Labor market integration of immigrants provides a difficulty in many countries. As a consequence, immigrants are substantially overrepresented in welfare systems. Despite forming a substantial share of all welfare recipients, relatively little is known about the impact of welfare-to-work programs on labor market outcomes of this group. This paper explicitly focuses on immigrants and evaluates the effects of a major German welfare-to-work program, namely off-the-job short-term training, on the probability of exiting the welfare system by taking up employment. The core questions are: Are programs similarly effective for immigrants and natives? And, if differences in effects are observable between both ethnic groups, what are the causes of these differences? Are they due to observable differences in socio-demographic characteristics or are they due to unobservable differences which must be attributed to the immigrant status? To answer these questions, we use a sample of about 80,000 immigrants and 80,000 natives from comprehensive register data of the inflows into welfare in 2006. The effects of natives are estimated to benchmark the effects for immigrants. The data provide detailed information about socio-demographic characteristics, employment history, program participation and the outcome variable of interest; in addition, they enable identification of immigrants beyond the concept of citizenship. Four different types of training are distinguished: aptitude tests, job search training, skill provision, and combined training programs. For the estimation of the treatment effects, we employ propensity score matching estimators in a dynamic setting, where treatment effects vary conditionally on the preceding duration in welfare. To answer the question whether differences in effects are caused by differences in the composition of the native and immigrant population in the welfare system (e.g. due to differences in education or in the age structure) or due to an immigrant fixed effect we suggest and apply a matching based decomposition of differences in treatment effects. Our estimation results show that the considered training programs exhibit substantial effect heterogeneity. For aptitude tests we observe on average positive employment effects. While in the sample of women treatment effects are larger for natives than for immigrants, the picture is ambiguous for men depending on the timing of the training. Aptitude tests starting in the second or third quarter of welfare receipt generate larger treatment effects for immigrants, whereas native men benefit more from tests in the first quarter. The difference in treatment effects of natives and immigrants in the first quarter is mainly due to differences in observable characteristics between the two ethnic groups. Keeping all covariates constant, immigrants tend to benefit even more from aptitude tests than natives. Job search training is ineffective for men irrespective of being an immigrant or not. Native women benefit from this form of training, while immigrant females face negative treatment effects. The large difference in treatment effects of native and immigrant women cannot be explained by observable characteristics and must instead be attributed to an immigrant fixed effect. Holding everything else constant, immigrant females participating at job search training have a nearly 15 percentage point lower treatment effect than native participants. Even though the negative immigrant fixed effect fades away nine months after the program starts, this result gives cause for serious concern. Job search training decreases rather than increases employment chances of female immigrants. In contrast, female immigrants clearly benefit from skill provision, which is a program exhibiting positive effects in general when assigned early during the welfare spell. For this form of training the immigrant fixed effect increases over time and amounts to 14 percentage points one year after program start. Thus, when netting out observable differences between immigrants and natives, the former have on average a 14 percentage points larger treatment effect than the latter. For the combined training programs, we do not find statistically significant differences in any subgroup. These results reflect the finding that combined programs are rather ineffective for both ethnic groups and for both genders. The general ineffectiveness of combined training programs might be due to the characteristics of the targeted group, since combined programs are in particular assigned to those persons who were out of labor force for a substantial fraction of the final two years before treatment. These persons are likely to face multiple obstacles for employment uptake, which might not be remediable by combined training programs
