatively early in the preparation cycle, budgetlimitations compelled the Census Bureau to stretch out theentire agriculture census program, which led to postponingthe irrigation survey until 1994. This meant that the data-collection effort would begin at the end of 1994 andprocessing, tabulation, and publication would be done in1995.

Sample Design and Selection

The Census Bureau designed the sample for the 1994Farm and Ranch Irrigation Survey to produce a relativestandard error not exceeding 5 percent on estimatedirrigated acreage for the United States, at the State level forthe 27 leading irrigation States, and collectively for thecombined 21 other States. The agency calculated thatthese requirements could be met with a sample of approxi-mately 20,000 operations. The universe from which thesamplewas selected included all farms or ranches—excludingabnormals, horticultural specialty operations3, and farmsand ranches in Alaska and Hawaii—that reported usingirrigation in the 1992 agriculture census.

All farm operations eligible for the survey were stratifiedinto 34 stratum groups, each consisting of the irrigators inthe 27 major irrigation States (covering the Midwest andWestern regions) and the 7Water ResourcesAreas (WRA’s)defining the Eastern region. Stratum assignment withineach stratum group was based on the 1992 reportedirrigated acreage, and varied from stratum group to stratumgroup. The Census Bureau selected an independent sys-tematic sample of farms for each stratum of each stratumgroup.

The sample included all farms that reported a minimumnumber of irrigated acres in the 1992 agriculture census(i.e., ‘‘certainty’’ farms for irrigation survey purposes), asfollows:

3Horticultural specialty farms are those farms with a standard industrialclassification (SIC) code of 018, or that reported annual horticulture salesexceeding $1,999.


Minimum acresirrigated States

5,000 California, Nevada3,500 Arizona3,000 Florida, Kansas, Mississippi2,500 Arkansas, Colorado, Georgia,

Idaho, Michigan, Texas, Wyoming2,000 Missouri, Montana, Nebraska,

Oklahoma, Oregon, Washington1,500 All other States

A total of 1,175 certainty irrigators were selected for inthe survey.

The sample size assigned to each of the uncertaintystrata was defined as the maximum of a Neyman allocationand an allocation proportional to the individual stratum’sshare of the stratum group’s 1992 total irrigated acreage.The sample included 18,823 farms selected from thenoncertainty strata.

The Census Bureau selected the sample systematicallyand independently for each stratum of a stratum group,beginning from a random start. The final sample file for thesurvey consisted of 19,998 records, representing about 8.1percent of all farms and ranches in the conterminous 48States reporting irrigation in the 1992 census, and 33.1percent of the 47.4 million acres of irrigated land.

Report Form

The 1994 FRIS questionnaire, form 94-A62, was an8-1/2″ x 11″, 12-page booklet on white stock, with printingin black ink and shading in a blue wash. The report formincorporated a cover letter asking for the respondent’scooperation on page 1, and general information on thesurvey, assistance available to respondents, and confiden-tiality, on pages 11-12. The main body of the report formconsisted of 20 sections. One section asked for the nameand telephone number of the person completing the form,while the remaining 19 requested detailed information on—

x Acreage.

x Land use.

x Whether any land was irrigated.

x Method of water distribution.

x Acres irrigated and estimated quantity of water used bysource.

x Acres harvested and crop yields.

x Irrigation frequency, method of water distribution, anduse of commercial fertilizers and pesticides in irrigationwater by selected crops.

x Number of irrigations wells, well depth and pumpingcapacity.

x Pumps other than well pumps.

x Energy used for pumping irrigation water by powersource.

x Maintenance and repair costs for irrigation equipmentand facilities.

x Expenditures for irrigation facilities.

x Irrigation practices.

x Other uses of irrigation water.

x Participation in Federal Government commodity pro-grams or wildlife programs.

x Improvements to irrigation systems to reduce energyand/or conserve water used in irrigation.

x Sources of irrigation information.

x Irrigated land in 1992, and (if no irrigation in 1994)reasons for discontinuing irrigating.

With the exception of Item 19 (‘‘Irrigated Land in 1992’’)all of the data were requested for calendar 1994.

DATA COLLECTION

General Information

The 1994 irrigation survey was carried out usingmailout/mailback enumeration and followup, supplementedby telephone followup to large irrigation operations. Theinitial mailout for the survey involved mailing 19,998 surveypackages to irrigators in January 1995, followed by a thankyou/reminder card approximately 4 weeks later, and threemail followups, each consisting of a cover letter, a reportform, and a return envelope. The Census Bureau com-pleted data collection in June 1995 and the publicationresults were released in November 1995.

Mailout and Mail Followup

Private contractors printed the report forms and associ-ated materials (Form 94-A62, Report Form, outgoing andreturn envelopes, for the initial mailout and the report formfollowups, and the reminder/thank you card), and deliveredthem to the Data Preparation Division (DPD) office inJeffersonville, IN, in early December 1994. The staff thereprinted the address labels for the mailout using the com-puterized address file compiled by the AGR, and theclerical staff at Jeffersonville assembled the mailout pack-ages. The initial mailout consisted of a Form 94-A62,Report Form, a brochure explaining the need for theirrigation data, and a return envelope. The initial mailout,and the second, third, and fourth followup mailings involvedcomplete report form packages, while the first mail followupused a card reminding addressees’ of the requested responsedate, and thanking them if they already had responded.


After eachmail response cutoff date the Census Bureau’sstaff generated an updated computerized list of nonrespon-dents, which was then used by the DPD staff to printaddress labels for the followup mailings. The Jeffersonvilleclerical staff attached the labels to followup mailing pack-ages during the 3-5 day intervals between the responsecutoff dates and the next mailout. The initial survey mailout,and the followup mailings were as follows:

Mailout Type Date Forms mailed

Survey mailout . . . . Report form 12/29/94 19,998First followup . . . . . Reminder/thank

you card01/13/95 19,998

Second followup . . Report form 02/10/95 11,800Third followup. . . . . Report form 03/16/95 8,300Fourth followup . . . Report form 04/21/95 6,716

Mail data collection was closed down at the end of thesecond week of June.

Telephone Followup

After the final followup mailing, the AGR staff prepared alist of the irrigation survey certainty cases that remainednonrespondent for possible telephone followup. Given thelimited size of the sample for the irrigation survey, theCensus Bureau considered the data from large-scale irri-gators critical to developing reliable statistical estimates,and subjected these cases to intensive followup. Clerks atJeffersonville, IN, researched telephone numbers for theselected cases and on May 10 began calling the certaintycases that remained nonrespondent (478 of the 1,175certainty cases were nonrespondent as of May 8). Thetelephone staff attempted to contact operators to obtainthe basic acreage, crop, and irrigation data required forthe survey. The telephone followup activities continueduntil June 2, by which time responses had been obtainedfrom all but 18 of the original nonrespondent list. AGR staffpulled the 1992 census reports for these cases to reviewthem and determine whether data could be projected forthem for incorporation into the irrigation survey estimates.

Results

The 1994 FRIS attained a final overall response rate of73.6 percent, representing 14,373 receipts, excluding 477undeliverable as addressed (UAA) cases. This was about1.6 percent below the final response rate obtained for the1988 survey. The Census Bureau employed statisticalestimation procedures to develop State, WRA, and nationalestimates for the over 248,000 irrigators and 47.4 millionacres of irrigated land in the 48 conterminous UnitedStates.

DATA PROCESSING

General Information

The 1994 FRIS report forms were returned to theJeffersonville office for data processing. All the forms werereviewed upon receipt and check-in to identify significant

inconsistencies, and to ensure that the data entries couldbe keyed to the data file. Remarks by respondents werereviewed for possible response by the agency or to makecertain accurate information was added to the data file, andany inconsistencies or obvious errors were corrected beforekeying. After the data were entered into the data file, the filewas subjected to a detailed computerized review and editat the Census Bureau’s headquarters in Suitland, MD.Before publication, the tabulations from the FRIS data filewere reviewed, using data from the 1992 census, toidentify inconsistencies or potential coverage problems.

Receipt and Check-In

The FRIS report forms, as well as UAA’s, were checkedin at the processing office using the wand/keyboard sta-tions. The returning report form packages then were slitopen and the report forms removed for check-in. Formswith attached congressional correspondence (notes indi-cating that a respondent intended to contact a congres-sional office counted as congressional correspondence)were pulled from the processing operation stream and sentdirectly to the AGR at Census Bureau headquarters inSuitland, where they were held and processed once thecorrespondence had been resolved. Materials sent to theattention of a particular analyst (i.e., the analyst’s nameappeared on the envelope, report form, or in attachedcorrespondence) were forwarded to the analyst. All otherchecked-in materials went to the agriculture processingunit for clerical review before data entry.

Correspondence and Telephone Assistance

The FRIS processing staff at the Jeffersonville officeincluded a correspondence and telephone assistance unitto handle respondent-originated correspondence (ROC)and telephone calls generated by the survey. The unitreceived ROC materials on a flow basis from the check-inarea and reviewed the items to determine action requiredfor resolution. In fact, only 10 ROC items were received bythe Jeffersonville office; 8 of these were resolved by clerksduring review of the enclosed materials. Clerks contactedthe two remaining cases by telephone for completion.

The initial mailing package, and all the followup mail-ings, included a toll-free telephone number for respondentsto use if they had questions about the FRIS or neededassistance in completing their report form. The Jefferson-ville telephone staff handled incoming calls, providingassistance to any respondents who needed help andanswering any questions callers had about requirements torespond, whether or not their farm qualified as an irrigationoperation for the purposes of the survey. The telephonestaff updated the FRIS check-in file using computer workstations and interactive processing systems. Whenever acallback was required (e.g., to obtain additional informationfrom the respondent, or to confirm that a report form hadbeen received) the telephone clerks asked for a telephone


number and the most convenient time to call. The Jeffer-sonville office received some 400 calls relating to the FRIS,approximately 170 from respondents requesting help incompleting their report forms. About 70 callers asked forinformation on whether they actually had to respond to thesurvey, while a further 60 claimed to have already filed areport form. (Clerks used the interactive systems to checkthe FRIS check-in file to determine whether questionnaireshad been checked in for the respondents who claimed tohave filed.) Approximately 70 callers said they no longerirrigated, and 15 more were outright refusals to respond.

Prekey Review

The Jeffersonville clerical edit staff received report formson a flow basis from the open and sort unit, and reviewedthem prior to data entry. The edit clerks pulled blank formsfrom the processing stream for remailing to the respondent(a cover letter requesting response was included whensuch forms had to be remailed). The prekey review involvedboth the report forms and interactive systems. The prekeyreview clerks reviewed the individual report forms andannotated them as needed. The reviewers began thereview of each form by checking item 3 (‘‘Was any land onthe farm or ranch you operated irrigated at any time in1994?’’) and item 19 (‘‘Irrigated Land in 1992’’). If therespondent answered ‘‘no’’ to both items, the clerk markedthe form as being out of scope. (Approximately 1,600report forms returned to Jeffersonville were determined tobe out of scope of the survey, and some 1,065 more surveyaddresses were determined to have discontinued farmingsince 1992.) Report forms from large farms that reportedirrigation in the 1992 census, but none in the FRIS, werefollowed up by telephone to clarify the apparent discrep-ancy. (The review clerks could make these followup callsdirectly from their work stations.)

After determining that a report form was inscope, theclerks conducted an item by item review of the form, usingwritten edit guidelines. The review checked for the com-pleteness and consistency of reporting (e.g., that acreageincluded land owned, land rented from others, minus landrented to others; that the individual land-use items addedup to the total acres in the place; that the methods of waterdistribution reported were consistent throughout), moved‘‘tenths’’ reported to the tenths column when necessary,converted decimals and fractions to the proper wholenumbers, deleted cents when reported in the dollar col-umn, converted bracketed and range data to a single itemas required, converted improper units to proper units whenrequired, and struck out unnecessary symbols (e.g., forfeet (’) or inches (″)).

Out-of-scope records were assigned an action code of‘‘57’’ using the interactive systems, and the report formswere routed to the analysts’ area where they were main-tained in a serialized file until FRIS processing was com-pleted. In-scope report forms were sent to the batchingarea to be grouped into work units for data keying.

Data Entry

Batch for data entry. Following the prekey review, thereport forms were routed to the Jeffersonville Data Sys-tems Branch (DSB) for data entry. The batches arrived,with data entry batch sheets attached, in plastic envelopes,and were distributed to individual keyers. The keyer(s)used the interactive minicomputer systems employed inthe general census for entering the data, beginning bylogging onto the agriculture/economic data entry batchingmenu and entering the appropriate user name, password,and three-digit operation code (FRI) required for the FRIS,then the batch information (batch size and identificationcodes).

Data entry. The keyer(s) followed procedures generallysimilar to those used for the agriculture census reportforms in entering the data from the individual reports. Eachkeyer opened the plastic envelope containing the batch tobe keyed and checked each report form for problems asthe data were entered. Keyers rejected report forms fordata entry and assigned ‘‘reject reason codes’’ for blankforms (code 01—items 1-19 on the report form were blank),maximum value failures (02—the data field(s) exceededthe maximum allowable value for that field), or the batchcontained more than 99 CFN’s with ‘‘good’’ data (03).Report forms also could be rejected and no reject reasoncode assigned for a CFN check-digit failure, or because theState code was invalid. Keyers pulled rejected report formsfrom the batches and referred them to a supervisor.

The interactive system assigned a series of screens forthe keyers to use in keying the data. Screens 1-2 wereused for keying the CFN and any information in the CensusUse Only (CUO) Box 022, and for name and addresscorrections. Screens 3-12 led the keyer through the rest ofthe report form. Individual keycodes were not entered,since the processing system already had these internallyprogrammed, but were automatically assigned to datafields based on their location on the screen. The keyerentered the data for each data field, then either used thefield release command to move on to the next data field, orpressed the ‘‘F6’’ function key, which enabled the keyer tospecify a particular field.

Quality control. Quality control procedures for the FRISdata keying operation were similar to those employed inthe agriculture census (see Chapter 6, Data Processing).Keyers were subject to the same two-stage verificationregime in training, and the work batches were reviewedand verified at the same rates, depending on the size of theindividual batches.


Work unit size Verification rate(percentage)

Samplinginterval

Less than 24 100 All24-44 12.5 845-74 6.67 1574-99 4.0 25

After completing keying or verification for a batch, thekeyer wrote a keyer/verifier ID number, the number ofreport forms and data fields keyed/verified, the number ofreport forms rejected, and the date on the Data Entry BatchCover Sheet, and routed the batches to a holding areawhere they were retained until computerized edit wascompleted.

Computer Edit and Final Edit

The individual data from all the report forms werepassed through a computerized edit review. An initialreview identified missing entries, entries outside accept-able ranges, and inconsistencies between predefined items.Based on the number and types of problems identified inthe initial review, analysts either prescribed computer editprocedures to correct individual data items, or initiatedcorrections of data items on a record by record basis.

Tabulation

The Census Bureau tabulated the FRIS data using theSAS software package. The tabulation program compiled35 tables containing detailed estimates of irrigation data

from farm operators reporting irrigated land in the 1992census and the 1994 FRIS. AGR analysts reviewed thenational and State tabulations as they were produced forinconsistencies and potential coverage problems, compar-ing the tables to 1992 census data. Any corrections neededwere made to the data file before running the final tabula-tions and releasing the statistics for publication.

PUBLICATION

Data from the 1994 Farm and Ranch Irrigation Surveywere published in the 1992 Census of Agriculture, Volume3, 1994 Farm and Ranch Irrigation Survey. The printedreport presented summary irrigation data for all States, withmore detailed tabulations for the 27 leading irrigationStates and for the 18 Water Resources Areas. The tablesshowed data for calendar 1994 on farms irrigated, landuse, quantity of water applied, land irrigated and method ofwater distribution, estimated quantity of water applied bysource, wells and pumps on farms; selected expendituresfor energy, equipment, and maintenance; selected cropsharvested; application of chemicals; other uses of irrigationwater; Standard Industrial Classification (SIC) code; mar-ket value of crops sold; water management systems used;participation in Federal commodity programs; energy andwater conservation improvements; sources of irrigationinformation used to reduce costs; and farms with dimin-ished crop yields resulting from irrigation interruption bycause. Most tables included 1988 and earlier historicaldata for comparison.

The survey data also were released on CD-ROM, andhighlights of the report were released online through theCensus Bureau’s CENDATATM service.


Chapter 10.

1992 Coverage Evaluation ---------------------------------------------------------------------- 110Introduction------------------------------------------------------------------------------------- 110Background and objectives ----------------------------------------------------------------- 110General procedures ------------------------------------------------------------------------- 110

Sample Survey Designs and Methodologies------------------------------------------------- 1101992 June Agricultural Survey (JAS) ------------------------------------------------------ 1101992 Classification Error Survey (CES) --------------------------------------------------- 111

Sample Survey Data Collection--------------------------------------------------------------- 111

Processing ------------------------------------------------------------------------------------- 111JAS file processing-------------------------------------------------------------------------- 111CES processing ----------------------------------------------------------------------------- 112

Estimation -------------------------------------------------------------------------------------- 112

Results and Publication ---------------------------------------------------------------------- 113Coverage error estimates------------------------------------------------------------------- 113Nonsampling error -------------------------------------------------------------------------- 113Publication ----------------------------------------------------------------------------------- 113

Classification and Regression Tree (CART) Model Evaluation-------------------------------- 113General Information --------------------------------------------------------------------------- 113

Statistical Methods----------------------------------------------------------------------------- 114General information ------------------------------------------------------------------------- 114Model Drop Survey analysis---------------------------------------------------------------- 114Farm proportion estimation ----------------------------------------------------------------- 114

Evaluation Results ----------------------------------------------------------------------------- 114

Research Studies -------------------------------------------------------------------------------- 115Film Optical sensing Device for Input to Computer (FOSDIC) Test ------------------------ 115General information ------------------------------------------------------------------------- 115

Test methodology ------------------------------------------------------------------------------ 115Evaluation of FOSDIC operations---------------------------------------------------------- 116Results --------------------------------------------------------------------------------------- 116

Special Inserts Evaluation--------------------------------------------------------------------- 116General information ------------------------------------------------------------------------- 116Study methodology-------------------------------------------------------------------------- 116Results --------------------------------------------------------------------------------------- 117Recommendation---------------------------------------------------------------------------- 117

Contents

Page


Coverage Evaluation and Research

1992 COVERAGE EVALUATION

Introduction

Background and objectives. The Census Bureau beganconducting regular coverage evaluations of the agriculturecensus in the 1945 program and first released the results ofthe evaluation study as part of the 1950 census publica-tions. Since then, the agency has routinely evaluated eachagriculture census for the accuracy and completeness ofthe farm count and for coverage of selected data items(e.g., land in farms, total value of agricultural products sold,and so on). The methodology used has remained relativelyunchanged—an area sample survey combined with a listsample survey—although specific techniques and sampledesigns have been refined and improved with each cen-sus.

The principal objectives of the 1992 Census of Agricul-ture coverage evaluation program were to provide—

x State estimates of the number of farms not on the maillist.

x Census region estimates of the number of farm opera-tors incorrectly classified, and of duplicate farms.

x Census division estimates of selected agricultural char-acteristics of undercounted farms.

The 1992 coverage evaluation program estimated that252,646 farms were not on the census mail list, 61,965farms were omitted from the census tabulations due toincorrect classification as nonfarms, 50,400 nonfarms werecounted as farms, and 23,505 extra farms were counteddue to respondents returning more than one census reportform.

General procedures. The Census Bureau established anagriculture census coverage evaluation unit at the DataPreparation Division (DPD) office in Jeffersonville, IN, inJanuary 1993. The unit consisted of clerks assigned fromthe DPD staff and an agriculture data analyst attached tothe DPD office from the Agriculture Division (AGR) at theSuitland, MD, headquarters. The coverage evaluation unitcarried out clerical and analytical review of the data fromthe census using other data from the U.S. Department ofAgriculture (USDA).

