Employment and Skill Reallocation over the Business Cycle: Evidence from Danish Data

Size: px
Start display at page:

Download "Employment and Skill Reallocation over the Business Cycle: Evidence from Danish Data"

Transcription

1 Employment and Skill Reallocation over the Business Cycle: Evidence from Danish Data A. Bertheau 1, H. Bunzel, 2 M. Hejlesen, 2 R. Vejlin 2 1 CREST-ENSAE 2 Aarhus University October 2017 Work in Progress. Comments welcome. 0 / 46

2 Summary Intro Data Worker Flows Skills Conclusion Appendix 1 / 46

3 Outline Intro Data Worker Flows Skills Conclusion Appendix 2 / 46

4 Goal Evidence on directional worker mobility between firms by "job-to-job" and "non-employment". 1. Reallocation to different firms 3 / 46

5 Goal Evidence on directional worker mobility between firms by "job-to-job" and "non-employment". 1. Reallocation to different firms 2. Reallocation of different workers to different firms 3 / 46

6 Why Few direct evidence on the reallocation process for heterogeneous workers/firms. 1. Non-employment flow varies by firm type: Cleansing effect Mortensen and Pissarides (94), Acemoglu and Hawkins (2013), 4 / 46

7 Why Few direct evidence on the reallocation process for heterogeneous workers/firms. 1. Non-employment flow varies by firm type: Cleansing effect Mortensen and Pissarides (94), Acemoglu and Hawkins (2013), 2. Job-to-job flow varies by firm type: Job ladder Burdett and Mortensen (98), Moscarini and Postel-Vinay (13) 4 / 46

8 Why Few direct evidence on the reallocation process for heterogeneous workers/firms. 1. Non-employment flow varies by firm type: Cleansing effect Mortensen and Pissarides (94), Acemoglu and Hawkins (2013), 2. Job-to-job flow varies by firm type: Job ladder Burdett and Mortensen (98), Moscarini and Postel-Vinay (13) 3. Flows for heterogeneous workers and firms: Sorting "Sorting". Lise and Robin (2017) 4 / 46

9 Preview of findings 1. Large firms are more cyclical sensitive than small firms using unemployment deviation from a HP trend 5 / 46

10 Preview of findings 1. Large firms are more cyclical sensitive than small firms using unemployment deviation from a HP trend 2. Using first difference change in unemployment, small shrink more relative to large. 5 / 46

11 Preview of findings 1. Large firms are more cyclical sensitive than small firms using unemployment deviation from a HP trend 2. Using first difference change in unemployment, small shrink more relative to large. 3. Net Poaching for large firms is negative/acyclic 5 / 46

12 Preview of findings 1. Large firms are more cyclical sensitive than small firms using unemployment deviation from a HP trend 2. Using first difference change in unemployment, small shrink more relative to large. 3. Net Poaching for large firms is negative/acyclic 4. High-wage firms are more cyclical sensitive than low-wage firms using both cyclical indicators 5 / 46

13 Preview of findings 1. Large firms are more cyclical sensitive than small firms using unemployment deviation from a HP trend 2. Using first difference change in unemployment, small shrink more relative to large. 3. Net Poaching for large firms is negative/acyclic 4. High-wage firms are more cyclical sensitive than low-wage firms using both cyclical indicators 5. Using labor productivity (Value added/hours), low productive firms are more cyclical sensitive in contractions (due to the non-employment margin) 5 / 46

14 Relevant literature and contributions 1. The role of firm heterogeneity on worker mobility Large vs small firms have distinct net employment flows over the cycle. Large firms do not grow by job-to-job Moscarini and Postel-Vinay 2012 [AER]. Haltiwanger Hyatt, Kahn McEntarfer 2017 [AEJ macro] 6 / 46

15 Relevant literature and contributions 1. The role of firm heterogeneity on worker mobility Large vs small firms have distinct net employment flows over the cycle. Large firms do not grow by job-to-job Moscarini and Postel-Vinay 2012 [AER]. Haltiwanger Hyatt, Kahn McEntarfer 2017 [AEJ macro] 2. Skill requirements over time/the cycle Recent direct evidence on upskillings of the workforce in recessions. Hershbein and Khan 2016, Modestino, Shoag and Ballance Haltiwanger, Hyatt and McEntarfer / 46

