CREATIVE DESTRUCTION & JOB MOBILITY: FLEXICURITY IN THE LAND OF SCHUMPETER

Transcription

1 CREATIVE DESTRUCTION & JOB MOBILITY: FLEXICURITY IN THE LAND OF SCHUMPETER Andreas Kettemann, University of Zurich Francis Kramarz, CREST-ENSAE Josef Zweimüller, University of Zurich OECD, Paris February 1, /01/2017 1/ 53

2 MOTIVATION Werner D., close to 50 years old, is a typical middle-class, white-collar worker in a multinational company, prime-earner of his family, financially burdened with the mortgage of his house. He hates his job, with all its side effects from high blood pressure to skin rash. Nevertheless, he will not quit. You don t throw 750,000 bucks (50,000 Euros) of severance pay out of the window. Der Standard, June 5, /01/2017 2/ 53

3 MOTIVATION High firing costs (typically large severance payments) are main source of low labor mobility and segmented labor markets. Workers are reluctant to quit, firms are reluctant to hire workers on permanent contracts. Southern Europe: two-tiered labor markets with permanent and temporary contracts ( insiders and outsiders ). Outsiders bear burden of adjustment in bad times. One possible reform: replace severance pay by occupational pensions. 31/01/2017 3/ 53

4 Severance-pay versus occupational-pension systems Severance pay system ) one-time payment by the firm to the worker ) CONDITIONAL ON A LAYOFF (a firing cost). Occupational pension system ) monthly payment by the firm into a pension account owned by the worker. Transfered to the new employer ) BOTH AFTER A LAYOFF AND AFTER A QUIT. In severance pay systems, workers eligible for a severance pay have an incentive to wait (for a layoff). Do severance-pay systems generate less job mobility? 31/01/2017 4/ 53

5 This paper evaluates Austria s 2003 severance-pay reform ( Abfertigung Neu ) Reform abolished old severance-pay (SP) system in favor of a new occupational-pension (OP) system. Reasons for introducing the new system: Align regulations with reality of modern labor markets lifetime jobs increasingly uncommon high financial burden of SPs often lead to bankcrupcy firms (and workers) exploit the system reduce inequalities (many workers were not eligible to SPs) 31/01/2017 5/ 53

6 The OLD system: severance pay (SP) Jobs starting before 2003 were subject to the mandatory SP system. After a layoff, firms had to pay a lump-sum amount to the worker. Quits and layoffs for cause (misconduct,...) were not eligible. Quits for retirement remained eligible (with 10+ years of tenure). Severance payment amount depended on tenure: 25 years tenure and more: 12 monthly salaries 20 to 24.9 years tenure: 9 monthly salaries 15 to 19.9 years tenure: 6 monthly salaries 10 to 14.9 years tenure: 4 monthly salaries 5to9.9yearstenure: 3monthlysalaries 3to4.9yearstenure: 2monthlysalaries 0to2.9yearstenure: noteligible 31/01/2017 6/ 53

7 The NEW System: Occupational pensions (OP) New labor contracts job starting on January 1, 2003 or later were subject to the new OP system. Ongoing contracts still covered under old SP scheme. Employer has to transfer 1.53 % of monthly salary to a pension account, on which the employee earns interest. If the job is terminated either through a layoff or a quit, the worker still owns the account and transfers it to a new employer; and can withdraw a limited amount from the account. Continuous OP contributions avoids liquidity problems for (small) firms confronted with lump-sum SP claims. 31/01/2017 7/ 53

8 This paper studies the effects of Austria s 2003 SP reform on job mobility (and its consequences for labor reallocation across firms and industries) Two changes are of main importance: 1. eligibility rules: voluntary quitters do not lose their claims 2. switch from a discontinuous SP scheme to a continuous OP scheme 31/01/2017 8/ 53

9 Severance-pay vs occupational-pension schedules Sev. Payment/Prev. Wage Tenure (Years) Before After (3% annual interest rate) 31/01/2017 9/ 53

