The COVID-19 pandemic closed schools at one point for over 1.6 billion children, with potentially long-term consequences. Angrist, Bergman, and Matsheng provide some of the first experimental evidence on strategies to minimize the fallout of the pandemic on learning. They evaluate two low-technology interventions to substitute schooling during this period: SMS text messages and direct phone calls. They conduct a rapid trial in Botswana to inform real-time policy responses, collecting data in multiple waves. The researchers find that phone calls and SMS messages result in cost-effective learning gains of 0.12 standard deviations. They cross-randomize targeted instruction, customizing instruction to a child's learning level using data collected during the trial. They find evidence that targeted instruction can be more effective than non-targeted instruction, especially for SMS messages which have no effect on their own if they are not targeted. Learning gains are robust to a variety of tests, such as randomized problems of the same proficiency and measures of effort on the test. Parents update their beliefs about their child's learning in tandem with progress and they feel greater self-efficacy to support their child's learning. The "low-tech" interventions tested have immediate policy relevance and could have long-run implications for the role of technology and parents as substitutes or complements to the traditional education system.
In addition to the conference paper, the research was distributed as NBER Working Paper w28205, which may be a more recent version.
Experimental Evidence from India
How should recipients of government transfers prove their identity? ID requirements must trade off reductions in fraud and corruption against the risk of denying benefits to legitimate beneficiaries. Muralidharan, Niehaus, and Sukhtankar evaluate reforms that integrated more stringent, biometric ID requirements into India's largest social protection program, using large-scale randomized and natural experiments. Theoretically, the effects depend on the underlying mechanics of corruption. Empirically, corruption fell but with substantial costs to legitimate beneficiaries: approximately 2 million of whom lost access to benefits at some point during the reforms. The results highlight that attempts to reduce corruption in welfare programs can also generate non-trivial costs in terms of exclusion and inconvenience to genuine beneficiaries.
Evidence from Clogged Courts of India
How do judicial institutions, such as sub-national courts, impact economic growth? Rao examines the effect of trial court capacity on local firms' performance by exploiting quasi-random variation in judge vacancies and mapping trial records for a third of such courts in India with court-level performance measures, bank lending, and firm outcomes. Rao finds that reducing judge vacancy increases local firms' labor use, production, and profitability through improved access to bank credit arising from better enforcement of debt contracts. Addressing judge vacancy would generate at least 12:1 benefit-cost ratio.
The mechanization of production has become a primary feature of modern agriculture and is central to agricultural labor productivity. Caunedo and Kala estimate the returns to mechanization and its impact on labor combining a randomized controlled experiment and a structural model of task-replacement. Treatment farmers were given subsidy vouchers to access agricultural equipment from nearby custom hiring centers (CHC). In addition, a subset of treatment farmers were given cash transfers. The voucher treatment increases overall mechanization hours, with an intent to treatment effect size of about 0.13 standard deviations (a treatment on the treated effect size of 0.36 standard deviations). The researchers find no significant improvements in output per acre due to mechanization on average. However, family labor decreases in response to the subsidy in capital, and farmers reduce hired labor in all farming processes, including those not directly affected by mechanization. The researchers document that family labor is mostly occupied in supervision activities, and that their lower engagement in farming is associated with higher non-agricultural income. The decline in supervision labor and the decline in hired labor across farming processes are interpreted as evidence of output standardization, which is beneficial in the presence of contracting frictions. The researchers use key elasticities from the experiment and our structural model to infer the marginal return to mechanizable tasks, which they estimate at 36% per season, and to estimate the relative importance of different channels in response to mechanization.
What roles do credit constraints and inattention play in the under-adoption of high-return technologies? Berkouwer and Dean study this question in the case of energy-efficient cookstoves in Nairobi. Using a randomized field experiment with 1,000 households, they estimate a 300% average annual rate of return to investing in this technology, or $120 per year in fuel savings -- around one month of income. Despite this, adoption rates are low: Eliciting preferences using an incentive-compatible Becker-DeGroot-Marschak mechanism, the researchers find that average willingness-to-pay (WTP) is only $12. To investigate what drives this puzzling pattern, the researchers cross-randomize access to credit with two interventions designed to increase attention to the costs and benefits of adoption. The first main finding is that credit doubles WTP and closes the energy efficiency gap over the period of the loan. Second, credit works in part through psychological mechanisms: around one-third of the total impact of credit is caused by inattention to loan payments. No evidence is found of inattention to energy savings. Private benefits and avoided environmental damages generate average benefits of $600 for each stove adopted and used for two years.
