Effects of Employment Protection Legislation on wages: a regression - - PowerPoint PPT Presentation

Prepared with SEVISLIDES

Effects of Employment Protection Legislation on wages: a regression discontinuity approach

Marco Leonardi (Universit` a di Milano and IZA) Giovanni Pica (Universit` a di Salerno and CSEF) LA VALUTAZIONE DELL’IMPATTO DI INTERVENTI PUBBLICI: METODI E STUDI DI CASO Firenze 18-19 Gennaio 2007

➪ ➲ ➪

Summary ➟ ➠ ➪

• Motivation ➟➠
• The evolution of Italian EPL ➟
• Identification: Regression Discontinuity Design ➟
• Endogeneity of treatment status ➟
• Data Description ➟
• Regression model ➟
• IV model ➟➠
• Results ➟
• Summary and preliminary conclusions ➟

➪ ➲ ➪ ➟ ➠

Motivation (1/2) ➣➟ ➪

• The paper estimates the effect of EPL on entry wages and on the

tenure-wage profile.

• Lazear (1990): the transfer part of EPL may be undone by an appropri-

ately designed wage contract = ⇒ EPL has no real effects (Mantained hp: no contractual nor markets imperfections).

• Guell (2003): in an efficiency wages framework, severance payments

raise wages.

• Wage effects of EPL also important to discern to GE effect of firing

costs (Ljungqvist (2002) and Garibaldi and Violante (2005)).

➪ ➲ ➪ ➟ ➣ ➥

1 27

Motivation (2/2) ➢ ➟ ➪

Empirical strategy:

• Focus on Italy and exploit two sources of variation of EPL: over time

(1990 reform) and among firms of different size (above/below 15 employees).

• Previous work on Italy: effects of EPL on employment, firm size and

job flows (Boeri and Jimeno (2003), Borgarello, Garibaldi and Pacelli (2002), Schivardi and Torrini (2004), Kugler and Pica (2006)).

➪ ➲ ➪ ➟ ➥ ➢

2 27

GE effects of the tax component of EPL (1/2) ➣ ➱ ➪

• What is the effect of the tax component of EPL on employment and

unemployment?

• Partial equilibrium analysis (Bentolila-Bertola 1990, Bertola 1990):

ambiguous effects

• Early general equilibrium analysis (Burda 1992, Hopenhayn and Roger-

son 1993, Saint-Paul 1995): employment down

• Later general equilibrium analysis (Alvarez and Veracierto 1998, Mortensen

and Pissarides 1999): employment up

➪ ➲ ➪ ➟ ➣ ➥ ➱

3 27

GE effects of the tax component of EPL (2/2) ➢ ➱ ➪

• Ljungqvist (2002) shows that:
• In models with employment lotteries, employment goes down due

to the diminished private return to work

• In matching models where workers’ relative share of surplus goes up

with firing costs, unemployment goes up

• In matching models with constant split of the match surplus (and in

search models), unemployment goes down (frictional unemployment down)

• In a two-tier wage system (Mortensen and Pissarides 1999), if the

relative split of the match surplus among the firm and unemployed workers is constant, unemployment goes down.

➪ ➲ ➪ ➟ ➥ ➢ ➱

4 27

Two-tier labour market ➪

• The firm does not incur any firing cost if it does not reach an agreement

in the first encounter. In contrast the firm’s threat point is weakened in future negotiations because of the firing cost τ. In this case: w1(p) = w(p) − βδτ w+(p) = w(p) + rβδτ

• The wage is reduced in the first period by the worker’s share δ of any

future lay-off tax, and future wages are increased by an amount equal to the net interest on the ’posted’ bond.

• This two-tier specification is equivalent to one with constant split of the

match surplus in any period (Ljungqvist 2002), where unemployment goes also down.

• Ljungqvist 2002 models the firing cost as a tax. Of course, if it was a

pure transfer it would have no real effects (Lazear 1990).

➪ ➲ ➪

GE effects of the transfer component of EPL ➱ ➪

• Garibaldi and Violante (2005) show that in a two-tier matching model:
• Under full wage rigidity transfers and taxes are equivalent (i.e. same

real effects).

• If only the wages of the outsiders are flexible EPL lowers unemploy-

ment.

• If only the wages of the insider are flexible EPL has an ambiguous

effect on unemployment.