10 Relation to previous literature Theoretical papers on severance pay Bentolila, Bertola (1990), Bertola (1990), Lazear (1990),..., Postel-Vinay, Turon (2014), Parsons (2014), Boeri, Garibaldi, Moen (2015) Emipirical literature Distressed firms and early leavers : Hamermesh and Pfann (2001), Schwerdt (2011),... (and earlier literature on effects of job loss: Jacubson, Lalonde and Sullivan 1993, Stevens 1997, etc. ) Firing costs and job durations: Garibaldi and Pacelli (2008),... Severance pay and retirement: Manoli and Weber (2016) UI and liquidity: severance pay is cash-on-hand and reduces liquidity constraints for unemployed workers. Chetty et al (2007), Chetty (2008) /01/ / 53

11 DATA EMPIRICAL STRATEGY 31/01/ / 53

12 Data Combine social security data (ASSD) with tax records (ATD, Austrian tax records); linked via anonymized social security ID. ATD covers all tax files (except public servants) over the period Has information on actual severance payment (which are subject to a reduced tax rate). ATD earnings are not top-coded. ASSD is a matched firm-worker data set, covers all private sector workers ( 85 percent of work force) over the period , tenure and eligibility for severance pay. ASSD earnings are top-coded. 31/01/ / 53

13 Percentage job separations receiving severance pay 31/01/ / 53

14 Empirical strategy: RDD RD Design: compare workers who started job immediately before January 1, 2003 (old system) to those who started on January 1, 2003 or immediately after (new system). Do workers under the old system wait for a layoff? Do workers under the new system leave earlier? Do they end up in new jobs more quickly? Basic sample: Workers who expect to be laid off (i.e. who expect a negative shock to their firm). Shock = mass layoff 31/01/ / 53

15 Empirical strategy: RDD Notice: no clean causal analysis Treatment (= shock to a firm) cannot be directly observed. Approximated by firms employment reduction, which is an outcome. The way firms (and workers in our sample) are selected might be affected by the reform. Sensitivity analysis: alternative definitions of a mass layoff, alternative definitions of baseline sample, alternative outcomes. Qualitatively, results are very robust. Use a structural model to check whether a standard DMP model (with on-the-job search) can replicate job mobility patterns under realistic parameter values (preliminary). 31/01/ / 53

16 Preview of results Workers under the new system: 1. are more likely to leave their firm in anticipation of a shock to their firm. 2. move more quickly to new jobs (smaller effect on J-U transitions). 3. have lower wage increases in job-to-job moves. 4. do not respond when their firm is not distressed. 5. are not less likely to wait for retirement. Effects are robust. 31/01/ / 53

17 RESULTS: Early leavers before a mass layoff 31/01/ / 53

18 Percentage early leavers : Old versus new system We look at firms with 30 workers that reduce employment within the next months by 33 percent or more ( baseline sample ) We select workers who (i) entered 3-4 years before the mass layoff (ii) started their job between Jan 1, 1997 and Dec 31, 2008 (iii) have at least 12 months of tenure in the distressed firm Focus on short/medium-tenured workers mitigates endogeneity problems Workers with tenure apple 12 months are mainly fixed-term contracts Sample size: 28,099 workers 31/01/ / 53

19 More job separations within 36 months after entry Exited by 36 months after entry q1 2000q1 2003q1 2006q1 2009q1 Start of Job discontinuity coef and t-stat.: (3.2461) 31/01/ / 53

20 Many more transitions to new jobs (J-J) Exited by 36 months after entry q1 2000q1 2003q1 2006q1 2009q1 Start of Job discontinuity coef and t-stat.: (3.9363) 31/01/ / 53

21 Somewhat more transitions to unemployment (J-U) Exited by 36 months after entry q1 2000q1 2003q1 2006q1 2009q1 Start of Job discontinuity coef and t-stat.: (1.8400) 31/01/ / 53

22 RDD analysis We estimate regressions of the form y i = (x i 0)+ 2x i + 31(x i 0)x i + " i where x i denotes the start date of the job (normalized so that 0 corresponds to the time of the reform) and y i is the outcome variable under consideration. We give more weight to observations close to the cutoff by using a triangular kernel (Porter, 2003; Hahn et al., 2001) Standard errors are clustered at the firm level 31/01/ / 53