16 Outline Intro Data Worker Flows Skills Conclusion Appendix 7 / 46

17 Data 1. New Labor Market Histories ( ) Entire labor market histories (include short spells), date of start and end of jobs, earnings and hours worked. Includes worker/establishment/firm ID 8 / 46

18 Data 1. New Labor Market Histories ( ) Entire labor market histories (include short spells), date of start and end of jobs, earnings and hours worked. Includes worker/establishment/firm ID 2. Individual ( ) Age, Educational attainment, occupation... 8 / 46

19 Data 1. New Labor Market Histories ( ) Entire labor market histories (include short spells), date of start and end of jobs, earnings and hours worked. Includes worker/establishment/firm ID 2. Individual ( ) Age, Educational attainment, occupation Firm ( ) Revenue, total labor costs, intermediate cost, capital... Restriction: Firm 2 workers; Worker age: [18,60]. 8 / 46

20 What we do Dissect worker mobility with threefold heterogeneity 1. We categorize hires (H) and separations (S) rates by job-to-job and non-employment Def H t S t = H tp S tp }{{} Net Poaching + H tn S tn }{{} Net Non Employment (1) 9 / 46

21 What we do 2. We categorize firms by size, average earnings, productivity Ranking: 2.1 Employment-weighted quintiles for high/medium/low average earnings and labor productivity. 2.2 Large: > 200 employees Medium: < 200 & > 50 employees Small < 50 employees Robustness: Firm fixed effect from AKM and age of firms 10 / 46

22 What we do 3. We categorize workers into educational attainment Education groups (Less than high school, Vocational training, Bachelor or related, Master and PhDs) 11 / 46

23 Outline Intro Data Worker Flows Skills Conclusion Appendix 12 / 46

24 Worker Flows: Aggregate Patterns 13 / 46

25 Hires and Separations for Large firms 14 / 46

26 Hires and Separations for Small firms 15 / 46

27 Hires and Separations for High-wage firms 16 / 46

28 Hires and Separations for Low-wage firms 17 / 46

29 Hires and Separations for High-productivity firms 18 / 46

30 Hires and Separations for Low-productivity firms 19 / 46

31 Cyclicality of employment We estimate on quaterly worker flows data Net Flows t,high Net Flows t,low = β 0 + β 1 Cycle t + β 2 year + γ s + ɛ t Cycle t is: 1. First difference of aggregate unemployment rate 2. Deviation from HP trend 20 / 46

32 Cyclical Analysis: Number of employees Table: Large Firms - Small Firms flows (top) and their decompositions (bottom). N= 104. Dependent Variables Unemp. HP filtered Diff. Net job flow.82*** -.10 Diff. Job-to-Job.14***.05** Diff. Non-Employment.68*** -.15* Dep variables Job-to-Job Non-Employment Unemp. HP filtered Unemp. HP filtered Hires Large -.33*** -.30*** -.13* -.05 Hires Small -.60*** -.50*** -.53**.01 Separations Large -44*** -32*** *** Separations Small -58*** -46***.16***.06** 21 / 46

33 Taking Stock Large firms are more cyclical sensitive in downturns, but effect is small/not significant and driven by net non-employment flow Small firms are more cyclical sensitive in contractions ( Unemp.), due to non-employment flow. Non-employment change is driven by both a stronger increase in separations (.16***) and a stronger decline of hires for small firms (-.53**) As opposed to the size ladder hypothesis, net hires by job-to-job decline more for small firms than for large. 22 / 46

34 Cyclical Analysis: average hourly earnings Table: High-wage - Low-wage Firms (top) and their decompositions (bottom). N= 104. Dependent Variables Unemp. HP filtered Diff. Net job -.33*** -.19*** Diff. Job-to-Job -.19*** -.13*** Diff. Non-Employment -.14* -.06* Dep var. Job-to-Job Non-Employment Unemp. HP filtered Unemp. HP Hires H wage -.50*** -.51*** -.21* -.12*** Hires L wage -.51*** -.32*** -.44**.01 Separations H wage -45*** -43***.17***.03 Separations L wage -65*** -37*** *** 23 / 46

