Presentation: A comparison of the contribution of labor reallocation to aggregate productivity growth: Canada and the United States

Cornell University ILR SchoolDigitalCommons@ILR

Labor Dynamics Institute Centers, Institutes, Programs

4-25-2012

Presentation: A comparison of the contribution oflabor reallocation to aggregate productivity growth:Canada and the United StatesCharles BerubeIndustry Canada

Benoit DostieInstitute of Applied Economics, HEC Montreal

Lars VilhuberCornell University, [email protected]

Follow this and additional works at: http://digitalcommons.ilr.cornell.edu/ldiThank you for downloading an article from [email protected] this valuable resource today!

This Article is brought to you for free and open access by the Centers, Institutes, Programs at DigitalCommons@ILR. It has been accepted for inclusionin Labor Dynamics Institute by an authorized administrator of DigitalCommons@ILR. For more information, please contact [email protected].

http://digitalcommons.ilr.cornell.edu/?utm_source=digitalcommons.ilr.cornell.edu%2Fldi%2F6&utm_medium=PDF&utm_campaign=PDFCoverPages

http://digitalcommons.ilr.cornell.edu/?utm_source=digitalcommons.ilr.cornell.edu%2Fldi%2F6&utm_medium=PDF&utm_campaign=PDFCoverPages

http://digitalcommons.ilr.cornell.edu?utm_source=digitalcommons.ilr.cornell.edu%2Fldi%2F6&utm_medium=PDF&utm_campaign=PDFCoverPages

http://digitalcommons.ilr.cornell.edu/ldi?utm_source=digitalcommons.ilr.cornell.edu%2Fldi%2F6&utm_medium=PDF&utm_campaign=PDFCoverPages

http://digitalcommons.ilr.cornell.edu/programs?utm_source=digitalcommons.ilr.cornell.edu%2Fldi%2F6&utm_medium=PDF&utm_campaign=PDFCoverPages

http://digitalcommons.ilr.cornell.edu/ldi?utm_source=digitalcommons.ilr.cornell.edu%2Fldi%2F6&utm_medium=PDF&utm_campaign=PDFCoverPages

https://securelb.imodules.com/s/1717/alumni/index.aspx?sid=1717&gid=2&pgid=403&cid=1031&dids=50.254&bledit=1&appealcode=OTX0OLDC

mailto:[email protected]

Presentation: A comparison of the contribution of labor reallocation toaggregate productivity growth: Canada and the United States

AbstractThis paper contributes to the literature by computing the contribution of labor reallocation to productivitygrowth for the United States and Canada, using several different decomposition specifications, and applying aconsistent approach to data from both countries. We compute the Baily, Hulten & Campbell (1992),Griliches & Regev (1995), Foster, Haltiwanger & Krizan (2001), and Baldwin & Gu (2006) decompositions,for both three-year and five-year time frames. We focus on the manufacturing sector for both countries, andassess the robustness of the conclusions to alternate specifications that the data from one or the other countryallow us to do.

Keywordsproductivity decomposition, Canada, United States, LBD, T2LEAP

CommentsSuggested Citation

Berube, C., Dostie, B. & Vilhuber, L. (2012, April). A comparison of the contribution of labor reallocation toaggregate productivity growth: Canada and the united states. Presented at CAED, Nurnberg.

Required Publisher's Statement

Copyright held by authors.

This article is available at DigitalCommons@ILR: http://digitalcommons.ilr.cornell.edu/ldi/6

http://digitalcommons.ilr.cornell.edu/ldi/6?utm_source=digitalcommons.ilr.cornell.edu%2Fldi%2F6&utm_medium=PDF&utm_campaign=PDFCoverPages

A comparison of the contribution of laborreallocation to aggregate productivity growth:

Canada and the United States

Charles Berube4 Benoit Dostie1 Lars Vilhuber2,3

1Institute of Applied Economics, HEC Montreal

2Center for Economic Studies/LEHD,

3 ILR, Cornell University

4Industry Canada

CAED 2012, Nurnberg

Berube, Dostie, Vilhuber Reallocation

Introduction Methodology Data Results Conclusion

Disclaimer

I Vilhuber’s research was supported by NSF Grant SES-0820349,SES-1042181, and SES-1042181.

