Paul Bingley SFI Copenhagen. Lorenzo Cappellari. Niels Westergaard Nielsen CCP Aarhus and IZA

Transcription

1 Flexicurity and wage dynamics over the life-cycle Paul Bingley SFI Copenhagen Lorenzo Cappellari Università Cattolica Milano and IZA Niels Westergaard Nielsen CCP Aarhus and IZA 1

2 Motivations Flexycurity (no firing costs + high unemployment insurance [UI] benefits) often advocated as a cure for sclerotic labour markets Some countries have gone half the way, cutting down rigidities iditi but without t introducing i UI. UI currently discussed in e.g. Italy. Cons : disincentives in job search for the unemployed Pros : allow longer/better search that should result in more stable post-unemployment t jobs Perhaps less known are the effects that the presence of such systems may exert on employed individuals We look at the latter by studying individual wage profiles in the presence of UI in Denmark 2

3 Motivations In particular, the life-cyclel wage profile conveys information on three aspects of the employment relationship that may be influenced by the fact that someone is covered by UI 1. Initial wage Remuneration of education 2. Wage growth Skill accumulation 3. Wage instability Employment instability We estimate models of life-cycle wage in which these three aspects are related to individual membership of UI schemes 3

4 This paper Uses register data (linked employer-employee) for danish men to characterise life-cycle wage profiles in the presence of UI. Distinguishes between long-term wage components and volatile shocks Long-term wage as a Random growth process that allows for entry wage heterogeneity and heterogeneity in wage growth Finds that when workers are covered by UI: 1. Workers at the top (bottom) of the initial wage distribution are shifted downward (upward) 2. High growth workers slow down relative to lower growth ones 3. wage instability (=shocks volatility) is higher Finding 1. speaks about a reduced signalling value of education Findings 2. and 3. are consistent with moral hazard effects associated with UI We do not have formal tests of moral hazard against selection, but discuss reasons why the former seems more plausible. 4

5 Outline of talk Institutional background Literature review Models and estimation Data Results Conclusion 5

6 Institutional background: the Danish flexycurity system The foundation of flexicurity in Denmark is a system with relatively low employment protection Essentially non-existent notice for laid-offs Coupled with high UI payments UI organized according to Bismarckian principles, which means that workers can voluntarily choose to join, paying contributions to insurance funds. Eligibility for unemployment benefits is conditional on the person s s having held a job for at least one year and of membership for the same period. Important: membership is time varying, typically workers start paying contributions after some years of labour market participation (see data) 6

7 Institutional background Flexibility to hire and fire in place since the mid 1970s Combined with effectively unlimited unemployment benefit duration until1993. Thereafter introduction and tightening of time limits and activation. In this paper we focus on the benefit side of flexicurity 7

8 Institutional background Daily UI =90%*(Wage in the last 65 days 8%)/65 Up to a maximum monthly payment of 1,800. Benefits are taxed, but the effective rate is low Low-wage workers have a replacement ratio of 90 percent (lower for higher paid) Laid-off workers are entitled to unemployment benefits from the third day of unemployment, for one year, without any other obligations than seeking work. 8

9 Institutional background Social assistance is available to those without work who are uninsured or those for whom unemployment insurance eligibility has expired, but at lower level (70% of UI benefit) Unemployment insurance funds are organised along occupational lines and run by unions, but union membership and fund membership are not bundled. Industry-level bargaining introduced in 1981 By 1987 (2003) only 34% (15%) of wages were centrally bargained. There was a minimum level of firm-level bargaining at around 4% until 1993, which increased to 21% by

10 Literature UI benefits and djob search Disincentive effects through higher reservation wages of the unemployed (e.g. Lalive et al 2006) Implication (macro): re-employment wage will increase (Andersen and Svarer 2007) Positive effects through longer search that allow better, more stable, post-unemployment matches (Tatsiramos, 2009; Caliendo et al., 2009; Centeno and Novo, 2009) 10

11 Individual wage dynamics Literature Long-term vs transitory components of earnings, widely used to study changing wage inequality (Gottschalk and Moffitt 2008) Interest in specifying the long-term component as a process with growth rate heterogeneity, the so-called Random growth model (Hause, 1979; Baker, 1997; Haider, 2001; Baker and Solon, 2003; Cappellari, 2004) Hause: Allows testing on-the-job training hypothesis Baker (1997) provides evidence that favours it against alternative models (random walks) Also interest in the evolution of the volatile component: relevant share of growing wage inequality in the US attributed to wage instability. 11

12 Wage models and estimation Canonical permanent-transitory wage model for cohort-time adjusted log-(hourly gross) wages w T ict = w P ict + wt ict ; E(w P T P T ict )= E(wT ic t )= E(wP ict,wt ict ) =0; i=1,,n; 1 t=tt 0c,, T c 12

13 wage models and estimation Base model Random growth permanent wage w P ict = π t λ c (α i +β i EXP it ); (α i,β i ) i (0,0; σ2 α, σ2 β, σ αβ ); Cov(w P ict w P ics EXP [ 2 i ) = [σ 2 α + σ2 β EXP itexp is + σ αβ (EXP it + EXP is )] π t π s λ 2 c 13

14 wage models and estimation w P σ αβ >0 (signalling+matching; σ αβ <0 (on the job training) training and education complements) w P Exp t =-σ 2 c σ αβ / σ β t c Exp 14