The 1992 coverage evaluation program used USDA’sNational Agricultural Statistics Service’s (NASS’s) 1992June Agricultural Survey (JAS) and the 1992 ClassificationError Survey (CES—carried out as part of the 1992 agri-culture census) for developing its coverage estimates. TheJAS is an annual national area sample survey designed tomeasure planted acreage and number of livestock byState. The JAS was a field canvass of selected areasegments, and the Census Bureau matched the segmentrecords to its census mail list. Matched and nonmatchedaddresses were mailed census report forms—the matchedones as part of the regular census mailing (the initialmailing in December 1992 and followups (as necessary))and the nonmatched cases as additional mailings. TheNASS conducted the JAS, and the Census Bureau usedthe 1992 JAS data to estimate the number and character-istics of farms not on the census mailing list.

The Census Bureau used the 1992 CES data to esti-mate the number of misclassified (i.e., farms incorrectlyclassified as nonfarms, and nonfarms incorrectly classifiedas farms) and duplicate farms in the census. The surveyused mail enumeration, with followup of nonrespondentsby telephone. The Census Bureau compared the CES farmstatus (i.e., classification as farm or nonfarm) to the censusfarm status to determine whether a given sample farm hadbeen correctly classified in the census.

Sample Survey Designs and Methodologies

1992 JuneAgricultural Survey (JAS). TheUSDA’s NationalAgricultural Statistics Service (NASS) conducts the JuneAgricultural Survey (JAS) to measure planted acreage ofcrops and number of livestock. The JAS results serve as abase for subsequent NASS surveys (e.g., the September,December, and March agricultural surveys). The areasample frame portion of the JAS is enumerated by per-sonal interview during the first 2 weeks of June each year,using June 1 as the data reference date.

The NASS samples from an area frame to conduct theJAS. The basic stratification divides all land for a State intosix to eight land-use strata—e.g., intensive cultivation,urban area, rangeland, etc. Cultivated land was dividedinto several strata based on the degree of cultivation.Primary sample units (PSU’s) were land parcels selectedwithin each land-use stratum, and each randomly selectedPSU was further divided into several additional samplingsubunits or segments. The size of the PSU’s varied, but a


typical one had six to eight segments. Segments weredrawn to have an average of three farms, and had toinclude at least one resident operator.

The JAS employed a two-stage, stratified probabilitysample of U.S. farm operations. The first stage of the actualsurvey was the selection of PSU’s for inclusion in thesample. The PSU’s were selected with probability of selec-tion proportional to the number of segments within eachPSU of the substratum. In the second stage, a segmentwas selected with equal probability from each PSU. Theland area within each selected segment was completelyenumerated so that the segment, rather than the individualfarm, was the sampling unit.

The Census Bureau’s JAS files were compiled fromthree files provided by the NASS—

x The JAS name and address file supplied by the NASS inJuly 1992. This file consisted of the names, addresses,and other identifier information for all JAS area seg-ments that had any indication of agricultural activity.

x The JAS detail file, containing all the requested supple-mental data items from the JAS, including identifyingand whole-farm agricultural data.

x A list of farm status changes compiled by the NASSbetween June and December, 1992, and transmitted tothe Census Bureau early in 1993. (This status-changelist was checked against the JAS name and address listand any needed corrections were made.)

Once the sample segments had been selected, fieldinterviewers canvassed them to establish who operatedthe land within the segment. Each separate farm operationwas identified as a tract, although a single farm mightinclude ‘‘tracts’’ in more than one segment, and each tractbecame a reporting unit. Unlike the 1987 census, the 1992agriculture census estimate from the JAS was based on aweighted segment estimator, which used a proportion ofdata from each farm operation in the segment, regardlessof the where the respective farm operator resided.

1992 Classification Error Survey (CES). The 1992 Clas-sification Error Survey (CES) was a reinterview of a sampleof 1992 agriculture census cases, using a different, shorterreport form to collect information to determine the true farmstatus of the operations contacted. The CES was designedto develop estimates at the national and census regionlevel of—

x Farms incorrectly classified as nonfarms.

x Nonfarms incorrectly classified as farms.

x Farms duplicated on the census mail list.

The survey used an independent, regionally stratified,systematic random sample of addresses from the censusmail list, excluding farms in Alaska and Hawaii, operationswith expected sales of $500,000 or more, multiunits, and

abnormals.1 The sampling rates within census regionswere based on the estimated proportion of farms misclas-sified or duplicated in the 1987 census and a specifiedcoefficient of variation.2 The universe from which thesample was drawn consisted of 3,447,112 mail list records.The sampling rates used in the census geographic regionswere as follows:

Region Sampling rate

Northeast 1 in 176Midwest 1 in 166South 1 in 157West 1 in 160

Sample Survey Data Collection

The NASS collected the data for the 1992 JAS by fieldcanvass of each segment in June 1992. The CensusBureau received the JAS data files, containing the names,addresses, and agricultural data on all area-segment resi-dents with any agriculture activity at the time of the survey,matched the file to the census mail-list development file,and assigned special processing codes to all JAS recordsnot found on the mail list. These codes identified the JASrecords once they were added to the census mail file.

The 1992 CES was a mail and telephone enumerationoperation. The Census Bureau obtained the report formcheck-in status for all sample addresses at two designatedcut-off dates using a unique CES evaluation code set in thecensus data base at the time of sample selection. Form92-A90, Evaluation of the 1992 Census of Agriculturereport forms were mailed to CES sample addresses in twowaves, the first in April 1993, the second in the followingJuly. These mailings totalled 16,804 sample survey cases.A reminder followup mailing was carried out 2 weeks aftereach initial mailout, using a post card to request response,and a second followup, including a report form, was done 2weeks later. The Census Bureau carried out a telephonefollowup to all CES cases still nonrespondent from eachmail group 8 weeks after that group’s initial mailing.

Processing

JAS file processing. The Census Bureau matched theJAS name and address list against the census mail list toidentify the JAS area sample records as either matched orunmatched to the census mail list, and assigned a censusfile number (CFN) and an evaluation code for censusprocessing to each JAS record. Any JAS records not on thecensus mailing list were added and included in the census

1Operations in Alaska and Hawaii, and multiunits (i.e., complexorganizational units with farming operations in more than one location)and abnormals (including Indian reservations, research stations, andother institutional farms) were considered inappropriate for the CESbecause of their unique characters, and because all were subjected tointensive followup in the census.

2A ‘‘coefficient of variation’’ is the standard deviation of a distributiondivided by the arithmetic mean, sometimes multiplied by 100.


report form mailout. The JAS name and address file andthe detail file were used to create a JAS-census control filewith the JAS identification and weighting information foreach record, and a JAS-census detail database containingthe identifying and agricultural detail data.

During census processing, the DPD identified reportforms for JAS-census cases using the pre-assigned evalu-ation codes. These cases were sorted out of the maincensus cases and microfilmed before being returned to thecensus processing cycle. All the JAS-census cases wentthrough normal census processing, including data entry,edit, and imputation. However, the Census Bureau with-drew the nonmatched JAS-census cases data before tabu-lation and analytical review of the aggregate census esti-mates.

JAS-census cases that did not respond to the censuswere followed up in the same manner as regular censuscases, except that after a specified cut-off date (based onthe processing schedule), all JAS-census nonrespondentswere telephoned for information. The Census Bureau usedJAS information to impute data for cases that did notrespond to the telephone followup.

The DPD’s agriculture coverage evaluation unit pro-cessed the JAS-census cases, comparing the data reportedin the census for each case with the data reported in theJAS to ensure valid matches. The coverage evaluation unitused the reported data for each case to independentlydetermine the farm status of unmatched JAS cases. Theunit reviewed the coverage classification codes assignedto each JAS sample case after each State closeout foranalytical review. These codes showed each JAS case’srelationship to census mail-list cases (i.e., whether therecord matched an address on the mail list) and the farmstatus assigned by the JAS, the census processing unit,and the coverage evaluation unit.

The Census Bureau used the JAS nonmatch records toestimate the total number and selected characteristics offarms not on the census mail list for States, divisions,regions, and the United States. The data required for theseestimates were transmitted to the Census Bureau’s Suit-land, MD, headquarters from the DPD office in Jefferson-ville, IN, on a State-by-State basis after closeout for finalanalytical review. The edited farm data for matched andunmatched records constituted the sample for derivingestimates of farms not on the mail list.

CES processing. The 92-A90 questionnaires were micro-filmed as they were checked in at the DPD processingoffice. The DPD staff also identified and microfilmed the1992 agriculture census report forms for the CES cases.Clerks reviewed the CES report forms to classify eachrecord as either a farm or nonfarm, based on reported data,and then the CES-derived farm status was compared to thecensus farm status to identify cases incorrectly classified.The clerks referred cases with apparent errors to analystsfor further review. The analysts attempted to reconcile anydifferences between the CES and census records, using

telephone followup as needed to confirm suspect informa-tion. The processing staff assigned coverage evaluationcodes to each record to identify its classification (and therelationship between the CES farm status and census farmstatus). Microfilm and data files were transferred to theCensus Bureau’s Suitland, MD, facility for editing andtabulation.

Estimation

The Census Bureau used the final data file from the JASin conjunction with CES data to produce coverage-errorand classification-error estimates for the census. The esti-mated true number of all farms in the United States is thecensus published farm count, plus the number of under-counted farms, minus the number of overcounted farms.

The undercount number can be divided into two majorcomponents—farms on the census mail list misclassifiedas nonfarms, and farms not on the mail list. The overcountcomprises those nonfarms misclassified as farms in thecensus, and farms duplicated in the enumeration.

The Census Bureau used statistical modeling to prepareits estimates for the undercount and overcount. The modelused an independent survey—in this case, the JAS—inconjunction with the census (adjusted for classification andduplication errors using CES results) to estimate the truetotal number of farms. The model assumed—

x Both the census and the JAS attempted to measure thesame statistical universe (i.e., all agricultural operationsin the United States that met the census farm definition)for the same period of time.

x The probability of a farm being on the census mail listwas independent of its probability of being included inthe JAS.

x The likelihood of being missed by either the census orthe JAS was the same for a a given size category.

x Census and JAS results could be matched withouterrors.

x Spurious events, such as nonexistent cases, had beeneliminated from both the census and the JAS.

x Sufficient data were collected about nonrespondents inboth the census and the JAS to permit accurate classi-fication.

x If post-stratification was done, the variable used wascorrectly recorded for all farms.

Estimates of classification error were based on the ratioof such errors identified by the CES. Sample counts ofsuch errors were expanded to represent all farms bymultiplying the ratio of all farms in the region to farmssampled in the region for study. Estimates of duplicationerror (overcount) were based on counts of farms deter-mined by review of CES data to have been reported more


than once in the census. (For details of the estimationmethodology, see the 1992 Census of Agriculture, Volume2, Subject Series, Part 2, Coverage Evaluation.)

At the region or United States level, the census farmcount and totals for selected farm characteristics wereadjusted for farm classification and duplication error. Stateand division level estimates of farms not on the mail listwere not adjusted for classification or duplication errorsbecause estimates for those factors were not available atthat geographic level. Division, region, and United Statesestimates were developed by summing lower-level esti-mates before adjusting for classification error.

Results and Publication

Coverage error estimates. The 1992 coverage evaluationprogram estimated net farm coverage error to be 11.1percent, plus or minus 0.7 percent, resulting in an esti-mated census coverage for the United States of 88.9percent of all farms. This was down from the 1987 cover-age estimate of 92.8 percent of all farms. The coverageerror, both undercount and overcount, was largest for smallfarms. The estimated gross undercount for farms with salesof less than $2,500 was 35.5 percent (plus or minus 1.5percent), while the estimated overcount was 4.7 percent(plus or minus 1.9 percent), yielding a net coverage error of30.8 percent. For farms with sales of $2,500 or more,estimated gross undercount was 6.6 percent (plus or minus0.4 percent), while overcount was 2.9 percent (plus orminus 0.5 percent), which resulted in a net coverage errorfor these farms of just 3.7 percent. The evaluation indicatedthat approximately 80 percent of undercounted farms werenot on the census mail list; the remainder were principallyfarms misclassified as nonfarms. Of the undercountedfarms, 67.8 percent had 49 acres or less, and two thirdswere livestock operations.

Nonsampling error. The coverage estimates are subjectto nonsampling error; which may result from flaws in thestatistical model design and/or report form design; incom-plete enumeration; inaccurate census and JAS list match-ing procedures; and incorrect or inaccurate data reportingand processing for the JAS, CES, and the census. TheCensus Bureau made extensive efforts to minimize non-sampling errors in the surveys used for the coverageevaluation through quality control and other verificationmeasures. Nevertheless, potential sources of nonsamplingerror remain, such as the failure to classify some 205 of thetotal JAS sample cases. The estimates were not adjustedspecifically to compensate for these cases, so there remainsa small potential bias in the estimates for farms not on thecensus mail list.

Publication. The Census Bureau published State-levelestimates of the number and characteristics of farms not onthe census mail list in the Volume 1, Geographic AreaSeries report for each State. The State-level estimates forfarms not the mail list, as well as the classification error

estimates, were released in a separate report, the 1992Census of Agriculture, Volume 2, Subject Series, Part 2,Coverage Evaluation. The report included text describingthe evaluation program and the statistical methods employedand charts and tables showing State estimates of farms noton the mail list; United States and regional estimates ofcensus farm coverage, farms by selected characteristicsand components of coverage, and number of farms mis-classified; and United States, regional, and divisional esti-mates of selected characteristics of missed farms.

CLASSIFICATION AND REGRESSION TREE(CART) MODEL EVALUATION

General Information

The initial mail list for the agriculture census typicallyincluded a proportion of addresses that did not meet thecensus farm definition. Collecting the required informationas economically as possible is a major objective of thecensus, hence the Census Bureau made every effort toreduce the number of nonfarm addresses on its agriculturecensus mail list. For the 1992 agriculture census, the initialmail list contained 3.78 million individual addresses; acombination of budget considerations and the need toattain the best possible coverage of farm operations com-pelled the Census Bureau to reduce the total size of themail list to approximately 3.55 million records. The meth-odology used to do this involved using a classification andregression tree (CART) statistical modeling procedure thatclassified records into groups of probable farm and non-farm operations. Those groups least likely to be farms weredropped until the list was reduced to the 3.55-million recordcutoff level. Subject matter analysts reviewed all recordsdropped from the mail list by the CART model and madechanges to derive the final dropped records. (For details ofthe 1992 CART modeling operation see Chapter 3, ‘‘Pre-paratory Operations.’’)

The Agriculture (AGR) Division staff evaluated the effec-tiveness of the CART methodology by examining—

x Associations between expected and observed farm pro-portions for all mail list CART groups.

x Differences between expected and observed farm pro-portions and possible variables that explain any differ-ences.

x Measures for comparing the 1992 and 1987 models.

x Differences between the original CARTdrops and analysts’adjustments, and comparisons to what was expected.

x The overall performance of the model, including mea-sures for correct record classification.

After completing the evaluation, the staff drew up a list ofrecommendations for improving the CARTmethodology foruse in the 1997 Census of Agriculture.


Statistical Methods

General information. The model drop survey evaluationused nonparametric comparisons (i.e., comparisons notinvolving estimating parameters of the statistical tests) insample paired tests to analyze the proportion of expectedfarms and observed farms in model groups on the final maillist. The statistical tests determined whether—

x There was a dependence between the expected andobserved paired observations (the results indicated theproportions were dependent).

x The expected and observed farm proportions had apositive or negative association (the proportions had astrong positive association).

Once dependence was established, the Census Bureautested the interchangeability or exchangeability of theexpected and observed farm proportions. The results indi-cated that there was not enough information available todetermine whether the respective proportions were inter-changeable. The Census Bureau also determined thatthere was a systematic difference between the expectedand observed proportions.

On the 1992 agriculture census mail list the observedfarm proportions were consistently greater than the expectedproportions for each model group. This was the opposite ofthe Census Bureau’s experience in the 1987 census, dueprimarily to differences in the overall proportion of farms onthe mail list from census to census.

In examining the differences between expected andobserved farm proportions, the Census Bureau’s evalua-tion tested for the possible impact of several other factorson the proportions, such as the State involved (the CARTmodeling was done by State) and the size of the modelgroup. Neither of these showed a significant impact onexplaining the variation in the differences.

Model Drop Survey analysis. In September and October1993, the Census Bureau carried out a Model Drop Evalu-ationSurvey using a sample of approximately 7,900 addressesdrawn from a national sample frame comprised of the229,180 addresses deleted from the census mail list by theCART procedures. The survey universe records weredivided into five strata based on expected farm probabilityof the record group assigned to it by the CART. A simplesystematic sample was selected for each strata. A total of5,892 responses were obtained by the survey, of which5,526 were classified as farm or nonfarm addresses.

Farm proportion estimation. The records used for themodel drop survey evaluation3 were classified in fourcategories based on CART model selection and the resultsof the subject-matter analysts’ review. The categorieswere—

x Cases selected for mailout by the CART, and statusconfirmed by analysts’ review.

x Cases selected for mailout by the CART, but changed tononmail by analysts.

x Cases dropped from the mailout by the CART, butrestored to the mailout list by analysts.

x Cases dropped from the mailout by the CART and statusconfirmed by analysts’ review.

Results from the Model Drop Survey and from the 1992agriculture census were combined to create the farmproportions for all four CART evaluation record categories.The expected farm proportion for each category wasestimated from the 1987 census mailout, while the observedfarm proportion was calculated using information on the1992 census mailout and the Model Drop Survey. Bycomparing the observed and expected farm proportions,analysts determined that the observed farm proportionswere higher than expected for all four record categories.The largest differences were in the groups of recordsoriginally dropped by the CART procedures but restored tothe mail list by analysts. The comparisons between recordsdropped from the mail list by the CART procedures, andthose dropped by subject-matter analysts showed that theanalysts’ adjustments improved the accuracy of the maillist. The proportion of records that proved to representfarms was larger (40.77 percent) among those recordsadded back to the mail list by analysts than in either theoriginal CART-dropped or analyst-dropped categories. As aresult, the proportion of farms on the mail list was esti-mated to have increased from approximately 58 percent toapproximately 58.4 percent after analysts’ adjustments,while the estimated proportion of farms among recordsdropped from the list fell dramatically, from just under 35percent, to just under 26 percent.

Evaluation Results

The general conclusion of the evaluation was that theCART methodology was successful in selecting whichaddresses should be included on the final census mail list.The CART methodology had been used in the 1987census, and the 1992 version proved to be an improve-ment over the earlier model, explaining twice as much ofthe variation in the observed farm proportion. The evalua-tion showed that 88.7 percent of the records on the mail listwere correctly assigned ‘‘farm’’ status. Approximately 64percent of the records dropped from the mail list werereceived from the National Agricultural Statistics Service(NASS), and over half of these records had a NASS farmsource only.

Despite the relative success of the CART model andother measures, approximately 26 percent of the recordsdropped from the mail list (229,810 in all) actually repre-sented farms. There was no significant difference in theproportions of farms dropped by analysts and those dropped

3Multiunits, abnormal farms, special list cases, and records selectedfor the census coverage evaluation were excluded from the CART model.All of these records were ‘‘certainty’’ cases, and were subject to intensivefollowup to ensure response.


by the statistical model (and not restored to the list byanalysts), but the overall figure was much higher thanexpected (the original estimate was that between 17 and18 percent of the dropped records would meet the censusfarm definition). The AGR staff suggested doing additionalresearchon improving theCARTmethodology, recommending—

x Improving CART input values.

x Increasing the minimum model group size.

x Developing additional steps in the CARTmethodology toreduce analyst adjustments, such as including 100 per-cent of all States with small farm counts (e.g., NewEngland), identifying groups for inclusion or exclusionbefore model application, and developing drop modelgroups by State.

x Selecting sample for the model drop survey by State.

x Investigating other possible methodologies for determin-ing expected farm status for mail list records.

RESEARCH STUDIES

Film Optical Sensing Device for Input toComputer (FOSDIC) Test

General information. Data entry—the transfer of informa-tion supplied by respondents on their report forms to theelectronic files that will be used for processing and tabu-lating the data—is one of the most expensive and laborintensive processing operations in any census. As part of ageneral effort to improve efficiency and reduce future costs,the Census Bureau evaluated an alternative method ofdata capture in the 1992 agriculture census in a test thattried to adapt the agency’s film optical sensing device forinput to computers (FOSDIC) system to the agriculturecensus processing requirements.

