Does a reduction in benefit payment duration increase employment? Evidence from the German “Hartz...

Preview:

Citation preview

Does a reduction in benefit payment duration increase employment?

Evidence from the German “Hartz Reforms”

ISEO Summer School 2012

Felix König - London School of Economics

1

Content

1. The unemployment crisis2. The German labour market and reform3. A simple search model4. The effect of the labour market reform

2

The unemployment crisis

Crisis German labour market Model Reform results Inequality

3

Unemployment duration as % of the total labour force

OECD Employment Outlook 2011

Rising long-term unemployment, even in flexible labour markets

Crisis German labour market Model Reform results Inequality

4

Unemployment during the crisis

OECD Employment Outlook 2011

Crisis German labour market Model Reform results Inequality

5

Questions of this presentation

1. What is the reason for the German success?

1. What can be done to tackle high unemployment rates?

2. Is Merkel’s labour market reform agenda for Europe non-sense?

Crisis German labour market Model Reform results Inequality

6

The German labour market

7

“Sick man of Europe”

German unemployment rate

OECD quarterly Data

in %

Peter Hartz

Crisis German labour market Model Reform results Inequality

8

The “Hartz Reforms“Crisis German labour market Model Reform results Inequality

9

A job miracle?German unemployment rate

OECD quarterly Data

in %

Falling unmployment since Q1 2006

Peter Hartz

Crisis German labour market Model Reform results Inequality

10

A job miracle?Crisis German labour market Model Reform results Inequality

11

Does this mean Mrs. Merkel is right?

Crisis German labour market Model Reform results Inequality

12

Duration of unemployment benefit

Three changes allow for identification

Crisis German labour market Model Reform results Inequality

13

A difference in difference framework

Reasons for being in the treatment group:

1. Older than a comprable individual

1. More work experience than a comprable individual

1. Timing of work experience at an unfortunate point in time

Crisis German labour market Model Reform results Inequality

14

Difference in Difference assumptions

beforeyit1 = ci +λt + uit

after

yit+1 = ci + λt+1 + α + uit+1

beforeyjt1 = cj + λt +ujt

after

yjt+1 = cj + λt+1 + ujt

Affected Non-affected

∆ yit+1 = λt+1 – λt + α + ∆ uit+1 ∆ yjt+1 = λt+1 – λt + ∆ ujt+1

Need common trend for identification

Crisis German labour market Model Reform results Inequality

15

Common trend assumptionCrisis German labour market Model Reform results Inequality

16

In a DID regression set up

yit = θt + Xit'β + δ After + γ Treated+μ(A*T)estimates -8.4 2.1 -1.7White standard errors (0.7) (0.5) (0.7)

Covariates: monthly time dummies, age dummies, education dummies, work experience dummies and a gender dummy

Unemployment duration in months

Crisis German labour market Model Reform results Inequality

17

Effect of UI payment duration reform

• A one-week reduction in BPD leads to 1.32 days reduction in the unemployment duration

• The average unemployment spell length decreased by about 9 days

• This amounts to 37,5 % of the decline in spell duration in Germany since 2006

Crisis German labour market Model Reform results Inequality

18

So cutting benefit payments is a success story?

19

Labour market flows

Elsby, Smith, Wadsworth (2010)

Change in unemployment not necessarily due to employment

Crisis German labour market Model Reform results Inequality

20

A simple search model

Behaviour changes as benefit expiry approaches

Objective function:(roughly following Mortensen 1977)

FOCs evaluated in the initial period of unemployment (U = UT):

2.

1.

c: search cost in units of consumption s: search intensity wR: reservation wage U: value func. of being unemployed b: benefitsλ(s): rate of job offer arrivals ( λ’>0 & λ’’<0 ) u: instantaneous utility W: value func. of working s*: threshold of search activity, below it a person is considered inactive

Stylised time pattern of the exit rate

T Time

Exit rate

Job finding rate

Job finding + becoming inactive

Exit rate from unemployment

T Benefits expire

Crisis German labour market Model Reform results Inequality

22

Estimating the exit rateThe hazard rate is the probability that an individual leaves unemployment in a given time interval (months in this case):

Crisis German labour market Model Reform results Inequality

23

Does this mean Mrs. Merkel is wrong?

Crisis German labour market Model Reform results Inequality

24

Conclusions about spell duration

• Unemployment duration was reduced by 1.7 months

• The time to find a new job fell by 1.3 months

• But only 17% of the unemployed affected

Effect on the treated

Crisis German labour market Model Reform results Inequality

25

Unemployment duration

-24 days

Total change

-7days

Matchingefficiency

Inactivity

-2 days

Residual

-15 days

Reform

Refrences• Card, Chetty, & Weber. (2007). The Spike at Benefit Exhaustion. American

Economic Review , 97 (2), 113-118.

• Mortensen. (1977). Unemployment Insurance and Labour Supply Decision. Industrial and Labor Relations Review , 30 (4), 505-517.

• OECD Employment Outlook 2011

• Socio-Economic Panel (SOEP), data for years 1984-2009, version 26, SOEP, 2010

• van-Ours, & Vodopivec. (2006). How shortening the Potential Duration of Unemployment Benefits Affects the Duration of Unemployment: Evidence from a Natrual Experiment. Journal of Labour Economics , 24 (2), 351-378.

26

Reply to questions:• Question: Maybe the affectedgroup takes different jobs compared to the unaffected(e.g. due to training differences among age groups). The common trend assumption might therefore fail.• Reply:

1. The responsiveness to shocks before 2006 can be tested directly. Such worries are unwarranted according to previous slides.2. For differential shocks that only occur after 2006 can only do indirect testing. Consider Industries with greatest share of job creation for the two groups. Four of the top five job creation industries are identical. No other industry creates more than 5% of jobs. Hence differential developments across industries are unlikely to bias the results.

27