Cash or condition?: evidence from a cash transfer experiment

Replication paper
Original publication: The Quarterly Journal of Economics
Original researchers: Sarah Baird, Craig McIntosh, Berk Ozler
Replication plan:

Reimao’s Replication plan

Replication researchers: Maira Reimao
Current status: Completed Replication Study

The Original Study

With the increasing popularity of conditional cash transfers has come the question of whether their generally positive results arise from the cash transfer itself or from the conditionality tied to it. “Cash or Condition: Evidence from a Cash Transfer Experiment” looks into this issue with respect to transfers made to girls and their parents, and their effect on schooling, test scores, and early marriage and pregnancy. This particular study is part of a broader randomized control trial (RCT) in the Zomba district of Malawi. Here, the authors focus on girls 13-22 years who were never married and were enrolled in school in 2008. There are three relevant arms for this paper: a group receiving unconditional cash transfers (UCT); a group receiving cash transfers conditional on school attendance (CCT); and a control group. In the two treatment groups, transfers were randomly assigned to both the girls and their household.

Baird et al. find that the CCT increased school enrolment and some test scores, while the effect of the UCT on these education outcomes was not statistically significant. In contrast, the CCT had no effect on the incidence of marriage or pregnancy by the end of the intervention, while the UCT significantly decreased the likelihood of both outcomes. This surprising finding is driven by the mechanics of the transfers on school drop-outs: in the CCT arm, drop-out girls do not receive the transfer, but in the UCT arm they continue to receive it. The recurring income shields this latter group from early marriage and pregnancy. These findings fit into the growing literature and call for comparing interventions (often CCTs) to the default of giving cash, but indicate that they might not always be comparable, as CCTs and UCTs may achieve different things. CCTs are better at encouraging compliance with the condition itself and achieving outcomes closely linked to it, while UCTs provide benefits – and potentially positive outcomes – to households that may be unable or unwilling to comply with conditions.

The Replication

The research behind “Cash or Condition” is valuable not just for its contribution to the debate on the benefits and effectiveness of UCTs versus CCTs, but also for its place in the broad discussion on using unconditional transfers as the benchmark for other transfer programs. Its results support the idea that, without conditionalities, transfers can uniquely benefit the most vulnerable households in ways that might not be comparable to a more rigid design or transfer. After a pure replication exercise to verify these original results, the replication of this study will focus on three areas of investigation. First, it will explore whether there are differentiated effects for girls who have a stronger attachment to school (i.e., those who are actually of school age or who are close to the right age for their grade level), in contrast to those with a history of drop-outs or grade repetition. Second, it will study the role that an additional transfer that was made to girls assigned to the UCT arm and technically eligible for secondary school may have had, as girls in the CCT arm did not receive this additional boost directly. Lastly, it will investigate in more detail the concentration of the positive effect of UCTs in deterring marriages and pregnancy on girls who drop out of school by addressing the endogeneity of this group. Other robustness checks are discussed in more detail within the full replication plan, and may depend on data availability.