The Census Bureau has used the FOSDIC system asthe primary data-capture system in the decennial censusessince 1970, as well as in demographic surveys. For the1992 agriculture census, the FOSDIC equipment was usedin combination with data keying using a computer terminalconnected to a microfilm access device (MAD—the com-plete system was referred to as either the FOSDIC/MAD orFOSDIC/Key system). The FOSDIC system employs asimple optical mark recognition design; the system recog-nizes a mark in a given answer position (it does notrecognize characters).

The objectives of the Census Bureau’s test were toevaluate the potential of the FOSDIC/Key system to—

x Reduce processing time. TheCensusBureau expectedthe FOSDIC/Key system to reduce overall processingtime by eliminating the need for keyers to handle thepaper report forms and by reducing the number ofkeystrokes required to key data to the electronic file.

x Improvedataquality. TheagencyexpectedtheFOSDIC/Keysystem to have reduced error rates compared to con-ventional data entry.

x Reduce costs. Less data keyed and faster data key-ing would mean costs would be reduced. (However,adopting the FOSDIC/Key system involved additionalactivities, such as microfilming, purchasing and recondi-tioning microfilm readers, etc.).

x Indirect benefits. The Census Bureau expected thatthe experience gained from the FOSDIC/Key test couldbe applied to other systems, such as the development ofdigitized imaging for write-in responses, high-speed datacapture systems, and so on.

Test methodology. The AGR conducted the FOSDIC/Keysystem test in conjunction with the 1992 agriculture cen-sus. A special report form—the form 92-A0202(F)—wasadapted from the standard sample report form design usedin report form Region 2.4 The agriculture FOSDIC formrequired modifications to the FOSDIC equipment and scanprogram so that the reference marks used could be recog-nized by the system.

To simplify processing, theAGR decided that the samplereport forms for a given State in Region 2 (Illinois, Nebraska,and Ohio) would be either all FOSDIC or all non-FOSDIC,except for Iowa, in which half the sample mail list wouldreceive the FOSDIC form and half the non-FOSDIC form.The report forms were mailed out as part of the regularcensus report form mailings and followup operations. Atotal of 96,833 FOSDIC and 64,295 non-FOSDIC question-naires were mailed to addresses in the test area. Alto-gether, 81,104 FOSDIC and 53,364 non-FOSDIC reportforms were returned; 49,919 of the FOSDIC forms requiredkeying.

The States that were mailed non-FOSDIC forms servedas control States for use in comparing overall responserates. Data for all the FOSDIC report forms returned weredouble-keyed—i.e., the data were keyed using both theconventional keying system (see Chapter 6, Data Process-ing, for details of the data entry operation) and theFOSDIC/Key system. The AGR staff used information fromthe two keying systems for all tests on cost, time, quality,and response rates. Originally, the two systems were to bein operation simultaneously, but this proved impossible andthe FOSDIC/Key section began operations only after theconventional keying section had already completed dataentry for the States involved.

4The Census Bureau decided that the agriculture census FOSDICreport form should resemble, as closely as possible, the standard samplereport form. Hence, the agriculture FOSDIC report form did not have theFOSDIC index marks (black printed squares that serve as referencemarks by the FOSDIC equipment to locate answer positions) used on thedecennial census and other report forms. Instead, the horizontal dottedlines on the regular agriculture census report form were converted intosolid horizontal arrows for use as FOSDIC reference points. SpecialFOSDIC markings also were printed on the borders of the form to identifyquestionnaire pages.


Evaluation of FOSDIC operations. Once all the data fromthe report forms from the test areas had been entered inthe data file the AGR staff reviewed the performance of theFOSDIC/Key system and compared it to the conventionalsystems for comparable report forms to determine if theformer represented any significant advantage over conven-tional equipment and procedures.

To determine whether processing time could be reducedusing FOSDIC, AGR staff compared the average time andnumber of keystrokes per document required using thealternative systems. The results of this comparison showedthat the FOSDIC/Key system had a significant advantagein terms of the average time required to key and verifyreport forms—the average time requirement reduction rang-ing from about 21 percent (13 percent when adjusted forthe time needed to microfilm the report forms) for keying to23.5 percent for verification. The FOSDIC/Key system alsoproduced an approximate 37 percent reduction in theaverage number of keystrokes needed per document.

Cost comparisons were less favorable, although a reduc-tion in total keying time for the FOSDIC/Key systemtranslated into considerable savings in keying costs. How-ever, part of this reduction was offset by the cost ofmicrofilming, scanning, and reconditioning the MAD units.The total projected data entry savings for the agriculturecensus if the FOSDIC/Key system was used for datacapture was just $9,000.5

In terms of data quality, the FOSDIC/Key system proveda disappointment; the error rate for the FOSDIC/Keysystem was 2.2 percent, compared to 1.2 percent for thepaper system.

Results. The Census Bureau analysis of the FOSDIC/Keytest results showed that while the new system offeredsubstantial advantages in certain areas of processing—e.g.,average keystrokes required per document for dataentry—therewasnoclear advantages tousing theFOSDIC/Keysystem for keying future agriculture census report formdata. Furthermore, the agency considered that moreadvanced data capture technologies, such as electronicimaging, may be available in the near future. The overallrecommendation resulting from the test was that the Cen-sus Bureau not use the FOSDIC/Key system in the 1997Census of Agriculture, and that any resources available beinvested in developing more advanced systems.

Special Inserts Evaluation

General information. The Census Bureau has routinelyused instructional inserts in selected agriculture censusreport form packages. For the 1992 census, these insertsprovided special instructions for 10 different types of farm

operations—multiunits, abnormal farms(i.e.,Indian reserva-tions, institutions, farms, and grazing associations), cattlefeedlots, nursery and greenhouse operations, poultry con-tractors, bee and honey producers, fish and aquacultureoperations, and laboratory animal producers. Each insertinformed the recipients that their operation was considereda farm for census purposes, identified the sections withinthe report forms they received that should be completed,and provided information on how to report the data (e.g.,laboratory animal producers were to report any animalssuch as mice, rats, cavies (short-tailed, rough-haired SouthAmerican rodent), dogs, etc., writing in the name of eachspecies). The Census Bureau decided to test the effective-ness of the special inserts, evaluating particularly whetherusing the inserts improved the response rate, the in-scoperate, and data response sufficiently to justify the extra costsinvolved in using them as part of the 1992 agriculturecensus.

Study methodology. The special insert evaluation studyused a sample drawn from the census mail list. Since mostof the farm types receiving the inserts had a relatively smallnumber of farms, the sample was drawn from the threemajor types of insert recipients—poultry contractors, cattlefeedlots, and nursery and greenhouse producers. Cattlefeedlots were selected for the study because they weresubject to intensive followup anyway (thus no additionalcosts would be incurred in adding them to the sample),poultry contractors were added because of the poor datareceived previously, while nursery and greenhouse opera-tions were included because of their size.

The three farm industries were divided into two catego-ries, those addresses that would receive the appropriateinsert, and those that would not. The Census Bureauassigned numeric package codes (1-6) to each category.Codes ‘‘1’’ and ‘‘2’’ were assigned to the cattle feedlots, ‘‘3’’and ‘‘4’’ to the nursery and greenhouse operations, and ‘‘5’’and ‘‘6’’ to poultry contractors. The agency assigned thecodes sequentially through each group, then selectedthose packages with odd numbered codes to receive theinserts. The total number of records mailed with andwithout inserts in each farm industry was as follows:

Total Number of Records

Farm type Receivedinsert

Did notreceiveinsert

Cattle feedlots . . . . . . . . . . . . . . . . . . . . . . . . . 4,805 4,812Nursery and Greenhouse . . . . . . . . . . . . . . . 28,253 28,238Poultry contractors . . . . . . . . . . . . . . . . . . . . . 17,482 17,489

The AGR staff evaluated the responses using uneditedcensus data for sample records provided by the agriculturecensus processing operation. The research compared datafrom records for farms that received the inserts to data fromfarms that did not. The evaluation involved the followingfive activities:

x Analysis of response rates.

5Adoption of the FOSDIC/Key system, however, would obtain addi-tional savings in future censuses—assuming it proved a reliable alterna-tive to conventional systems—through the availability of indexed microfilmfor use in edit resolution and table review.


x Analysis of farm proportions.

x Comparison of data capture rates (i.e., presence orabsence of data in each key code) .

x Analysis of early response (i.e., farms responding to thefirst census mailing).

x Cost analysis.

The staff conducted each analysis separately for eachfarm type and computed estimates of the five specificareas of interest for the United States and for the leadingStates for each farm type.6 The staff developed its esti-mates and the associated variances using a simple ran-dom sampling methodology.

Results. Analysis of response rates showed there was nosignificant difference in responses between those receivingthe inserts and those that did not for any of the three majorfarm types used in the study. The response rate for cattlefeedlots receiving the insert, for example, was 99.04 per-cent, while those not receiving the insert responded at therate of 98.96 percent.

Similarly, the farm proportion (i.e., the number of respon-dents that met the census farm definition and were inscope)showed little or no difference for cattle feedlots and nurseryand greenhouse producers. For poultry contractors, therewas a significant difference of 1 percent between thoserecords that received the special insert and those that didnot—71.76 percent of respondents receiving the insertwere inscope, compared to 70.68 percent that were inscopeand had not received the insert.

Overall, the review of the presence or absence of data inkey codes showed little variation between categories.None of the three major farm types showed a significant

increase in the presence of any given section that could becorrelated with the use of the inserts. There were, however,some improvements at the State level; e.g., cattle feedlotsin Texas and Minnesota showed improvements of 3 per-cent and 2 percent, respectively, in the presence of data ofwith the use of the inserts. Nursery and greenhouseoperations in Florida also improved in this area by about 1percent with the use of the insert. On the other hand,contracted poultry in Arkansas showed an overall improve-ment in data presence without the insert.

The cost analysis of the use of the special insertsshowed that using them for the three major farm types, inaddition to bee and honey producers, laboratory animalproducers, and fish and aquaculture operations, cost theCensus Bureau approximately $12,000 ($2,000 for designand printing, and about $5,000 each for postage andadditional clerical work). Using inserts with multiunit andabnormal farms also represents additional cost to thecensus of about $2,000.

Recommendation. The special insert study showed thatusing the inserts for the three major farm types did not leadto any significant improvement in response rate or farmrate, and only very small improvement in the presence ofdata in each key code. The improvements realized did notseem to justify the additional costs associated with the useof the inserts.

The analysts involved in the study recommended the1997 Census of Agriculture—

x Drop the use of the inserts for the three major farm typescovered by the study—cattle feedlots, nursery and green-house operations, and poultry contractors.

x Include a 50-percent sample study of insert effective-ness for bee and honey producers, fish and aquacultureoperations, and laboratory animal operations.

x Continue to use special inserts for multiunit and abnor-mal operations.

x For new commodities, such as maple sap and Christmastrees, the instructional inserts will be sent to all opera-tors.

6The cutoff levels for State calculation varied by farm type. For cattlefeedlots, the evaluation developed estimates for eight States: Colorado,Illinois, Iowa, Kansas, Minnesota, Nebraska, South Dakota, and Texas.For nursery and greenhouse operations, estimates were prepared for 10States: California, Florida, Michigan, New York, North Carolina, Ohio,Oregon, Tennessee, and Texas. The major poultry contractor States wereAlabama, Arkansas, Georgia, and North Carolina.


Chapter 11.

Introduction --------------------------------------------------------------------------------------- 120General Information --------------------------------------------------------------------------- 120The Composition Systems -------------------------------------------------------------------- 120Quantity of 1992 Statistics Published -------------------------------------------------------- 120

Publication Media -------------------------------------------------------------------------------- 121General Information --------------------------------------------------------------------------- 121

Printed Reports -------------------------------------------------------------------------------- 121Volume 1, Geographic Area Series ------------------------------------------------------- 121Volume 2, Subject Series ------------------------------------------------------------------- 122Volume 3, Farm and Ranch Irrigation Survey (1994) ------------------------------------- 122

Other Publishing Media ----------------------------------------------------------------------- 122Public-use computer tape------------------------------------------------------------------- 122Compact disc-read only memory (CD-ROM) ---------------------------------------------- 122On-line access------------------------------------------------------------------------------- 123

Publication Process ------------------------------------------------------------------------------ 123General Information --------------------------------------------------------------------------- 123

Preliminary Preparation ----------------------------------------------------------------------- 123Text and graphics components ------------------------------------------------------------- 123Statistical tables ----------------------------------------------------------------------------- 123

Final Production Processing ------------------------------------------------------------------ 124Electronic Media Products -------------------------------------------------------------------- 124Pricing and Distributing Reports -------------------------------------------------------------- 124

Comparability of the Data------------------------------------------------------------------------ 125

Contents

Page


Publication Program

INTRODUCTION

General Information

The Census Bureau’s mission is to collect and dissemi-nate statistics on the population and economy of the UnitedStates, not only to various government offices, but to thegeneral public. Consequently, each census includes anextensive publication program designed to make censusdata available to users as economically as possible. The1992 agriculture census publication program provided thatdata be published in a variety of formats—printed reports,computer tapes, and compact-disc read only memory(CD-ROM)—accessible to the largest audience possible.

The Census Bureau issued detailed county- and State-level tabulations in the Volume 1, Geographic Area Seriesprinted reports—one for each State, Puerto Rico, the VirginIslands of the United States, and Guam, and the UnitedStates. The data file used for the Volume 1 reports formedthe core of the statistical data from the 1992 census, andalso was disseminated on computer tape and CD-ROM.The file included statistics on all agricultural operations thatmet the census farm definition (e.g., in the 50 States, anyplace from which $1,000 or more of agricultural productswere sold, or normally could be expected to have beensold, during the census year) as well as additional detaileddata for farms with annual sales of $10,000 or more.

The Composition Systems

The two principal components of the census data pub-lications are statistical tables and explanatory text andgraphics. The vast bulk of material published for the 1992census was in the form of statistical tables, and for the1992 agriculture census publications, the Agriculture Divi-sion (AGR) introduced the Tabulation And Disclosure Sys-tem (TADS) for developing and reviewing statistical tableselectronically. Prior to the 1992 census, the Administrativeand Publication Services Division (APSD1) used its CensusElectronic Publication System (CEPS) and a commerciallyavailable electronic graphics system (EGS) in conjunction

with the Table Image Processing System (TIPS) II forpreparing materials for publication. The introduction of theTADS required modification of the TIPS II system for usewith the DEC hardware employed by the TADS, which alsocould provide some of the graphics components. Thesesystems together provided computerized composition capac-ity for the census materials.

The CEPS enabled the publication staff to developautomated page layout for text, tables, and graphics; codetext and table files directly on the publication sponsor’selectronic file; merge graphics files and text as needed,and translate files from a variety of microcomputer andword processing systems. The TIPS II had been developedby the Census Bureau’s APSD and Systems SoftwareDivision to be used to produce large numbers of statisticaltables in printed reports. The system actually was acomputer program that controlled the computer-output-to-microform video composition system (COMp80) or theGovernment Printing Office’s (GPO’s) Videocomp system.Modified for use with the tables developed using the TADS,the TIPS II created tape files that were sent to GPO for usein producing photographic negatives of the publicationpages. The negatives and tapes then were returned toAPSD for review, and after each page had been approved,they were assembled and sent for publication.

Photo-offset reproduction was used for the printed reportsin the standard 8″ x 11″ page size.

Quantity of 1992 Statistics Published

Budget constraints led the Census Bureau to bothstretch out the publication program for the 1992 census,and to reduce the total volume of material published to theextent possible without reducing the usefulness of thecensus data. The 1987 census publication program hadincluded 2-page advance reports for each county andState, but no advance reports were released for the 1992census, which reduced the total volume of published pagesby 6,200. The 54-part Volume 1, Geographic Area Series,State and County Data, seriesAC92-A-1 to -54, set includedfinal State and county (or equivalent) detailed data for the50 States, Puerto Rico, Guam, the Virgin Islands of theUnited States, and a United States. The set contained over24,800 pages of tabular data and text, compared to approxi-mately 23,000 pages in the 1987 reports. The Volume 2,Subject Series, for 1992 comprised five reports:

1The APSD was retitled the Administrative and Customer ServicesDivision (ACSD) in the summer of 1995. Since the bulk of the 1992agriculture census publications were prepared and published while thedivision was still called APSD, that title will be used throughout thisHistory.


1. Agricultural Atlas of the United States, AC92-S-1, with204 pages of maps and charts illustrating nationalagricultural statistics.

2. Coverage Evaluation, AC92-S-2 (not printed; issued onCD-ROM and online).

3. Ranking States and Counties, AC92-S-3, with 111pages of tables showing the comparative ranking ofStates and counties for selected subjects.

4. History, AC92-S-4.

5. ZIP Code Tabulation of Selected ITems From the 1992Census of Agriculture, AC92-S-5, showed tabulationsof basic data items by five-digit ZIP Code areas, andwas available only on CD-ROM.

The Volume 3, Farm and Ranch Irrigation Survey (1994),AC92-FRIS-1, contained 178 pages of irrigation data forthe United States, States, and for 18 water resourcesareas.

PUBLICATION MEDIA

General Information

The Census Bureau has traditionally published its majordata findings in printed reports. The introduction of com-puters, and their use by both public and private data users,led the Census Bureau to begin publishing data in elec-tronically readable form, first on reels of computer tape,and later, as technology developed, on flexible diskettes,compact disc-read only memory (CD-ROM), and online.The 1992 agriculture census data were published in printedreports, computer tape, CD-ROM, and (for some tabula-tions) online. The Government Printing Office (GPO) soldthe printed reports, while the Census Bureau was the onlyprimary source for its own data on electronic media.

Printed Reports

Volume 1, Geographic Area Series , Series AC92-A-1 to-54. This volume comprised separate reports for the UnitedStates, the 50 States, the Commonwealth of Puerto Rico,the Virgin Islands of the United States, and Guam. EachState or area report included an introductory text giving ageneral description of the enumeration, two chapters ofdetailed statistical tables—chapter 1 shows State or areadata and chapter 2 the detailed tables for counties orcounty-equivalents—and appendixes with information oncoverage and error rates, a facsimile of a typical question-naire, and definitions. (Part 51, United States, showednational-level data in chapter 1, with summary State sta-tistics in chapter 2.)

Table 1, in chapter 1, in each State or area reportincluded historical highlights of the State’s agriculture fromthe 1959 through the 1992 censuses, followed by 51 tablesof detailed statistics covering every item collected fromagricultural operators in the subject State or area, including—

x Farms.

x Land in farms, land use, and irrigated land.

x Crop production and value of sales.

x Livestock and poultry inventories and sales.

x Selected characteristics of farms operated by females,and by persons of Spanish origin and specified racialgroups (Black, Asian and Pacific Islander, and AmericanIndian).

x Tenure and operator characteristics.

x Value of machinery and equipment.

x Agricultural chemicals and fertilizer used.

x Total production expenses and selected farm expenseitems.

x Commodity Credit Corporation (CCC) loans.

x Government payments and other farm-related income.

x Characteristics of farms by Standard Industrial Classifi-cation (SIC) code.

x Value of direct sales.


x Injuries and deaths.

The State tables usually showed comparable 1987 data.Tables 46 through 52 also showed 1992 State data cross-tabulated by various farm classifications, such as size offarm (based on sales and/or acreage), tenure of operator,type of organization, SIC code, and so on.

Chapter 2 presented county-level highlights and 39additional tables of selected data for all counties or county-equivalents for most of the subjects covered in the Statetables, plus—

x Market value of agricultural products sold.

x Net cash return from agricultural sales, Governmentpayments, other farm-related income, direct sales, andCommodity Credit Corporation (CCC) loans.

x Characteristics of farms with annual sales of $10,000 ormore.


x Major and minor crops.

x Value of land and buildings.


The first Volume 1 report (Delaware) appeared in Decem-ber 1994, and the last (for the Virgin Islands of the UnitedStates) in March 1995.

