This article presents a discussion about methodological questions related to social programs evaluation, taking as an example the case of the National Worker Qualification Plan - PLANFOR. The present study used a quasi-experimental design containing three groups: two experimental groups and one control group. The research was realized from 2001 to 2004, with data collection before and after the program, and comprising 360 organizations, one per city, from all Brazilian regions. For data collection, it was used an instrument with quantitative indicators (social and economic). Four data analysis strategies were compared: analysis of variance (ANOVA) of repeated measurement (pre- and post-test scores), ANOVA with gain scores, ANOVA after subjects' categorization on their pre-test (blocking), and analysis of covariance (ANCOVA). Alternative explanations of the observed effects were considered. Finally, the suitability of these strategies to the evaluation of social programs was discussed.
social program evaluation; worker qualification; methodological discussion
