Andreas Ammermüller Bernhard Boockmann Michael Meier Thomas Zwick
-
Upload
bruno-stokes -
Category
Documents
-
view
24 -
download
1
description
Transcript of Andreas Ammermüller Bernhard Boockmann Michael Meier Thomas Zwick
1
The Effects of Hiring Subsidies for Older Workers on Unemployment Durations in Germany
Andreas Ammermüller
Bernhard Boockmann
Michael Meier
Thomas Zwick
Centre for European Economic Research (ZEW), Mannheim
2
Outline
(1) Introduction
(2) Description of hiring subsidies
(3) The data
(4) Estimation approach and implementation
(5) Results
(6) Conclusions
3
Two questions should be distinguished:
1. Do hiring subsidy programmes causally lead to earlier exit from unemployment to employment in the group of eligible persons as compared to the situation in which no subsidies are available?
2. Do subsidised hirings causally lead to more unsubsidised employment? Forslund et al. (2004), Sianesi (2003), and Hujer et al. (2002)
In this paper, the focus is on the first question
Introduction
4
• Static model with perfect competition on the labour market: hiring subsidy lowers net wages paid by the employer and increases demand for subsidised employees
• Reaction depends on wage elasticity: dlnL/ds = /(+) , where L is subsidised employment
• If elasticity is low, deadweight effects (Buslei and Steiner, 1999; Hujer and Caliendo, 2003; Meyer, 1995a) occur
• Empirically, wage elasticities differ between male/female and single/married and East/West Germany
Conditions for the effectiveness of hiring subsidies
5
• Hiring subsidies may provide too low an incentive and / or may not be known among eligible firms and workers implementation study
• In addition to deadweight effects, hiring subsidies may be ineffective due to substitution or a displacement effect
• Since our approach is based on individual-level data, we concentrate on deadweight effects. The particular interest in the deadweight effect is whether the programme is effective for the targeted group
More reasons why hiring subsidies may fail to affect the number of hirings
6
• Deadweight effects involve a counterfactual that must be
estimated
• In this paper we use changes to the eligibility rules as „natural“ variation
• Before 2002, the Integration Supplement for older workers was only available for hiring long-term unemployed workers. Taking effect on January 1st, 2002, this condition was dropped
• Taking effect on January 1st, 2004, the EGZ subsidy for older workers was integrated into the framework of the general EGZ, so that workers aged 50+ lost preferential treatment
Estimating deadweight effects
7
• One of the major instruments of German active labour market policy
• Legal basis: German Social Code (SGB), Volume III
• EGZ are paid to the employer as a percentage of standardised labour costs (maximum 50 per cent) for up to 24 months (other limits for workers with specific disadvantages)
• If the employment relationship is terminated before a minimum period after the expiration of the subsidy, the employer is legally obliged to refund parts of the subsidy
• No legal claim to EGZ either by the worker or the employer
Description of hiring subsidies (EGZ) in Germany
80
30,000
60,000
90,000
120,000
150,000
180,000
Jan-00 Jul-00 Jan-01 Jul-01 Jan-02 Jul-02 Jan-03 Jul-03 Jan-04 Jul-04 Jan-05 Jul-05
All
East
West
Employment contracts subsidised
by Integration Supplements
9
• EGZ is regarded as an important instrument by employment agencies
• There is substantial scope for decision-making concerning the allocation of the subsidy
• In the majority of cases, an initial contact between a worker and an employer already existed
• It will often be difficult for placement officers to decide whether the company would refrain from hiring without EGZ
• The implementation study strongly confirms the notion that deadweight effects are a major issue
Implementation of the programme
10
• Evaluation is based on the Integrated Employment Biographies (IEB)