I Part of the research was done while one of the researchers was aSpecial Sworn Status researcher of the U.S. Census Bureau at theCenter for Economic Studies. Research results and conclusionsexpressed are those of the authors and do not necessarily reflect theviews of the Census Bureau.

I Other data used in this paper are confidential data housed at aStatistics Canada Research Data Center, accessed through the OttawaHead Office by arrangement with Industry Canada.

I All results presented in this paper have been screened to insure that noconfidential data are revealed.



Disclaimer (2)

This is work in progress, and results are very preliminary!



Outline

Introduction to the paper

Methodology

Data

Results

Conclusion



CAED/COST link

Cross-national analysis

I We started several years ago with Canadian survey(WES), French and US administrative data (CAED2009).

I This paper is with US and Canadian administrative data(combined in the US with firm survey data: ASM)

I “Multi-site, multi-author” replication approachI start with same code, sit in front of respective secure

terminalsI ... then see what happens....



CAED/COST link








CAED/COST link




I “Multi-site, multi-author” replication approach

I start with same code, sit in front of respective secureterminals

I ... then see what happens....



CAED/COST link





terminals

I ... then see what happens....



CAED/COST link








Replication issues

Slew of issuesI lack of common variable names

I code divergence/creepI different merge/match issuesI different variable definitions - obvious and subtle - that

affect the outcomesI access issues



Replication issues

Slew of issuesI lack of common variable namesI code divergence/creep

I different merge/match issuesI different variable definitions - obvious and subtle - that




Replication issues

Slew of issuesI lack of common variable namesI code divergence/creepI different merge/match issues

I different variable definitions - obvious and subtle - thataffect the outcomes

I access issues



Replication issues

Slew of issuesI lack of common variable namesI code divergence/creepI different merge/match issuesI different variable definitions - obvious and subtle - that

affect the outcomes

I access issues



Replication issues

Slew of issuesI lack of common variable namesI code divergence/creepI different merge/match issuesI different variable definitions - obvious and subtle - that




Background

I A large literature attempts to quantify how factorreallocation contributes to productivity growth.

I Some numbers:

I Quarterly job reallocation rates are about 12% in the US(Abowd and Vilhuber 2011)

I Worker reallocation rates are about 2-3 times larger(Abowdand Vilhuber 2011)

I Labor productivity growth is an important contributor

I 50% of labor productivity growth is dependent on laborreallocation (Foster, Haltiwanger, and Krizan 2001, USdata)

I ... or is it higher: 70% (Lentz and Mortensen 2008, Danishdata)



Background


I Some numbers:

I Quarterly job reallocation rates are about 12% in the US(Abowd and Vilhuber 2011)







Background


I Some numbers:I Quarterly job reallocation rates are about 12% in the US

(Abowd and Vilhuber 2011)







Background



(Abowd and Vilhuber 2011)I Worker reallocation rates are about 2-3 times larger(Abowd

and Vilhuber 2011)






Background




and Vilhuber 2011)I Labor productivity growth is an important contributor





Background









Background









Our approach

Here: evidence on the evolution of labor productivitydecomposition ...

I ... for two countries: Canada and the United States...

I ... using several different labor productivity decompositionmethodologies, applied homogeneously to both datasets

I ... caveats at the end



Our approach


I ... for two countries: Canada and the United States...I ... using several different labor productivity decomposition

methodologies, applied homogeneously to both datasets

I ... caveats at the end



Our approach


I ... for two countries: Canada and the United States...I ... using several different labor productivity decomposition

methodologies, applied homogeneously to both datasetsI ... caveats at the end



Productivity

Aggregate productivity

Pt =∑j∈J

θjtpjt (1)

θjt represents the firm’s market share (share of labor or share ofsales), and pjt is the individual firm’s productivity.



