Effectiveness of sequences of classroom training for welfare recipients: what works best in West Germany?

ABSTRACT

Sequences of active labour market programmes (ALMPs) may be part of an intensified activation strategy targeting hard-to-place unemployed individuals. Such sequences are very common among welfare recipients in Germany, but most studies only evaluate either single ALMPs or unemployed individuals’ first ALMP. I analyse the effects of different sequences of classroom training for West German men and women on different labour market outcomes. Using rich administrative data and a dynamic causal model, I can control for dynamic selection problems that occur during a sequence. The results show that two classroom trainings are more effective than two periods of welfare receipt in helping welfare recipients find regular employment, especially among West German women. Moreover, immediately assigning individuals to classroom training is more effective than waiting and assigning them to classroom training in the second period. However, in some cases, avoiding participation in multiple programmes is preferable.

Abbreviations: ALMP, active labour market programme; CIA, Conditional Independence Assumption; CSR, Common Support Requirement; DATET, dynamic average treatment effect on the treated; IEB, Integrated Employment Biographies; IPW, inverse probability weighting; LHG, UBII-Receipt History (Leistungshistorik Grundsicherung); MSB, mean standardized absolute bias; SUTVA, Stable Unit Treatment Value Assumption; UBII, unemployment benefit II; UBI, unemployment benefit I; WDCIA, Weak Dynamic Conditional Independence Assumption

KEYWORDS:

• Sequences
• dynamic causal model
• activation
• classroom training
• Germany

JEL CLASSIFICATION:

• C13
• I38
• J68

Acknowledgments

I particularly thank Joachim Wolff, Regina T. Riphahn, Katrin Hohmeyer and Eva Kopf for their very helpful comments. Furthermore, I would like to thank Michael Lechner and his team for very helpful suggestions. Financial support by the Graduate Programme (GradAB) of the Institute for Employment Research (IAB) and the University of Erlangen-Nuremberg is gratefully acknowledged.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ By law, application training lasts up to 2 weeks, aptitude test up to 4 weeks and skill training up to 8 weeks. This yields an average duration of 32.6 days which I round down to 30 days. The average duration of classroom training in the sample also corresponds to 30 days.

² Data from local authorities (zugelassene kommunale Träger) are not included because of data collection problems. In these 69 districts, the Federal Employment Agency did not administer UBII. For approximately 13% of unemployed welfare recipients between 2005 and 2008, UBII was administered by local authorities (Department for Statistics of the Federal Employment Agency 2014).

³ I also conduct some further data preparations: I do not consider individuals already participating in an ALMP at the entry date, individuals with programme combinations starting on the same day and individuals participating in more than six programmes. Furthermore, I delete a few individuals with missing covariate values.

⁴ I use 1 month because the individual window is 30 days. Otherwise, the intermediate outcomes would extend into the pretreatment period (before period 1).

⁵ See Tables A1 and A2 in the appendix for some descriptive statistics and gross outcomes.

⁶ See Table A3–A9 in the appendix for the results of the probit estimations.

⁷ Not all of the covariates must be included in all of the probit models of the three matching steps. Thus, this may lead to poorer final matching quality, as I consider all MSBs of all covariates.

⁸ Table A10 in the appendix provides the number of observations for the subpopulation CT before and after the CSR for the final matching.

⁹ I calculate t-tests on the means for each covariate that is included in the probit models between the treated individuals and (matched) controls before and after matching. Thus, I can check the final matching quality of the considered sequences at the end for each subgroup and covariate (Tables A11–A18 in the appendix). In summary, the t-tests on the means of single covariates generally do not show any significant (at the 1% level) differences between the treated and control groups after matching. However, as expected, for West German women for (CT,CT) versus (UBII,UBII), some significant differences at the 1% level remain after matching, but the mean differences and t-values mostly decline. In particular, some sociodemographic variables (nationality, children and partner), time since last ALMP before entry; intermediate variables on children and household income; and some variables concerning the labour market history of the partner show significant differences, which might lead to biased results. However, as the sequences are not directly matched but are matched only via three dynamic matching steps to the subpopulation of classroom training, the t-tests on the means for the final matching are only an approximation of the final matching quality.

¹⁰ See Table A2 in the appendix.

Additional information

Funding

This study was financially supported by the Graduate Programme (GradAB) of the Institute for Employment Research (IAB) and the University of Erlangen-Nuremberg.