➪ ➲ ➪ ➟ ➱

6 27

The evolution of Italian EPL ➟ ➪

Individual dismissals are costless either in case of misconduct or for eco- nomic reasons. In case of dismissal, the worker has the right to go to

• court. The judge decides whether the dismissal is unfair.

The Italian legislation:

• Statuto dei Lavoratori, 1970
• Firms > 15 employees: reinstate the unjustly dismissed worker and

pay foregone wages

• Firms ≤ 15 employees: exempt
• L. 108, 1990
• Firms > 15 employees: no change
• Firms ≤ 15 employees: pay between 2.5 and 6 months pay

Although, after 1990, EPL is still stricter in firms above 15 employees, the reform narrows the gap between firms above and below 15 employees.

➟ ➠ ➪ ➲ ➪ ➟

7 27

Identification: Regression Discontinuity Design ➟ ➪

• RDD: a dichotomous treatment that is a deterministic function of a

single variable.

• Randomized experiment in a neighborhood of the threshold.
• Compare wages paid by firms just below 15 employees to wages paid

by firms just above 15 employees before and after the reform:

• Wages of new hires.
• Wages of displaced workers.
• Identification assumptions:
• Any other variable affecting wages is continuous in firm size.
• The average outcome for individuals in firms marginally above the

threshold is a valid counterfactual for individuals in firms just below the threshold

➟ ➪ ➲ ➪ ➟

8 27

Wage-firm size in 86-94: new hires ➱ ➪

Wages of new hires in years 86−89

Firm size Firms below 15 Firms above 15 1 15 30 490 670

Wages of new hires in years 91−94

Firm size Firms below 15 Firms above 15 1 15 30 490 670

Wages of new hires in years 95−97

Firm size Firms below 15 Firms above 15 1 15 30 490 670

➟ ➪ ➲ ➪ ➟ ➱

9 27

Wage-firm size in 86-94: displaced workers ➱ ➪

Pre−reform wages of displaced workers

Firm size Firms below 15 Firms above 15 1 15 30 490 670

Wages of displaced workers in years 91−94

Firm size Firms below 15 Firms above 15 1 15 30 490 670

Wages of displaced workers in years 95−97

Firm size Firms below 15 Firms above 15 1 15 30 490 670

➟ ➪ ➲ ➪ ➟ ➱

10 27

Endogeneity of treatment status ➟ ➪

• Are otherwise identical firms (and workers) randomly assigned to treat-

ment?

• If there are benefits to receiving the treatment, those who gain se-

lect themselves into the treatment group. Therefore both firms and workers may sort above and below the threshold according to their preferences.

• Firms’ sorting
• Workers’ sorting

➟ ➪ ➲ ➪ ➟

11 27

Firms’ sorting ➱ ➪

• Firms keeping their size just below 15 before the reform to avoid strict

EPL rules, may have increased their size because of the reform.

• Previous work (Schivardi and Torrini (2004) and Borgarello, Garibaldi

and Pacelli (2003)) shows that the 15 employees threshold reduces by 2% firms propensity to grow.

• Simple inspection of the distribution of firm size before and after the

reform is not insightful.

• Regression analysis shows evidence of sorting.

Better firms are less affected by the threshold.

• Solution: IV strategy.

➟ ➪ ➲ ➪ ➟ ➱

12 27

Firm size distribution ➱ ➪

.05 .1 .15 .2 Density 5 15 25 Firm Size

Pre−Reform Firm size distribution

.05 .1 .15 Density 5 15 25 Firm Size

Post−Reform Firm size distribution

➟ ➪ ➲ ➪ ➟ ➱

13 27

Evidence on firms’ sorting ➱ ➪

• 0.008
• 0.007
• 0.005

(0.008) (0.011) (0.011)

• 0.014
• 0.021
• 0.022

(0.009) (0.011) (0.012)

• 0.013
• 0.029
• 0.032

(0.009) (0.012) (0.012)

• 0.049

0.057

• (0.007)

(0.008)

• 0.002

0.009

• (0.015)

(0.017)

• 0.016

0.034

• (0.016)

(0.018)

• 0.035

0.054

• (0.018)

(0.02)

• 0.280
• (0.025)
• 0.083
• (0.105)
• 0.062
• (0.114)
• 0.017
• (0.118)
• 0.090
• (0.037)
• 0.105
• (0.086)
• 0.360
• (0.177)
• 0.455
• (0.19)

