A preliminary evaluation of the sample sizes used for the evaluation of the Education, Health, and Nutrition Program (PROGRESA) of Mexico

The aim of PROGRESA is to provide support for families living in conditions of extreme poverty in small communities in rural Mexico in order to broaden their opportunities and capabilities to attain higher levels of well-being. The Program carries out actions conducive to raising their living standards by improving opportunities for education, health and food. Good evaluation of a program such as PROGRESA requires a systematical approach that integrates closely underlying models of the behaviors of those to whom the project is directed, data on those behaviors, and estimation of program effects based on those models and on those data. PROGRESA recognized the importance of collecting good data as part of the evaluation and therefore undertook a baseline survey in March 1998 (ENCEL98M) that is to be followed up with subsequent surveys at approximately half-year intervals. This research report focuses solely on the question of the sample sizes and how they relate to alternative ways of phasing in the current evaluation control households/localities into treatment. The evaluation sample in ENCEL98M is composed of the 24,077 households that are divided as indicated in Table 1. For certain questions of interest in the evaluation of PROGRESA, however, the household is not the relevant unit. For example for evaluation of some of the educational aspects of the program, what is relevant may be teenage girls in “poor” households. This Research Report considers these sample size questions. Section 2 summarizes the general considerations regarding sample size, treatment effect sizes, significance level, power, and relative size of treatment and control samples. Section 3 gives information on the sample sizes that currently are available in the evaluation sample for questions concerning a number of central aspects of PROGRESA. Section 4 gives the required sample sizes under alternative assumptionsfor a number of selected variables from ENCEL98M that are likely to be central in the evaluation of PROGRESA. Section 5 concludes.Non-PRIFPRI1; PROGRESAFCN