Volume 2, Subject Series. The Volume 2, Subject Series,consisted of five reports, series AC92-S-1 through -5.

x Agricultural Atlas of the United States, AC92-S-1, pre-sented a profile of American agriculture in a series ofU.S. dot and multicolor pattern maps. Clear mylar over-lays showing State and county outlines, in a pocketinside the back cover, enabled data users to use eventhe half-page size maps to see county-level geographicinformation. (Issued September 1995.)

x Coverage Evaluation, AC92-S-2, (available in print, onCD-ROM, and online) offered State level estimates ofthe number of farms not on the census mailing list;national and regional estimates of the number of opera-tions incorrectly classified, and the number of duplicatefarms; together with national, regional, and census geo-graphic division estimates of selected agricultural char-acteristics for missed farms.

x Ranking of States and Counties, AC92-S-3, showed theranking of the leading 20 States and counties (thenumber varied from table to table) for selected itemsfrom the 1992 Census ofAgriculture. Most tables includedcomparative statistics from the 1987 census, as well ascumulative total percentages of products or other itemsfor the highest ranked States and counties. (IssuedDecember 1995.)

x History, AC92-S-4, described the major census opera-tions and reproduced selected data-collection forms andmaterials.

x ZIP Code Tabulation of Selected Items From the 1992Census of Agriculture, AC92-S-5, available only onCD-ROM, this file contained farm counts by ZIP Code forselected items from the 1992 census, such as number offarms, land in farms, land in farms by size, market valueof agricultural products sold, and market value of prod-ucts sold by size, livestock inventory, selected crops,and other data items.

Volume 3, Farm and Ranch Irrigation Survey (1994) ,AC92-RS-1 (issued February 1996). The irrigation surveyreport presented statistics about on-farm irrigation prac-tices collected from a sample of irrigated farm operationsidentified in the 1992 census in the 48 conterminous States(excluding abnormal farms and horticultural specialty opera-tions). The survey data supplement the basic irrigationinformation collected in the census. The survey reportshowed information for the 18 water resources regions(WRR’s) of the 48 conterminous States, as well as for the27 leading irrigation States. The major data items shownincluded—

x Acreage in 1994.

x Land use and acres irrigated by category of land use.

x Acres and yields for irrigated and nonirrigated crops.

x Expenditures in 1994 for maintenance and repair ofirrigation equipment and facilities.

x Method of water distribution used in 1994.

x Source of water used.

x Energy use.

x Irrigation practices in 1994.

Other Publishing Media

Public-use computer tape. Despite the popularity of theCD-ROM products, the Census Bureau found that a marketstill existed for agriculture census data on computer tapeand released the major 1992 files on that medium as wellas on CD-ROM. The tape publication program involved thefinal county and State data tabulations. The county tapescontained final data available for each State and county orequivalent in the 50 States. The final State data werereleased in twofiles: theAfilehadsimpler tabulations—essentiallytables 1-45 of the Volume 1, Geographic Areas Series,printed reports; the B file showed more detailed tabulationsand cross tabulations drawn from tables 46-52 of theVolume 1 reports.

The Census Bureau released the final county data filefor selected States (that is, the buyer could request countydata for selected States (up to 40 on a single high-densitytape cartridge)), and a complete consolidated set on twohigh-density reels. The final State files were released ontwo tape reels, one each for the A and B files.

The Census Bureau’s computer tapes were available in9-track, 6,250-bpi (bits per inch) recording density, and ineither Extended Binary Coded Decimal Interchange Code(EBCDIC) or the American Scientific Code for InformationInterchange (ASCII) recording languages. Price for thedata tapes covered production, documentation, handling,and postage costs. Users could purchase the completefiles, or tapes containing only the data for specified States.

Compact disc-read only memory (CD-ROM). The rigidcompact disc (CD) used for electronic data file publicationin the 1992 census can hold up to 600 million characters(600 megabytes) of information—the equivalent of about1,500 flexible diskettes, or 4 high-density (6250 bpi) com-puter tape reels. The Census Bureau first released agricul-ture census data on CD-ROM in 1985, when the completefinal 1982 census data file was issued as part of the datacontent of two CD’s produced for test purposes. TheCensus Bureau issued two CD-ROM’s for the 1987 agri-culture census, one containing the State aggregate dataand the county-level data for the 1978, 1982, and 1987censuses; while the second had selected data from thevolume 2, 3, and 4 reports.


The Census Bureau originally planned to issue only twoCD-ROM’s for the 1992 census as well, the first coveringthe first 25 or so States for which the Volume 1 reports hadbeen released, while the second would cover all 50 Statesand the U.S. summary report for theVolume 1, GeographicAreas Series. However, soon after processing began, theCD-ROM program was expanded. The basic State andcounty data were issued on three discs—1992 Census ofAgriculture, Geographic Area Series, series CD-92AG-1Athrough -1C. Disc 1A included the first 27 State filesprocessed, covering basic State and county data, with croptables for selected crops only. Discs 1B and 1C, issued asa set, contained the basic data for all States and counties,plus the national-level summary data and cross-tabulationsby tenure of operator, size of farm, type of organization,SIC classification, value of products sold, governmentpayments, and occupation of operator. In addition to thethree discs covering the geographic areas data, the Cen-sus Bureau issued CD-92AG-ZIPS, with selected data atthe five-digit ZIP Code level for the United States andPuerto Rico, and planned to release two more—

1. CD-92AG-SPECIALTY, showing data for—

x The 1994 Farm and Ranch Irrigation Survey, togetherwith comparable data from the 1988 irrigation survey.

x Congressional districts (103rd Congress).

x 1992 Public Use Files, containing sample data on 119selected items.

x The 1992 census coverage evaluation, with coverageestimates for the 1992 Census of Agriculture.

2. CD-92AG-ATLAS, containing the base image files (BMF’s)for the 1992 Agricultural Atlas of the United States(CD-92AG-ATLAS will be issued if funding is avail-able).

On-line access. The Census Bureau’s Data User ServicesDivision (DUSD) administered the agency’s onlineservice—CENDATATM. The service is available to usersthrough two commercial vendors—CompuServe and DIA-LOG. The system carried highlights of the 1992 Geo-graphic Area Series, Volume 2, Subject Series, and Vol-ume 3, Farm and Ranch Irrigation Survey (1994) reports. Inaddition, selected agriculture census highlights were madeavailable through a third vendor, theAgriData Network. Theintroduction of the Internet system and its use by theCensus Bureau led to the agency releasing selected dataon the Internet system.

PUBLICATION PROCESS

General Information

The Census Bureau began general editorial and com-position work on the agriculture census publications inDecember 1989 in order to finish as much of the prepara-tory work associated with the data publication operations

before the completion of the census. The 1992 publicationplans included the introduction of the Tabulation and Dis-closure System (TADS), which enabled the AGR staff todevelop and review statistical tables electronically (fordetails of the TADS, see Chapter 6, ‘‘Data Processing’’).The AGR was responsible for developing the statisticaltables and wrote the text for the 1992 census data vol-umes, while the APSD prepared these materials for publi-cation. The AGR and DUSD cooperated in the productionof the electronic data files (i.e., computer tapes and com-pact disc-read only memory (CD-ROM)).

Preliminary Preparation

Text and graphics components. The APSD staff pro-duced the text components of the census reports on theCEPS and used the EGS for the graphics materials. TheAGR staff wrote standardized text (e.g., introductory mate-rial, definitions of terms, abbreviations, and symbols, etc.)for the census publications and sent it to the APSD onflexible diskette for editing and incorporation into the reportfile. Once the text was finalized, the APSD staff entered therequired codes into the electronic file containing the text forprocessing the text itself through the CEPS or, for textrelated to graphic illustrations, the EGS, and generatedlaser proofs (paper facsimiles of the planned final printedpages) for review and approval by the AGR staff. TheAPSD staff processed any corrections or changes andrecycled the laser proofs until all were approved, thencopied the electronic files containing the final text to themainframe computer at the Suitland, MD, headquarters forlater merging with other publication components.

The APSD staff prepared the variable text (i.e., textcontaining segments dependent on the results of the datacollected in the census) ‘‘shells’’ in advance, using thesame general procedures employed for the standard text,and held them on the CEPS for final production processing.

The census publications used three principle kinds ofgraphics—artwork for the report covers and title pages,and bar and piecharts displaying summary statistical dataand maps. The APSD and AGR staffs prepared the coverartwork and title pages on the EGS. When finalized, theartwork was transferred to the CEPS for merging with otherpublication components (i.e., titles and text) to produce thecovers and title pages. AGR analysts also used the EGS toprepare the basic chart ‘‘shells’’—without plot points andother variables while the Geography Division supplied theAPSD with negatives of the various maps required. (EachVolume 1 report included a map (showing counties (orequivalents) of the subject State or territory.) These mate-rials also were held for the final assembly before eachreport was sent to the printer.

Statistical tables. AGR staff used the TADS to prepareand review statistical tables in electronic data files, but theAPSD still had to develop the table layout parameters(TLP’s, or table ‘‘shells’’—essentially the table boxheadsand stubs, minus the variable geography and data) for the


statistical tables using electronic files supplied by the AGR.The APSD staff followed the same general proceduresemployed in preparing the standard texts for the reports,except that they used the TIPS II on the mainframecomputer for developing the TLP’s. The staff edited thetable manuscript and prepared the TLP computer files,processed them through the TIPS II, and submitted theresulting tables layouts to AGR for review and approval.Once AGR approved the tables, the final TLP file was heldin the mainframe computer for merging with the appropri-ate State or national data files and other publicationmaterials during the final production processing.

Final Production Processing

During final production processing, AGR supplied anyadditional information needed for the variable text compo-nents and charts to APSD. The AGR supervised EconomicProgramming Division (EPD) staff in executing variousAPSD programs to link the appropriate publication compo-nents and generate electronic files for transmission to theGPO for production of printing negatives. GPO sent theprinting negatives to the Census Bureau, where APSDprinting and editorial staff and AGR editors reviewed themfor conformity to Census Bureau publishing and printingstandards. All the 1992 census publications were printed,through the GPO, by private contractors, after APSDprepared the required printing documents and submittedthe various packages of materials to the contractor printersfor publication.

The APSD staff inspected the printed materials receivedfrom the contract printer to ensure that the individualreports were complete and met print quality standards.Inspectors randomly selected 1-5 individual copies (depend-ing on the total size of each shipment, with up to 5 copiespulled from each carton for the largest shipments) fromeach carton of volume 1 reports for checking by an APSDprinting specialist. While the overall quality of printing washigh, the inspection showed problems with the reports forseveral States (e.g., one contractor, printing the datareports for both South Carolina and Florida, had inadvert-ently intermingled pages from each State in each of thereports). When required, the reports for a State werereturned to the contractor involved for reprinting.

Electronic Media Products

In addition to disseminating selected data online throughthe CENDATATM, the DUSD also reproduced electronicmedia products (for the agriculture census, computer tapesand CD-ROM discs) from master data tapes supplied bytheAGR. TheAGR and DUSD staffs prepared the technicaldocumentation (TD) for each data file. The TD’s containedan abstract and detailed technical descriptions (e.g., geo-graphic coverage, identification numbering system, coding)of the file, file indexes and a data dictionary describing thecontents and record layout of the entire file.

The Census Bureau released all of the 1992, Volume 1,Geographic Area Series electronic data files on computertape and CD-ROM, as well as highlights online. The ZIPCode Tabulation of Selected Items From the 1992 Censusof Agriculture (Part 5) from the Volume 2, Subject Seriesreports was released on CD-ROM only, while highlights ofPart 3, Ranking States and Counties, were prepared andissued online. Selected tables from Volume 3, Farm andRanch Irrigation Survey (1994) were also made availableonline.

Pricing and Distributing Reports

The Census Bureau of the Census collected and tabu-lated the agriculture census and survey data, and preparedand released the electronic data files. Copies of thecomputer data tapes for all files were supplied to the Statedata centers (SDC’s), and the Census Bureau distributedCD-ROM discs containing the agriculture data to about 150Federal depository libraries and to any SDC’s or businessand industry data centers (BIDC’s) that requested them.The Census Bureau also sold census tapes and CD-ROMdiscs to the public.

The Census Bureau also prepared the text and tables,and did all the other preparatory work for each of theprinted publications, but printing, pricing, and sales of thecensus printed reports was the responsibility of the Super-intendent of Documents at the Government Printing Office(GPO). The Census Bureau paid all production costs, aswell as for printing copies for its own use. The GPO setprices for individual publications sold through its offices,based on a price schedule established by the GPO basedon the actual production (that is, printing), postage, andhandling costs of the reports purchased.

The GPO also handled distribution of Census Bureaupublications, including the agriculture census reports, tosome 1,500 Federal Government and Census depositorylibraries.2 About 10 percent of the Federal depositorylibraries—usually those at large universities—received all,or nearly all Census Bureau publications. The remainderreceived only those reports or products they specificallyrequested.

The Census Bureau distributed copies of its principaldata publications to its own 12 regional offices, to theInternational Trade Administration’s (ITA’s) district offices,and to over 300 SDC’s, as well as to the BIDC’s, and theiraffiliates (over 1,000 in all) in the 50 States, the District ofColumbia, Puerto Rico, Guam, and the Virgin Islands of theUnited States. (The SDC’s and BIDC’s received the prin-cipal publications for their own States, while the affiliateswere sent only those reports they requested.)

2The Federal depository libraries received a selection of the publica-tions from a variety of Federal departments and offices, including theCensus Bureau. Census depository libraries—about 130 in all, usuallylocal public libraries—typically received the census publications for theirlocal area or State.


The Data Preparation Division (DPD) office in Jefferson-ville, IN, maintained a supply of the Census Bureau’svarious publications, but the Superintendent of Documentshad primary responsibility for the storage and sales ofcensus printed reports.

The Census Bureau publicized the availability of its datapublications (print and electronic) through the GPO, the ITAand its district offices, and other organizations. The agencyissued press releases about new publications (for example,a separate press announcement describing the report, andincluding selected statistical highlights, was released foreach State report for the Volume 1, Geographic AreasSeries) and distributed order forms, publication schedules,a Monthly Product Announcement, and an annual CensusCatalog and Guide. Most of the publicity materials andcatalogs included brief descriptions of the products, as wellas actual or estimated publication dates, and purchaseprice. The DUSD issued a monthly newsletter, Census andYou, with articles on agency activities and publications,including the agriculture census, and the DUSD and AGRstaff jointly produced the Guide to the 1992 Census ofAgriculture and Related Statistics, containing a descriptionof the 1992 agriculture census publications and informationon data from other related Census Bureau activities.

COMPARABILITY OF THE DATA

A major objective of the data tabulation and publicationprogram is to maintain a reasonable degree of compara-bility of agriculture census data from census to census.Data comparability will be affected by changes in the pricestructure of the general economy, changes in the defini-tions of data concepts being measured, the introduction ofnew procedures for collecting and processing the data, andby the methods adopted for estimating and/or imputingdata.

The census farm definition used in the 1992 census hasbeen in place since 1974, and the basic data published inthe censuses are directly comparable in that sense. Acre-age, production, and inventory totals throughout the period1974-1992 are generally comparable, although dollar totalsfor expenses and sales are given in current dollars for eachcensus, unadjusted for inflation or deflation. Data forcensuses carried out prior to 1974 are not directly compa-rable to post-1974 enumerations because of changes inthe farm definition.


Appendix A.Provisions of Title 13, United States Code,Relating to the 1992 Census of Agriculture

CHAPTER 1.—ADMINISTRATIVE

Subchapter I—General Provisions

1. DefinitionsAs used in this title, unless the context requiresanother meaning or unless it is otherwise provided—(1) ‘‘Bureau’’ means the Bureau of the Census;(2) ‘‘Secretary’’ means the Secretary of Commerce;

and(3) ‘‘respondent’’ includes a corporation, company,

association, firm, partnership, proprietorship,society, joint stock company, individual, or otherorganization or entity which reported information,or on behalf of which information was reported,in response to a questionnaire, inquiry, or otherrequest of the Bureau.

2. Bureau of the CensusThe Bureau is continued as an agency within,and under the jurisdiction of, the Department ofCommerce.

5. Questionnaires; number, form, and scope ofinquiriesThe Secretary shall prepare questionnaires, andshall determine the inquiries, and the number, form,and subdivisions thereof, for the statistics, surveys,and censuses provided for in this title.

6. Information from other Federal departments andagencies; acquisition of reports from other gov-ernmental and private sources(a) The Secretary, whenever he considers it advis-

able, may call upon any other department, agency,or establishment of the Federal Government, orof the government of the District of Columbia, forinformation pertinent to the work provided for inthis title.

(b) The Secretary may acquire, by purchase orotherwise, from States, counties, cities, or otherunits of government, or their instrumentalities, orfrom private persons and agencies, such copiesof records, reports, and other material as may berequired for the efficient and economical conductof the censuses and surveys provided for in thistitle.

(c) To the maximum extent possible and consistentwith the kind, timeliness, quality and scope of thestatistics required, the Secretary shall acquireand use information available from any sourcereferred to in subsection (a) or (b) of this sectioninstead of conducting direct inquiries.

7. Printing; requisitions upon Public Printer; publi-cation of bulletins and reportsThe Secretary may make requisitions upon the Pub-lic Printer for miscellaneous printing necessary tocarry out the provisions of this title. He may furtherhave printed by the Public Printer, in such editions ashe deems necessary, preliminary and other censusbulletins, and final reports of the results of theseveral investigations authorized by this title, andmay publish and distribute such bulletins and reports.

8. Authenticated transcripts or copies of certainreturns; other data; restriction on use; disposi-tion of fees received(a) The Secretary may, upon written request, furnish

to any respondent, or to the heir, successor, orauthorized agent of such respondent, authenti-cated transcripts or copies of reports (or portionsthereof) containing information furnished by, oron behalf of, such respondent in connection withthe surveys and census provided for in this title,upon payment of the actual or estimated cost ofsearching the records and furnishing such tran-scripts or copies.

(b) Subject to the limitations contained in sections6(c) and 9 of this title, the Secretary may furnishcopies of tabulations and other statistical mate-rials which do not disclose the information reportedby, or on behalf of, any particular respondent,and may make special statistical compilationsand surveys, for departments, agencies, andestablishments of the Federal Government, thegovernment of the District of Columbia, the gov-ernment of any possession or area (includingpolitical subdivisions thereof) referred to in sec-tion 191(a) of this title, State or local agencies, orother public and private persons and agencies,upon payment of the actual or estimated cost ofsuch work. In the case of nonprofit agencies ororganizations, the Secretary may engage in joint

APPENDIX A A-1HISTORY—1992 CENSUS OF AGRICULTURE

statistical projects, the purpose of which areotherwise authorized by law, but only if the costof such projects [is] shared equitably, as deter-mined by the Secretary.

(c) In no case shall information furnished under thissection be used to the detriment of any respon-dent or other person to whom such informationrelates, except in the prosecution of allegedviolations of this title.

(d) All moneys received in payment for work orservices enumerated under this section shall bedeposited in a separate account which may beused to pay directly the costs of such work orservices, to repay appropriations which initiallybore all or part of such costs, or to refund excesssums when necessary.

9. Information as confidential; exception

(a) Neither the Secretary, nor any other officer oremployee of the Department of Commerce orbureau or agency thereof, may, except as pro-vided in section 8 of this title—

(1) use the information furnished under the pro-visions of this title for any purpose other thanthe statistical purposes for which it is sup-plied; or

(2) make any publication whereby the data fur-nished by any particular establishment orindividual under this title can be identified; or

(3) permit anyone other than the sworn officersand employees of the Department or bureauor agency thereof to examine the individualreports.

Nodepartment, bureau, agency, officer, or employeeof the Government, except the Secretary in car-rying out the purposes of this title, shall require,for any reason, copies of census reports whichhave been retained by any such establishment orindividual. Copies of census reports which havebeen so retained shall be immune from legalprocess, and shall not, without the consent of theindividual or establishment concerned, be admit-ted as evidence or used for any purpose in anyaction, suit, or other judicial or administrativeproceeding.

(b) The provisions of subsection (a) of this sectionrelating to the confidential treatment of data forparticular individuals and establishments, shallnot apply to the censuses of governments pro-vided for by subchapter III of chapter 5 of thistitle, nor to interim current data provided for bysubchapter IV of chapter 5 of this title as to thesubjects covered by censuses of governments,with respect to any information obtained thereforthat is compiled from, or customarily provided in,public records.