Productivity growth

Productivity growth

∆Pt ,t−k =∑j∈Jt

θjtpjt −∑

j∈Jt−k

θjt−kpjt−k (2)



BHC decomposition

I BHC decomposition (Baily, Hulten, and Campbell 1992)

∆Pt =∑i∈Ct

θit−1∆pit

+∑i∈Ct

∆θitpit−1 +∑i∈Ct

∆θit ∆pit

+∑i∈Et

θitpit −∑i∈Xt

θit−1pit−1 (3)

= Within + Between + Cross + Entry − Exit

I where Jt = {Ct ,Et} and Jt−k = {Ct ,Xt}



BHC decomposition


∆Pt =∑i∈Ct

θit−1∆pit

+∑i∈Ct


∆θit ∆pit

+∑i∈Et


θit−1pit−1 (3)





BHC decomposition


∆Pt =∑i∈Ct

θit−1∆pit

+∑i∈Ct


∆θit ∆pit

+∑i∈Et


θit−1pit−1 (3)

= Within

+ Between + Cross + Entry − Exit




BHC decomposition


∆Pt =∑i∈Ct

θit−1∆pit

+∑i∈Ct


∆θit ∆pit

+∑i∈Et


θit−1pit−1 (3)

= Within + Between

+ Cross + Entry − Exit




BHC decomposition


∆Pt =∑i∈Ct

θit−1∆pit

+∑i∈Ct


∆θit ∆pit

+∑i∈Et


θit−1pit−1 (3)

= Within + Between + Cross

+ Entry − Exit




BHC decomposition


∆Pt =∑i∈Ct

θit−1∆pit

+∑i∈Ct


∆θit ∆pit

+∑i∈Et


θit−1pit−1 (3)

= Within + Between + Cross

+ Entry − Exit




BHC decomposition


∆Pt =∑i∈Ct

θit−1∆pit

+∑i∈Ct


∆θit ∆pit

+∑i∈Et


θit−1pit−1 (3)

= Within + Between + Cross + Entry

− Exit




BHC decomposition


∆Pt =∑i∈Ct

θit−1∆pit

+∑i∈Ct


∆θit ∆pit

+∑i∈Et


θit−1pit−1 (3)





BHC decomposition


∆Pt =∑i∈Ct

θit−1∆pit

+∑i∈Ct


∆θit ∆pit

+∑i∈Et


θit−1pit−1 (3)





GR decomposition

I GR decomposition (Griliches and Regev 1995)

∆Pt ,t−k =∑j∈C

θj∆pj

+∑j∈C

∆θj(pj − PJ

)+

∑j∈E

θjt(pjt − PJ

)−∑j∈X

θjt−k(pjt−k − PJ

)

= Within + Between + Entry + Exit

I within-firm productivity growth is weighted by the averagemarket shares between period t and t − k

I between effect weighted by the difference between firm’saverage productivity vs. average productivity of all firms

I productivity differences for entrants/exiters are weighted bythe contemporaneous market share of the firm



GR decomposition



θj∆pj

+∑j∈C

∆θj(pj − PJ

)+

∑j∈E

θjt(pjt − PJ

)−∑j∈X


)