15 wage models and estimation Base model (Non-stationary) AR(1) transitory wage w T ict = τ t μ c v it v it =ρv it-1 + ε it ε ) it it (0; σ 2 ε) v it0c i (0; σ 2 0) Cov(w T ict w T ics) ={d 0c σ d dc [σ 2 ε ct + Var(v it-1 )ρ 2 ] +d 1 [Cov(v it-1 v i t-s ) ρ]}τ t τ s μ 2 c 15

16 wage models and estimation Extended model: Permanent wage w P ict = λ c π t ( (α i + β i EXP it + γ i F it + δ i F it EXP it ); (α 0); ( i, β i γ i δ i )~[(0,0); (σ 2 α σ2 β σ2 γ σ2 δ σ αβ σ αγ σ βδ )] Cov(w P ict wp ics EXP i F i ) = [σ2 α + σ2 β EXP it EXP is + σ )+σ 2 αβ (EXP it + EXP is σ γ F it F is + σ 2 δ F it EXP it F is EXP is + σ αγ (F it + F is ) + σ 2 βδ (F it EXP it + F is EXP is )] π t π s λ c 16

17 wage models and estimation Extended model: Instability σ 2 εct = σ2 ε exp(ψf ct ) 17

18 wage models and estimation GMM ( Minimum distance ) estimation based on the following identifying restriction E[a i - ω(θ, ( X i )]=0 Note 1: identifying restrictions do not require orthogonality between UI membership and wage residuals. In effect we model that correlation Note 2: estimation of the parameter on wage instability based on variation in membership across cohorts and time, while time and cohort effects in wage covariances controlled for by the non-parametric shifters. 18

19 Data Register data on gross hourly wage for the Danish labour force between 1980 and 2003 We focus on prime age men, aged 21-55, who are full time private sector employees. Revolving-by-cohort panel design (with min 10 year per-cohort) Youngest born 1973 Oldest born 1943 (not 1934 due to left censoring on labour market experience) 19

20 Data Table 1: Descriptive statistics Year Number of observations Average hourly Standard Age Unemployment Labour wage (DKron, 2000 prices) deviation hourly wage Insurance market experience Sample Labour force Sample Labour Sample Labour Sample Labour Sample Labour Sample (N=811651) (N= ) force force force force All years

21 Data Table 3: Unemployment insurance by birth cohort, proportions covered at least once Cohort U.I. % Cohort U.I. % All cohorts

22 Data Figure 1: Earnings covariances at various lags.1.15 Born 1945 Born 1950 Born Born 1960 Born 1965 Born Graphs by cohort Earnings Covariances at Various Lags cov0 cov1 cov3 cov5 22

23 Results Figure 2: Predicted variance components.1.15 Born 1945 Born 1950 Born Born 1960 Born 1965 Born Graphs by cohort Predicted Variance Components Total Variance Permanent Variance Instability 23

24 Table 4: Core parameter estimates for the earnings model Baseline Permanent component Transitory component Coeff. S.E. Coeff. S.E. σ 2 α σ 2 ε σ β σ σ αβ ρ t c =4.25 Results Main model with unemployment insurance Permanent component Transitory component Coeff. S.E. Coeff. S.E. σ 2 α σ 2 ε ψ σ αβ σ σ 2 γ ρ σ 2 δ σ 2 β ψ t c =2.03 β σ αγ σ βδ

25 Results w P Exp 25

26 Interpretation Results 1. Changes in the distribution of entry wages are consistent with a reduced signalling value of education. Note: at entry typically not covered, result essentially a backward projection 2. Changes in the distribution of growth rates consistent with moral hazard: once insured lose incentives es to accumulate skills 3. Changes in instability consistent with moral hazard: e.g. once insured lose incentives to avoid shirking 26

27 Results Alternative ti interpretations t ti based on adverse selection: Result 2: skill accumulation acts asself-insurance, and when the potential for wage growth starts diminishing, individuals become members of UI scheme. But typically membership occurs well before reaching wage peak. Result 3: Those with high h unemployment risks chose membership. But the vast majority is insured, it is not a matter of (time-invariant) heterogeneity. Result 3: Become insured when lay-off announced. We do not observe it in the data 27

28 Robustness checks Is the result on instability driven by inherent differences in employment/wage instability across different groups? Manual vs Non-manual (some period of notice) Manufacturing (more performance related pay) vs Other industries σ 2 εct = σ2 ε exp(ψ 1F 1 ct + ψ 2 F 2 ct ) 28

29 Robustness checks Table 5: Core parameter estimates for the earnings model with group specific instability insurance coefficients i A) Industry based groups Permanent Transitory component component Coeff. S.E. Coeff. S.E. σ 2 α σ 2 ε σ 2 β ψ σ αβ ψ σ 2 γ σ σ 2 δ ρ σ αγ σ βδ

30 Robustness checks Table 5: Core parameter estimates for the earnings model with group specific instability insurance coefficients i B) Occupation based groups ( ) Permanent Transitory component component Coeff. S.E. Coeff. S.E. σ 2 α σ 2 ε σ 2 β ψ σ αβ ψ σ 2 γ σ σ 2 δ ρ σ αγ σ βδ

31 Concluding remarks The properties of the wage profile interact with individual membership of UI scheme 1. Changes in the distribution of entry wages 2. Changes in the distribution of wage growth 3. Changes in wage instability Signalling (in principle) and moral hazard interpretations This paper shows that UI schemes may affect the behaviour of employed individuals, id not only search effort of the unemployed. 31