12. Mechanical and electronic developmentThe Secretary is authorized to have conductedmechanical and electronic development work as hedetermines is needed to further the functions andduties of carrying out the purposes of this title andmay enter into such developmental contracts as hemay determine to be in the best interest of theGovernment.

Subchapter II—Officers and Employees

21. Director of the Census; dutiesThe Bureau shall be headed by a Director of theCensus, appointed by the President, by and with theadvice and consent of the Senate. The Directorshall perform such duties as may be imposed uponhim by law, regulations, or orders of the Secretary.

22. Qualifications of permanent personnelAll permanent officers and employees of the Bureaushall be citizens of the United States.

23. Additional officers and employees(a) The Secretary may establish, at rates of com-

pensation to be fixed by him without regard tothe Classification Act of 1949, as many tempo-rary positions as may be necessary to meet therequirements of the work provided for by law.Bureau employees who are transferred to anysuch temporary positions shall not lose theirpermanent civil service status by reason of thetransfer. The Secretary may make appointmentsto such temporary positions in conformity withthe civil service laws and rules.

(b) In addition to employees of the Department ofCommerce, employees of other departmentsand independent offices of the Governmentmay, with the consent of the head of the respec-tive department or office, be employed andcompensated for field work in connection withthe work provided for by law without regard tosection 301 of the Dual Compensation Act.

(c) The Secretary may utilize temporary staff, includ-ing employees of Federal, State, or local agen-cies or instrumentalities, and employees of pri-vateorganizations toassist theBureau inperformingthe work authorized by this title, but only if suchtemporary staff is sworn to observe the limita-tions imposed by section 9 of this title.

24. Special employment provisions(a) The Secretary may utilize the services of non-

temporary employees of the Bureau (by assign-ment, promotion, appointment, detail, or other-wise) in temporary positions established for anycensus, for not to exceed the period during

A-2 APPENDIX A HISTORY—1992 CENSUS OF AGRICULTURE

which appropriations are available for that cen-sus. Whenever the Secretary determines thatthe services of an employee which have beenutilized under this section are no longer requiredin such a temporary position, he may, withoutregard to the provisions of any other law, returnthe employee to a continuing position, with rankand compensation not less than that which heheld in his last permanent position in the Bureau:Provided, That no employee shall, by reason ofhis service in a temporary position under thissubsection, lose the protection of any law orregulation with respect to his separation, sus-pension, furlough, or reduction in rank or com-pensation below the level held in his last per-manent position in the Bureau. Service by anontemporary employee in a temporary posi-tion under this subsection shall be creditable forstep increases (both periodic and longevity)under title VII of the Classification Act of 1949,as amended, as though it were a continuation ofservice in his last permanent position.

(b) As used in this title with respect to appoint-ments or positions, ‘‘temporary’’ shall be con-strued to mean not in excess of one year, or notin excess of the specific period during whichappropriations are available for the conduct of aparticular census,whichever is longer.Noemployeeof the Bureau who holds only a temporaryappointment within the meaning of this sectionshall be considered as other than strictly tem-porary for purposes of any other provision oflaw relating to separations, suspensions, orreductions in rank or compensation.

(c) The enlisted men and officers of the uniformedservices may be appointed and compensatedfor service in temporary enumerator positionsfor the enumeration of personnel of the uni-formed services.

(d) The Secretary may fix compensation on a piece-price basis without limitation as to the amountearned per diem, and payments may be madeto enumerators for the use of private automo-biles on official business without regard to sec-tion 4 of the Travel Expense Act of 1949, asamended (5 U.S.C. 837), but at rates not inexcess of the rates provided by that Act.

(e) The Secretary may authorize the expenditure ofnecessary sums for travel expenses of personsselected for appointment for attendance at train-ing courses held by the Department of Com-merce with respect to any of the work providedfor by law.

(f) Notwithstanding any other provision of law pro-hibiting the expenditure of public money fortelephone service, the Secretary, under suchregulations as he shall prescribe, may authorizereimbursement for tolls or charges for telephone

service from private residences or private apart-ments to the extent such charges are deter-mined by the Secretary to have been incurred tofacilitate the collection of information in connec-tion with the censuses and surveys authorizedby this title.

25. Duties of supervisors, enumerators, and otheremployees(a) Each supervisor shall perform the duties imposed

upon him by the Secretary in the enforcementof chapter 5 of this title in accordance with theSecretary’s orders and instructions.

(b) Each enumerator or other employee detailed toserve as enumerator shall be charged with thecollection in his subdivision of the facts andstatistics called for on such schedules as theSecretary determines shall be used by him inconnection with any census or survey providedfor by chapter 5 of this title.

26. Transportation by contractThe Secretary may contract with field employees forthe rental and use within the continental limits of theUnited States of means of transportation, other thanmotorcycle, automobile, or airplane, and for therental and use outside of the continental UnitedStates of any means of transportation, which meansmay be owned by the field employee. Such rentalcontracts shall be made without regard to section 4of the Travel Expense Act of 1949, as amended (5U.S.C. 837). The rentals shall be at rates equivalentto the prevailing rental rates of the locality. Therental contracts within the continental United Statesmay be entered into only when the use by the fieldemployee of such other means of transportation issafer, more economical, or more advantageous tothe Government than use of his motorcycle, auto-mobile, or airplane in conducting the census.

CHAPTER 5.—CENSUSES

Subchapter II—Population, Housing, Agriculture,Irrigation, and Unemployment

142. Agriculture and Irrigation(a) The Secretary shall in 1979, in 1983, and in

every fifth year beginning after 1983, take acensus of agriculture.

(b) In conjunction with the census to be takenunder subsection (a) of this section in 1979, in1988, and every tenth year beginning after1988, the Secretary shall take a census ofirrigation.

(c) The data collected in each of the censusestaken under this section shall relate to the yearimmediately preceding the year in which suchcensus is taken.


Subchapter V—Geographic Scope, Preliminaryand Supplemental Statistics, and Use ofSampling

191. Geographic scope of censuses(a) Each of the censuses authorized by this chap-

ter shall include each State, the District ofColumbia, the Virgin Islands, Guam, the Com-monwealth of the Northern Mariana Islands,and the Commonwealth of Puerto Rico, andas may be determined by the Secretary, suchother possessions and areas over which theUnited States exercises jurisdiction, control, orsovereignty. Inclusion of other areas over whichthe United States exercises jurisdiction orcontrol shall be subject to the concurrence ofthe Secretary of State.

(b) For censuses taken in the Virgin Islands, Guam,the Commonwealth of the Northern MarianaIslands, or any possession or area not specifi-cally designated in subsection (a) of this sec-tion, the Secretary may use census informa-tion collected by theGovernor or highest rankingFederal official, if such information was obtainedin accordancewith plans prescribed or approvedby the Secretary.

(c) If, pursuant to a determination by the Secretaryunder subsection (a) of this section, any cen-sus is not taken in a possession or area overwhich the United States exercises jurisdiction,control, or sovereignty, the Secretarymay includedata obtained from other Federal agencies orgovernment sources in the census report. Anydata obtained from foreign governments shallbe obtained through the Secretary of State.

193. Preliminary and supplemental statisticsIn advance of, in conjunction with, or after thetaking of each census provided for by this chap-ter, the Secretary may make surveys and collectsuch preliminary and supplementary statisticsrelated to the main topic of the census as arenecessary to the initiation, taking, or completionthereof.

195. Use of samplingExcept for the determination of population forpurposes of apportionment of Representatives ofCongress among the several States, the Secretaryshall, if he considers it feasible, authorize the useof the statistical method known as ‘‘sampling’’ incarrying out the provisions of this title.

CHAPTER 7.—OFFENSES AND PENALTIES

Subchapter I—Officers and Employees

211. Receiving or securing compensation for appoint-ment of employees

Whoever—

(1) receives or secures to himself any fee, reward,or compensation as a consideration for theappointment of any person as supervisor,enumerator, clerk, or other officer or employeeof the Department of Commerce or bureau oragency thereof, referred to in subchapter II ofchapter I of this title; or

(2) in any way receives or secures to himselfany part of the compensation paid to anyperson so appointed— shall be fined notmore than $3,000 or imprisoned not morethan five years, or both.

212. Refusal or neglect of employees to performdutiesWhoever, being an employee referred to in sub-chapter II of chapter I of this title, and having takenand subscribed the oath of office, neglects orrefuses, without justifiable cause, to perform theduties enjoined on such employee by this title,shall be fined not more than $500.

213. False statements, certificates, and information(a) Whoever, being an officer or employee referred

to in subchapter II of chapter I of this title,willfully and knowingly swears or affirms falselyas to the truth of any statement required to bemade or subscribed by him under oath by orunder authority of this title, shall be guilty ofperjury, and shall be fined not more than$2,000 or imprisoned not more than five years,or both.

(b) Whoever, being an officer or employee referredto in subchapter II of chapter I of this title—(1) willfully and knowingly makes a false cer-tificate or fictitious return; or(2) knowingly or willfully furnishes or causesto be furnished, or, having been such an officeror employee, knowingly or willfully furnished orcaused to be furnished, directly or indirectly, tothe Secretary or to any other officer or employeeof the Department of Commerce or bureau oragency thereof, any false statement or falseinformation with reference to any inquiry forwhich he was authorized and required to col-lect information provided for in this title—shallbe fined not more than $2,000 or imprisonednot more than five years, or both.

214. Wrongful disclosure of informationWhoever, being or having been an employee orstaff member referred to in subchapter II of chapterI of this title, having taken and subscribed the oathof office, or having sworn to observe the limitationsimposed by section 9 of this title, publishes orcommunicates any information, the disclosure ofwhich is prohibited under the provisions of section9 of this title, and which comes into his possession


by reason of his being employed (or otherwiseproviding services) under the provisions of thistitle, shall be fined not more than $5,000 or impris-oned not more than 5 years, or both.

Subchapter II—Other Persons

221. Refusal or neglect to answer questions; falseanswers

(a) Whoever, being over eighteen years of age,refuses or willfully neglects, when requestedby the Secretary, or by any other authorizedofficer or employee of the Department of Com-merce or bureau or agency thereof actingunder the instructions of the Secretary orauthorized officer, to answer, to the best of hisknowledge, any of the questions on any sched-ule submitted to him in connection with anycensus or survey provided for by subchaptersI, II, IV, and V of chapter 5 of this title, applyingto himself or to the family to which he belongsor is related, or to the farm or farms of whichhe or his family is History of the 1992 Censusthe occupant, shall be fined not more than$100.

(b) Whoever, when answering questions describedin subsection (a) of this section, and under theconditions or circumstances described in suchsubsection, willfully gives any answer that isfalse, shall be fined not more than $500.

(c) Notwithstanding any other provision of thistitle, no person shall be compelled to discloseinformation relative to his religious beliefs or tomembership in a religious body.

223. Refusal, by owners, proprietors, etc., to assistcensus employees

Whoever, being the owner, proprietor, manager,superintendent, or agent of any hotel, apartmenthouse, boarding or lodging house, tenement, orother building, refuses or willfully neglects, whenrequested by the Secretary or by any other officeror employee of the Department of Commerce orbureau or agency thereof, acting under the instruc-tions of the Secretary, to furnish the names of theoccupants of such premises, or to give free ingressthereto and egress therefrom to any duly accred-ited representative of such Department or bureauor agency thereof, so as to permit the collection ofstatistics with respect to any census provided for insubchapters I and II of chapter 5 of this title, or anysurvey authorized by subchapter IV or V of suchchapter insofar as such survey relates to any of thesubjects for which censuses are provided by suchsubchapters I and II, including, when relevant tothe census or survey being taken or made, the

proper and correct enumeration of all personshaving their usual place of abode in such pre-mises, shall be fined not more than $500.

224. Failure to answer questions affecting compa-nies, businesses, religious bodies, and otherorganizations; false answersWhoever, being the owner, official, agent, personin charge, or assistant to the person in charge, ofany company, business, institution, establishment,religious body, or organization of any nature what-soever, neglects or refuses, when requested bythe Secretary or other authorized officer or employeeof the Department of Commerce or bureau oragency thereof, to answer completely and cor-rectly to the best of his knowledge all questionsrelating to his company, business, institution, estab-lishment, religious body, or other organization, orto records or statistics in his official custody, con-tained on any census or other schedule or ques-tionnaire prepared and submitted to him under theauthority of this title, shall be fined not more than$500; and if he willfully gives a false answer to anysuch question, he shall be fined not more than$10,000.

225. Applicability of penal provisions in certain cases(a) In connection with any survey conducted by

the Secretary or other authorized officer oremployee of the Department of Commerce orbureau or agency thereof pursuant to subchap-ter IV of chapter 5 of this title, the provisions ofsections 221, 222, 223 and 224 of this titleshall apply—

(1) with respect to the answering of questionsand furnishing of information, only to suchinquiries as are within the scope of theschedules and questionnaires and of thetype and character heretofore used in con-nection with the taking of complete cen-suses under subchapters I and II of chap-ter 5 of this title, or in connection with anycensuses hereafter taken pursuant to suchsubchapters;

(2) only after publication of a determinationwith reasons therefor certified by the Sec-retary, or by some other authorized officeror employee of the Department of Com-merce or bureau or agency thereof withthe approval of the Secretary, that theinformation called for is needed to aid orpermit the efficient performance of essen-tial governmental functions or services, orhas significant application to the needs ofthe public, business, or industry and is notpublicly available from nongovernmentalor other governmental sources;


(3) in the case of any new survey, only afterpublic notice, given by the Secretary orother authorized officer or employee of theDepartment of Commerce or bureau oragency thereof at least thirty days in advanceof requesting a return, that such survey isunder consideration.

(b) The provisions for imprisonment provided bysection 222 of this title shall not apply inconnection with any survey conducted pursu-ant to subchapter II of chapter 3 of this title, orto subchapter IV of chapter 5 of this title.

(c) The provisions of sections 221, 222, 223, and224 of this title shall not apply to any censusesor surveys of governments provided for bysubchapters III and IV of chapter 5 of this title,nor to other surveys provided for by subchap-ter IV of such chapter which are taken morefrequently than annually.

(d) Where the doctrine, teaching, or discipline ofany religious denomination or church prohibitsthe disclosure of information relative to mem-bership, a refusal, in such circumstances, tofurnish such information shall not be an offenseunder this chapter.

Subchapter III—Procedure

241. Evidence

When any request for information, made by theSecretary or other authorized officer or employeeof the Department of Commerce or bureau oragency thereof, is made by registered or certifiedmail or telegram, the return receipt therefor orother written receipt thereof shall be prima facieevidence of an official request in any prosecutionunder such section.


Appendix B.

Origins of the Census---------------------------------------------------------------------------- 2The First Censuses: 1790-1840 -------------------------------------------------------------- 2The Agriculture Census ----------------------------------------------------------------------- 2

The 20th Century--------------------------------------------------------------------------------- 3Program Development------------------------------------------------------------------------- 3The establishment of a permanent Census Bureau--------------------------------------- 3Agriculture census programs --------------------------------------------------------------- 3Enumeration dates, timing, and methodologies ------------------------------------------- 4Sampling------------------------------------------------------------------------------------- 4Agricultural services census ---------------------------------------------------------------- 5Follow-on surveys --------------------------------------------------------------------------- 5

Report Form Content and Format ------------------------------------------------------------ 5Content -------------------------------------------------------------------------------------- 5Format --------------------------------------------------------------------------------------- 5

Processing and Publishing the Data---------------------------------------------------------- 6Processing----------------------------------------------------------------------------------- 6Publication ----------------------------------------------------------------------------------- 7

The Farm Definition ------------------------------------------------------------------------------ 7

Contents

Page

APPENDIX B B–1HISTORY—1992 CENSUS OF AGRICULTURE

Appendix B.Historical Notes

ORIGINS OF THE CENSUS

The First Censuses: 1790-1840

The Constitutional Convention in 1787 set many prece-dents, among which was the incorporation in the governinginstrument of the new Nation of a requirement for a periodiccount of the population of that Nation for purposes of theequitable distribution to each State of taxes and represen-tation in the House of Representatives. Article I, section 2,of the United States Constitution required an enumerationof the ‘‘whole Number of free Persons, including thosebound to Service for a Term of Years, and excluding Indiansnot taxed, three-fifths of all other Persons...within threeYears after the first Meeting of the Congress of the UnitedStates, and within every subsequent Term of ten years....’’

The first census was carried out by the new Governmentin 1790, but even some members of the ConstitutionalConvention had recognized that a periodic and universalenumeration of the country could be used to gather muchuseful information on more than the population. JamesMadison, when a member for Virginia of the first House ofRepresentatives, introduced a bill to use the census tocollect information on occupations and other economicsubjects, as well as basic data on the age, sex, and race ofthe population. The House approved the idea, but theSenate rejected it, and Madison had to content himself witha census collecting information only on the number ofinhabitants, their age, sex, and whether slave or free.Nevertheless, the rapid growth of the new Nation sooncaused the Federal Government and others to look forsomemeans of measuring that growth, and in 1810—duringthe first administration of President James Madison—itemson the kind, quantity, and value of goods manufacturedwere included in the third national enumeration.1From itsfounding until well into the 19th Century, the United Stateswas primarily an agricultural nation—95 percent of thepopulation counted in the first census lived on the land. The1820 census was the first to include any question onagriculture, and it asked only how many persons in eachhousehold were engaged in agriculture. (Approximately 70percent of the total population at the time were so engaged.)

The 1840 census introduced separate schedules ofquestions relating to mining, agriculture, commerce, manu-factures and trades, and navigation. The agriculture sched-ule included questions on cereals and other crops, and onlivestock, and the results were published with the rest ofthe census data. Even this expanded information wasconsidered unreliable and lacking in detail, and there weredemands that the census collect more detailed information,and that the Government pay more attention to ensuringthe accuracy of the data collected and published.

The Agriculture Census

The dissatisfaction with the 1840 census promptedCongress to give particular attention to the organizationand data content of the 1850 census. A select committee ofthe House of Representatives recommended that the Fed-eral Government establish a permanent census office. TheSenate shared many of the concerns expressed by theHouse, but declined to support the measure, so a perma-nent census office had to wait another half-century. Never-theless, the 1850 census assumed much of the specializedorganization that has since characterized the censuses,and is often considered the first ‘‘modern’’ enumeration.

The increased specialization began at the top, where theCongress transferred the responsibility for supervising thecensus from the Department of State to the newly formedDepartment of the Interior. The census was organized intosix subject areas for data collection, including agriculture,each with a separate list of questions. The agriculturequestionnaire, or ‘‘schedule,’’ asked for the name of theperson(s) in each household who operated a farm, andmade relatively detailed inquiries on acreage and agricul-tural activities, including quantities produced of selectedproducts; the value of farm implements and machinery,livestock, animals slaughtered, and homemade manufac-tures; and the cash value of the farm. The 1850 censuspublications included the total number of farms for theUnited States (1.4 million) and each State; acreage (294million acres under cultivation); and total value of farms,buildings, livestock, machinery, and equipment (nearly $4billion).

The censuses became more detailed as the centuryprogressed. The census law of 1879 provided for theappointment by the President (with Senate confirmation) ofup to 150 local supervisors for the 1880 census (at least 1for each State or territory) as well as for employing

1The Federal Government published a separate report covering thenondemographic data collecting in the third census: A Statement of theArts and Manufactures of the United States of America, for the Year 1810.This document may be considered the forerunner of the Census Bureau’senormous economic statistical publishing program.

B–2 APPENDIX B HISTORY—1992 CENSUS OF AGRICULTURE

specialists to collect data on certain census subject matterareas, such as manufactures and mining. The local super-visors were responsible for the actual data collection, aswell as for hiring suitable enumerators. This enlarged staffpermitted closer supervision of the enumeration and thus,it was hoped, greater accuracy. The 1880 agriculturecensus schedule included new items on tenure, weeks ofhired labor, costs for building and maintaining fences, andcost of fertilizer purchased, and used specialized question-naires to collect detailed production information on selectedcrops and livestock.

The agriculture census expanded further for 1890, whenthe agriculture schedule doubled in length—including newinquiries on agricultural organizations, floriculture, andirrigation—and again used special questionnaires for selectedoperations. The population census also collected agricul-tural data, asking for information on farm mortgages.

