papers

We benchmark a multi-dimensional child nutrition intervention against an unconditional cash transfer of equal cost. Randomized variation in transfer amounts allows us to estimate impacts of cash transfers at expenditure levels equivalent to the in-kind program, as well as to estimate the return to increasing cash transfer values. While neither the in-kind program nor a cost-equivalent transfer costing $124 per household moves core child outcomes within a year, cash transfers create significantly greater consumption than the in-kind alternative. A larger cash transfer costing $517 substantially improves consumption and investment outcomes and drives modest improvements in dietary diversity and child growth.

We present results of an experiment benchmarking a workforce training program against cash transfers for underemployed young adults in Rwanda. 3.5 years after treatment, the training program enhances productive time use and asset investment, while the cash transfers drive productive assets, livestock values, savings, and subjective well-being. Both interventions have powerful effects on entrepreneurship. But while labor, sales, and profits all go up, the implied wage rate in these businesses is low. Our results suggest that credit is a major barrier to self-employment, but deeper reforms may be required to enable entrepreneurship to provide a transformative pathway out of poverty.

We use a randomized experiment to compare a workforce training program to cash transfers in Rwanda. Conducted in a sample of poor and underemployed youth, this study measures the impact of the training program not only relative to a control group but relative to the counterfactual of simply disbursing the cost of the program directly to beneficiaries. While the training program was successful in improving a number of core outcomes (productive hours, assets, savings, and subjective well-being), cost-equivalent cash transfers move all these outcomes as well as consumption, income, and wealth. In the head-to-head costing comparison cash proves superior across a number of economic outcomes, while training outperforms cash only in the production of business knowledge. We find little evidence of complementarity between human and physical capital interventions, and no signs of heterogeneity or spillover effects.

This paper reports on a two-tiered experiment designed to separately identify the selection and effort margins of pay-for-performance (P4P). At the recruitment stage, teacher labor markets were randomly assigned to a 'pay-for-percentile' or fixed-wage contract. Once recruits were placed, an unexpected, incentive-compatible, school-level re-randomization was performed, so that some teachers who applied for a fixed-wage contract ended up being paid by P4P, and vice versa. By the second year of the study, the within-year effort effect of P4P was 0.16 standard deviations of pupil learning, with the total effect rising to 0.20 standard deviations after allowing for selection.

Despite widely documented shortfalls of teacher skills and effort, there is little systematic evidence of rates of teacher turnover in low-income countries. I investigate the incidence and consequences of teacher turnover in Rwandan public primary schools over the period from 2016--2019. To do so, I combine the universe of teacher placement records with student enrollment figures and school-average Primary Leaving Exam scores in a nationally representative sample of 259 schools. Results highlight five features of teacher turnover. First, rates of teacher turnover are high: annually, 20 percent of teachers separate from their jobs, of which 11 percent exit from the public-sector teaching workforce. Second, the burden of teacher churn is higher in schools with low learning levels and, perhaps surprisingly, in low pupil-teacher-ratio schools. Third, teacher turnover is concentrated among early-career teachers, male teachers, and those assigned to teach Math. Fourth, replacing teachers quickly after they exit is a challenge; 23 percent of exiting teachers are not replaced the following year. And fifth, teacher turnover is associated with subsequent declines in learning outcomes. On average, the loss of a teacher is associated with a reduction in learning levels of 0.05 standard deviations. In addition to class-size increases, a possible mechanism for these learning outcomes is the prevalence of teachers teaching outside of their areas of subject expertise: in any given year, at least 21 percent of teachers teach in subjects in which they have not been trained. Taken together, these results suggest that the problem of teacher turnover is substantial in magnitude and consequential for learning outcomes in schools.

In 2013, Tanzania introduced “Big Results Now in Education”, a low-stakes accountability program that published both nationwide and within-district school rankings. Using data from the universe of school performance from 2011-2016, we identify the impacts of the reform using a difference-in-differences estimator that exploits the differential pressure exerted on schools at the top and bottom of their respective district rankings. We find that BRN improved learning outcomes for schools in the bottom two deciles of their districts. However, the program also led schools to strategically exclude students from the terminal year of primary school.