= Within + Between + Entry + Exit






GR decomposition



θj∆pj

+∑j∈C

∆θj(pj − PJ

)+

∑j∈E

θjt(pjt − PJ

)−∑j∈X


)= Within

+ Between + Entry + Exit






GR decomposition



θj∆pj

+∑j∈C

∆θj(pj − PJ

)+

∑j∈E

θjt(pjt − PJ

)−∑j∈X


)= Within + Between

+ Entry + Exit






GR decomposition



θj∆pj

+∑j∈C

∆θj(pj − PJ

)+

∑j∈E

θjt(pjt − PJ

)−∑j∈X


)= Within + Between + Entry

+ Exit






GR decomposition



θj∆pj

+∑j∈C

∆θj(pj − PJ

)+

∑j∈E

θjt(pjt − PJ

)−∑j∈X


)= Within + Between + Entry + Exit






GR decomposition



θj∆pj

+∑j∈C

∆θj(pj − PJ

)+

∑j∈E

θjt(pjt − PJ

)−∑j∈X








GR decomposition



θj∆pj

+∑j∈C

∆θj(pj − PJ

)+

∑j∈E

θjt(pjt − PJ

)−∑j∈X








GR decomposition



θj∆pj

+∑j∈C

∆θj(pj − PJ

)+

∑j∈E

θjt(pjt − PJ

)−∑j∈X








FHK decomposition

I FHK version (Foster, Haltiwanger, and Krizan 2001)


θjt−k ∆pj +∑j∈C

∆θj(pjt−k − PJt−k

)+

∑j∈C

∆θj∆pj +∑j∈E

θjt(pjt − PJt−k

)−

∑j∈X

θjt−k(pjt−k − PJt−k

)

I contribution of firm’s pi , i = t , t − k now relative toeconomy/sector-wide Pt−k



FHK decomposition

I FHK version (Foster, Haltiwanger, and Krizan 2001)


θjt−k ∆pj +∑j∈C

∆θj(pjt−k − PJt−k

)+

∑j∈C

∆θj∆pj +∑j∈E

θjt(pjt − PJt−k

)−

∑j∈X

θjt−k(pjt−k − PJt−k

)I contribution of firm’s pi , i = t , t − k now relative to

economy/sector-wide Pt−k



Baldwin-Gu decomposition

I If new entering firms are taking market share away fromboth exiting and existing firms, then


θj∆pj +∑j∈C

∆θj(pj − PD

)+

∑j∈X

θjt−k(PN − pjt−k

)+ (SN − SX ) (PN − PD)

I between-effect: relative to the average for firms withdeclining market share, PD.

I productivity of new entrants PN measured relative to theproductivity of exiting firms pjt−k → new entrants displacingexiting firms.

I contribution to productivity growth of new entrants thatrecoup market shares from declining firms.

I SN market share of j ∈ E , SX market share of j ∈ X






θj∆pj +∑j∈C

∆θj(pj − PD

)+

∑j∈X


)+ (SN − SX ) (PN − PD)










θj∆pj +∑j∈C

∆θj(pj − PD

)+

∑j∈X


)+ (SN − SX ) (PN − PD)










θj∆pj +∑j∈C

∆θj(pj − PD

)+

∑j∈X


)+ (SN − SX ) (PN − PD)










θj∆pj +∑j∈C

∆θj(pj − PD

)+

∑j∈X


)+ (SN − SX ) (PN − PD)







Data



Canadian data - T2/LEAP

T2/Longitudinal Employment Analysis Project (LEAP)

I two main sources of administrative data:

I the Longitudinal Employment Analysis Program (LEAP),containing information on the employment of firms with(paid) employees

I Corporate Tax Statistical Universe File (T2SUF): covers allcompanies filing income tax, provides financial information

I excludes unincorporated businesses, non-employersI employment variable not directly measured: computed by

Statistics Canada as ratio of labor expenditures to thetypical worker’s average annual remuneration, adjusted forindustry, province, and firm size (Average Labor Unit, ALU,Baldwin and Gu (2011))

I No value-added: productivity measured as sales (receipts)per worker





I two main sources of administrative data:I the Longitudinal Employment Analysis Program (LEAP),

containing information on the employment of firms with(paid) employees






















I excludes unincorporated businesses, non-employers

I employment variable not directly measured: computed byStatistics Canada as ratio of labor expenditures to thetypical worker’s average annual remuneration, adjusted forindustry, province, and firm size (Average Labor Unit, ALU,Baldwin and Gu (2011))
























US data

ASM+CMI Available: 1973-2009 (1987-2009 used)

