Ive Marx Centre for Social Policy University of Antwerp, Belgium

JOB SUBSIDIES AND CUTS IN EMPLOYERS’ SOCIAL SECURITY CONTRIBUTIONS:

THE VERDICT OF EMPIRICAL EVALUATION STUDIES

Ive Marx

Centre for Social Policy

University of Antwerp, Belgium

As a percentage of total public spending on active labour market programmes

As a percentage of GDP

Belgium 53.8 0.69

Canada 6.0 0.03

Denmark 10.7 0.17

Finland 27.7 0.26

France 26.2 0.34

Germany 17.8 0.21

Ireland 47.0 0.53

Italy _ 0.29

Netherlands 24.0 0.38

Spain 39.3 0.33

Sweden 13.3 0.19

United Kingdom 7.3 0.03

United States 8.1 0.01

Spending on targeted employment subsidies, 2001

Source: OECD (2003)

Our purpose

• To bring together empirical findings:

– What is known about the impact on employment ?

– What is known about subsequent mobility trajectories ?

Types of employer subsidies

• Roughly two types:

– permanent but typically modest subsidies/ employer SS reductions applying to low-paid workers

– temporary but typically more substantial subsidies/SS reductions for employers hiring people deemed at high risk of structural labour market exclusion

Measure Target Group Description

Vermeend-Moor Act (Nl) long-term unemployed reduced social contributions for a period of 4 years;

one-off recruitment bonus

RAP (Nl) long-term unemployed reduced social security contributions for a period of 4 years;

recruitment bonus (increases with period of unemployment)

Workstart (UK) long-term unemployed temporary wage subsidy

Jobstart (Aus) long-term and quasi long-term unemployed

temporary wage subsidy

Employment Incentive (Ireland) long-term unemployed temporary wage subsidy

Jongerenbanenplan (Belgium) unemployed youngsters temporary, regressive reduction in social contributions

VLW (Nl) long-term unemployed reduced social security contributions for a period of 4 years

SPAK (Nl) low-paid permanent reduction in social contributions

CIE (France) long-term unemployed, unskilled, welfare recipients, disabled

temporary reduction in social security contributions

Partial overview of reviewed measures

Gross impact: take up

• There appears to be a connection between the scope, the generosity and the duration of initiatives.

• Relatively generous measures with a relative broad scope involving a substantial subsidy or reduction generally generate a greater response.

• In the case of more strongly targeted measures, take up rates are often considerably lower, often well below 50 per cent

Reasons for non take up

• Ignorance or inadequate information

– particularly a factor in small and medium-sized companies– multitude, temporary nature and complexity also factors

• (Perceived) bureaucratic fuss and cost

• Voucher schemes more effective for targeting very specific groups – downside: stigmatisation

Gross to net impact: intervening factors

• Deadweight losses

• Substitution effects

• Displacement effects

Three evaluation methods

• Employers surveys

• Time series analysis

• Experiments

Net impact

Employer survey based evidence

Deadweight losses

• Losses which occur when subsidised individuals would have found employment had the subsidy not been in place

• Estimated losses are found to be consistently and considerably higher than what is assumed in theoretical models and simulations– Snower (1994) in ‘Converting unemployment benefits

into employment subsidies’ assumes a deadweight cost of 25 per cent

Name (country) Deadweight Substitution (Displacement) Sum(deadweight+substitution

Reference

Targeted Measures

Vermeerd-Moor Act(Netherlands)

80-85% ( 28% ) > 80% De Koning et al. (1995)

RAP (Netherlands 42% 47% ( 36% ) > 89% De Koning et al. (1995)

Workstart (UK) 55% 25% ( 33% ) > 80% Atkinson and Meager (1994)

Jobstart (UK) 67-79% - - > 67% OECD (1993); Byrne (1994)

Employment Incentive (Ireland)

70% 21% ( 4% ) 95% OECD (1993); NERA (1995)

Voordeelbanenplan (Belgium)

53% 36% - > 89% Van der Linden (1995)

VLW (Netherlands) 27-60% 37-63% - 57-87% NEI (1999)

CIE (France) 19-39% 37-63% 57-87 Belleville (2001)

General Measures

SPAK (Netherlands) 93% - - 93% NEI (1999)

Employer survey based findings: deadweight losses

Substitution effects

• Targeted groups need to be clearly demarcated and this inevitably leads to distortions at the margins, i.e. subsidized workers being recruited at the expense of non-subsidised employees

• Empirical studies generally find substitution to be significant


Reference

Targeted Measures


80-85% ( 28% ) > 80% De Koning et al. (1995)





70% 21% ( 4% ) 95% OECD (1993); NERA (1995)


53% 36% - > 89% Van der Linden (1995)



General Measures


Employer survey based findings: substitution losses

Displacement effects

• job losses which occur through distortion of competition, i.e. job losses caused by the fact that enterprises that do not receive subsidies lose market share.

• difficult to estimate using employer surveys; respondents are usually unable to attribute an increase or decrease in market share to this single factor.


Reference

Targeted Measures


80-85% ( 28% ) > 80% De Koning et al. (1995)





70% 21% ( 4% ) 95% OECD (1993); NERA (1995)


53% 36% - > 89% Van der Linden (1995)



General Measures


Employer survey based findings: displacement estimates


Reference

Targeted Measures


80-85% ( 28% ) > 80% De Koning et al. (1995)





70% 21% ( 4% ) 95% OECD (1993); NERA (1995)


53% 36% - > 89% Van der Linden (1995)



General Measures


Summary of employer survey based findings

Employer surveys: methodological limitations

• Response rates tend to be low

• Estimates rely on statements by an interested party – the employer

Net impact

Time series based evidence

Time series analysis

• The approach taken in these studies is to ascertain whether the introduction of a particular measure coincided with additional job growth (or slower job destruction) that could not be attributed to any other measurable factor.

