Publications by PSE researchers

This paper analyzes the implications of attrition for the internal and external validity of the results of four randomized experiments and proposes a new method to correct for attrition bias. We find that not including those found during the intensive tracking can lead to a substantial overestimation or underestimation of the intention-to-treat effects, even when attrition without such tracking is balanced. We propose to correct for attrition using inverse probability weighting with estimates of weights that exploit the similarities between missing individuals and those found during an intensive tracking phase.

Interventions targeting improvements in human capital are often motivated by their potential to break the intergenerational transmission of poverty from parents to children. This study contributes to the thin evidence base on these links by examining outcomes for children of former program beneficiaries of a conditional cash transfer (CCT) program, capitalizing on randomized variation in the timing and CCT’s impact on maternal human capital. We estimate intent-to-treat (ITT) differential effects on early childhood anthropometric and cognitive outcomes for 0–3-year-old children of program beneficiaries [N=366], as well as effects on key domains including nutrition, health, stimulation and the home environment. We find that moderately higher schooling for mothers (19–22 years old) who were the original program beneficiaries did not translate into improvements in anthropometrics or cognitive outcomes for their children. We also find no effects on behaviors commonly thought to be affected by higher education such as investments in nutrition and preventive health, or stimulation. Early program beneficiary mothers, however, had worse mental health outcomes and were more likely to use violent disciplinary practices such as spanking, threatening and punishing. Findings demonstrate the complexity of intergenerational mechanisms across genetic, biological, environmental and behavioral factors, and also suggest the importance of maternal mental health as a mechanism influencing child outcomes.

CONTEXT Agricultural innovations are inclusive when they are used by any member of society who wants to use them. Conversely, agricultural innovations that can only be used by a specific privileged group within society can be characterized as “exclusive”. OBJECTIVE The first objective of this paper is to examine the inclusivity of agricultural innovations in Ethiopia, using national representative data and considering a wide portfolio of innovations resulting from the collaborative research between CGIAR and its national partners. Second, we also examine how measurement error may affect how we characterize the inclusivity of agricultural innovations. METHODS We use nationally-representative survey data from Ethiopia (collected in 2018/19) in which best-practice measures of the adoption of a large number of agricultural innovations were embedded, including the adoption of CGIAR-related improved maize varieties measured using two different approaches: subjective, self-reported survey data; and objective DNA fingerprinting of crop samples taken from the same farmers' plots. A rich set of household variables is also collected in the survey, which allows characterizing the types of farmers that are adopting different innovations, and the extent to which conclusions regarding the inclusivity of innovations depends on the measurement of the latter. RESULTS AND CONCLUSIONS Many innovations are not disproportionately more likely to be adopted by male, larger, richer, or more connected farmers. When using self-reported data on adoption of improved maize varieties, adoption appears positively correlated with having larger landholdings and households with lower female participation in agriculture, and negatively correlated with poorer households (being among the bottom 40% of consumption distribution). Substituting survey responses with the results of DNA fingerprinting these correlations disappear, with farm size, gender and poverty status no longer predictive of adoption. SIGNIFICANCE The results suggest the potential value of offering a menu of innovations to farmers to increase inclusivity, as it allows each farmer to be a critical consumer of potential innovations and select those that best correspond to their own needs and constraints. We also highlight how important data quality is in ensuring we have correct information about inclusive innovation.

This paper presents results from an RCT in 140 schools in Madagascar that targets both hygiene practices and menstrual stigma. First, we show that a bundle of interventions (sanitation infrastructure, menstrual products, and teacher sensitization) leads to substantial (0.15 SD) improvements in learning tests and school marks, without affecting attendance or health. These learning benefits appear to be driven by reduced stress and an improved psychosocial environment in treatment schools, where girls’ heart rate at endline is lower (-0.12 SD), severe bullying is less common (-0.08 SD), and a measure of network integration is higher (+0.24 SD). Second, we evaluate the additional effect of nominating and coaching “young girl leaders” – school girls willing to speak out against menstrual stigma – to spread positive messages about hygiene and menstruation. The combined program generates substantial improvements in hygiene knowledge and behavior (0.33-0.56 SD) and in menstrual stigma (0.74 SD), and the Young Girl Leader component significantly increases the impact on all of these dimensions.

Conditional cash transfer programs are the anti-poverty program of choice in many developing countries, aiming to improve human capital and break the intergenerational transmission of poverty. A decade after a randomized 3-year CCT program began, earlier exposure during primary school ages when children were at risk of dropout led to higher labor market participation for young men and women and higher earnings for men. Results highlight the roles of the different program components with variation in timing of access to nutrition, health and education investments translating into substantial differential effects on learning for men and reproductive health outcomes for women.

Since the 1960s, the increased availability of modern seed varieties in developing countries has had large positive effects on households’ well-being. However, the effect of related land use changes on deforestation and biodiversity is ambiguous. This study examines this question through a randomized control trial in a remote area in the Congo Basin rainforest with weak input and output markets. Using plot-level data on land conversion combined with remote sensing data, we find that promotion of modern seed varieties did not lead to an increase in overall deforestation by small farmers. However, farmers cleared more primary forest and less secondary forest. We attribute this to the increased demand for nitrogen required by the use of some modern seed varieties, and to the lack of alternative sources of soil nutrients, which induced farmers to shift towards cultivation of land cleared in primary forest. Unless combined with interventions to maintain soil fertility, policies to promote modern seed varieties may come at the cost of important losses in biodiversity.

