Active labour market policies in Germany: do regional labour markets benefit?

ABSTRACT

This article examines on a regional level whether active labour market policies (ALMP) improve the matching process. To take the fact of heterogeneous search effectiveness during programme participation into account, we distinguish between current and former programme participants. Our findings based on a regional augmented matching function show that higher search effectiveness due to ALMP is not outweighed by indirect effects on nonparticipants. The total number of matches in a region increases with a higher share of former programme participants among the jobseekers. However, these effects largely differ between programme types.

Abbreviation OECD (Organisation for Economic Co-operation and Development)ALMP (active labour market policy)

KEYWORDS:

• Active labour market programmes
• evaluation
• unemployment
• search theory

JEL CLASSIFICATION:

• C23
• H43
• J64
• J68

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplemental material

Supplemental data for this article can be accessed here.

Notes

¹ Crépon et al. (2013) is one of the few articles using an experimental setting. As their results show, the labour market outcomes of the nontreated differ with regard to the programme intensity in a region.

² Following Nickell and Layard (1999), we define search effectiveness as ‘the ability and willingness of the unemployed to make themselves available for unfilled vacancies’.

³ The approximation $ln(1+(s_{\bar{P}}-1)\bar{P}/S)\approx (s_{\bar{P}}-1)\bar{P}/S$ is only valid for values of $\bar{P}/S$ close to zero. In our case, we have a mean value for $\bar{P}/S$ of 0.1. Hence, the above transformation is a close approximation.

⁴ Dauth, Hujer and Wolf (2014) have applied our idea to the Austrian labour market.

⁵ For more details, see vom Berge, König and Seth (2013).

⁶ We need to aggregate the three offices in Berlin to one unit.

⁷ If a participant is dismissed within this period for reasons that are attributable to the employer, then the employer can be asked to reimburse part of the subsidy. However, this option does not seem to be applied very often (Boockmann et al. 2012, 740).

⁸ This group comprises people with a non-German nationality as well as Ethnic Germans who have German nationality.

⁹ As a dynamic panel data model is specified, the number of times-series observations is actually 18.

¹⁰ All presented estimates are derived using the System GMM estimator with all available lags of the dependent variable as instruments. When the number of instruments is reduced, the results are qualitatively and also quantitatively almost the same. The same is true when using the Arellano–Bond GMM estimation procedure instead of the System GMM. Results are available upon request from the authors. See the online appendix for the full results.

¹¹ As explained in Section III, due to the short duration of the two kinds of training measures, we do not distinguish between current and former programme participants.

¹² The average time interval is 54 days for short-term vocational training, 222 days for long-term vocational training, 12 days for in-firm training measures, 32 days for classroom training measures and 172 days in the case of wage subsidies.

¹³ When reclassifying group $Q$, we face the potential risk of focusing on a nonrepresentative group of $Q$s, namely those who seem to have the most difficulties (i.e. need the longest time) to find a new job if we let the time length between programme completion and job start become too large. Hence, we ran an additional regression where the time span after programme completion was reduced to 90% of the average programme duration. These results are almost identical to the results shown in Table 2.