The 1900 agricultural census introduced a question onthe race of the farm operator, but was otherwise similar tothe previous enumeration. The 1900 census saw theagriculture census enter the age of automated data pro-cessing when the census staff used punchcards andelectric tabulating machines to process and tabulate thestatistics (the punchcard tabulating equipment had firstbeen used for processing in the 1890 population census).This equipment tabulated the results of the 1890 popula-tion census, and was adapted for the agriculture enumera-tion by the development and addition of an automatic sorter(required because of the large number of crop cards usedin processing the agriculture data) and the use of a newkeypunch machine.

THE 20TH CENTURY

Program Development

The establishment of a permanent Census Bureau.The first 12 U.S. censuses used temporary organizations,established a few months before the official census dateand disbanded once the data had been collected andpublished. It soon was evident that the elaboration of thecensus operation, and of the kinds and volume of datatabulated and published, required more time and resourcesthan could be provided by a staff and office organized anddisbanded for each census. The results of the 1880 censusdissatisfied many data users, and led to a campaign byinterested professional and commercial organizations—e.g.,the National Board of Trade, the American StatisticalAssociation, and the American Economic Association—fora permanent census office. The 12th census, for 1900, wasthe last carried out on the old ad hoc basis. In 1902 theCongress authorized the establishment of the permanentcensus office within the Department of the Interior. The newunit, later designated the Bureau of the Census, moved tothe newly created Department of Commerce and Labor inthe following year, and when the Department was dividedin 1913, was assigned to the Department of Commerce,where it has since remained.

Agriculture census programs. By 1880, the variouscensuses’ statistics had proved so useful that the formerchairman of theAmerican Statistical Association suggestedexpanding the decennial census by adding a mid-decadeenumeration. Many of the same persons and associationspushing for a permanent statistical office also urged this, aswell as the addition of specialized censuses to the agricul-ture program. In 1910, the agriculture census programbegan to expand, adding related enumerations, as follows—

x In 1910, the Congress directed the Census Bureau tocarry out a decennial census of irrigation as part of theagriculture census.

x In 1920, the Congress required the Census Bureau toconduct decennial censuses of drainage. (The legalrequirement for a census of drainage lasted into the1980’s, but in 1985, lack of interest among data usersprompted Congress to remove it from the census law.)

x Special censuses of horticultural specialties also wereadded to the agriculture census program, although atirregular intervals, i.e., as part of the 1890, 1930, 1950,1959, 1969, 1978, and 1987 censuses.

The ‘‘special’’ censuses usually were carried out for theyear following the general agriculture census’s referenceyear (the 1994 Farm and Ranch Irrigation Survey—originallyscheduled for 1993—was postponed a year due to budgetconstraints), using operator lists drawn from the censusrespondent lists.

Congress also considered the question of mid-decadeenumerations, and, in 1919, after wavering back and forthon the question for some years, authorized a mid-decadeagriculture enumeration beginning with a census for 1925(in addition to the decennial censuses for years ending in‘‘0’’). Subsequent agricultural censuses have been on a5-year cycle in years ending in ‘‘0’’ and ‘‘5,’’ collecting datafor the preceding crop year. This system continued until1954, when Congress codified the Census Bureau’s opera-tions in Title 13, United States Code—Census. Title 13established the new reference years for the agriculturalcensus—years ending in ‘‘4’’ and ‘‘9’’ (again collecting datafor the preceding year), although the agricultural (andirrigation and drainage) censuses, ostensibly done in yearsending in ‘‘9,’’ actually continued to be carried out as part ofthe decennial census operation.

The advantages of obtaining agricultural and economiccensus data for the same year, providing a single ‘‘snap-shot’’ of the national economy, were obvious to mostobservers, and it was not long before there was consider-able support for simultaneous agricultural and economiccensuses. The Department of Commerce and the CensusBureau recommended in 1972 that the dates of the agri-culture census be changed to coincide with those of theeconomic censuses; in 1976 Congress enacted Public Law94-229, requiring agriculture censuses for 1978, 1982, andevery fifth year thereafter, making them concurrent, for1982 and after, with the economic censuses.


Enumeration dates, timing, and methodologies. Priorto the 1950’s, the agriculture census was usually carriedout during the first months of the year following thereference year. The Census Bureau enumerated agricul-ture at the same time as the population censuses indecennial census years—e.g., in April of the census yearfor 1910, and from 1930 through 1950 (the 1920 censusdate was January 1 of that year)—to make use of the fieldand office staffs already in place. The long gap between theend of the reference year and the beginning of the cen-suses in these years presented a significant problem forrespondents trying to answer the census accurately. Themid-decade censuses for 1925 through 1945 were con-ducted in January of the year following the reference year,to collect the data on the entire calendar year as soon aspossible after its end.

The 1920 and the mid-decade enumerations provideddata for reference periods corresponding to those used inthe U.S. Department of Agriculture crop and livestockestimates (issued as of December 1 and January 1,respectively), but winter weather made canvassing, par-ticularly in rural areas, very difficult. In addition, there wereproblems fitting a fairly long interview into the busy work-day of the average farmer. The Census Bureau tested amethodology intended to mitigate these problems as partof the 1920 census. Local mail carriers, who were assumedto have sufficient knowledge about the people and farmson their routes to permit them to check the accuracy ofresponses, as well as avoid at least part of the problem oflong interviews, were employed to deliver the question-naires, pick them up when completed, and carry out apreliminary review of the responses. The plan had mixedresults, particularly in the rural test area, and was notadopted as a general enumerative technique.2 For the firstmid-decade agriculture census in 1925, the Census Bureaumade a special effort to hire as enumerators peopleidentified with agriculture. The presidents of State agricul-tural colleges recommended persons to act as supervisors,forest rangers from the Department of Agriculture enumer-ated farms near national forests, employees of the Bureauof Reclamation canvassed irrigation projects, and Indianagents collected data for farms on Indian reservations. Inlater canvasses, farmers and farmers’ wives frequentlywere hired to carry out the canvass.

In the 1954 agriculture census, the Census Bureauinaugurated an attempt to collect the required data as nearas possible to the end of the crop year of the referenceyear; canvassing began in November of the reference year.This improved the quality of the data, and the CensusBureau continued to use this procedure for the 1959 and1964 censuses, but weather still presented a seriousproblem for canvassing.

The Census Bureau adopted mailout/mailback method-ology as the rule for the 1969 agriculture census, permitting

a return to a January data collection. The Census Bureauhad used this procedure for the 1963 Economic Censuses,and planned to use it in major portions of the 1970population and housing censuses as well. Employed in theagriculture census, mailout/mailback allowed farmers tocomplete their report forms at their own convenience, withfull access to their own records, and gave them an oppor-tunity to review and correct their own report forms. TheCensus Bureau compiled an address list of potential agri-cultural operations from the previous agriculture censusrecords and from administrative records supplied by theInternal Revenue Service, the Social Security Administra-tion, and the U.S. Department of Agriculture. The mailout,at the end of December 1969, requested inventory andproduction data for that year as of December 31. TheCensus Bureau asked farmers to complete the forms andmail them back; field enumeration was retained only forPuerto Rico, Guam, the Virgin Islands of the United States,the Trust Territory of the Pacific Islands, and AmericanSamoa, and for certain followup activities.

While far more convenient for farmers and more eco-nomic than using a field staff, the mailout/mailback censushad problems of its own, particularly the need for acomplete address list and for timely response from farm-ers. The Census Bureau found mail-list compilation par-ticularly difficult with respect to small farms, which were notlikely to be included in many of the administrative recordsthe agency used to assemble its list. Adequate responserates required multiple followup mailings—six or seven in atypical census—as well as telephone enumeration of par-ticularly large or important nonrespondent operations, overa period of several months.

Despite these problems the overall coverage obtainedby mailout/mailback was only slightly lower than from theold canvassing methodology. The cost savings realized, aswell as the convenience (and hence reduced resistance toresponse for the agricultural operators), were consideredto outweigh the drawbacks of the technique.

Sampling. The Census Bureau introduced sampling forspecific information in the 1940 agriculture census, when itsampled the collected data for a series of special tabula-tions. The 1945 agriculture census used sampling as anenumeration method. While the census collected county-level data in a conventional canvassing operation, selectedState-level estimates were obtained from an area-segmentsample3 that included approximately 7 percent of all farmsin the United States.

2It was tried again in the 1950 census, when mail carriers delivered thequestionnaires to rural route boxholders, who were asked to complete theforms and hold them until an enumerator arrived to pick them up.

3An area-segment sample typically involved identifying particulargeographic area segments, usually expected to have a specified averagenumber of agricultural operations, and sampling the total number ofsegments identified. (For the 1945 census, a 1-in-18 sample of areasegments (each expected to contain five agricultural operations) in allcounties was selected, and this, together with 50,000 large farms selectedfor certain inclusion, made up the national sample.) Once a sample wasselected, the cooperating agency (USDA’s National Agriculture StatisticsService (NASS) and its predecessor, the Statistical Reporting Service(SRS) frequently collaborated with the Census Bureau, making USDA’sJune Enumerative Survey area sample available for census evaluation,and other purposes) canvassed the famers in the segments to collect thedata needed.


Later agriculture censuses sampled to reduce overallrespondent burden by collecting selected data only fromcertain farm operations, and developed estimates for allfarms based on the sample data. The technique for thegeneral census used a short form that included the coreitems requested of all farms, and a standard form (used fora sample of farms) that asked both the core items andadditional questions on areas of special interest. Thesample generally included all farms expected to meetspecified value of sales or acreage limits, plus a randomsample of all other farms on the census lists.

The Census Bureau also made extensive use of sam-pling in its coverage evaluation program to estimate thecompleteness of the enumeration and to adjust censusstatistics to compensate for nonrespondent operations.

Agricultural services censuses. Increasing interest inbusinesses providing services to agricultural operations ledthe Census Bureau to develop a census of agriculturalservices for the 1969 agriculture census program. Thisspecialized enumeration aimed at collecting business infor-mation (i.e., volume of sales, payroll, number of employ-ees, and so on) at the establishment level on specifiedtypes of service operations, such as veterinary services,soil treatment operations, animal and livestock servicesother than veterinary, soil and crop services, landscapers,farm labor management firms, and the like. The CensusBureau prepared mail lists from administrative records(e.g., Internal Revenue Service (IRS) individual, partner-ship, and corporation tax records) and enumerated theservice establishments by mail, publishing tabulations forthe United States and for States, with limited data availablefor counties. The services’ enumeration was repeated forthe 1974 and 1978 agriculture censuses, but following the1978 operation Congress withdrew funding and ended theprogram.

Follow-on sample surveys. Follow-on surveys enabledthe Census Bureau to select the most efficient samples tocollect detailed data on subjects of special interest withoutadding greatly to the overall respondent burden. A follow-onsample survey allowed the Census Bureau to draw asample from the pool of agriculture census farms, andrequested operators to provide more detailed informationon their agricultural activities than it was possible to ask inthe general census.

The Census Bureau carried out the first agriculturefollow-on survey after the 1954 census, with a mail samplesurvey of farm expenditures. Since then, every agriculturecensus program except that for 1982 included at least onefollow-on survey (the original plans for the 1982 censuscalled for several, but these were canceled because ofbudget constraints). The 1978 program included samplesurveys of farm finances, farm and ranch irrigation prac-tices, farm energy use, and a census of horticulturalspecialties. The 1987 follow-on program consisted of ahorticultural specialties census, a farm and ranch irrigationpractices sample survey, and an agriculture economics and

land ownership survey. The 1992 Census of Agricultureprogram included only one follow-on survey, the 1994 Farmand Ranch Irrigation Survey.

The follow-on operations drew samples from the agri-culture census itself, so the census data collection opera-tion and the subsequent identification of sample operationsfor any follow-on survey had to be completed before thelatter could get under way. Thus the follow-on surveys (orcensuses) usually were carried out for the calendar yearfollowing the census reference year; e.g., the 1987 follow-onprogram data collection activities all took place in 1989,and requested data for calendar year 1988.

Report Form Content and Format

Content. In the 19th century, the agriculture censusschedules asked for simple production quantities and totalsales values for selected products, with relatively few itemson such things as machinery and equipment, or fertilizersused. Changes to these schedules generally were restrictedto changes in the kinds of crops and livestock for whichdata were requested. The 1900 census introduced ques-tions on the race and tenure of farm operators, and fromthen until today, the agriculture census collected a consid-erable amount of social and economic information, alongwith the crop and livestock data. During the period betweenthe World Wars, questions were added on such things asthe availability of electricity, telephone service, and pavedroads, as well as the degree of mechanization of farmoperations, and nonfarm employment and income. Therace, sex, and ethnic background of farm operators becameimportant objects of the census questionnaire after WorldWar II, and the 1978 and later agriculture census reportforms asked for the respondent’s sex, and whether he orshe was of Spanish/Hispanic origin.

Business organization, off-farm income, and participa-tion in a variety of Federal Government agricultural pro-grams became increasingly important to data users aswell. Some information on organization and income hadbeen requested since the 1920’s, and following World WarII, items were added on participation in various Federalloan and land conservation programs. For 1974, the cen-sus questionnaire introduced an item on farm credit anddebt, and for the 1987 census, additional inquiries onproduction expenses. The 1992 census added questionson hired farm labor by number of days worked, landlords,sales of products to individuals, injuries and deaths occur-ring on the farm, and additional detailed crop breakdowns(e.g., wheat by type).

Format. The agriculture censuses employed a field can-vass of farm operators until the 1969 census. During the19th century, enumerators used pages in large ledger typebinders for collecting the agriculture data, but the CensusBureau began using separate agricultural questionnaires inthe 1900 census, and has continued to do so since. (The1945 enumeration reverted to the binder format as awartime measure.)


The growing demand for more detailed data, and theopposing demand to reduce respondent burden led tocompromises in every census, and to experiments duringthe 1940 and later censuses in tailoring report forms toreflect the special characteristics of agriculture in variousparts of the country. The tailored forms typically had twosets of questions, one asking for basic information of allfarm operations, and a second, varying from area to area,covering the crops and livestock produced there. Thisspecialization of report forms reached its peak in the 1964census, when there was a separate questionnaire for eachState, Puerto Rico, Guam, and the Virgin Islands.

The agency eliminated the specialized forms for the1969 census, when it designed two questionnaires—ashort form asking for basic information, and a standardform that included additional items. Tailored report forms4

were used only for Hawaii, Puerto Rico, and the outlyingareas. The census used the standard form for farms in the50 States expected to have $2,500 or more in agriculturalproducts sold during the census year, while the short formwent to smaller farms. A similar format was used for the1974 census, while the Census Bureau employed a varia-tion of the system in the 1978 census, with a somewhatlonger basic questionnaire used for all farms, but with a‘‘sample’’ questionnaire that included all the basic itemsplus six additional sections of inquiries for a sample ofabout 20 percent of all farms. Larger page size and otherformat changes enabled the Census Bureau to collect thedata needed while reducing overall response burden byover 30 percent.

Standardization simplified the Census Bureau’s job interms of designing, printing, mailing, and processing thequestionnaire, but respondents still were unhappy aboutquestions irrelevant to their own operations. For the 1982census, the Census Bureau reintroduced regionalized ques-tionnaires for 12 geographic regions of the country, plusseparate questionnaires for the outlying areas, and withsample and nonsample forms5for each region. The agricul-ture census ‘‘regions’’ did not coincide either with theCensus Bureau’s census geographic regions or with theUSDA’s regions, but were simply groupings of States inwhich the Census Bureau expected to find similar types ofcrops and livestock operations. The nonsample report formcontained all the items requested of all farmers, while thesample version contained both the ‘‘core’’ items requestedon the nonsample form and additional questions, and wasused for a sample of about 20 percent of all farms. TheCensus Bureau continued to use this system of region-alized/sample and nonsample report forms for the 1987and 1992 censuses, although the number of regions was

raised to 13. (In the 1982 census, the Census Bureau therewere 10 multi-State regions, while Florida and Hawaii eachmade up a region of their own. For 1987 and 1992, Alaskawas designated a separate region as well.)

Processing and Publishing the Data

Processing. Processing the census data during mostof the 19th century was a fairly straightforward operation;the enumeration staff returned completed schedules to thecensus office and the clerical staff tabulated and compiledthe data by hand. The introduction of mechanical punch-card and electric tabulating equipment (first used in the1890 population census, and for the 1900 agriculturecensus) was a major methodological and technologicalchange, so much so that a comparable transformation inprocessing waited until the advent of the electronic com-puter and automated data processing systems half acentury later. Technical improvements to the equipmentcontinued throughout the intervening decades, (e.g., the1940 agriculture census introduced automated editing ofthe census punchcards) however, the basic systems intro-duced for processing at the turn of the century remained inplace until after World War II.

The Census Bureau played a major role in the develop-ment of modern computer technology. Its staff drew up thespecifications and cooperated in the design of the ‘‘Univer-sal Automatic Computer,’’ better known as UNIVAC, thefirst general purpose electronic computer system, whichwas installed at the Census Bureau’s Philadelphia fieldoffice in 1951 for use in processing the 1950 populationcensus. The system was moved to the Suitland headquar-ters in time for the 1954 agriculture census. Even with thenew system, a large clerical staff was required to manuallyedit the individual report forms before the data were keyedto punchcards for computer processing. The 1964 censusintroduced ‘‘string’’ punching, which saved time in keypunching and computer processing. This technology reducedthe total number of punchcards needed to transfer the datato magnetic tape, and used computerized programs toperform much of the editing and tabulating work. For the1969 census, the Census Bureau’s Data Preparation Divi-sion (DPD) in Jeffersonville, IN, began keying the agricul-ture data directly to small magnetic tape reels, ‘‘pooling’’(i.e., consolidating) the data on standard computer tapereels, and shipping the tapes to the main computer facilityat Suitland, MD, for processing. High-speed printers pro-duced copies of tables for review and correction, and evenfor photo-offset reproduction for publication. For the 1974census, computer disks replaced the small tape reels, andthe Jeffersonville office transmitted the data to Suitlandelectronically via telephone datalink. For the 1978 census,individual bar code address labels and laser ‘‘reading’’equipment facilitated automated check-in, while in the1982 census the data were keyed directly to computer diskonce again, but there was no clerical edit before keying,since the edit programs developed by the Census Bureaumade manual editing unnecessary. The 1982 census also

4The Census Bureau also produced separate report forms for theagricultural services census, and the decennial censuses of irrigation,drainage, and horticultural specialities carried out as part of the 1969program.

5The sample forms were further specialized by the use of ‘‘must’’report forms. ‘‘Must‘‘’’ forms were used for very large or special opera-tions, and were identical to the other sample forms in content. The CensusBureau used a different shading color for ‘‘must‘‘’’ forms to facilitateidentification of these cases during clerical processing.


saw the first use of the interactive data base system, whichallowed analysts access to the entire data file to resolveproblems. This system was expanded further for 1987,using minicomputer systems to edit the tabulations and toprepare the actual tables, making it possible to dispensewith the paper printouts required in the earlier systems.

Publication. The agriculture census data traditionally waspublished in printed reports, containing tabulations at thenational, State, or even county level (as appropriate foreach report series), with occasional use of illustrations andgraphics. Since the turn of the century, this conventionalsystem was modified successively to include individualreports for each State and county, special reports onselected subjects, greatly increased use of graphics andthe development of a graphics report, and the adoption ofelectronic and other publishing media.

The Census Bureau employed relatively simple geogra-phy for publishing census data. Until the 1987 censusreports were issued, the standard area reports covered thecountry as a whole, census geographic regions (andoccasionally census divisions), States, and counties. Forthe 1987 enumeration, theCensus Bureau produced selectedstatistics at the five-digit ZIP-Code level as well; this wasthe first time agriculture census data were published for alevel below the county since the first farm enumeration. Inthe 1992 census, the Bureau continued to publish ZIP-Code level data, and also produced tabulations of selecteddata for congressional districts.

The early census reports sometimes included selectedmaps and an occasional chart, but these were very limitedin scope. The Census Bureau produced the first GraphicSummary, showing farm tenure and land use, as part of the1945 agriculture census publication program, and for 1969,introduced computer generated maps as well as additionalcharts and graphs. Renamed the Agricultural Atlas of theUnited States for 1987, the graphics report became aregular and popular part of the census publication program.