35 Taking Stock High-wage firms are more cyclical sensitive using any cyclical indicator Stronger decline of high-wage firms relative to low-firms in contractions ( Unemp.) compared to downturns (HP) Contrary to Haltiwanger et al (2017), non-employment flow are contributing to relative decline of high wage firms (-14**) Using data from 1999, we get non significant negative effect of non-employment (-.12) Using sample of private firms w/out missing productivity, we get a significant positive effect of non-employment flow (.26*), with overall decline of HW firms Significant effect of separation by high-wage firms 24 / 46

36 Cyclical Analysis: Value added / Hours Worked Table: High-productivity - Low-productivity Firms (top) and their decompositions (bottom). N= 48. Independant Variables Unemp. HP filtered trend Diff. Net job flow.29* -.14*** Diff. Job-to-Job -.19** -.14*** Diff. Non-Employment.48*** -.00 Dep var. Job-to-Job Non-Employment Unemp. HP filtered Unemp. HP Hires H VA/H -.54** -.46*** -.31** -.05 Hires L VA/H -.65** -.41*** -1.12** -.08 Separations H VA/H -.43* -.38*** Separations L VA/H -.73** -.47*** / 46

37 Taking Stock Low-productivity firms are more cyclical sensitive using Unemp. Net poaching and net non-employment have opposite effect on net differential flows, but net non-nonemployment effect outweights net poaching effect Stronger drop of hires (-1.12***) through non-employment for low-productivity firms compared to low-wage firms. Layoffs response is similar for high-wage/high-productivity firms 26 / 46

38 Outline Intro Data Worker Flows Skills Conclusion Appendix 27 / 46

39 Reallocation of heterogeneous workers between heterogeneous firms Which workers high-wage firms poach in booms? Which workers do they layoffs in recessions? Net Poaching Propensity = H p(x,y ) S p (X,Y ) Emp(X ) (2) 28 / 46

40 High-wage Net Poaching flows 29 / 46

41 Less educated workers have a higher propensity to be reallocated towards high-wage firms High wage - Low wage differential is more pronounced for low educated workers Table: High-wage - Low-wage firms differential flows across worker groups Dep variables <High-School Vocational Bachelor Master + Net Job Flows -.11*** -.09*** -.05* -.04 Job-to-job -.05*** -.06*** -.02*** -.01 Non-employment * / 46

42 High-wage net non-employment flows 31 / 46

43 Low-wage non-employment flows 32 / 46

44 Outline Intro Data Worker Flows Skills Conclusion Appendix 33 / 46

45 Conclusions 1. Job-to-job and nonemployment flows are both component of productivity enhancing reallocation but at different phases of the cycle 34 / 46

46 Conclusions 1. Job-to-job and nonemployment flows are both component of productivity enhancing reallocation but at different phases of the cycle 2. Less-educated workers are more affected by fluctuations to match with higher productivity firms 34 / 46

47 Outline Intro Data Worker Flows Skills Conclusion Appendix 35 / 46

48 Table: Firm-to-firm transitions in spell data ( ) # in Unweighted Weighted cluster Percent Cumulative percent Percent Cumulative percent > / 46

49 Cyclical indicators: unemployment rate 37 / 46

50 Net job flows from poaching (H tp S tp ) 38 / 46

51 Net job flows from non-employment (H n S n ) 39 / 46

52 Hires and Separations for Middle size firms 40 / 46

53 Net poaching by age 41 / 46

54 Hires and Separations for Middle-wage size firms 42 / 46

55 Hires and Separations for Middle-productivity size firms 43 / 46

56 Net job growth rate 44 / 46

57 AKM w ijt = x it β + θ i + ψ J(i,t) + ɛ ijt We include experience, experience squared, # months of education and year fixed effects. 45 / 46

58 Cyclical Analysis: robustness on average hourly earnings Table: Using data from N = 48 High wage Firms - Low wage Firms Dependent Variables First difference Deviation from HP trend Net Job Flows -.34*** -.14*** Net Job-to-Job Flows -.21** -.14*** Net Non-Employment Table: Using non-missing labor productivity data. N = 48 High wage Firms - Low wage Firms Dependent Variables First difference Deviation from HP trend Net Job Flows *** Net Job-to-Job Flows -.30** -.22*** Net Non-Employment.26* -.14** 46 / 46