An "ex ante"evaluation of the effects of reforms to an Italian labour market policy

My thesis is about "ex ante" policy evaluation, i.e. the estimate of the impact of a public intervention on a defined outcome measure prior to its implementation. The policy regime I analyze is an Italian labour market program targeted to dismissed employees, called "Liste di mobilità'' (literally, "Mobility lists'', LM hereafter), which includes both a "passive'' component (monetary benefits to part of the unemployed workers) and an "active'' one (money transfer for the firm hiring them). Length of the period in the LM and entitlement to benefits vary according to the age of the worker and the size of the firm at time of dismissal. However, the amount of the unemployment subsidy (for people entitled to them) is proportional to the last wage earned. Linked administrative panel data set for the Veneto region (a large region in the Northeastern Italy) are used. Information about people ever entered in the LM (including labour market history, socio-demographic chacteristics, entitlement to receive monetary benefits and characteristics of the firms where they have been employed) is available. My interest is on the effect of possible changes to the existing policy regime on the probability of re-employment in the 36 months subsequent to enrollment in the LM. Since it deals with potential reforms to the current policy, I need to perform an "ex-ante" policy evaluation. I will briefly review the job-market economic theory about wage determination, describing why similar workers (from the point of view of attractiveness for firms) may earn different wages. Basically, I use such random variability in wages, hence in the amount of monetary benefits received, to mimic a variation in the policy regime. If such policy equivalent variation (PEV) exists and workers are otherwise similar, identification of the effects of not-yet-implemented policy reforms is possible. Specifically, I consider two alternative identification strategies and test their validity. First, I compare individuals basing on the generalized propensity score (an extension of the procedure of propensity score matching or subclassification to non-binary cases; then, I consider the hypothesis that the selection bias pattern is the same in the group of treated and non-treated, using a "difference in differences'' approach.