• The value of these estimations depends on the thoroughness with which one tests for other potential explanatory factors, such as cyclical movements of the economy.

Time series analysis: additional problems

• Usually there are multiple policy changes at the same time the effects of which are difficult to disentangle

– e.g. subsidies are sometimes coupled with training or additional job seeking support

• Never a ‘perfect’ counterfactual / reference group available

Type of measure Findings Source

Targeted Jobs Tax Credit (US)- target group: youths, public

assistance and SSI recipients, veterans, certain ex-convicts

- tax credit for employer amounting to 50 % of first and 25 % of second year earnings

- Net employment effect of 7.7 per cent or 3 percentage points

Katz (1996)

SPAK (Netherlands)- permanent reduction in employers’

social security contributions for low-paid workers (up to 115 % minimum wage)

- Zero net employment effect Muhlau and Salverda (2000)

(France)- permanent reduction in employers’

social security contributions (from 36.5 % in 1993 to 21.8 per cent in 1996) for minimum wage workers

- Small but statistically insignificant effect

Kramarz and Philippon (2000)

Employment effects:

Findings from time-series/DDD analysis

Net impact

Experimental evidence

A classic: Dayton experiment• A controlled experiment conducted in 1980-1981 by the US Department of

Labor among a group of welfare recipients

• One group was given a tax credit voucher, which entitled the employer hiring the vouchered job applicant to a tax credit (of the magnitude of the earlier discussed Targeted Tax Job Credit)

• A second group was given a voucher that entitled the employed to a direct cash subsidy

• A third control group was not given a voucher, even though they qualified in principle.

• The individuals were assigned to one or the other group randomly, so that the groups were comparable in terms of composition.

– All three groups got two weeks of job search training– Efforts were undertaken to make sure that the three groups did not receive

different treatments from administrators or trainers

Group Sample size

Number placed in jobs

Percentage placed in jobs

Tax Credit Voucher

247 32 13.0

Direct Rebate Voucher

299 38 12.7

Control 262 54 20.6

Total 808 124 15.3

Source: Burtless (1985)

Findings from experimental research:

Dayton Voucher Experiment

Experimental research: issues

• In theory the superior method since only treatment and not (unobserved) composition effects explain differential outcomes

• But….

– a degree of contamination is difficult to prevent (in reality it is difficult to deal seperately but equally with treatment and control groups)

– experimental studies are expensive and time-intensive– ethical issues: is it ethical to deny potentially positive ‘treatment’

to equally entitled people for the sake of science ?

From subsidised to regular work

Evidence on mobility

Evidence on mobility

• Number of studies available up until recently • Data requirements are substantial

– Collecting adequate longitudinal data is costly and methodologically complex.

• Increasingly studies use administrative data (for example from social security databases)– These allow longer term analysis (say over several

years) at reasonable cost. – The problem is that such data bases often do not

contain information on all relevant variables

Type of measure Findings Source

JOBSTART (Austr.)- employer subsidy for recruitment of unemployed persons

- improvement of employment chances 6 months after subsidised job- no correction for selection bias

Byrne (1994)

Special Youth Employment Training Scheme (Austr.)- employment subsidy coupled with training

- positive effect on employment chances two years after- correction for selection bias

Richardson (1998)

Voordeelbanenplan (BE)- targeted, temporary reduction in social security contributions

- employment prospects 24, 30 and 36 months after leaving subsidised employment comparable to non-subsidised unemployed- no correction for selection bias

Bollens et al. (1996)

Voordeelbanenplan (BE)- targeted, temporary reduction in social security contributions

- positive yet statistically insignificant effect on employment chances- correction for selection bias

Cockx et al. (1998)

PEP (Germany)- Subsidised jobs (1 yr. max.) for long-term unemployed

- participants have higher probability of being in work than non-participants with similar observable characteristics

Eichler and Lechner (2002)

France- Subsidised jobs for long-term unemployed in non-market sectors

- few participants move to regular work after their spell of subsidized employment has ended

Bardaji (2001) (from OECD, 2003)

Slovak Republic- subsidised jobs for the unemployed, maximum duration expanded first from 6 to 9 months, then to 12

- short term subsidised jobs have positive effect on job finding rate, opposite effect as maximum duration got expanded (‘lock-in’)

Van Ours (2002)

Why limited mobility ?

• beneficiaries get locked into their subsidised jobs

• stigmatisation

• job experience acquired inadequate to escape from the ‘productivity trap’

Conclusion: prior remark

• very significant differences between the measures reviewed across a very wide range of dimensions

• same applies to evaluation methodologies

Conclusion

• The measured net employment effects of targeted subsidies and employer SS reductions tend to be consistently and considerably smaller than what ex-ante evaluations (i.e. theoretical models and simulations) invariably suggest

• The empirical evaluation literature remains ambiguous as to the longer term impact of spells in subsidized employment – exact design features seem to matter a lot

Policy implications

• More scepticism vis-a-vis ex-ante policy evaluation warranted

• Need for systematic empirical evaluation

• Evidence based policy adjustment

Documents

Ive Marx Centre for Social Policy University of Antwerp, Belgium