Center for Effective Global Action

Scarce information and human capital may make it difficult for residents of developingcountries to produce accurate forecasts, limiting responses to uncertain future eventslike air pollution. We study two randomized interventions in Lahore, Pakistan: 1)provision of air pollution forecasts; 2) general training in forecasting. Both reducedsubjects’ own air pollution forecast errors; the training effect suggests that modesteducational interventions can durably improve forecasting skills. Forecast receipt increased demand for protective masks and increased the responsiveness of outdoor time to pollution. Forecast recipients were willing to pay 60 percent of the cost of mobile internet for continued access.

Slack – the underutilization of factors of production – varies systematically with economic development. Using novel and detailed measures of the utilization of labor and capital from a large representative sample of firms in rural and urban Kenya, we show that utilization is increasing in firm size, market access, and economic activity. We present a model of firm capacity choice where indivisibility in at least one input is a key driver of slack. We embed the model in spatial general equilibrium, with features characteristic of low-income settings – including many small firms and high transport costs – and show that it rationalizes both the endogenous emergence of slack in steady-state and elastic aggregate supply curves. We empirically validate model predictions using reduced-form estimates of the general equilibrium effects of cash transfers from a large-scale RCT in Kenya. The parsimonious model replicates much of the experimental evidence, predicting a large real multiplier of 1.5, driven by expansion in low-utilization sectors and firms, and limited average price inflation. Counterfactual analyses indicate that multipliers are likely to be meaningfully smaller in lower slack settings, such as urban areas. We use the model to revisit the estimation of spatial spillovers in clustered RCTs and uncover non-trivial ’missing intercept’ effects on income and inflation. Additionally, we innovate methodologically by pre registering key elements of model estimation and validation. The findings suggest that input indivisibilities and slack are key features of developing country settings, and are quantitatively important for macroeconomic dynamics and policies.

We measure the impact of two main signals of tertiary-level human capital accumulation, college quality and certification, on hiring in India. Using a correspondence experiment, we send 16,944 resumes to 1412 job postings for recent engineering graduates at small and medium firms. In precisely estimated results, we find that these employers do not respond to signals of tertiary education quality. Specifically, there is no impact on callbacks of having graduated from a mid-tier college ranked in the top 300 relative to an unranked college outside of the top 1000, despite significant government investment in college rankings. There is also no impact of scoring in the highest as opposed to the lowest quartile of a post-tertiary certification test that has been taken by millions of graduating students. There is evidence that women modestly benefit in the first stage of hiring in this market, with this effect concentrated in some regions.

Can increasing women’s use of digital financial services raise their empowerment? We test this hypothesis using a randomized control trial with 152 female microfinance groups in Tanzania, where treated groups were randomly switched to repay their loan using mobile money instead of cash. This exogenous shift in women’s use of mobile money for loan repayment substantially increases their use for other types of transactions. Women’s control over their finances increases, they have higher levels of empowerment in the household and expenditures shift towards goods plausibly aligned with their preferences. These findings highlight the benefits of greater use of digital technologies for women.

The proliferation of mobile phones in low- and middle-income countries has suddenly and dramatically increased the extent to which the world’s poorest and most vulnerable populations can be observed and tracked by governments and corporations. Millions of historically “off the grid” individuals are now passively generating digital data; these data, in turn, are being used to make life-altering decisions about those individuals — including whether or not they receive government benefits, and whether they qualify for a consumer loan. This paper develops an approach to implementing algorithmic decisions based on personal data, while also providing formal privacy guarantees to data subjects. The approach adapts differential privacy to applications that require decisions about individuals, and gives decision makers granular control over the level of privacy guaranteed to data subjects. We show that stronger privacy guarantees typically come at some cost, and use data from two real world applications — an anti-poverty program in Togo and a consumer lending platform in Nigeria — to illustrate those costs. Our empirical results quantify the tradeoff between privacy and predictive accuracy, and characterize how different privacy guarantees impact overall program effectiveness. More broadly, our results demonstrate a way for humanitarian programs to responsibly use personal data, and better equip program designers to make informed decisions about data privacy.