N 104061 104061 97858 Post 1990 × Firms Fixed Effect (× 1000) Post 1990 × Firms Fixed Effect × Dummy 13 (× 1000) Post 1990 × Firms Fixed Effect × Dummy 14 (× 1000) Post 1990 × Firms Fixed Effect × Dummy 15 (× 1000) Dummy 15 Post 1990 Post 1990 × Dummy 13 Post 1990 × Dummy 14 Post 1990 × Dummy 15 Firms Fixed Effect (× 1000) Firms Fixed Effect × Dummy 13 (× 1000) Firms Fixed Effect × Dummy 14 (× 1000) Firms Fixed Effect × Dummy 15 (× 1000) Dependent Variable: employment growth dummy Dummy 13 Dummy 14

➟ ➪ ➲ ➪ ➟ ➱

14 27

Workers’ sorting ➱ ➪

• Do workers move differently before and after the reform?
• Conditional on moving no evidence of sorting: Graph.
• Find evidence that better workers move more from small to large

firms after the reform: table

• To minimize workers’ sorting we take a sample of exogenously dis-

placed workers.

➟ ➪ ➲ ➪ ➟ ➱

15 27

Workers’ unconditional probability of moving ➱ ➪

Dependent Variable: mover dummy 0.026 0.015 0.004 0.000 (0.001) (0.001) (0.001) (0.001) 0.031 0.022 0.004 0.000 (0.001) (0.001) (0.001) (0.001) 0.011 0.000 0.004 0.001 (0.001) (0.001) (0.001) (0.001)

• 0.008

0.0001 0.041 0.041 (0.001) (0.002) (0.001) (0.001)

• 0.029
• 0.025
• (0.005)
• (0.004)
• 0.043
• 0.014
• (0.012)
• (0.009)
• 0.008
• 0.017
• 0.004
• 0.007

(0.002) (0.002) (0.001) (0.002)

• 0.011
• 0.019
• 0.006
• 0.007

(0.002) (0.002) (0.001) (0.002) 0.007 0.003

• 0.006
• 0.008

(0.002) (0.002) (0.001) (0.002)

• 0.007
• 0.003
• (0.006)
• (0.005)
• 0.028
• 0.001
• (0.007)
• (0.005)
• 0.003
• 0.003
• (0.007)
• (0.005)
• 0.001
• 0.014
• (0.015)
• (0.012)
• 0.058
• 0.001
• (0.016)
• (0.012)
• 0.057
• 0.002
• (0.016)
• (0.012)

N 1603117 1412427 1603117 1412427 P >15 P ≤ 15 Workers FE × Dummy 1990 × Small Firm Dummy (× 1000) Workers FE × Dummy 1991 × Small Firm Dummy (× 1000) Small firm dummy × Dummy 1991 Dummy 1991 Small firm dummy Small firm dummy × Dummy 1989 Workers Fixed Effect (× 1000) Workers Fixed Effect × Small firm dummy (× 1000) Workers Fixed Effect × Dummy 1989 (× 1000) Small firm dummy × Dummy 1990 Dummy 1989 Dummy 1990 Workers Fixed Effect × Dummy 1990 (× 1000) Workers Fixed Effect × Dummy 1991 (× 1000) Workers FE × Dummy 1989 × Small Firm Dummy (× 1000)

➟ ➪ ➲ ➪ ➟ ➱

16 27

Workers’ conditional probability of moving ➱ ➪

av_emplpre (mean) dummypre Linear prediction Linear prediction 1 15 35 1 av_emplpost (mean) dummypost Linear prediction Linear prediction 1 15 35 1 av_emplpre (mean) dummypre_15 Linear prediction Linear prediction 1 15 35 1 av_emplpost (mean) dummypost_15 Linear prediction Linear prediction 1 15 35 1

➟ ➪ ➲ ➪ ➟ ➱

17 27

Data Description ➟ ➪

• Dataset from Italian Social Security records:

each firm and worker employed in the private sector located in Vicenza and Treviso.

• North-eastern part of Italy: small firm size and tight labour markets.

In 2000 GDP per capita was 22.400e, 20% higher than the national average.

