The Evaluation of Start-Up Subsidies for the Unemployed and the Role of Unobserved Characteristics for Matching Estimators

Start-up subsidies for the unemployed have become an important part of Active Labor Market Policy (ALMP) in many countries. Previous evaluation results show pre-dominantly (very) positive results indicating that these programs are an effective way to increase employment probabilities and income of participants. Most of the studies are using matching estimators based on the conditional independence assumption (CIA) to estimate these effects and are prone to bias if there are unobserved factors affecting the selection process into the programs. From the entrepreneurship literature we know that entrepreneurs are different , e.g., with respect to personality traits, non-cognitive skills and risk preferences. Since most of the previous evaluation studies are based on administrative data, information on such variables is not available. This raises the question whether the effects are potentially over-estimated. We have access to data which allow us to model the selection process with and without usually unobserved personality characteristics and to examine the consequences for the estimated propensity scores and treatment effects. We show that openness to new experiences and internal/external locus of control have a significant influence on selection into treatment (and labor market outcomes). Our empirical findings also indicate that neglecting personality traits in the program evaluation leads to slight over-estimation of the average treatment effects on the treated which still remain positive and significant even after controlling for relevant personality traits.