• The IEB are composed of four separate data bases:
• Employment Register (BeH)
• Benefit Claimants Register (LeH)
• Programme-Participants Comprehensive Data Base (MTG)
• Job Applicant Files (ASU)
The data
11
The paper uses two legal changes in eligibility as natural variation
• First, before 1-1-2002 eligibility was limited to individuals aged 50+ who were either long-term unemployed (in the legal definition) or had been unemployed for more than 6 consecutive months; this criterion was dropped in 2002
• Hence, workers 50+ and not fulfilling the criterion are used as the treatment group
• Second, on 1-1-2004 EGZ subsidies for older workers were integrated into the general EGZ framework; workers 50+ lost preferential status
• Hence, workers 50+ are used as the treatment group
Estimation approach and implementation
12
Definition of age groups:
o workers aged 50 to 50+6 months at the time of entering unemployment (treatment group)
o workers aged between 49 and 49+6 months (control group)
• Treatment and control group are observed before and after the legal changes
• We observe individuals belonging to a 3-months entry cohort during a period of 180 days after entry in unemployment
Estimation approach and implementation
13
Time frame for the difference-in-differences analysis
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
2001/2003 2002/2004
Entry into unemployment
Entry into unemployment
Policy change
14
Three difference-in-difference estimators
(1) time varying unconditional effect obtained from Kaplan-Meier-Survivor functions (DD1):
(2) estimation of a PH model
and calculation of the treatment effect on the survivor function (DD2):
1 , 1 , 0 , 1 , 0( ) ( ) ( ) ( ) ( ) .h t h t k t k tDD S S S S
2 0 2 3 0 2ˆ ˆ( ) ( ) exp exp( ´ ) ( ) exp exp( ´ ) .h h h
t i iDD S d d z S d z
0 1 2 3( ) ( ) exp( ´ ).h ht t id d d z
15
Three difference-in-difference estimators
(3) time varying conditional effect calculated from the difference-in-differences of the baseline hazard rate of the Cox partial likelihood model (DD3)
3 , 1 , 0 , 1 , 0ˆ ˆ ˆ ˆ( ) ( ) ( ) ( ) ( ) .h t h t k t k tDD S S S S
16
• The most important assumption underlying the difference-in-differences estimator is that all differences in the changes of the outcome variable between the treatment and the control group are due to the treatment
• A number of reasons why this could by invalid:
1. Other programmes affecting treatment or control group differently
2. Other influences on particular age groups
3. Anticipation effects
Validity of the DD estimators
17
Note: Number of individuals in parentheses, t-statistics estimated robustly.
Exits into subsidised employment
2001 2002 2003 2004
Treatment Group(Age 50 to50+6 months)
1.16(6209)
3.72(7320)
3.03(14406)
0.94(16389)
Control Group (Age 49 to 49+6 months)
0.82(6453)
1.14(6947)
1.25(14374)
0.52(17181)
18
DD1-effect on employment, 2002
0,7
0,75
0,8
0,85
0,9
0,95
1
1 30 59 88 117 146 175
Days
Sh
are
Un
emp
loye
d
Control group, after treatment
Treatment group, after treatment
Control group, before treatment
Treatment group, before treatment
19
0.7
0.75
0.8
0.85
0.9
0.95
1
1 30 59 88 117 146 175
Days
Sh
are
un
em
plo
ye
d
Kontrollgruppe, nach Änderung Treatmentgruppe, nach Änderung
Kontrollgruppe, vor Änderung Treatmentgruppe, vor Änderung
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
1 30 59 88 117 146 175
Tage
Arb
eit
slo
sig
ke
its
an
teil
Kontrollgruppe, nach Änderung Treatmentgruppe, nach Änderung
Kontrollgruppe, vor Änderung Treatmentgruppe, vor Änderung
0.7
0.75
0.8
0.85
0.9
0.95
1
1 30 59 88 117 146 175
Tage
Arb
eits
losi
gke
itsa
nte
il
Kontrollgruppe, nach Änderung Treatmentgruppe, nach Änderung
Kontrollgruppe, vor Änderung Treatmentgruppe, vor Änderung
0.7
0.75
0.8
0.85
0.9
0.95
1
1 30 59 88 117 146 175
Tage
Arb
eits
losi
gke
itsa
nte
il
Kontrollgruppe, nach Änderung Treatmentgruppe, nach Änderung
Kontrollgruppe, vor Änderung Treatmentgruppe, vor Änderung
