Does the PAL help the poor? A study on the differential effectiveness of a food support program

Presenter: Paola Gadsden, National Institute of Public Health, Mexico

Abstract

In the last decade Conditional Cash Transfer (CCT) programs have become a common policy instrument in developing countries to increase human capital investment among the poor. Notwithstanding the already available evidence about their positive impacts, most of these studies concentrate on average treatment effects; however, by focusing only on mean impacts relevant policy questions are ignored.
The experimental evaluation data of a food support program in Mexico, called Programa de Apoyo Alimentario (PAL), offered a unique opportunity to investigate treatment heterogeneity. The objective of this study was to evaluate the presence of differential impacts of PAL in total and food consumption across socioeconomic status (SES) and household composition groups.

The evaluation data was collected in two stages. First, 208 rural communities were randomly selected from the pool of eligible communities in Southern Mexico. Second, a random sample of 33 households was drawn within each community. After the baseline survey, communities were randomly assigned to one of four groups: a monthly food basket (FB) without education, a FB with education, a cash transfer (CT) equivalent to the cost of the FB (about 13 USD/month) with education and a control group. Two years later the follow-up survey was conducted in the same households. A difference-in-difference household fixed effects model was used to estimate treatment effects by socioeconomic status (SES) (households above and below the median total consumption at baseline) and household composition (households with and without children under-5 years of age).

The largest impact of the program was found in households with low SES. Total and food consumption increased by 16% and 23% (P<0.001) respectively in the FB and CT groups with low SES. In high SES households, the impact on total consumption was 9% (P<0.001) in the FB and 6% (P<0.05) in the CT group; food consumption increased by 15% (P<0.001) and 9% (P<0.001) in the FB and CT group respectively. Regarding household composition, the largest impact was found in households without children under 5. Total consumption increased 21% and 17% in the FB and CT groups respectively. The impact on food consumption was 26% in the FB and 23% in the cash group (both P<0.01). In contrast, PAL’s impact on total consumption in households without under fives was 10% (P<0.001) for the FB and 9% (P<0.05) for the cash group. In terms of food consumption, the impact was 17% (P<0.001) and 13% (P<0.001) for the FB and CT groups respectively. Significant differences between the FB and CT groups were only found in food consumption in the group of household with high SES and without children under 5.

Results from this study demonstrate the existence of a heterogeneous impact of PAL by socioeconomic status and household composition. The program is effectively reaching the poorest households, but it may not reach households with children under 5, which is a vulnerable group. Impact evaluations should not only focus on mean impacts but also explore differential treatment effects by potential treatment modifiers, such as SES and household composition.