In this paper, we survey the microeconometric literature on the quantitative evaluation of public interventions in the labour market, focussing upon the measure of programme effects for the participants to the programme (“the effect of the treatment on the treated”). We briefly summarize the main statistical problem that evaluation studies are confronted with, that is endogenous selection. We distinguish estimation methods according to the use of controlled experiment, natural experiment or non experimental data. We continue by describing an example of programme evaluation where we study the effects of public training schemes for youths, on their subsequent employment probabilities. We use non experimental data on youth employment histories at the beginning of the nineties in France and we contrast the results of the evaluation between urban and rural areas. The main result is that the effects of training schemes on employment probabilities are not significant in both areas. We continue by discussing the benefits and costs of experimental and non experimental data according to the type of public policies and their intended range and by presenting some examples of limitations, when using controlled or natural experiments, that have been described in the literature. Despite these limitations, we conclude by insisting upon the need to more closely relate the data collection process to any programme evaluation.