DD1-effect on employment, 2002
Men, West Men, East
Women, West Women, East
20
DD1-effect on employment, 2004
0,7
0,75
0,8
0,85
0,9
0,95
1
1 30 59 88 117 146 175
Days
Sh
are
un
em
plo
ye
d
Control group after treatment
Treatment group, after treatment
Control group, before treatment
Treatment group, before treatment
21
DD1-effect on employment, 2004
0,7
0,75
0,8
0,85
0,9
0,95
1
1 30 59 88 117 146 175
Days
Sh
are
un
em
plo
ye
d
Control group, after change Treatment group, after change
Control group, before change Treatment group, before change
0,7
0,75
0,8
0,85
0,9
0,95
1
1 30 59 88 117 146 175
Days
Sh
are
un
emp
loye
d
Control group, after change Treatment group, after change
Control group, before change Treatment group, after change
0,7
0,75
0,8
0,85
0,9
0,95
1
1 30 59 88 117 146 175
Days
Sh
are
un
em
plo
ye
d
Control group, after change Treatment group, after change
Control group, before change Treatment group, before change
0,7
0,75
0,8
0,85
0,9
0,95
1
1 30 59 88 117 146 175
Days
Sh
are
un
emp
loye
d
Control group, after change Treatment group, after change
Control group, before change Treatment group, after change
Men, West Men, East
Women, West Women, East
22
All workers Men West Men East Women West
Women East
After treatment (2002)
1,00 0,98 1,01 1,07 0,88
(-0,01) (-0,28) (0,07) (0,9) (-0,87)
Treatment group (age 50 to 50+6)
0,95 1,06 0,92 0,94 0,64
(-1,15) (0,93) (-1,01) (-0,91) (-2,94)
Interaction term 1,03 (0,6)
0,98 (-0,22)
1,05 (0,41)
0,99 (-0,15)
1,64 (2,32)
PH test, p-level 0,57 0,90 0,92 0,31 0,99
DD2Number of obs.
0.6823760
-0,438870
1,253895
-0,268440
6,882555
LR chi2 367,75 131,82 87,81 84,29 31,72
Conditional DD Results (DD2-Effect), 2002
23
All Men West Men East Women West Women East
After treatment(2004)
0,99(-0,51)
0,88 (-2,7)
0,97 (-0,48)
1,06 (1,07)
1,18(2,02)
Treatmentgroup (age 50to 50+6)
0,97 (-1,12)
0,89 (-2,39)
0,99 (-0,17)
1,02 (0,36)
1,04 (0,46)
Interaction term 0,95 (-1,23)
1,10(1,39)
0,99 (-0,06)
0,80 (-2,88)
0,85 (-1,43)
PH test, p-level 0,24 0,13 0,96 0,13 0,04
DD2
Number of obs.
-0,74
57449
1,48
20490
-0,10
10154
-2,84
18253
-1,92
8552
LR chi2 954,51 333,36 363,27 139,51 80,05
Conditional DD Results (DD2-Effect), 2004
24
Conditional DD Results (DD3-Effect), 2002
25
Conditional DD Results (DD3-Effect), 2004
26
Conditional DD Results (DD3-Effect), 2002
West East
Men
Women
27
Conditional DD Results (DD3-Effect), 2004
West East
Men
Women
28
Evidence on deadweight effects
• How does the number of jobs created with the help of hirings subsidies compare to the number of subsidies disbursed?
• Compare number of exits into subsidised and unsubsidised jobs – if the latter declines by as much as the first increases, crowding out is complete
• Competing risks framework using differences of differences to the cumulative incidence functions of exit into subsidised and unsubsidised jobs
• All results are unconditional on covariates
29
DD of the Cumulative Incidence Function, 2002
30
DD of the Cumulative Incidence Function, 2004
31
Conditional DD Results (DD3-Effect), 2002
West East
Men
Women
32
Conditional DD Results (DD3-Effect), 2002
West East
Men
Women
33
• We have used a natural experiment design to answer the question whether hiring subsidy programmes are effective
• Two changes were compared: extension of eligibility for workers aged 50+ in 2002 and abolishment of preferential treatment for the same group in 2004
• Application of three DiD estimators shows that employment effects for workers as a whole are insignificant and small compared to the number of subsidised hirings
• Competing risks framework suggests that the EGZ is not very effective in the group of treated individuals and deadweight effects are important
• However, for East German Women (in 2002) and Women in both parts of Germany (in 2004), some effects are significant and large relative to subsidised hirings deadweight effects much smaller here
Conclusions