I CM: quinquennial census of firms (years in 2 and 7)I CM: sampled in Business Register, includes ASM

establishments in CM yearsI ASM: Certainty sample for large firms, size-stratification for

smaller firmsI ASM: about 50,000 establishments per yearI ASM: panel for 5 years, sampled in CM, refreshed based

on expansion of the frame through tax recordsI ASM/CM: information on employment, wages, sales,

value-added



US data

ASM+CMI Available: 1973-2009 (1987-2009 used)I CM: quinquennial census of firms (years in 2 and 7)

I CM: sampled in Business Register, includes ASMestablishments in CM years

I ASM: Certainty sample for large firms, size-stratification forsmaller firms

I ASM: about 50,000 establishments per yearI ASM: panel for 5 years, sampled in CM, refreshed based


value-added



US data

ASM+CMI Available: 1973-2009 (1987-2009 used)I CM: quinquennial census of firms (years in 2 and 7)I CM: sampled in Business Register, includes ASM

establishments in CM years

I ASM: Certainty sample for large firms, size-stratification forsmaller firms



value-added



US data



smaller firms



value-added



US data



smaller firmsI ASM: about 50,000 establishments per year

I ASM: panel for 5 years, sampled in CM, refreshed basedon expansion of the frame through tax records

I ASM/CM: information on employment, wages, sales,value-added



US data




on expansion of the frame through tax records

I ASM/CM: information on employment, wages, sales,value-added



US data





value-added



US data

LBDI longitudinal research file (Miranda and Jarmin 2002)

I corrects linkages in Business RegisterI contains link id to ASM, CM, employment



US data

LBDI longitudinal research file (Miranda and Jarmin 2002)I corrects linkages in Business Register

I contains link id to ASM, CM, employment



US data

LBDI longitudinal research file (Miranda and Jarmin 2002)I corrects linkages in Business RegisterI contains link id to ASM, CM, employment



ASM-CM schema



ASM-CM schema



ASM-CM schema



ASM-CM schema



T2LEAP schema



T2LEAP schema



Overview of LBD data



ASM-CM-LBD schema



Methodology for US

Matching methodology using LBD

I Define births/deaths/continuers in LBD

I Match to records in ASM/CM as feasibleI Create panel weight to match birth/death rates in LBD

(here: by ten size-classes)



Methodology for US


I Define births/deaths/continuers in LBDI Match to records in ASM/CM as feasible

I Create panel weight to match birth/death rates in LBD(here: by ten size-classes)



Methodology for US


I Define births/deaths/continuers in LBDI Match to records in ASM/CM as feasibleI Create panel weight to match birth/death rates in LBD

(here: by ten size-classes)



Data definitions

Imputations

I We impute missing sales based on data for surroundingyears for the same firm.

I When employment is missing, we assume the plant isinactive (dead)

Adjustments

I Productivity = (real value of) sales/workerI Trim top and bottom 2% of productivity by removing from

the panel.



Comparing

Common characteristicsI Long time series

I Data accessible in restricted access environmentsI Results need to pass disclosure review



Comparing

Common characteristicsI Long time seriesI Data accessible in restricted access environments

I Results need to pass disclosure review



Comparing

Common characteristicsI Long time seriesI Data accessible in restricted access environmentsI Results need to pass disclosure review



Comparing

DifferencesI Canada is administrative data, US is administrative linked

to survey sample data

I Canada is firm-based (no establishments), US isestablishment-based

I Canada is all industries, US Economic Censuses/AnnualSurveys are by-sector (no unified dataset)

I Entity adjustment: flow adjustment in Canada,name/location linkage in US

I Productivity not quite the same: Sales receipts vs. value ofshipments

I Employment not measured the same way (averageemployment in year imputed in Canada, point-in-timeemployment in the US)

(could be adjusted in the US)



Comparing


to survey sample dataI Canada is firm-based (no establishments), US is

establishment-based

I Canada is all industries, US Economic Censuses/AnnualSurveys are by-sector (no unified dataset)







Comparing



establishment-basedI Canada is all industries, US Economic Censuses/Annual

Surveys are by-sector (no unified dataset)