The Census Bureau issued agriculture census data oncomputer tape—in two standard computer languages—forthe first time as part of the 1964 publication program,although only tapes of the preliminary data were offered.For the 1969 and following censuses, the Bureau providedthe final census data on computer tape, while preliminarydata were available only on tape for the 1978 enumeration.As computer use became more widespread, data usersindicated that they needed both the preliminary and finalagriculture census data on computer tape, and urged theagency to expand its data publication in machine-readableformat to include new media. For the 1982 census, theCensus Bureau issued preliminary and final data files oncomputer tape and the preliminary data on flexible dis-kettes as well. Conventional computer tape files werethose for which the user had to have access to a main-frame computer and the necessary programming andservice staffs. Flexible diskettes could be used on the

rapidly proliferating mini- and microcomputer systems,although they had limited data capacity (e.g., the 1982agriculture preliminary data file required over 100 dis-kettes).

For the 1987 census, the Census Bureau droppedflexible diskettes in favor of developing data files for sale oncompact disc-read only memory (CD-ROM). The CD-ROMformat employed rigid plastic disks virtually identical tothose introduced for audio recordings, and a single CDROMhad a data capacity comparable to four high-density com-puter tapes. Moreover, while special ‘‘readers’’ were requiredto use the new product, the equipment could be added to astandard mini- or even microcomputer system at minimalexpense, while giving the user access to an enormousamount of data. The entire 1987 agriculture census datafile could be contained on a single CD-ROM disk with roomto spare. The Census Bureau, after producing two testdisks to evaluate the capabilities of the new medium,adopted it for future censuses and issued the final 1987agriculture census data file on a single CD-ROM. The basicState and county data for the 1992 agriculture census werereleased on three CD-ROM’s, the first containing selecteddata for the first 27 States processed, and the second andthird, issued as a set, containing data for all States, plus thenational summary data, and detailed cross-tabulations.

THE FARM DEFINITION

The first official definition of what constituted a farm forcensus purposes was used for the 1850 census, when anyplace that had $100 or more in total value of sales ofagricultural products qualified. Since 1850, acreage anddollar value limits were added, altered, or removed, while arequirement evolved that the land on the place be (1)involved in, or connected with, agricultural operations, and(2) under the day-to-day control of a single management(either by an individual, partnership, corporation, or otherorganization).

The important point was, of course, the involvement withagricultural operations, which—again for census pur-poses—were the production of livestock, poultry, and ani-mal specialties, and their products, and/or crops, includingfruit, and greenhouse and nursery products. The land didnot need to be a single contiguous tract to comprise asingle farm, but had to be operated as a single economicunit (although exceptions were allowed; see the section onthe 1950-1954 definition below).

The changes in the various criteria used in the farmdefinition, by census, were—

1850-1860 No acreage requirement, but a minimum of$100 in sales of agriculture products.

1870-1890 Any place of 3 or more acres, involved withagricultural production, qualified as a farm.Places with less than 3 acres were consid-ered farms, if they had a minimum annualvalue of agricultural product sales of $500.


1900 No acreage or minimum sales requirement,and cranberry marshes, greenhouses, andcity dairies were included, if they requiredthe full-time services of at least one per-son.

1910-1920 A minimum of 3 acres, with $250 or morein total value of sales, unless the individualoperation required the full-time services ofat least one person.

1925-1945 The requirement of the full-time services ofat least one person was deleted; otherwisethe definition was unchanged.

1950-1954 Places of less than 3 acres qualified asfarms if they had, or normally would havehad, sales of $150 or more in agriculturalproducts during the census year. Placesthat began operating for the first time asfarms in 1954 also were included. Parcelsoperated by sharecroppers, and tenantfarms, counted as separate farms, eventhough the landlord handled the entireholding as a single unit. (Land retained andoperated by the landlord also was countedas a separate unit.)

1959-1974 The acreage requirement was raised to 10acres or more, with at least $50 or more inagricultural product sales. A place of lessthan 10 acres qualified as a farm if it hadsales of $250 or more during the censusyear.

1978- The acreage requirement was dropped andany place that had, or normally would havehad, $1,000 or more in total agriculturalproduct sales during the census year quali-fied as a farm.

The farm definitions used in Puerto Rico and the outlyingareas differed from that employed in the 50 States. InPuerto Rico, the definition generally required 3 or morecuerdas (a cuerda equals approximately .97 acres) and/orspecified numbers of livestock, poultry, or fruit or nut trees.The outlying areas’ definitions were similar, although inAmerican Samoa a variety of different landholding arrange-ments had to be taken into consideration in definingindividual agricultural operations. The Census Bureau andthe Government of Puerto Rico agreed to change the farmdefinition to give greater emphasis to product sales in the1982 census, when a farm was any place with $500 ormore in annual sales of agricultural products, or any placeof 10 cuerdas or more with $100 in sales of agriculturalproducts. The 1987 agriculture census (actually conductedin 1990 for the Commonwealth of the Northern Marianasand American Samoa) in the other outlying areas intro-duced a similar change, dropping the crop, livestock, andacreage requirements in favor of a minimum of $100 inannual sales of agricultural products.


Appendix C.Report Forms, Letters, and Envelopes Printed forthe 1992 Census of Agriculture

Table 1. 1992 Census of Agriculture Report Forms Printed (by Region and Type)

Region1 A04 Screener Form A01 (Nonsample) A02 (Sample) A03 (Must)

Total . . . . . . . . . . . . . . . 1,816,000 5,288,000 2,795,9002 466,000

01 . . . . . . . . . . . . . . . . . 160,300 462,000 213,000 39,00002 . . . . . . . . . . . . . . . . . 279,900 1,098,000 426,000 70,50003 . . . . . . . . . . . . . . . . . 93,900 258,000 168,000 28,00004 . . . . . . . . . . . . . . . . . 429,800 803,000 459,000 75,00005 . . . . . . . . . . . . . . . . . 59,800 118,000 56,500 17,50006 . . . . . . . . . . . . . . . . . 208,100 741,000 307,000 51,50007 . . . . . . . . . . . . . . . . . 79,100 617,000 214,000 38,000

08 . . . . . . . . . . . . . . . . . 73,500 463,000 244,500 52,00009 . . . . . . . . . . . . . . . . . 82,900 147,500 107,000 27,00010 . . . . . . . . . . . . . . . . . 97,200 200,500 89,000 26,00011 . . . . . . . . . . . . . . . . . 151,500 380,000 122,500 41,50012 . . . . . . . . . . . . . . . . . 33,40013 . . . . . . . . . . . . . . . . . 11,00014 (General) . . . . . . . . 100,000 125,000

1The States in each region were as follows: 01. Connecticut, Delaware, Maine, Maryland, Massachusetts, New Hampshire, New Jersey, New York,Pennsylvania, Rhode Island, Vermont, West Virginia; 02. Illinois, Indiana, Iowa, Kansas, Nebraska, Ohio; 03. Michigan, Wisconsin; 04. Alabama, Georgia,Kentucky, North Carolina, South Carolina, Tennessee, Virginia; 05. Florida; 06. Arkansas, Louisiana, Mississippi, Missouri, Oklahoma; 07. Texas; 08.Minnesota, Montana, North Dakota, South Dakota; 09. Colorado, Nevada, New Mexico, Utah, Wyoming; 10. Idaho, Oregon, Washington; 11. Arizona,California; 12. Hawaii; 13. Alaska.

2Sample total includes 220,000 report forms (91,000 -A202 and 129,000 -A202(F)) used in the film optical sensing device for output to computer(FOSDIC) test in Illinois, Iowa, Nebraska, and Ohio, not included in the count for region 02.

Facsimilies of representative screener and report formsare included in appendix F. Other printed materials orderedfor the data collection operation mailings are shown intables 2 and 3.

Facsimiles of the standard instructions sheet, the remindercard, and principal followup letters are included in appendixG.

Table 2. Information Sheets and Followup Letters

Form number Description Quantity

92-A01(I) . . . . . Instruction sheet (screener and non-sample questionnaires) 6,584,000

92-A02(I) . . . . . Instruction sheet (sample and mustquestionnaires) 2,705,000

92-A01(L1) . . . . Transmittal letter (initial mail out) 4,401,00092-A01(L2) . . . . Reminder card 4,200,00092-A01(L3) . . . . Followup letter 2,231,00092-A01(L4) . . . . Followup letter 1,405,00092-A01(L5) . . . . Followup letter 55,00092-A01(L6) . . . . Followup letter 965,00092-A01(L1A) . . Transmittal letter (UAA’s) 162,000

Table 3. Outgoing and Return Envelopes

Formnumber Description Quantity

92-A7.1 . . . . Outgoing envelope (initial mailout) 4,535,50092-A7.2 . . . . Outgoing envelope (followup) 2,322,00092-A7.3 . . . . Outgoing envelope (followup) 1,461,00092-A7.4 . . . . Outgoing envelope (followup) 44,00092-A7.5 . . . . Outgoing envelope (followup) 1,005,00092-A7B . . . . Outgoing envelope (UAA) 212,00092-A7C . . . . Outgoing envelope (correspondence) 71,00092-A7 . . . . . Outgoing envelope (reserve) 200,00092-A8A(SC) Return envelope (screener

questionnaires) 1,702,00092-A8A(N) . Return envelope (nonsample

questionnaires) 5,100,00092-A8A(S) . Return envelope (sample questionnaires) 2,375,00092-A8A(M) . Return envelope (must questionnaires) 423,00092-A8A . . . . Return envelope (reserve) 200,000

APPENDIX C C–1HISTORY—1992 CENSUS OF AGRICULTURE

Appendix D.1992 Census of AgricultureMailout and Followup Mailings

Table 1. 1992 Census of Agriculture Mailout: December 8-18, 1992

Form Type Quantity

Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3,551,407

Nonsample (forms 92-A0101 to -A0111) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2,131,699Sample (total (excluding multiunits and abnormals)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,000,442General Sample (forms 92-A0201 to -A0213) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 865,226Must Cases (forms 92-A0301 to -A0311) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135,216

Multiunit (forms 92-A0301 to -A0311, forms 92-A0201 and -A0212) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5,727Abnormal (forms 92-A0301 to -A0311, 92-A0201, and 92-A0213) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,899Screener (forms 92-A0401 to -A0411) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 411,640

Table 2. First Followup: Reminder/Thank You Card(Form 92-A01(L2)): January 6-8, 1993

Date Quantity

January 6-8, 1993 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3,543,7811

1The Form 92-A01(L2), Reminder Card was mailed to all addresses inthe mail file except multiunits and abnormals.

Table 3. Geographic Segments Used in the 1992 Census of AgricultureMail-Followup Operations

Segment States

1 . . . . . . . . . Must cases for all States2 . . . . . . . . . Arizona, California, Colorado, Idaho, Nevada, New Mexico, Oregon, Utah, Washington,

Wyoming3 . . . . . . . . . Texas4 . . . . . . . . . Michigan, Minnesota, Montana, North Dakota, South Dakota, Wisconsin5 . . . . . . . . . Connecticut, Delaware, Maine, Maryland, Massachusetts, New Hampshire, New Jersey,

New York, North Carolina, Pennsylvania, Rhode Island, Virginia, West Virginia6 . . . . . . . . . Alabama, Georgia, Kentucky, South Carolina, Tennessee7 . . . . . . . . . Arkansas, Louisiana, Mississippi, Missouri, Oklahoma81 . . . . . . . . . Florida, Kansas, Indiana, Iowa9 . . . . . . . . . Alaska, Hawaii, Illinois, Nebraska, Ohio

1FOSDIC test cases for Iowa were mailed as part of the segment 9 mailout.

APPENDIX D D–1HISTORY—1992 CENSUS OF AGRICULTURE

Table 4. Second Followup (Report Form): February 11-24, 1993

SegmentCheck-incloseout Total Nonsample

Generalsample Must Screener

Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,521,702 858,117 398,505 66,957 198,123

1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 8 66,957 66,9572 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 9 198,493 98,596 53,935 45,9623 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 10 136,677 98,668 31,917 6,0924 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 11 203,721 123,297 61,000 19,4245 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 15 170,131 98,326 55,302 16,5036 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 16 224,585 104,440 56,040 64,1057 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 17 213,524 130,953 56,996 25,5758 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 18 163,983 115,502 38,397 10,0849 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 21 143,631 88,335 44,918 10,378

Table 5. Third Followup (Report Form): March 18-29, 1993

Segment Check-in/closeout Total Nonsample


Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1,102,924 610,123 295,772 48,782 148,247

1,3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 15 148,550 71,398 23,526 48,782 4,8442 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 16 136,771 66,535 38,075 32,1614 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 17 146,540 87,479 45,319 13,7425 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 18 123,443 69,886 41,119 12,4586 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 19 166,585 74,515 42,717 49,3537 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 22 157,798 94,691 43,286 19,8218 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 23 117,510 81,736 27,890 7,8849 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 24 105,727 63,903 33,840 7,984

Table 6. Fourth Followup (Letter): April 23- May 7,1993

Geographic SegmentCheck-incloseout Quantity

Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 856,191

1,3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apr. 21 113,4792 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apr. 20 105,5734 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apr. 22 116,0365 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apr. 26 94,3886 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apr. 27 132,4387 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apr. 28 123,7488 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apr. 29 88,8399 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . May. 4 81,690

D–2 APPENDIX D HISTORY—1992 CENSUS OF AGRICULTURE

Table 7. Fifth Followup (Report Form): May 27-June 3, 1993

SegmentCheck-in/closeout Total Nonsample


Total . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 722,874 401,852 191,766 25,841 103,415

1,2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . May 24 113,017 41,553 24,234 25,841 21,3893,4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . May 25 168,177 108,037 47,146 12,9945,6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . May 26 188,469 92,799 52,148 43,5227,8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . May 27 182,757 116,217 46,712 19,8289 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . May 28 70,454 43,246 21,526 5,682

Table 8. Mailing Undeliverable As Addressed (UAA)

MailingCheck-

in/closeoutMailoutDate Quantity

Total . . . . . . . . . . . . . . . . . . . . . . . . . . . 33,393First . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 4 Feb. 9-16 32,735Second . . . . . . . . . . . . . . . . . . . . . . . . Mar. 1 Mar. 5 658

APPENDIX D D–3HISTORY—1992 CENSUS OF AGRICULTURE

Appendix E.Publications in Selected Series

The major publications series for the 1992 Census ofAgriculture were the Volume 1, Geographic Area Series,AC92-A, the Volume 2, Subject Series (in 5 parts), AC92-S,and the Volume 3, 1994 Farm and Ranch Irrigation Survey,AC92-FRIS.

The Volume 1, Geographic Area Series, consisted of areport for the United States, and individual reports for eachState, Puerto Rico, Guam, and the U.S. Virgin Islands.Each report included an introductory text and two chaptersof detailed statistical tables, the first for the State/area, andthe second covering the county or country equivalents(e.g., parishes, municipios, election districts). The CensusBureau issued the Volume 1 reports in printed volumes,online (selected highlights only), computer tape, and onthree compact disc-read only memory (CD-ROM) discs.(Data for Puerto Rico, Guam, and the U.S. Virgin Islandswere not released on CD-ROM. Reports for these areaswere available in printed reports and online highlights, anda computer tape file for Puerto Rico also was available.)

The Volume 2, Subject Series, consisted of five reports—Part 1, Agricultural Atlas of the United States; Part 2,Coverage Evaluation; Part 3, Ranking of States andCounties; Part 4, History; and Part 5, ZIP Code Tabulationsof Selected Items. Parts 1-4 all were available as printedreports, while Part 5 was released on CD-ROM and theCensus Bureau plans to issue Parts 1 and 2 together withcongressional district data and a 1992 Public Use Micro-data File on CD-ROM as well. Highlights of Parts 3 and 5also were available online.

Volume 3, 1994 Farm and Ranch Irrigation Survey,showed statistics about on-farm irrigation practices col-lected from a sample of 1992 agricultural operations (exclud-ing Alaska, Hawaii, abnormal farms, and horticultural spe-cialty operations) who reported using irrigation during thecensus year. The publication reported data for 18 waterresources regions of the 48 conterminous States, and forthe 27 leading irrigation States. The report was released asa printed volume, on CD-ROM, and highlights were madeavailable online.

PRINTED REPORTS

Volume 1, Geographic Area Series, Series AC92-A

Geographic Area Report No.(AC92-A-) Pages Price Published

Alabama . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 520 28.00 Nov. 16, 1994Alaska . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 242 8.50 Dec. 1, 1994Arizona . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 300 16.00 Oct. 27, 1994Arkansas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 536 28.00 Nov. 9, 1994California . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 526 28.00 Oct. 25, 1994

Colorado . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 484 26.00 July 19, 1994Connecticut . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 284 16.00 June 16, 1994Delaware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 272 15.00 Jan. 13, 1994Florida . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 500 28.00 June 6, 1994Georgia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 830 37.00 Dec. 12, 1994

Hawaii . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 260 15.00 Dec. 12, 1994Idaho . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 424 23.00 July 7, 1994Illinois . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 628 24.00 June 2, 1994Indiana . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 620 33.00 Mar. 4, 1994Iowa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 620 32.00 Apr. 5, 1994

Kansas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 644 34.00 June 1, 1994Kentucky . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 712 35.00 Aug. 23, 1994Louisiana . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 484 27.00 Sept. 6, 1994Maine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 316 18.00 June 16, 1994Maryland . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 344 19.00 May 11, 1994

Massachusetts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 296 17.00 June 16, 1994Michigan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 596 31.00 June 6, 1994

APPENDIX E E–1HISTORY—1992 CENSUS OF AGRICULTURE

PRINTED REPORTS

Volume 1, Geographic Area Series, Series AC92-A—Con.

Geographic Area Report No.(AC92-A-) Pages Price Published

Minnesota . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 588 30.00 Nov. 16, 1994Mississippi . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 540 30.00 Nov. 21, 1994Missouri . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 652 35.00 Apr. 14, 1994

Montana . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 472 25.00 Sept. 26, 1994Nebraska . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 618 32.00 Sept. 28, 1994Nevada . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 310 18.00 Sept. 19, 1994New Hampshire . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 288 16.00 June 16, 1994New Jersey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 332 19.00 July 26, 1994

New Mexico . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 392 21.00 Nov. 18, 1994New York . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 498 27.00 July 27, 1994North Carolina . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 604 34.00 Nov. 29, 1994North Dakota . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 480 26.00 Oct. 26, 1994Ohio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 606 32.00 Apr. 29, 1994

Oklahoma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 578 30.00 Nov. 28, 1994Oregon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 412 22.00 Feb. 9, 1994Pennsylvania . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 532 28.00 Aug. 24, 1994Rhode Island . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 244 14.00 June 16, 1994South Carolina . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 430 23.00 Dec. 12, 1994

South Dakota . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 506 27.00 Oct. 25, 1994Tennessee . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 608 31.00 Aug. 22, 1994Texas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 1,088 40.00 Dec. 2, 1994Utah . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 362 19.00 Sept. 1, 1994Vermont . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 292 16.00 June 16, 1994

Virginia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 630 33.00 May 11, 1994Washington . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 396 22.00 Mar. 25, 1994West Virginia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 448 24.00 May 13, 1994Wisconsin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 546 29.00 Mar. 16, 1994Wyoming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 316 18.00 July 19, 1994

United States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 516 29.00 Dec. 14, 1994Puerto Rico . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 273 18.00 Jan. 31, 1995Guam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 68 5.00 June 5, 1995Virgin Islands of the United States . . . . . . . . . . . . . . . . . . . . . . . . . . 54 44 3.75 Mar. 28, 1995

Volume 2, Subject Series , Series AC92-S

Part TitleReport No.(AC92-S-) Price Published

1 Agricultural Atlas of the United States AC92-S-1 20.00 Jan. 19962 Coverage Evaluation AC92-S-2 (X) May 19963 Ranking of States and Counties AC92-S-3 11.00 Feb. 19954 History AC92-S-4 . /

Volume 3, 1994 Farm and Ranch Irrigation Survey , Series AC92-FRIS

Title Report Series No. Price Published

1994 Farm and Ranch Irrigation Survey AC92-FRIS 12.00 Feb. 1996

E–2 APPENDIX E HISTORY—1992 CENSUS OF AGRICULTURE

Appendix F.Chronology of Major Activities

Census Began Completed

1992 CENSUS OF AGRICULTURE OF THE UNITED STATES

1990 Census PretestInitial mailout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nov. 1990 Nov. 1990Field interviews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dec. 1990 Feb. 1991Telephone interviews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dec. 1990 Feb. 1991

Mail List DevelopmentProcure source lists . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . June 1990 July 1992Mail list model development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jan. 1991 Oct. 1992List production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apr. 1991 Sept. 1992Sample selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oct. 1992 Oct. 1992Final list processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oct. 1992 Oct. 1992

Census Mail Preparation and MailoutPrint report forms and envelopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 1992 Aug. 1992Preparation for postal delivery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 1992 Dec. 1992Initial mailout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dec. 1992 Dec. 1992First (reminder card) followup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jan. 1993 Jan. 1993Second followup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 1993 Feb. 1993Third followup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 1993 Mar. 1993Fourth followup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apr. 1993 May 1993Fifth followup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . May 1993 June 1993Citrus caretaker enumeration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oct. 1991 Dec. 1992

Precomputer ProcessingReceipt and check-in . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jan. 1993 Sept. 1993Correspondence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jan. 1993 June 1993Data keying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 1993 Dec. 1993Edit review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 1993 Jan. 1994Clerical review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . June 1993 Jan. 1994

Computer ProcessingCensus edit/item imputation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 1992 Aug. 1993Disclosure analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Aug. 1993 Mar. 1994Analytical review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . May 1993 Mar. 1994Data tabulation production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . July 1993 May 1994

PublicationVolume 1, Geographic Area Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Nov. 1993 Oct. 1994Volume 2, Subject Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Part 1, Agricultural Atlas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X) Jan. 1995Part 2, Coverage Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X) May 1996Part 3, Ranking States and Counties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X) Mar. 1995Part 4, History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X)Part 5, ZIP Code Tabulation of Selected Items From the 1992 Census ofAgriculture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X) 1Dec. 1995

1992 Coverage EvaluationDevelop estimation methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apr. 1990 Mar. 1992Receive and process USDA June Agriculture Survey (JAS) data . . . . . . . . . . . . . . . July 1992 Jan. 1995See footnotes at end of table.

