ICES III June 2007 The Redesign of Agriculture Surveys by Laurie Reedman and Claude Poirier.

ICES III June 2007

The Redesign of Agriculture Surveys

by

Laurie Reedman and Claude Poirier

Outline

BackgroundCurrent SituationPrioritiesScopeIssuesNext Steps

Mandate of the Agriculture Statistics Program

Estimates of agriculture production for crops, horticulture, livestock and animal products, as well as revenues and expensesTo conduct the Census of Agriculture (CEAG) every 5 yearsTo manage the statistical system of Canada's agriculture sector from data collection to publication Ensure quality outputs for economic analysis and policy making in Canada

Current Situation

Large regular surveys: Crops, Livestock, Hogs, Atlantic, Farm Financial, Fruit and Vegetables, Greenhouse, Sod and Nursery

Smaller regular surveys: Potato Area and Yield, Potato Prices, other prices

Irregular surveys: cost recovery surveys on the environment, farming practices, risk management

Administrative data

Farm Register (FR)

Priorities

Reduce response burdenIndividual; whole population

Improve robustness

Standardize methods and adopt “best practices”

Coverage

Efficient use of internal resources

Efficient use of the farming community’s capacity to respond

Scope

The surveys that use a static frame for the 5 year period between censuses:

Crops

Livestock

Atlantic

Farm Financial

The methodology of survey design

Small Farm Exclusion Threshold

Want to reduce burden on the many small farms that do not have much impact on survey estimates

Propose a method to compensate for the under-coverage that would result from excluding the small farms from the regular survey sampling

Who are the small farms?

Current small farm threshold is $10K reported for the sale of agriculture products on CEAG

21% of all farms and 0.6% of total sales

Other small farm thresholds could be:$25K, 39% of all farms, 2.4% of total sales

$50K, 53% of all farms, 5.6% of total sales

The bottom 5% of sales in each province, 50% of farms

What do the small farms contribute?

Say threshold is $25K in sales on CEAG 2006 …2% of hogs in Canada4% of field crop area in Manitoba9% of the field crop area in Atlantic Canada10% of program payments in Alberta22% of total farm capital in New Brunswick30% of sheep in Alberta35% of beef cattle in OntarioNearly 100,000 acres in different varieties of lentils, beans, dry peas and chick peas in Saskatchewan and Alberta

How to estimate for the small farms if not through regular surveys

Admin sources (tax) do not have commodity data, not adequateCEAG 2006Annual Farm Update Survey (FUS)

Sample is drawn from tax records, producer lists and the margins of the FR to detect farms not already in the active populationExpand scope to also represent the small farmsAugment questionnaire to cover more commoditiesIncrease sample size to provide reliable estimates

Factors in Decision Making

CEAG and/or FUS can adequately estimate livestock variables, the major crops and many components of the Farm Financial Survey (FFS)CEAG questionnaire does not have the varieties of lentils, beans, dry peas and chick peasUnlikely that the FUS questionnaire would have detailed commoditiesSmall farms are part of the target population for some FFS concepts

Decision for 2006 Redesign

Risk of under coverage is too high …Crops and FFS are not ready to raise the small farm exclusion threshold Not feasible to redesign FUS just for Livestock and AtlanticDecision:

keep small farm threshold at $10K for all surveyspilot redesign of FUS, to demonstrate its ability to measure the under coveragestratum boundary at $25K

Stratification and Sample Allocation

Reduce sample sizes, ensure reliable estimates for domains of interest

As few strata as necessary

As few take-all strata as necessary

Use generalized software

Stratify once for the 5 year period

Crops Survey

Estimate acreage of crops, production and yield at provincial as well as sub-provincial level, 6 surveys annuallySize classes based on total field crop areaKey crops are barley, corn for grain, oats, soybeans, winter wheat and hayTarget sample size is 16,000

Crops Survey continued

Allocated sample to the provinces proportional to the square root of number of farmsMultivariate allocation to strata, using key variablesCalculated theoretical coefficients of variation (CVs) and also selected a random sample and verified that there were no deviations in the estimates

Livestock Survey

Estimates totals of different types of cattle, sheep and hogs, at provincial level, 2 surveys annuallySize classes (counts of animals) within farm typeKey variables are total cattle, beef cows, total pigs, sows, total sheep, and also milk cows in some provincesTarget sample size 10,000

Atlantic Survey

Estimates both crops and livestock variables in Atlantic provinces, 2 surveys annuallyChallenge to measure crops and livestock with one sample, farms tend to be mixedSize classes within farm typeKey variables are total cattle, total pigs and total field crops, and potatoes in Prince Edward IslandTarget sample size 1,200

Farm Financial Survey

Estimates financial activity and farm characteristics at provincial level, 1 survey annually

Size classes (total assets) within farm type

Sample is allocated based on total farm revenue

Sample size is usually 18,000

Summary of Sample Allocation

Population Size

Sample Size

CVs on key variables

CVs on other variables

Crops Survey

155,000 16,000 1-2% 3-15%

Livestock Survey

100,000 10,000 1-2% 3-15%

Atlantic

Survey

6,000 1,200 1-4% 5-30%

Financial Survey

179,000 18,000 1-3% 4-30%

Large or Complex Farms

Group of people dedicated to collecting and maintaining data pertaining to the biggest and most influential farmsManage the response burdenProfiling once each yearControl number of times they are contacted, carry-forward information for some survey occasions

Frame Maintenance

Changes in stratification variablesMinimized by having a robust stratification

Births from the Farm Update SurveySame probability of selection as rest of frame

Updates from Farm RegisterAre they independent, is there a risk of bias?

DeathsAre they independent, can we drop them?

Partnerships, buy-outs, splits

Sample Co-ordination

Permanent random numbers

Moving, growing sampling windows

What to do about strata with high sampling fractions

What to do about births

What to do about irregular surveys

What to do about special requests, for example, when more sample is needed to improve precision for a particular domain

Next Steps

Confirm all assumptions and decisions with CEAG 2006 data

Create new survey frames

Select samples

Monitor performance on first few survey occasions, evaluate performance

Estimation system, review of the Farm Register

Redesign FUS

Examine target population definition

Thank-you

For more information, or

to obtain a French copy

of this presentation,

please contact:

Pour de plus amples

informations ou pour obtenir

une copie en français du

document, veuillez

contacter:

Laurie ReedmanEmail / Courriel: [email protected] number / Téléphone: 613-951-7301