Evaluation aléatoire et expérimentations sociales

Randomized evaluation of public policies began to be applied to major social programs in France since 2007, more than forty years after the first similar work in the United States. They have been experiencing rapid development. These experimental methods are to evaluate the effects of a policy by comparing a test group with a control group, made by random assignment. In this paper whose purpose is to provide an introduction to the application of these methods to social experiments, we emphasize that their specificities are several and cannot be reduced to i) the principle of random assignment. They also have other peculiarities that distinguish them from other evaluation methods: ii) the fact that the program is experimental, iii) the particular partnership between the researcher and the institution experimenter, iv) the fact that researchers have to built an explicit protocol and a system of observation and information processing before the experiment beginning. We show that each of these four singularities is a source of both methodological advantage and potential limitation that should be emphasized.