Impacts of Conditional Cash Transfer Programmes on Educational Outcomes in Developing Countries. A Meta-analysis

Main findings

Headline Findings: a summary statement

The authors conclude that CCTs have a positive and significant impact on enrolment and attendance, with larger effect in contexts where baseline enrolment and attendance are poorest. They find that programs with larger transfer amounts and less frequent payments give larger effects on these outcomes, and transfers conditional on student achievement have larger effects in secondary schools.

Evidence Base

The authors included 42 impact-evaluation studies in the synthesis, reporting effects for 19 programs in fifteen different countries. Thirty-one of these studies are located in Latin America and the Caribbean, three in South Asia, five in East Asia and the Pacific (including South- East Asia), two in Sub-Saharan Africa and one in Europe and the CIS. Twenty-eight of these studies reported the effect of CCTs on enrolment, nineteen on attendance and nine on drop-out rates.

Implications for policy and practice

The authors’ first finding from the meta-analysis is that in all studies CCTs have a positive and significant impact on enrolment and attendance, and a negative impact on drop-out rates, with larger effect sizes seen in secondary schools than in primary schools. There was an average increase in enrolment of 5 per cent in primary schools (95 per cent CI: 3.7-6.6 per cent) and 6 per cent in secondary schools (95 per cent CI: 4.44 – 7.3 per cent). In relation to baseline enrolment rates however, this represents a 6 per cent increase in enrolment in primary schools and 10 per cent increase in secondary schools. There was an average increase in attendance of 2.5 per cent in primary schools (95 per cent CI: 1.6 – 3.34 per cent) and 8 per cent in secondary schools (95 per cent CI: 6.64 – 9.49 per cent). In relation to baseline attendance, this represents a 3 per cent increase in primary schools and 12 per cent increase in secondary schools.. For both primary and secondary schools, effect sizes are statistically larger when baseline attendance levels are lower, consistent with the observation that schooling-effect sizes are generally larger when there is a lower starting point. Drop-out rates were reduced by twice as much in secondary schools compared with primary schools, although this is reported with greater uncertainty as there were fewer reported observations in the included studies.

The authors found a positive relationship between the size of the cash-transfer and enrolment effects in both primary and secondary schools, in contrast to previous single-study findings that suggested decreasing returns relative to the size of the transfer. Programs where payments were made less frequently, for instance on a quarterly basis, saw larger effect sizes than those that paid on a monthly basis. Conditions imposed on secondary-school attainment, rather than attendance, were also positively associated with larger enrolment and attendance impacts.

Programmes that introduced educational supply-side reforms alongside the cash transfers saw significantly larger effects in primary-school enrolment. This effect was not observed for secondary-school enrolment. The authors hypothesise that this is because there are greater supply-side constraints in primary schools, where enrolment is already high in many cases.

Implications for further research

•  The authors recommend that clear reporting standards are introduced for CCT monitoring and evaluation, given their increasing popularity around the world.
• There were relatively few studies that reported the impact of CCTs on dropout rates and thus the effect size for this outcome was calculated with greater uncertainty. This could therefore be an area that deserves future research.

Background

Conditional cash transfers (CCTs) have become an increasingly popular social-protection intervention in recent years, particularly in Latin America, with the goal of ending inter-generational poverty. They typically involve a regular transfer of cash to a family, conditional on investment in a child’s schooling or health, for example enrolling children in school and ensuring regular attendance. The transfers are sometimes accompanied by supply-side support, such as textbook provision, recognising that many schools in low- and middle-income countries face considerable supply-side constraints. Many CCT programmes have been accompanied by at least one impact evaluation, and thus there is a reasonably large base of evidence, making this area highly suitable for a systematic review.

Research objectives

The authors aim to estimate the average effect size of conditional cash transfers on enrolment, attendance and drop-out rates of primary- and secondary-school children, and to investigate the programme characteristics and factors that cause heterogeneity in treatment effects.

Methodology

The authors systematically searched the published and grey literature for impact-evaluation studies that assessed the effects of a CCT programme on school enrolment, attendance and/or drop-out in a developing country. The authors searched for studies published in English and Spanish in more than 25 electronic databases, including EBSCO, EconLit, Eldis, British Library for Development Studies (BLDS), Google Scholar, International Bibliography of the Social Sciences (IBSS), the Cochrane Central Register of Controlled Trials and the World Bank. To be included, studies needed to use a treatment-comparison research design. Non-randomised studies were eligible for inclusion only if they reported the relevant pre-treatment characteristics of the treatment and comparison groups.

The authors used meta-analysis to estimate average effect sizes for enrolment, attendance, and drop-out rates in both primary and secondary schools, and meta-regression to analyse the relationship between the estimates of effect sizes and programme characteristics such as transfer size, type of conditionality and the frequency of disbursement.

Quality assessment

The authors have clear inclusion criteria, search a reasonably comprehensive set of databases and use appropriate methods to avoid bias in the selection of included studies as well as in data extraction. They also examine the likely explanatory factors behind differences in program outcomes, both descriptively and using meta-regression. However the review has some major limitations. The authors did not search reference lists of included studies or contact experts in the field for additional studies. Although the authors retrieved articles in English and Spanish, they did not search for studies written in other languages. Finally, there is no reported assessment of the risk of bias of included studies, and therefore there is no indication of the quality of the included literature and how this could bias their conclusions. This is of particular concern due to the relatively broad range of included studies.

N.B. The authors are currently working on a updated paper which will contain an updated universe of studies and a quality assessment of all included studies.

Link for the paper: http://www.prgs.edu/content/dam/rand/pubs/working_papers/2012/RAND_WR921-1.pdf