Heterogeneous treatment effects in conditional cash transfer programmes: assessing the impact of Progresa on agricultural households

Abstract

The success of Mexico's conditional cash transfer programme (Progresa) has sparked a wave of similar programmes across the developing world, and the highly successful social experiment in Progresa has created demand for experimental evaluations among development partners, multilateral agencies and governments as a way to assess development policy. But existing evaluations do not consider the possibility of heterogeneous treatment effects due to either multiple programme participation or the special circumstances of agricultural households when production and consumption decisions are not separable. This article shows that the impact of Progresa on health check-ups is significantly smaller among participants of Procampo, a programme that is linked to agricultural production. This differential impact may be due to Procampo conditionality or the fact that the shadow price of time is different between these households and other Progresa beneficiaries. The authors' conclusion is that conditional cash transfer programmes must consider multiple programme participation and non-separable agricultural households when designing programmes and assessing impacts

Keywords:

• cash transfers
• agricultural households
• Mexico
• programme evaluation
• poverty

Acknowledgements

This paper reflects the opinions of the authors and not those of UNICEF or the African Development Bank, its Board of Directors or the countries they represent.

Notes

1. This is the Spanish acronym for the Programme for Education, Health and Food (Programa de Educación, Salúd y Alimentación). In 2002 Progresa changed its name to Oportunidades. We use the acronym Progresa because that was the name of the programme during the period of our analysis.

2. In their evaluation of Nicaragua's Red de Proteccion Social, Maluccio and Flores (2005) show that programme impacts are larger among the poorest participants, but they do not consider AHs in their investigation of heterogeneous treatment effects.

3. This is the Spanish acronym for Programme of Direct Payments to the Countryside.

4. Specifically, the value of the product foregone by not working on the farm is marked up by the shadow price of credit. Hired labour to replace family labour is also marked up by this internal price.

5. The March 1998 baseline was a smaller instrument because socio-economic, demographic and schooling information were collected a few months earlier in the ENCASEH. These two surveys are typically combined to obtain full information at baseline for the households and individuals in the sample.

6. Initially, Progresa classified as eligible about 52 per cent of households. Afterwards, due to perceived bias against certain kinds of poor households (especially elderly with no children), criteria of eligibility were revised and the programme was extended to cover 78 per cent of households. Since these households were declared eligible at a later date, most of them started receiving cash transfers some time after the initial households so that the impact of Progresa on their consumption could be different. We exclude these households from our sample, as do Hoddinott and Skoufias (2004).

7. We choose this age group because the objective of Progresa is human capital development of children, notwithstanding the conditions imposed on adults, and because most of the energy on programme compliance is focused on children.

8. In the data we observe households reporting Procampo transfers in some rounds and not others, probably due to the timing of agricultural production. We define an AH as those who report positive Procampo receipts in at least three out of the four survey rounds; among these households 95 per cent report growing one or more of the eligible staple crops.

9. It is important to note that the intent to treat is the potential transfer available to the family if it chooses to participate in the programme, not the actual transfer the household is eligible to receive based on observed behaviour. Similarly, for Procampo we use land available and not actual land cultivated.

10. Note that reported consumption expenditure includes the monetary value of gifts and own production.

11. The conditional logit is identified only through observations that change outcomes over the period, since observations that do not change do not contribute to the likelihood function; in these data the conditional logit is thus estimated over a very selected sub-set of the sample. These results are available from the authors upon request.

12. The eligibility score is based on a set of approximately 12 variables covering household demographic structure, assets and housing conditions. Diaz and Handa (2006) use these variables to construct the balancing score in their propensity score matching analysis of Progresa.

13. While all Procampo households own land (median size two hectares), less than one-half of the matched sample owns any land (median size is less than one hectare).