Debit cards reduce the travel distance to access bank accounts and can increase financial inclusion. We show that in Mexico, cash transfer beneficiaries who already received their transfers in bank accounts and subsequently received debit cards reduce their median distance to access the account from 4.8 to 1.3 kilometers. They also report being less likely to forgo important activities (childcare, work) to withdraw their transfers. Using account level data, we find a strong correlation between the reduction in travel distance and financial activity: beneficiaries facing the largest reductions in distance increase both their number of withdrawals and their savings balances.

We present results from a field experiment using price incentives to address two obstacles to women's access to formal land ownership. First, the price of formal land titles is often prohibitive, restricting access for both men and women. Second, when formal titles are issued, men are recorded as the sole owners of household land in the vast majority of cases. In theory, these problems may be in tension: requirements to grant women equal ownership may reduce the overall household demand for titling. In practice, we find no such trade-off: when residents of low-income, unplanned settlements in Dar es Salaam were offered price discounts for formal land titles, demand rose significantly. Price elasticity of demand was unchanged when price discounts were conditional on registering a woman as (co-)owner of household land. Furthermore, conditional price discounts achieved near gender parity in land ownership.

The low demand for microinsurance products observed in recent studies is characterized by two stylized facts: first, demand is negatively correlated with measures of risk aversion, in contrast to simple models; and second, demand is positively associated with measures of trust. In this paper we develop a model of the demand for indemnity insurance that explains both of these stylized facts, based on the idea that individuals have limited trust in the credibility of the insurer's promise to deliver a payment in the event of an indemnified loss. The model has the further testable implication that, when trust in insurer credibility is heterogeneous across individuals, those with low trust are more sensitive to variation in premium costs. We test and find support for the predictions of this model using data from a randomized, controlled trial of composite health microinsurance product among tea growers in Nyeri, Kenya. To do so we combine experimental variation in prices with measures of risk preferences and trusting behavior from a laboratory experiment conducted in the field at baseline. The results suggest scope for policies that increase insurance purchases by improving potential clients' confidence in the enforceability of the insurance contract.

Community-based monitoring of public services provides a possible solution to accountability problems when state oversight is limited. However, the mechanisms through which such policies can be effective are not well understood. Are community-monitoring interventions successful because they improve information alone, or do they also need to overcome collective action problems? We investigate this question by implementing a combined field and lab experiment in 100 Ugandan primary schools, which randomly assigns schools and their Management Committees (SMCs) either to standard community-based monitoring, to a participatory variation that addresses coordination problems, or to a control group. We find substantial impacts of the participatory treatment on pupil test scores as well as pupil and teacher absenteeism, while the standard treatment has small and insignificant effects, and we develop a test using randomization inference to show that differences in these outcomes between treated groups are statistically significant. Combining this evidence with SMC member behavior in laboratory games, we find evidence that improved collective action explains these differences. The results have implications for the design of community-based monitoring policies, and help to explain their variable effectiveness across contexts.

Adoption rates of apparently high return agricultural technologies are puzzlingly low in many developing countries. While recent work has suggested that these high average returns are consistent with substantial heterogeneity, relatively little is known about the mechanisms that relate adoption decisions to heterogeneity in realized returns. In this paper, I derive distinct predictions for models in which this relationship is driven by precautionary savings and by learning -- mechanisms with distinct policy implications. These are tested using unique data on the Cocoa Abrabopa Association, which provided a package of fertilizer and other inputs on credit to cocoa farmers in Ghana. Exploiting the expansion of the program to identify impacts, I find high but heterogeneous returns. Low experienced returns among adopters are associated with low program retention rates, and patterns of returns and adoption support the hypothesis that disadoption by farmers who experience low returns is driven by learning rather than precaution. This learning mechanism is also seen in the relationship between measures of subjective expectations and output realizations. Taking into account subjective uncertainty in the learning process, structural estimates suggest that farmers' choices are consistent with reasonable levels of risk aversion.

This paper tests the external validity of a simple Dictator Game as a laboratory analogue for a naturally occurring policy-relevant decision-making context. In Uganda, where teacher absenteeism is a problem, primary school teachers’ allocations to parents in a Dictator Game are positively but weakly correlated with their time allocations to teaching and, so, negatively correlated with their absenteeism. Guided by a simple theoretical model, we find that the correlation can be improved by allowing for (a) variations in behavioural reference points across teachers and schools and (b) the positive effect of some School Management Committees on teacher attendance.