This comment discusses the scope of the results obtained by M. Bléhaut and R. Rathelot (BR) and their consequences for the use of matching. The BR paper demonstrates indisputably that observable variables do not suffice to neutralise selection bias, the presence of attrition and displacement effects simultaneously. Thus, even though it does not provide definitive evidence in this respect, the paper casts serious doubts on the ability of matching to solve the selection bias problem. These doubts are compounded by results of similar research work, whose key lessons are summarised. This comment draws the consequences of these findings for the use of matching for public policy evaluation, particularly in the field of employment.
matching; randomised experiment; selection bias; attrition; displacement effect; public policy evaluation;
Économie et Prévision, 2015, pp. 183–191