Comparing




Surveys are by-sector (no unified dataset)I Entity adjustment: flow adjustment in Canada,

name/location linkage in US






Comparing





name/location linkage in USI Productivity not quite the same: Sales receipts vs. value of

shipments





Comparing






shipmentsI Employment not measured the same way (average

employment in year imputed in Canada, point-in-timeemployment in the US)




Comparing







employment in year imputed in Canada, point-in-timeemployment in the US)




Comparing







employment in year imputed in Canada, point-in-timeemployment in the US) (could be adjusted in the US)



Results



Previous results

Enormous literatureThe literature is enormous. Many studies provide somesummary of previous studies.

I Within-plant contribution between 0.79-1.2 (Foster,Haltiwanger, and Krizan 2001)



Comparing to FHK2001



Comparing to Baldwin and Gu (multiple)

to come



US and Canada: FHK



US and Canada: BHC



US and Canada: GR



US and Canada: BG



Variations and robustness checks

I Variations in k (3 years, 5 years, 1 year?) [easy in Canada,not in US] (already noted in Foster, Haltiwanger, andKrizan (2001))

I Robustness to firm birth/death adjustments

I Missing data [currently very simple (simplistic) impute]I Importance of measuring at firm level [only way in Canada,

only in EC years in US]





I Robustness to firm birth/death adjustments

I Missing data [currently very simple (simplistic) impute]I Importance of measuring at firm level [only way in Canada,






I Robustness to firm birth/death adjustmentsI Missing data [currently very simple (simplistic) impute]

I Importance of measuring at firm level [only way in Canada,only in EC years in US]





I Robustness to firm birth/death adjustmentsI Missing data [currently very simple (simplistic) impute]I Importance of measuring at firm level [only way in Canada,




Preliminary conclusions and speculation

I Much stronger role for within-reallocation, little role forentrants/exiters in Canada, in all periods

I ... due to measurement at firm vs. establishment level?I ... due to fundamentals?

I decreasing role of cross effect (all), between effect(GR,BG) in Canada?

I In the US, positive net effect of entry/exit, but secularincrease in role of entrants/decrease in role of exiters?





I ... due to measurement at firm vs. establishment level?

I ... due to fundamentals?I decreasing role of cross effect (all), between effect

(GR,BG) in Canada?I In the US, positive net effect of entry/exit, but secular

increase in role of entrants/decrease in role of exiters?
























Thank you.


References

References

Abowd, J. M. and L. Vilhuber (2011).National estimates of gross employment and job flows from the quarterly workforce indicators with demographic

and industry detail.Journal of Econometrics 161, 82–99.

Baily, M., C. Hulten, and D. Campbell (1992).Productivity dynamics in manufacturing plants.Brookings Paper on Economic Activity. Microeconomics, 187–249.

Baldwin, J. R. and W. Gu (2011, January).Firm dynamics and productivity growth: a comparison of the retail trade and manufacturing sectors.Industrial and Corporate Change, 1–29.

Foster, L., J. C. Haltiwanger, and C. J. Krizan (2001).Aggregate productivity growth. lessons from microeconomic evidence.In C. R. H. . E. R. D. . M. J. Harper (Ed.), New Developments in Productivity Analysis, NBER Chapters, pp.

303–372. National Bureau of Economic Research, Inc.

Griliches, Z. and H. Regev (1995).Firm productivity in Israeli industry 1979-1988.Journal of Econometrics 65(1), 175–203.

Lentz, R. and D. T. Mortensen (2008).An empirical model of growth through product innovation.Econometrica 76(6), 1317–1373.

Miranda, J. and R. Jarmin (2002).The longitudinal business database.Discussion Paper CES-WP-02-17, U.S. Census Bureau, Center for Economic Studies.


References

The end

$Id: Presentation-CAED2012-reallocation-appendix.tex 1083 2012-04-25 17:11:40Z vilhu001 $


Presentation: A comparison of the contribution of labor reallocation to aggregate productivity growth: Canada and the United States

Documents