• Sample Selection rules:
• Permanent male workers in years 1986-1994 (year 1990 removed)
• Firms between 10 and 20 employees
• Dependent variable: Weekly Wage = Yearly Wage / N. Paid Weeks

Descriptive statistics for displaced male workers

➟ ➪ ➲ ➪ ➟

18 27

Descriptive statistics: displaced male workers ➱ ➪

Table 1: Descriptive statistics (Displaced Workers) Variables Small firms Large firms Small firms Large firms Diff-in-Diff 6.271 6.267 6.372 6.431

• 0.063

(0.237) (0.237) (0.234) (0.284) [0.013] 36.948 36.680 38.311 39.299 (9.37) (9.727) (8.995) (9.163) 0.142 0.111 0.168 0.204 (0.35) (0.314) (0.374) (0.403) 12.342 17.742 12.386 17.930 (1.707) (1.402) (1.715) (1.499) 0.821 0.973 2.634 2.558 (0.96) (1.01) (2.391) (2.333) N 1601 1019 2178 1300 Pre-reform Post-reform Log Wage Age Notes: Only firms between 10 and 20 workers are included. Standard deviations in parentheses. White collars % Firm size Tenure

➟ ➪ ➲ ➪ ➟ ➱

19 27

Regression model ➟ ➠ ➪

• If firms and workers are exogenously assigned to the treatment status,
• δ2 is the estimated causal effect of EPL on entry wages, and

α3 the estimated causal effect of EPL on the tenure-wage profile: log wijt = β′Xijt + δ0Post + δ1DS

jt + δ2

• DS

jt × Post

• +

+α0Tenijt + α1

• Tenijt × DS

jt

• + α2
• Tenijt × Post
• +

+α3

• Tenijt × DS

jt × Post

• + uijt

DS

jt

= 1 [firm size ≤ 15 in year t] Post = 1 [year ≥ 1991]

• The matrix Xijt includes a polynomial in firm size.
• Xijt may also include time effects, worker characteristics (a quadratic

in age; white collar/blue collar dummy) and firm characteristics (ge-

• graphical location, industry, net job-creation rate NJCjt = 2(ejt −

ejt−1)/(ejt + ejt−1).

➟ ➪ ➲ ➪ ➟➠

20 27

IV model (1/2) ➣➟ ➪

• Firms keeping their size just below 15 before the reform to avoid strict

EPL rules, may have increased their size because of the reform.

• Instruments: size dummies in 1987, 1988 and 1989:

log wijt = β′Xijt + δ0Post + δ1DS

jt + δ2

• DS

jt × Post

• +

+α0Tenijt + α1

• Tenijt × DS

jt

• + α2
• Tenijt × Post
• +

+α3

• Tenijt × DS

jt × Post

• + vijt

DS

jt

= γ′

0Xijt + γ1Post + γ′ 2DS jpre + γ′ 3

• DS

jpre × Post

• + γ4Tenijt

+γ′

5

• Tenijt × DS

jpre

• + γ6
• Tenijt × Post
• +

+γ′

7

• Tenijt × DS

jpre × Post

• + νijt
• All terms interacted with DS

jt (i.e. DS jt × Post, Tenijt × DS jt and Tenijt ×

DS

jt×Post) are also instrumented using the interaction with DS jpre, which

is correlated with size in other years but not affected by the reform.

➟ ➪ ➲ ➪ ➟ ➣ ➥

21 27

Results ➟ ➪

• 1. OLS and IV on the sample of new hires.
• 2. OLS and IV on the sample of displaced workers.
• 3. Results by sector and occupation.

➟ ➠ ➪ ➲ ➪ ➟

22 27

New Hires: OLS and IV Estimates ➱ ➪

(1) (2) (3) (4) (5) Dependent Variable: log wage 0.038 0.020 0.032 0.033 0.015 (0.003) (0.003) (0.003) (0.003) (0.003) 0.059 0.091 0.020 0.033 0.039 (0.004) (0.005) (0.012) (0.013) (0.012) 0.013 0.015 0.169 0.038

• 0.183

(0.005) (0.005) (0.123) (0.132) (0.117)

• 0.003
• 0.002

0.043 0.025 0.001 (0.005) (0.004) (0.02) (0.021) (0.019)

• 0.015
• 0.002
• 0.011
• 0.011

0.002 (0.003) (0.003) (0.004) (0.004) (0.003)

• 0.0003
• 0.0001

0.009 0.010 0.011 (0.003) (0.003) (0.004) (0.004) (0.004)

• 0.002
• 0.002
• 0.008
• 0.008
• 0.009

(0.003) (0.003) (0.005) (0.005) (0.005) N 122954 122954 89248 87989 87989 Job creation/job destruction NO NO NO YES YES Additional controls NO YES NO NO YES Notes: Only firms between 10 and 20 workers are included. Robust standard errors in parentheses OLS IV Post 1990 × Tenure Small firms × Tenure Small firms × Post 1990 × Tenure Table 4: New Hires in years 1986-1994 (excl. 1990). OLS and IV estimates. Post 1990 × Small firms Post 1990 Small firms Tenure