Chalendard, Fernandes, Rijkers, and Raballand present a new methodology to detect collusion in customs and apply it to Madagascar's main port. Manipulation of assignment of import declarations to inspectors is identified by detecting deviations from random assignment prescribed by official rules. Deviant declarations are more at risk of tax evasion, yet less likely to be deemed fraudulent by inspectors, who also clear them faster. An intervention in which inspector assignment was delegated to a third party validates the approach, but also triggered a novel manifestation of manipulation that perpetuated systemic corruption. Tax revenues for deviant declarations would have been 27% higher had their assignment not been tampered with.
People's value of their own time is essential to the evaluation of public policies affecting recipients' time use: these evaluations should account for time spent on the intervention rather than other work or leisure. Using choice data collected from farming households in western Kenya, Agness, Baseler, Chassang, Dupas, and Snowberg find that households exhibit behavioral biases resulting in non-transitive preferences over their own time. Consequently, the data imply a range of possible values of time rather than a single number. The researchers discuss possible interpretations of these preferences, including loss aversion and self-serving biases, and the welfare implications of these different interpretations. A reasonable rule of thumb is to value time at 60% of the local wage. The researchers suggest steps experimenters can take to obtain and interpret estimates of value of time in the field.
Brooks, Donovan, Johnson, and Oluoch-Aridi deliver an unconditional cash transfer equal to one month's average profit to a randomly selected group of Kenyan female microenterprise owners in May 2020 at the outset of exponential growth in COVID-19 cases. Firm profit, inventory spending, and food expenditures increased relative to a control group. Entrepreneurs recovered about one-third of the profit lost during the crisis. The transfers caused greater business activity by inducing previously closed businesses to re-open. PPE spending and precautionary management practices increase to mitigate this effect, but only among those who perceive major health risk from COVID-19. The results suggest cash transfers promoted economic stabilization during the crisis, but may work against the public health goal of reducing interpersonal interaction.
Using a randomized experiment with an automobile manufacturing firm in China, Cai and Wang measure the effects of letting workers evaluate their managers on worker and firm outcomes. In the treatment teams, workers evaluate their supervisors monthly. The researchers find that providing feedback leads to significant reductions in worker turnover and increases in team-level productivity. In addition, workers report higher levels of happiness and positive mood. The evidence suggests that these results are driven by changes in the behavior of managers and an overall better relationship between managers and workers.
In addition to the conference paper, the research was distributed as NBER Working Paper w27680, which may be a more recent version.
Despite advantages in technology and human capital, multinational firms may operate less effectively than their local competitors in markets plagued by corruption and conflict. Rexer studies the effects of divestment to local firms in the context of a two-decade indigenization drive in Nigeria's turbulent oil sector, during which the share of local production grew substantially. Local takeover considerably increases oilfield output and reduces the share of nonproducing assets. Local firms increase output by mitigating conflict risk: oil theft, maritime piracy, and violence by criminal-militant groups all fall following local takeover. However, since local firms are less efficient, divestment leads to increased operational oil spills and gas flaring, magnifying the environmental externalities of oil production. A simple bargaining model illustrates that organized crime operates a protection racket, local firms' lower bargaining costs allow them to buy protection more cheaply, explaining their superior output performance despite lower technical efficiency. Rexer finds evidence that connections to high-level politicians and the security forces drive local firms' advantage in reducing criminal activity.
Despite the growing evidence suggesting that lockdown measures increased parental stress and child maltreatment, how to counteract such consequences of COVID-19 remains an open question. Dinarte, Amaral, Perez, and Rivera estimate the causal effect of parental stress management and social norms around child maltreatment on the quality of parent-child interactions in El Salvador. Using an individual-level randomized control trial design, they test the impact of a high-dosage and free SMS intervention delivered over three months to 4,718 parents. The intervention is a remote-delivery adaptation of the in-person intervention shown to have successfully reduced stress and violent behavior in other populations. Caregivers in the treatment group receive three SMS/WhatsApp messages per week for three months. The messages contain information, exercises, videos, and links to a blog with additional content. The blog will store all content and exercises from previous weeks, allowing caregivers to access the material on-demand. Despite the growing evidence suggesting that lockdown measures increased parental stress and child maltreatment, how to counteract such consequences of COVID-19 remains an open question. The researchers estimate the causal effect of an intervention providing parental stress management and positive parenting techniques on attitudes about child maltreatment and the quality of parent-child interactions. Using an individual-level randomized control trial design, they test the impact of a free high-dosage SMS intervention delivered over three months to parents residing with children aged eight years or younger in El Salvador. The intervention is a remote-delivery adaptation of an in-person intervention shown to have successfully reduced stress and violent behavior in other populations. Caregivers in the treatment group receive four SMS/WhatsApp messages per week for three months. The messages contain information, exercises, videos, and links to a blog with additional con-tent, which caregivers can access on-demand.