How do inaccurate priors about the distribution of interest rates affect search and outcomes in consumer credit markets? Consumer credit markets feature large amounts of within-borrower price dispersion in interest rates; if consumers are unaware of the extent of this price dispersion, they may shop less and take out loans at higher interest rates than they would otherwise. We conducted a randomized controlled trial with 112,063 loan seekers in Chile where we showed treated participants a price comparison tool that we built using administrative data from Chile’s financial regulator. The tool shows loan seekers a conditional distribution of interest rates based on similar loans obtained recently by similar borrowers, using data on the universe of consumer loans merged with borrower characteristics. We also cross-randomized whether we asked participants their priors about the distribution of interest rates. We find that consumers thought interest rates were lower than they actually were, and the price comparison tool caused them to increase their expectations about the interest rate they would obtain by 56%. Consumers also underestimated price dispersion, and our price comparison tool caused them to increase their estimates of dispersion by 69%. The price comparison tool did not cause people to search or apply at more institutions, but it did cause them to receive 13% more offers and 11% lower interest rates, and to be 28% more likely to negotiate with their lender and 4.7% more likely to take out a loan. In contrast, merely asking participants their expectations about interest rates led them to search at 4% more institutions and obtain 9% lower interest rates.

This randomized experiment on a social media platform in Pakistan measures the impact of two treatments that use ex-ante moderation to control misinformation about the COVID-19 pandemic compared to standard ex-post moderation. One treatment never posts misinformation, while the other rebuts it. We also disseminate official information about the pandemic on the platform. The treatments reduce daily users by 19%. This reduces exposure to official information by 29% more than exposure to misinformation. A conceptual framework posits that this can be explained by the fact that, in this setting, official information is more trusted and disseminated than misinformation.

We measure the impact of three components of air pollution on daily labor productivity in call centers in five Indian cities.  We find that a one standard deviation increase in fine particulate matter (PM_2.5), a pollutant that has been the primary focus of the literature on the harms of air pollution, has a large negative effect of 0.15σ on productivity.  Notably, we find a comparable negative effect for a one standard deviation increase in carbon monoxide of 0.14σ as well as a negative effect of 0.09σ from ozone. In summing air pollution harms across our sample, carbon monoxide is responsible for more than half of the total productivity lost, which is more than double the losses attributable to PM_2.5. Our results underscore the importance of considering components of air pollution beyond particulate matter.  To that end, we illustrate the potential productivity impacts of a national policy in India that targets PM_2.5 compared to a counterfactual policy that also targets carbon monoxide and ozone.

How can we encourage the adoption of new digital financial services? We use an RCT with 115 microfinance groups in Ghana to understand the respective roles of individual incentives to adopt a new technology and endorsement of the technology from a trusted peer. We study this in the context of mobile banking services, a technology allowing transfers between a mobile phone and bank account, dramatically lowering the costs of accessing the bank account. We find that while individual incentives increase adoption of mobile banking services by 50%, adding endorsement by a peer doubles the impact of the individual subsidy alone. Peer endorsement is particularly effective at increasing confidence in dealing with fraud and peer interaction around mobile banking. The increased use of mobile banking services leads to 30% higher savings in the linked bank account 6 months later. Our study highlights the importance of peers in encouraging technology adoption and facilitating formal financial inclusion.

Globally, the number of forcibly displaced individuals has surpassed 100 million, with approximately80 percent situated in low- or middle-income countries. In these resource-constrained settings the magnitude and protracted nature of displacement poses an increasing challenge. Recent policy discourse has begun to shift away from emergency response interventions towards those that promote investments inthe human capital and self-reliance of displaced populations. This paper reviews evidence of the impacts and costs of eleven interventions designed to improve the livelihoods of forcibly displaced people in low-and middle-income settings. The study team finds suggestive evidence that graduation-style approaches and cash transfers can improve people’s self-employment, wages, engagement in paid work and wellbeing. However, too few studies have been conducted among the exact populations and settings of interest to discern clear strategies for adapting interventions for success in every context. To address this gap, the authors introduce a primitive taxonomy of contextual factors, which encourages more robust discussion of how context impacts interventions. It will be important for future research to both evaluate a broader range of interventions, and to more systematically identify the effects of context to advance our understanding of how to improve the economic wellbeing of displaced people.