Measuring Labor Input Dale Jorgenson, Mun Ho, Jon Samuels Harvard University World KLEMS Conference, Harvard University August 19, 2010.

Measuring Labor Input

Dale Jorgenson, Mun Ho, Jon Samuels

Harvard University

World KLEMS Conference, Harvard University

August 19, 2010

Topics

- Measurement Issues and Methodology

- Data and Implementation

- Results

- Contribution of labor input to productivity revival

- Criticisms of this method

Information Technology and the American Growth ResurgenceJorgenson, Ho and Stiroh (2005); Chapter 6

New Data on U.S. Productivity Growth by IndustryJorgenson, Ho and Samuels (2010)

Issues in Measuring Labor Input

- Number of workers, or Hours worked, are not suitable units of measure for heterogenous labor

- Wide range of market wages indicate wide range of productivities

- A wage-weighted index have been growing faster than simple sum of hours, productivity residual using hours will overstate the growth of TFP. -Need tractable method of handling this great heterogeneity

HoopsHype.com Salaries

Top NBA Salaries 1. Kobe Bryant LA Lakers $24,806,250 2. Rashard Lewis Orlando $20,514,000 3. Kevin Garnett Boston $18,800,000 4. Tim Duncan San Antonio $18,700,000 5. Michael Redd Milwaukee $18,300,000 6. Pau Gasol LA Lakers $17,822,187 7. Andrei Kirilenko Utah $17,822,187 8. Yao Ming Houston $17,686,100 9. Gilbert Arenas Washington $17,730,694 10. Dirk Nowitzki Dallas $17,300,000 11. Vince Carter Orlando $17,300,000 12. Zach Randolph Memphis $17,333,333 13. Carmelo Anthony Denver $17,149,243 14. Amare Stoudemire New York $16,800,000 15. Dwight Howard Orlando $16,509,600 16. Kenyon Martin Denver $15,959,099 -. Elton Brand Philadelphia $15,959,099 18. Predrag Stojakovic New Orleans $15,336,000 19. Chris Paul New Orleans $14,940,152 -. Deron Williams Utah $14,940,152

Methodology for a tractable measure of labor input

-Cross classify workers in each industry by demographic characteristics * In Jorgenson, Gollop & Fraumeni (1987): sex, class, age, education, occupation * Now: sex, class, age, education

-Define industry labor input as a Tornqvist index of the demographic components

Gender 2 Male; FemaleClass 2 Employees; Self-employed and unpaidAge 7 16-17; 18-24; 25-34; 35-44; 45-54; 55-64; 65+Education 6 0-8 years grade school

grade 9-12 no diplomaHigh School graduatesome College no Bachelors degreeBachelors degreemore than BA degree

Classification of demographic groups for each industry

2x2x7x6 = 168

Index of labor input for industry j, Ljt as Tornqvist index of components

, , ,

ln lnjt scaejt scaejts c a e

L v L

Lscaej scaej

scaej Lscaej scaej

scae

P Lv

P L

1, 12 [ ]ljt ljt lj tv v v

, 1

ln ln jtjt

j t

LL

L

scae: sex, class, age, education j: industry j or aggregate economy

Index Ljt, cont. Constant Quality Index

scaejt scae scaejtL Q H

Assume labor input is proportional to hours worked:

Qscae is the quality of hours of group scae, fixed for all t.Thus input index becomes:

, , , , , ,

ln ln lnjt scaejt scaejt scaejt scaejts c a e s c a e

L v L v H

, , ,jt scaejt

s c a e

H H Compared to simple hours:

Index Ljt, cont.

L Lt scaej scaej jt jt

scae

value P H P L

Price of industry labor input is simply value/Lj

after choosing a normalization like:

Quality of industry labor input is labor input index divided by hours worked:

jt

jtLjt H

LQ j scaej

scae

H H

,2000 ,2000 20001.0;Lj jP L value

Decomposing the labor input index

Partial indices of labor input. E.g. first-order index by age

ln ln ln( )aa a a saec

a a s e c

L v H v H

How much of the quality change is due to changes ..in educational attainment? ..in the aging of the labor force? …

Contribution of age to labor quality

ln ln lna aQ L H

Data

Need number of workers, hours and compensation to fill matrices of dimension (2sex, 2 class, 7age, 6educ, 70indus).

Total of 11760 cells.

Household survey data (hours/week, weeks/year, wages/year, demographics, industry) Census of Population. - every 10 years - 1% percent sample (1 million workers) Current Population Survey, Annual Supplement (ASEC) - every year, 1964+ - about 100,000 households

Establishment survey data Bureau of Economic Analysis tabulations of total employment, total compensation, wages for 72 industries; annual hours for 18 industries

Implementation

-Begin with Census microdata (1% sample, ~1 mil. workers) to populate EMP, HOURS, COMP matrices for benchmark years

-From CPS annual microdata, construct marginal matrices: EMP, HOURS, COMP matrices of lower dimension (e.g. indus x edu, sex x age x edu, …)

-Interpolate between benchmark years using these annual marginal matrices

-Scale to industry totals in the National Accounts

Data Issues

-Change from SIC to NAICS classification

(CPS 2003+ and Census 2000 uses NAICS)

-Change in education classification in 1992

-Small sample size in CPS (use fewer industries)

-Household data is “top-coded” for wages-Workers in multiple jobs (multiple industries)-Estimating wages for self employed-No data on fringe (non-wage) benefits by person

Table 6.8: Labor Characteristics by Industry, year 2005.

Compen-sation

($/hour)Hours: % aged 16-35

Hours: % females

Females: % college educated

Males: % college

educatedIT Producing

Computer & elec prod. 48.6 27.8 30.4 50.1 66.2Telecom equip 49.7 24.6 30.4 36.5 57.8Electronic components 49.8 26.6 34.0 30.5 56.6Software publishing 40.3 41.1 38.1 72.4 78.4Information, data proc. 43.6 41.9 40.2 53.1 65.6Computer sys. design 48.9 36.8 27.2 65.9 71.2

IT UsingWholesale trade 35.7 26.6 25.8 26.1 28.9Banks, credit interm. 40.0 29.9 61.8 24.8 61.8

Non-ITConstruction 26.3 30.8 8.3 19.2 10.3Hospitals 30.2 23.4 75.3 39.3 49.2

72-industry median 34.7 24.3 33.1 27.7 28.9

December 23, 2000 issue

Labor Contributions to Aggregate Growth

Criticisms of this methodology

- Equation of wages with marginal product is not valid with non-competitive markets and discrimination

- Small sample sizes for many industries give poor estimates of cell averages

- Education is not directly productive and merely a “signal”

- Intensity of work effort is not recognized

Summary

- Simple sum of hours understate labor contribution, overstate TFP growth

- Our labor input index – an aggregate over hours by demographic groups, weighted by wages – is a tractable measure with the use of U.S. Census microdata.

- The growth of labor quality was about 0.4% per year, or, ¼ of the labor contribution to GDP growth is due to labor quality and ¾ due to hours growth.