The researchers conducted a baseline survey with 2,344 fathers and 2,374 mothers in our treatment and control groups. they collected various measures on parental investments and mental health through validated instruments and the use of vignettes. The survey was conducted before the intervention when caregivers were already two months into the lockdown. The baseline shows that, when compared to fathers, mothers exhibit twice as high above-average scores for depression, anxiety, and stress. For fathers, above-normal anxiety affects over 20 percent of men. For 30 percent of parents, physical punishment is considered a useful tool for disciplining children. Fathers are more likely to tolerate physical punishment. Unemployment at the start of the quarantine increased for both parents but, more so, for fathers. On average, both caregivers increased the time spent with children by 2 hours per week.
Parents have currently undergone two months and a half of intervention, and compliance with treatment – individuals who open the messages – remains stable at 70 percent. Incentivized content knowledge surveys reveal an increased knowledge of stress management and positive parenting techniques among treated caregivers compared to the control group. The researchers hypothesize that such improvement in information on managing stress and positive parenting improves mental health and pa-rental time investments. They expect such improvement to lower child maltreatment. Together, these findings could inform how to adapt interventions into a remote delivery format to mitigate the short and long-run negative consequences of stay-at-home orders and COVID-19-related stress.
Misconduct -- market actions that are unethical and indicative of fraud or wrongdoing -- is a significant yet poorly understood issue that underlies many economic and financial transactions. Does misconduct in markets matter? When and how does reputation acts as a discipline against market misconduct? Annan and Sanoh design a field experiment to study the impact of two-sided anti-misconduct information programs on markets, which they deploy on the local markets for mobile money (Human ATMs) in Ghana. They show that, at baseline, these markets are characterized by substantial imperfect information, consumer mistrust, and vendor misconduct. The information programs led to a large reduction in misconduct (-21 pp = -72%) and as a result, an increase in overall market activity, firm sales and consumer welfare. The researchers develop a signaling framework between vendors and consumers that shows the treatment effect is due to a combination of more accurate consumers' beliefs about misconduct and increased reputation concerns. Together, the results indicate a potentially significant source of local financial market frictions, where market activities are underprovided due to misconduct and difficulty in building reputation. Social sanctions through reputational impacts can promote formal local markets when formal sanctions are weak.
Using an RCT, Banerjee, Duflo, MaKelway, Sharma, Schilbach, Vaidyanathan, and Grela study the effects of old-age pensions (OAP) in Tamil Nadu, India during the COVID pandemic on food security, health, and family support among a sample of 570 elderly. Old-age pensions differ from most other short-term assistance programs adopted to mitigate the economic and health impacts of COVID in that the payments are permanent. Furthermore, we know little about the causal effects of pensions in general.
In 2018, the researchers identified 1,124 elderly individuals who were likely eligible for the state pension but not receiving it from a census of 61,954 households conducted in partnership with the Tamil Nadu government. The researchers randomly assigned half to receive assistance filing for the OAP in early 2019. By 2020, treatment assignment had increased pension receipt by 35 percentage points off a base of 8 percent. They conducted phone surveys with 570 elderly in the sample in July 2020, providing data on how the elderly are faring during the pandemic and allowing us to investigate whether the OAP helped shield the elderly from the consequences of the pandemic.
The researchers find three sets of results. First, pensions improved an index of food security by 0.4 standard deviations. Second, they find no detectable effects on mental health or access to medical care. They do find suggestive evidence of reduced activities that would have violated the stay-at-home order, suggesting that financial resources can help people comply with government regulations to protect themselves and others. Finally, rather than crowding out private support, pensions increased financial support the elderly received from their family by 0.4 standard deviations, suggesting that family assistance for the elderly during crises is responsive to long-run financial considerations.
Historical states with low capacity often delegated tax collection to local elites, despite the risk of mismanagement. Could this strategy raise revenues without undermining government legitimacy in fragile states today? Bergeron, Tourek, Weigel, and Balan provide evidence from a randomized policy experiment assigning neighborhoods of a Congolese city -- spanning 45,162 properties -- to tax collection by state agents or by city chiefs. Chief collection raised property tax compliance by 3.3 percentage points, increasing revenues by 43%. Although chiefs collected more bribes, the researchers found no evidence of mismanagement or backlash on other margins. Results from a hybrid treatment arm in which state agents consulted with chiefs before collection suggest that chief collectors achieved higher compliance by using local information to more efficiently target households with high payment propensities, rather than by being more effective at persuading households to pay conditional on having visited them.
