POVERTY, INEQUALITY AND SPILLOVER IN MEXICO'S EDUCATION, HEALTH, AND NUTRITION PROGRAM

This report provides an evaluation of the community-level effects of the Programa Nacional de Educacion, Salud, y Alimentacion (PROGRESA) using household-level data from various rounds of PROGRESA’s evaluation sample (the Encuesta de Evaluacion de los Hogares [ENCEL] surveys). These surveys, along with the Encuesta de CaracterRsticas Socioecon\micas de los Hogares (ENCASEH) 1997 survey, are a valuable source of information on household- and community-level characteristics before and after the implementation of the program. Other reports in the evaluation series have focused on the direct effects of PROGRESA, using the control and treatment groups in the ENCELS. The objective of this report is slightly different, in that it explores the possible spillover effects of the program on the wider community. Hence, instead of focusing on program effects at the individual level, the study focuses on aggregate community-level indicators of well being such as poverty, inequality, and school and health care attendance rates, in order to assess the impact of PROGRESA at this level. Using the data sets mentioned above, the authors developed five indicators with which to measure the potential impact of PROGRESA at the community level: (1) changes in rates of relative poverty; (2) changes in inequality; (3) school continuation rates; (4) changes in nutrition surveillance rates; and (5) changes in prices (inflation). Exploiting the longitudinal aspect of t he evaluation data, the authors constructed “difference in differences” estimators and used regression techniques to isolate community-level program effects. The main results from this analysis are as follows. Poverty. Although relative poverty increased in the evaluation sample between March and October, the increase was significantly less in PROGRESA communities relative to control ones. For the relative poverty line set at the 25th percentile of consumption in March, the difference in changes in poverty rates was 4 percentage points between treatment and control localities. The same result is found for higher-order poverty measures (poverty gap and squared poverty gap). In all cases, the increase in poverty was significantly less in PROGRESA communities relative to non-PROGRESA localities. Inequality. The two inequality indicators used in the evaluation were the coefficient of variation and the standard deviation of the log of consumption. For both measures, there was a decline in inequality in the survey between March and October, and for both indicators, the decline was greater in PROGRESA localities relative to controls. The results are even stronger when the richest 1 percent of households is excluded from the sample. School Continuation. School continuation rates between the school years 1997/98 and 1998/99 were constructed for five age groups, and by sex and beneficiary category. Significant spillover effects appear to exist among children ages 11–12, especially girls. In other words, nonbeneficiary children in this age group living in PROGRESA communities have significantly higher continuation rates then non-beneficiary children in non-PROGRESA communities. This spillover effect is especially strong for girls, where continuation rates are almost 10 percentage points higher for non-eligible girls in PROGRESA communities relative to non-eligible girls in non-PROGRESA localities. Nutrition Surveillance Rates. Community-level nutrition surveillance rates were constructed for preschool children by beneficiary status. These data indicate important spillover effects in terms of the health care behavior of nonbeneficiary households. Estimates of mean changes in nutrition surveillance rates for preschool children show that six months after the program, there was no difference in mean changes in surveillance rates among nonprogram children in treatment and control localities. However, one year after program inception (in May 1999), the increase in mean rates of surveillance was nearly 7 percentage points higher among nonbeneficiary children in PROGRESA localities, compared to this same group in treatment localities. These results indicate not only the strong presence of possible spillover effects of PROGRESA, but also that these effects take some time to manifest themselves. Inflation. To assess whether PROGRESA is having an inflationary impact in the locality, we compare prices of 10 commodities from the ENCASEH and ENCEL98O, and 33 products from the ENCEL98M and ENCEL98O surveys. There was only one significant price increase found between ENCASEH and ENCEL98O (jitomate), and this occurred in both treatment and control localities. Seven significant prices increases were found in PROGRESA localities between the March and October ENCEL surveys, but five of these increases were also found in control localities. Multivariate analysis of difference in differences in prices also indicated no program-related inflationary pressure. On the contrary, between March and October 1998, there is evidence that mean increases in prices were actually higher in control localities relative to treatment, even after controlling for possible intervening factors such as availability of a Diconsa store, drought, and insect diseases