➟ ➠ ➪ ➲ ➪ ➟ ➱

23 27

Displaced workers: OLS and IV Estimates ➱ ➪

(1) (2) (3) (4) (5) Dependent Variable: log wage 0.015

• 0.004

0.013 0.010 0.000 (0.008) (0.009) (0.01) (0.01) (0.01) 0.160 0.135 0.219 0.194 0.106 (0.018) (0.016) (0.04) (0.037) (0.038)

• 0.020
• 0.023
• 0.357
• 0.208

0.037 (0.021) (0.019) (0.196) (0.21) (0.188)

• 0.040

0.002

• 0.132
• 0.098
• 0.002

(0.021) (0.018) (0.061) (0.059) (0.053)

• 0.008

0.009

• 0.010
• 0.006

0.006 (0.008) (0.009) (0.011) (0.011) (0.011) 0.030 0.035 0.033 0.033 0.029 (0.01) (0.008) (0.013) (0.013) (0.012)

• 0.030
• 0.038
• 0.027
• 0.029
• 0.032

(0.01) (0.009) (0.015) (0.015) (0.013) N 6098 6098 4702 4638 4638 Job creation/job destruction NO NO NO YES YES Additional controls NO YES NO NO YES Table 6: Displaced workers in years 1986-1994 (excl. 1990). OLS and IV estimates. Post 1990 × Small firms Post 1990 Small firms Tenure Notes: Only firms between 10 and 20 workers are included. Robust standard errors in parentheses OLS IV Post 1990 × Tenure Small firms × Tenure Small firms × Post 1990 × Tenure

➟ ➠ ➪ ➲ ➪ ➟ ➱

24 27

Results by sector and occupation ➱ ➪

• 0.002
• 0.129
• 0.081

(0.053) (0.017) (0.013)

• 0.032
• (0.013)
• 0.102
• 0.078
• 0.032

(0.131) (0.032) (0.023)

• 0.008
• (0.03)
• 0.143
• 0.143
• 0.055

(0.115) (0.036) (0.025)

• 0.025
• (0.04)
• 0.234
• 0.119
• 0.124

(0.157) (0.033) (0.029) 0.068

• (0.049)
• 0.041
• 0.124
• 0.062

(0.053) (0.019) (0.014)

• 0.035
• (0.014)
• Log wage

(Displaced workers) Accession dummy Separation dummy Dependent Variable Notes: The dependent variables in columns 3 and 4 are, respectively, a dummy that takes the value of 1 if a 42124 42124 19015 19015 27306 27306 N 3894 Small firms × Post 1990 × Tenure All Sectors Manufacturing Constructions Post 1990 × Small Firm Small firms × Post 1990 × Tenure Post 1990 × Small Firm Small firms × Post 1990 × Tenure Post 1990 × Small Firm White Collars Blue Collars N 1189 N 634 N 673 Post 1990 × Small Firm Small firms × Post 1990 × Tenure 129595 129595 N 4638 156901 156901 Post 1990 × Small Firm Small firms × Post 1990 × Tenure

➟ ➠ ➪ ➲ ➪ ➟ ➱

25 27

Summary and preliminary conclusions ➲ ➪

• We identify the causal effect of EPL on entry wages and on the tenure-

wage profile.

• RD Design:

compare the change in mean wages paid by firms just below 15 to the change in mean wages paid by firms just above 15, before and after the 1990 reform.

• As endogenous sorting of workers and firms may bias the results:
• Use plant closures to select exogenously displaced workers.
• Use size dummy in 1987, 1988 and 1989 as instruments for the

small/large dummy.

• The 1990 EPL reform lowered males the tenure-wage profile by about

3%.

➟ ➪ ➲ ➪

26 27

Prepared with SEVISLIDES

Effects of Employment Protection Legislation on wages: a regression discontinuity approach

Marco Leonardi (Universit` a di Milano and IZA) Giovanni Pica (Universit` a di Salerno and CSEF) LA VALUTAZIONE DELL’IMPATTO DI INTERVENTI PUBBLICI: METODI E STUDI DI CASO Firenze 18-19 Gennaio 2007

➪ ➲ ➪