APPENDIX F F–1HISTORY—1992 CENSUS OF AGRICULTURE

Census Began Completed

Classification Error Survey (CES)Design report form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Aug. 1991 June 1992Select CES sample . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oct. 1992 Oct. 1992Mailout and mail followup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Feb. 1993 July 1993Telephone followup to nonrespondents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . June 1993 Sept. 1993

Produce combined JAS/CES estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oct. 1994 Sept. 1995Publish 1992 Coverage Evaluation report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X) May 1996

1992 CENSUS OF AGRICULTURE IN PUERTO RICO

Negotiate special agreement with the Commonwealth of Puerto Rico . . . . . . . . . . May 1990 June 1991Design and print report form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . May 1991 Nov. 1992Sample design and selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jan. 1991 Sept. 1992Prepare mail list . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mar. 1991 Mar. 1993Data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . July 1993 Oct. 1993Data keying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Aug. 1993 Oct. 1993Data processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sept. 1993 Mar. 1994Publication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X) May 1994

1992 CENSUS OF AGRICULTURE IN THE U.S. VIRGIN ISLANDS

Negotiate special agreement with the Virgin Islands Government . . . . . . . . . . . . . . Apr. 1990 Apr. 1990Design and print report forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . June 1991 Nov. 1992Data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . June 1993 Aug. 1993Data keying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oct. 1993 Dec. 1993Tabulation and editing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dec. 1993 Feb. 1994Publication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X) Apr. 1994

1992 CENSUS OF AGRICULTURE IN GUAM

Negotiate special agreement with the Government of Guam . . . . . . . . . . . . . . . . . . . Apr. 1990 June 1991Design and print report forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oct. 1991 Nov. 1992Data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . July 1993 Sept. 1993Data keying . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Oct. 1993 Dec. 1993Tabulation and editing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dec. 1993 June 1994Publication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X) July 1994

1994 FARM AND RANCH IRRIGATION SURVEY

Design report forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sept. 1993 Dec. 1993Sample design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . June 1993 Aug. 1994Sample selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sept. 1994 Nov. 1994Data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Jan. 1995 June 1995Data tabulation and editing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Apr. 1994 Dec. 1995Publication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X) June 1996Electronic data release . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X) Nov. 1995Printed publication release . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (X) Feb. 1996

(X) Not applicable.1The ZIP Code Tabulation of Selected Items From the 1992 Census of Agriculture was issued on CD-ROM only.

F–2 APPENDIX F HISTORY—1992 CENSUS OF AGRICULTURE

Appendix G.Report Forms

CHANGES IN THE STANDARD REPORT FORMS

Background Information

The Secretary of Commerce is responsible for thedesign and content of all census data collection forms andusually delegates this task to the Director of the Bureau ofthe Census. The Census Bureau regularly consults inter-ested organizations and agencies, including the U.S. Depart-ment of Agriculture, the Bureau of Economic Analysis, andothers, for comments and suggestions regarding the reportforms. The Census Bureau’s own advisory committee onagricultural statistics reviews the suggestions, comments,and data requests and makes its own recommendations onthe priorities to be assigned to the various items forinclusion in the final report forms. (See chapter 2 for moreinformation on consultation on the census.)

1992 Census of Agriculture Report Forms

General design. The 1992 general report forms reintro-duced the booklet format last used in the 1974 census. Thecensus retained the regionalized census report forms,identifying 9multi-State regions, plusAlaska, Florida, Hawaii,and Texas as separate regions for a total of 13. Four reportforms were used for each region—two nonsample versions(one with a screener question and one without) and twosample versions (one for ‘‘must’’ cases and one for ‘‘cer-tainty’’ and general sample cases—see chapter 2 fordetails and definitions of ‘‘must’’ and ‘‘certainty’’ cases). Allof the report forms used identical formats, but employedtailored items to list crops and livestock common withineach region.

The census used sample and nonsample versions ofeach regionalized form except for Alaska and Hawaii,where only sample forms were employed. The nonsamplequestionnaires contained the items asked of all respon-dents, while the sample versions included additional ques-tions asked only of a sample of farmers. Both sample andnonsample versions used the booklet format; the bookletmeasured 8-1/2″ x 11″, consisting of white stock with blackprinting and shading in a different color for each region.The sample and nonsample forms were 12 and 10 pageslong respectively, including the information sheet. Thesample questionnaire contained all the sections on thenonsample form, plus additional sections on—

x Use of fertilizers and chemicals.

x Total production expenses, including interest expensefor secured and unsecured loans.

x Machinery and equipment (inventory and value).

x Market value of land and buildings.

x Income from farm-related sources.

The Census Bureau also employed a ‘‘screener’’ form,which was, essentially, the nonsample form with the firstsection split into two subsections—the first asked recipi-ents whether they had any agricultural activities during1992, while the second covered the acreage items thatcomposed the full section in the other report forms.

The Census Bureau assigned form numbers to thequestionnaires by type of form (i.e., nonsample-screener,nonsample, sample-nonmust, sample-must) and by geo-graphic region as follows:

Nonsample Forms 92-A0101 through92-A01111

Nonsample (screener)(the -A0414 was ageneralized form)

Forms 92-A0401 through92-A0411 and 92-A0414

Sample (nonmust) Forms 92-A0201 through92-A0213

Sample (must) Forms 92-A0301 through92-A0311

1No nonsample questionnaires were used in Alaska or Hawaii, each ofwhich constituted a separate ‘‘region.’’

In addition, in Illinois, Iowa, Indiana, and Ohio, theCensus Bureau tested form 92-A0202(F) for use with filmoptical sensing device or input to computer (FOSDIC)equipment. The FOSDIC form was identical in content tothe standard sample form except for the addition of opticalrecognition and alignment marks printed on the pages.

Major changes in data requested. The overall datacontent of the 1992 census report forms was similar to thatof 1987—the bulk of the data collected on agriculturaloperations obviously concerned the basic information onnumber of farms, acreage in farmland and various crops,

APPENDIX G G-1HISTORY—1992 CENSUS OF AGRICULTURE

inventories of livestock, value of sales of agricultural prod-ucts, and so on. Specific data requirements and prioritieschange over time, and the content of the report forms hasto be adjusted to reflect these changes. The major changesto the ‘‘all farms’’ sections of the report forms for the 1992agriculture census were as follows:

Section Change

Section 1. ACREAGE IN 1992 . . . . . . . . Item asking for landlords’ andrenters names, addresses, andacres rented from or to eachwas deleted; item added askingfor ‘‘Number of landlords’’ wasadded to the section.

Section 2. Were any of the followingCROPS harvested from‘‘THIS PLACE’’ in 1992

‘‘Winter wheat for grain’’ and‘‘Spring wheat for grain’’ wereadded to the crop list in thesection.

Section 7. Were any CROPS har-vested from ‘‘THIS PLACE’’in 1992

‘‘Ginseng’’ was separately listedinthissection,and‘‘Canola(pounds)’’includedon thewrite-in list included‘‘other (industrial) rapeseed.’’

Section 17. Did you or anyone elsehave any HORSES, BEES,FISH, GOATS, OTHERLIVESTOCK, or ANIMALSPECIALTIES on thisplace in 1992?

‘‘Hybrid striped bass’’ and ‘‘craw-fish’’ were added to the write-inlist.

Section 20. (NEW) During 1992 didyou grow or raise anycrops, livestock, or live-stock products that weresold DIRECTLY to indi-vidual consumers forHUMAN consumption—roadside stands, farmersmarkets, pick your own,door to door, etc.?

New section requesting prod-ucts sold and gross value ofsales.

Section 27. (NEW) HIRED FARM orRANCH LABOR—Did youhave any paid workersdoing agricultural labor onthis place in 1992?

New section (data on labor wascollected in the section on pro-ductionexpensesfor1987)request-ing number of hired laborersworking 150 days or more, andless than 150 days.

Section 28. (NEW) Were there anyINJURIES or DEATHSconnected with farm orranch work on or for thisplace in 1992?

New section asked for numberor injuries or work-related deathsfor family members and hiredworkers

Puerto Rico and the Outlying Areas

Form 92-A1(PR), Puerto Rico. The Census Bureau con-sulted with data users in Puerto Rico, including the Depart-ment of Agriculture, the College of Agriculture of theUniversity of Puerto Rico, the Agriculture Extension Ser-vice, and the Planning Board, to consider changes to theagriculture census report form for the 1992 enumeration.Based on the advice and recommendations received, theCensus Bureau made the following changes to the reportform content:

x Dropped the 1987 Section 21 (PRODUCTS FOR HOMECONSUMPTION).

x Dropped the item on tobacco from, and added plantainsand bananas (and requested individual sales data for allmajor crops) to Section 3 (Was any COFFEE, PINE-APPLES, PLANTAINS, or BANANAS harvested FORSALE from this place during the last 12 months?).

x Added item on number of cuerdas planted for individualfruit trees to Section 5 (Were there a combined total of10 or more FRUIT TREES or COCONUT PALMS on thisplace for commercial production in the last 12 months?).

x Dropped value of sales of individual vegetable cropsfrom Section 6 (Were any VEGETABLES or MELONSharvested FOR SALE from this place in the last 12months?).

x Dropped items on molasses grass, milo, and elephantgrass from Section 8 (Were any GRASSES grazed orcut, or were any SEEDLINGS or OTHER CROPS har-vested FOR SALE on this place in the last 12 months?).

x Added Section 15, Did you or anyone else raise anyFISH or AQUACULTURE PRODUCTS FOR SALE onthis place during the last 12 months?

Form 92-A1(G), Guam. The report form used in the 1992agriculture census for Guam was similar to that used in the1987 enumeration, with modifications based on recommen-dations and suggestions from data users on Guam. The1992 report form incorporated the following major changes:

x Deleted items on (1) land held under land permits andgovernment-furnished land, (2) value of sales for indi-vidual crops (sales were collected by type of crop only,e.g., total value of sales in 1992 of fruits, nuts, or nurserycrops), and (3) value of sales of each type of animal(sales were requested for cattle and calves (includingcarabaos), hogs and pigs, poultry, and other livestock).

x Dropped the 1987 distinction between ‘‘crops harvested’’and ‘‘crops sold,’’ and asked for crops harvested for sale(usually acres and pounds) in all crops sections.

x Added Section 6, Was any LAND in this place IRRI-GATED or was any water used for other agriculturalpurposes at any time during 1992?

x Added item on race of operator (senior partner or personin charge) to Section 13 (CHARACTERISTICS of theFARM OPERATOR).

Form 92-A1(VI), Virgin Islands of the United States. Thechanges made to the report form for the 1992 agriculturecensus in the Virgin Islands were similar in some instancesto those made for Guam. The items on value of sales forindividual crops and types of animals were dropped fromthe respective sections, and sales value was requestedonly for types of crops (e.g., field and forage crops,vegetables, etc.) and for specified kinds of animals (e.g.,

G-2 APPENDIX G HISTORY—1992 CENSUS OF AGRICULTURE

cattle and calves, hogs and pigs, and other livestock). Theitems on ‘‘crops harvested’’ also were replaced by ‘‘cropsharvested for sale.’’ In addition, the following significantchanges were made—

x Dropped 1987 Section 1 (OPERATOR).

x Dropped the item on production for home consumptionfrom section on land in agriculture (Section 1 (LAND INAGRICULTURE) for 1992).

x Split the 1987 section on crops harvested into threesections for 1992—Section 2 (Did you harvest anyFIELD OR FORAGE CROPS FOR SALE in the last 12months?), Section 3 (Did you harvest any VEGETABLES

FOR SALE in the last 12 months?), and Section 4 (Wereany FRUITS, NUTS, or NURSERY CROPS grown orharvested FOR SALE in the last 12 months?).

x Dropped the questions on whether the place had elec-tricity and/or running water from the section on equip-ment and facilities (Section 7 in 1987, Section 11 in1992).

x Dropped question on place of birth of the operator fromsection on operator characteristics (Section 8 in 1987,Section 12 in 1992).

x Added Section 8, Did you or anyone else raise any FISHor other AQUACULTURE PRODUCTS (shrimp, etc.) inthe last 12 months?

APPENDIX G G-3HISTORY—1992 CENSUS OF AGRICULTURE

Selected Facsimiles


1992 Census of Agriculture, Report Form 92-A0203 ------------------------------------------ 6Cover Letter, Form 92-A01(L1)------------------------------------------------------------------ 18Information Sheet Form 92-A02(I) -------------------------------------------------------------- 201992 Census of Agriculture, Report Form 92-A0403 ------------------------------------------ 241992 Census of Agriculture, Report Form 92-A0202(F), FOSDIC Test Form ---------------- 34Information Sheet Form 92-A01(I) -------------------------------------------------------------- 46

Reminder/Thank You Card Form 92-A01(L2) -------------------------------------------------- 50Followup Letter Form 92-A01(L3) --------------------------------------------------------------- 51Followup Letter Form 92-A01(L4) --------------------------------------------------------------- 52Followup Letter Form 92-A01(L5) --------------------------------------------------------------- 53Followup Letter Form 92-A01(L6) --------------------------------------------------------------- 551992 Census of Agriculture, Report Form 92-A0215, Citrus Caretakers --------------------- 571992 Classification Error Study, Report Form 92-A90 ----------------------------------------- 65

Outlying Areas

1992 Census of Agriculture for Puerto Rico, Report Form 92-A1(PR)------------------------ 731992 Census of Agriculture for Guam, Report Form 92-A1(G)-------------------------------- 811992 Census of Agriculture for the Virgin Islands of the United States,Report Form 92-A1(VI) ------------------------------------------------------------------------- 85

Followon Operations

1994 Farm and Ranch Irrigation Survey, Report Form 94-A62 ------------------------------- 89

Page

APPENDIX G G–5HISTORY—1992 CENSUS OF AGRICULTURE


Facimilies not available.

1992 CENSUS OF AGRICULTURE

Results of the 1992 Census of Agriculture are publishedin a series of reports that provide data at the national,State, and county (or equivalent) levels for the UnitedStates. Data also are available for Puerto Rico, Guam, andthe Virgin Islands of the United States.

VOLUME 1. GEOGRAPHIC AREA SERIES(AC92-A-1 TO -54)

National, State, and County Data are published indetailed national and State tables for the United States,and in State and county tables for each State. Thesereports include data on number and size of farm; cropproduction; livestock, poultry, and their products; tenure,age, and principal occupation of operator; type of organi-zation; value of products sold; government payments plusmarket valueof agricultural products sold; productionexpenses;direct marketing; number of hired employees; land use;irrigation; and the standard industrial classification of farms.

U.S. Summary and State Data Report (AC92-A-51)Chapter 1. National-level dataChapter 2. State-level data

State and County Data Reports (AC92-A-1 to -50)Chapter 1. State-level dataChapter 2. County-level data

Outlying Areas Data (AC92-A-52 to -54) are publishedfor the regions and municipios of Puerto Rico, the electiondistricts on Guam, and the islands of the Virgin Islands ofthe United States.

VOLUME 2. SUBJECT SERIES (AC92-S-1 TO -5)

Agricultural Atlas of the United States (AC92-S-1)graphically illustrates a profile of the Nation’s agriculture atthe county level in a series of dot and multicolor patternmaps. The maps provide displays on sizes and types offarms, land use, farm tenure, market value of agriculturalproducts sold, crops harvested, livestock inventories, andother characteristics of farms.

Coverage Evaluation (AC92-S-2) provides national-and regional-level estimates on the completeness of thecensus, in terms of both the number of farms missed andselected characteristics of those farms.

Ranking of States and Counties (AC92-S-3) ranks theleading States and counties for selected items from the1992 census. Tables show cumulative percent of U.S.totals and most include comparative data from the 1987census.

History (AC92-S-4) provides a detailed description ofthe planning and conduct of the 1992 Census of Agricul-ture. It explains the history of the agriculture census, farm

definition, data collection and processing, data dissemina-tion, coverage evaluation and research, and the census ofagriculture in outlying areas. Also included are facsimiles ofthe report forms, information sheets, and letters sent to thepublic.

ZIP Code Tabulations of Selected Items from the1992 and 1987 Censuses of Agriculture (AC92-S-5)provides tabulations by five-digit ZIP Code for selecteditems from the 1992 and 1987 censuses. Data itemsinclude number of farms, land in farms, farms by size,market value of agricultural products sold by size of sales,livestock inventory, cropland harvested, selected crops,and other data.

VOLUME 3. 1994 FARM AND RANCHIRRIGATION SURVEY (AC92-RS-1)

This report provides data collected from a sample ofirrigated farm operations identified in the 1992 Census ofAgriculture. Data items include acres irrigated, land use,yields of specified crops, methods of water distribution,quantity of water used by source, expenses, and otherirrigation practices. Data are available for each of the 17Western States, 10 other leading irrigation States, and the18 water resources regions of the conterminous UnitedStates.

PRINTED REPORTS

All of the reports listed above are available as printedreports, except Volume 2, Part 5, ZIP Code Tabulations. Toobtain order forms or further information concerning anyagriculture census printed reports, write the Bureau of theCensus, Agriculture and Financial Statistics Division, Edi-torial and Information Staff, Washington, DC 20233-8300or call 1-800-523-3215.

ELECTRONIC PRODUCTS

x Computer Tapes —Public-use computer tapes containthe same summary statistics published in the Volume 1reports.

x Compact Disc-Read Only Memory (CD-ROM) —Datafrom Volume 1, Volume 2 parts 1 and 5, and Volume 3are available on CD-ROM.

x Online Access —Highlights from Volume 1, Volume 2parts 3 and 5, and Volume 3 are available on the ARINetworkR and on CENDATATM through DIALOGR andCompuServeR.

For information regarding electronic data products, write toCustomer Services, Bureau of the Census, Washington,DC 20233-8300 or call 301-763-4100. Internet users maycontact the Census Bureau’s home page at http://www.census.gov/.

Publication Program