23 The reform effect on job mobility, RD estimates 31/01/ / 53

24 How does the effect build up with tenure? 31/01/ / 53

25 RDD validity: Density of new hires, by quarter 31/01/ / 53

26 RDD validity: Firms work force composition 31/01/ / 53

27 RDD validity: Are covariates smooth? 31/01/ / 53

28 RDD validity: Placebo reforms 31/01/ / 53

29 ROBUSTNESS: Alternative mass layoff definitions 31/01/ / 53

30 ROBUSTNESS: Matched controls 31/01/ / 53

31 ROBUSTNESS: RDD diff-in-diff, matched controls 31/01/ / 53

32 ROBUSTNESS: By (i) gender, (ii) alternative mass-layoff definition 31/01/ / 53

33 ROBUSTNESS: By (i) mass layoff size, (ii) hiring dates, (iii) firm size before mass layoff 31/01/ / 53

34 FURTHER RESULTS: Wage growth for job changes Only job-to-job transitions (fewer observations) Outcome: Log earnings (new firm) - Log earnings (mass-layoff firm) 31/01/ / 53

35 FURTHER RESULTS: Waiting for retirement? Job starters older than 45 years Outcome: Still employed in the firm 10 years later? 31/01/ / 53

36 Summary of reduced form analysis Austrian severance-pay reform increased turnover on the labor market Substantially higher job mobility of workers expecting a layoff Also more transitions to unemployment (misclassification?) No effects on worker in non-distressed firms (matched controls) No effects on workers close to retirement Effects are quite robust 31/01/ / 53

37 ASTRUCTURALMODEL Standard search and matching (DMP) framework with on-the-job search 31/01/ / 53

38 Structural model Search and matching model with on-the-job search. Firm-worker pairs are subject to stochastic productivity shocks. Vacant firms draw initial productivity from the unconditional distribution and meet unemployed or employed workers with endogenous probability. Workers start out ineligible to severance pay and turn eligible with constant probability. Wages are fixed by Nash bargaining. 31/01/ / 53

39 Structural model Firms can dissolve matches but have to pay to eligible workers. Exogenous dissolution of matches with probability. Employed workers receive outside offers from vacant firms with endogenous probability f ( ) and decide whether to accept them. Workers incur moving costs if there is an U-J or J-J transition. 31/01/ / 53

40 Structural model In equilibrium Workers decide whether to accept (outside) job offers. Firms decide whether or not to terminate the employment relationship. Firms open vacancies so that the free-entry condition holds. The distribution of productivity among firms with eligible and non-eligible workers is stationary. 31/01/ / 53

41 Structural model: Targeted moments 31/01/ / 53

42 Structural model: job-to-job moves by workers 31/01/ / 53

43 Structural model: firm exit behavior 31/01/ / 53

44 Structural model: Estimation Model is estimated by Simulated Method of Moments. Targeted moments are Monthly exit rate to new jobs Monthly exit rate to unemployment Macro moments: probability of firm closure, unemployment rate, labor market tightness, job-finding rates, hiring cost share. 31/01/ / 53

45 Structural model: Fixed parameters 31/01/ / 53

46 Structural model: Estimated parameters 31/01/ / 53

47 Model versus data 31/01/ / 53

48 Model versus data 31/01/ / 53

49 Counterfactuals: Mimicking a Southern European labor market SCENARIO I: basic model, vary parameters Immediate access to severance pay = 2months(insteadof = 36) Higher severance pay = 3 monthly earnings (instead of = 2) (Wage rigidity) SCENARIO II: extended model, with temporary jobs 31/01/ / 53

50 Counterfactual I: Immediate access to severance pay 31/01/ / 53

51 Counterfactual II: Model with temporary jobs 31/01/ / 53

52 Conclusions I The Austrian 2003 policy change abolished a severance pay system in favor of an occupation pension system Evaluation using RDD for identification (compare workers by start date before and after January 1, 2003 Workers in distressed firms are more likely to move to new jobs No effects on workers in non-distressed firms, no effect on job mobility close to retirement 31/01/ / 53

53 Conclusions II Structural search and matching model with on-the-job search Model predicts job separations and J-J transitions quite well Counterfactual policy experiments suggest that both (i) high level of severance pay and (ii) immediate access reduce job creation and willingness to accept outside job offers Model generates high unemployment equilibrium unemployment with generous severance pay (But: no positive role of severance pay in the model) 31/01/ / 53