Self-employment grants vs. subsidized employment: Is there a difference in the re-unemployment risk?

Self-employment grants and employment subsidies are active labor market programs that aim at helping unemployed workers to escape unemployment by becoming self-employed or being hired at an initially reduced cost for the employer. In Sweden in the 1990’s the participation rate in the self-employment program increased from virtually none to almost same as in the employment subsidy program. The advancement of the self-employment program is likely to be a result of (i) a change in the labor market program policy, and (ii) an increase in the supply of skilled unemployed workers. The justification for the policy change is unclear, however. The literature indicate that a rather specific group of unemployed workers may benefit from self-employment programs; Neither are there any strong reasons to believe in general that self-employment should be preferable to conventional employment through subsidies.
We examine, ex post, the justification for the policy change by comparing the post-program duration of employment for the two programs. In addition, we focus in some detail on the outcome for female workers and workers of foreign citizenship. The reason for this is the explicit policy to direct those workers to self-employment.
The data we study are the inflow to the two programs from June 1995 to December 1996. The program participants are followed to March 1999. The data contain detailed spell and background information on 9,043 unemployed workers who participated in the self-employment program and 14,142 who participated in the employment subsidy program.
The second explanation, see (ii), for the increase in self-employment program implies a  potentially serious selection problem. We discuss how the selection process may bias the effect estimate in the non-linear duration model that we use. Simulations help us to determine the magnitude of the selection bias in our application. Moreover, we exploit the existing behavioral heterogeneity across labor market offices to reduce the selection bias.
We find that the risk of re-unemployment is more than twice as high for the subsidized employment program compared with the self-employment program. The large positive effect is, however, limited to male and female workers of Swedish origin. We thus conclude that the policy change in general has been successful, though we note that directing immigrant workers to selfemployment is unlikely to improve the situation for this group of unfortunate workers on the Swedish labor market.

Keywords: Empirical Bayes methods, Employment duration, Program evaluation, Proxy variables, Selection bias, Simulations