We examine the effects of a subsidy program in Haiti that provided smallholders subsidies for inputs (rice seeds, fertilizer, pesticides, and specific labor tasks) using a randomized control trial. The program led to lower input use and lower yields in the year subsidies were received, and the decline in input use and yields persisted through the following year. Using data from a complementary information intervention in which randomly selected farmers were provided clarification regarding their future receipt of vouchers, we find evidence suggesting that incorrect expectations of future transfers partially explain the disappointing outcomes. In addition, instead of increasing input use, the subsidies seem to have led farmers to pay off their loans and take fewer new ones. In complex post-emergency environments such as the one in which this program took place, input subsidies may need to be avoided, as they require considerable information to optimally design and careful coordination by many actors to achieve the expected gains.

Background and objectives Early Childhood Development is high on the policy agenda in Côte d’Ivoire, where the government has identified it as part of its overall approach to improve human capital outcomes. This paper describes a multi-partner approach to piloting, monitoring, adaption, testing and scaling of parental training for ECD. It discusses the learnings from the pilots, and present early evaluation results from two RCTs, focusing on parental participation in trainings and acceptability of messages, with the objective to inform national scaling strategies. As such, this paper illustrates how “MEL systems contributed to ensuring that positive early childhood development (ECD) outcomes were improved as interventions were seeking to achieve scale,” one of the research questions outlined in the call description for the special issue. The paper further provides a real-world example of “How MEL systems can support contributions and buy-in from a variety of stakeholders as ECD interventions (seek to) achieve impacts at scale (e.g., through the public system)? Methods Five training approaches to improve caregivers’ knowledge and practices around nutrition, preventive health, stimulation, and disciplining were piloted at small scale between 2018 and 2020. An intensive process evaluation was embedded to identify strengths and weaknesses, adapt through an iterative phase, and ultimately make recommendations for their scale up against 11 defined criteria. In early 2021, the two most promising approaches were scaled through two clustered randomized control trials to more than 150 villages each. A cost-effectiveness study was designed in consultation with government stakeholders, centered around targeting different caregivers and decision makers in the household and the extended family and on enhancing community interactions around ECD. Results The evaluation of the five pilots identified one model recommended to be scaled, and one other model to scale after further adaptations. Monitoring and evaluation data from the two models at scale show high levels of participation and acceptability of core messages. Experimental variations involving community champions and fathers increase participation. Conclusion The iterative and multi-partner process led to two models of parenting training that have wide acceptability. Future work will analyze impacts on cognitive and socio-emotional outcomes, together with cost analysis.

We take stock of the major changes in methodology for studying the impacts of international agricultural research, focusing on the period 2006–2020. Impact assessment of agricultural research has a long and recognized tradition. Until the mid-2000s, such assessments were dominated by a model of demand for and supply of agricultural products in partial equilibrium. The basic ideas for this approach were sketched out by Griliches more than half a century ago. We describe the implications of heightened standards of evidence for good practice in three domains of research design: causal inference, valid measurement, and statistical representativeness. We document advances in each of these domains and review recent evidence that demonstrates the lessons that can be learned from adopting these practices, emphasizing the importance of evidence at-scale, the need to consider portfolios of innovations at a national level, and the challenges of accounting for innovations that are promoted as bundles.

We present experimental evidence on a programme aimed at improving households' risk management through income diversification. The intervention targeted rural Nicaraguan households exposed to weather variability and combined a one-year conditional cash transfer with vocational training or a productive investment grant. Both complementary interventions provided protection against weather shocks two years after the programme ended. Households that received the productive investment grant also had higher average consumption levels. The complementary interventions facilitated income smoothing and diversification of economic activities. Relaxing capital constraints induced investments in non-agricultural businesses, while relaxing skills constraints increased wage work and migration in response to shocks.

Measures of cognitive, noncognitive, and technical skills are increasingly used in development economics to analyze the determinants of skill formation, the role of skills in economic decisions, or simply because they are potential confounders. Yet in most cases, these measures have only been validated in high-income countries. This paper tests the reliability and validity of some of the most commonly used skills measures in a rural developing context. A survey with a series of skills measurements was administered to more than 900 farmers in western Kenya, and the same questions were asked again after three weeks to test the reliability of the measures. To test predictive power, the study also collected information on agricultural practices and production during the four following seasons. The results show the cognitive skills measures are reliable and internally consistent, while technical skills are difficult to capture and very noisy. The evidence further suggests that measurement error in noncognitive skills is non-classical, as correlations between questions are driven in part by the answering patterns of the respondents and the phrasing of the questions. Addressing both random and systematic measurement error using common psychometric practices and repeated measures leads to improvements and clearer predictions, but does not address all concerns. The paper provides a cautionary tale for naïve interpretations of skill measures. It also points to the importance of addressing measurement challenges to establish the relationship of different skills with economic outcomes. Based on these findings, the paper derives guidelines for skill measurement and interpretation in similar contexts.

We examine the effects of a subsidy program in Haiti which provided smallholders subsidies for modern inputs (rice seeds, fertilizer, pesticides and specific labor tasks) through a randomized control trial. The program led to lower input use and lower yields in the year subsidies were received, and the decline in input use and yields persisted through the following year. Using data from a complementary information intervention in which randomly selected farmers were provided clarifications regarding their status in the program, we find evidence suggesting that incorrect expectations of future transfers help explain the disappointing outcomes. In addition, instead of increasing input use, subsidies seem to have led farmers to pay off their loans and take fewer new ones. In a complex postemergency environment as the one in which this program took place, input subsidies may need to be avoided, as they require considerable information to optimally design and careful coordination by many actors to achieve the expected gains.