Hicks, Miguel, Walker, Kremer, and Baird exploit a randomized school health intervention that provided deworming treatment to Kenyan children and utilizes longitudinal data to estimate impacts on economic outcomes up to 20 years later. The effective respondent tracking rate was 84%. Individuals who received 2 to 3 additional years of childhood deworming experience an increase of 14% in consumption expenditure, 13% in hourly earnings, 9% in non-agricultural work hours, and are 9% more likely to live in urban areas. Most effects are concentrated among males and older individuals. Given deworming's low cost, a conservative annualized social internal rate of return estimate is 37%.
Platforms such as preprints websites have become an increasingly important way to rapidly disseminate new knowledge prior to peer review. While these platforms are accessible to scientists and audiences around the world, there is considerable uncertainty around the knowledge itself, particularly with articles produced by Chinese scientists. MacGarvie and Fry explore how readers allocate attention across preprints in a context of great urgency (the initial months of the COVID-19 pandemic), and the extent to which the community reduces uncertainty itself through endorsements on social media. They find that preprints with authors from Chinese institutions receive less attention than preprints with authors from the rest of the world. The researchers also document that self-organizing screening mechanisms such as twitter endorsements drive attention, although these are less common and no more effective for Chinese authored preprints, replicating the original attention gap. The results suggest that platforms designed to, or used to, promote unfettered access to early research findings do not necessarily lead to democratization of science in a context of high urgency and uncertainty.
Chowdhury, Schildberg-Hörisch, Schneider, and Sutter compare the effectiveness of a monetary treatment in containing COVID-19 with an information campaign among more than 3,000 participants from 150 villages right before the local onset of the pandemic. The phone-based information campaign stressed the importance of social distancing and hygiene measures, and illustrated the effect of exponential spread within a village. In the monetary treatment, they additionally offered an unconditional cash transfer (via mobile cash) worth 2-3 days of agricultural wages in rural areas.
In each treatment village, about 20 households were randomly selected for treatment. Another 20 households per village serve as the treatment village control group for investigating spillover effects, in addition to a pure control group. The intervention was followed by two household surveys two weeks and about three months after the intervention.
Preliminary findings indicate that both treatments improve knowledge and create the attitude of being decisive in fighting Covid-19 in the short-run, which translates into improved preventative and physical distancing behavior, and improved health (mainly measured by symptoms). This heads-up in knowledge of 2.5 months compared to control villages reduces the share of households reporting a deceased member three months after the intervention significantly from 1.6% to 0.8%, with identical reduction among untreated households in treatment villages. These numbers suggest that a country-wide implementation could have reduced Covid-19 related death by about 50%.
Global e-commerce platforms present new export opportunities for small and medium-sized enterprises (SME) in developing countries by significantly lowering the entry barriers of exporting. However, the impact of such increased export opportunities on SME growth and market dynamics is largely unknown. Bai, Xu, Liu, and Chen document a set of novel stylized facts about online exporters using detailed data from AliEpxress as well as a rich set of objective quality measures collected through real interactions with the sellers on the platform. Combined, these facts indicate the presence of search and information frictions and a potentially persistent impact of initial demand shocks on firm growth. To disentangle the role of demand from supply-side factors, the researchers conduct an experiment where they generate exogenous demand and information shocks to a set of small prospective exporters via randomly placed online orders and reviews. The experimental findings demonstrate the role of initial demand in the presence of search and information frictions. Motivated by these empirical findings, the researchers estimate an empirical model of the online market featuring these demand-side frictions. The model successfully rationalizes the experimental findings. The researchers use the model to quantify the role of search and information frictions. Counterfactual analyses show that initial demand shocks (or "luck"), as opposed to economic fundamentals, can lead to long-lasting impact on firm growth and market allocation. Alleviating information frictions and reducing the number of sellers can help to restore market efficiency and raise consumer welfare.
Fast-moving economic crises require responses from policymakers that are both timely and proportionate. Unfortunately, the indicators used to monitor require time to assemble and refine. Even in the most advanced economies, industrial production and employment statistics are collected monthly and subject to revision. Such delays impose extraordinary costs by leaving individuals and firms exposed to shocks without recourse. Even small improvements in the turn-around of economic indicators can yield significant social benefits in times of crisis. Cicala exploits the unique role electricity consumption plays in the activity of a modern economy, and the availability of consumption data in real-time from countries spanning a wide range of economic development. They collect hourly or daily electricity consumption data covering nearly half the global population, and present stark usage patterns reflecting the impact of the COVID-19 pandemic on economic activity. These data include state-level measurements from India, Mexico, and electricity zone-level estimates for much of Latin America. Cicala merges these data with detailed meteorological data to adjust for heating and cooling demands. The results are sobering. Economic shutdowns appear clearly in the data, and recoveries are heterogeneous. Electricity consumption for many states in India fell by about 40%, while the United States bottomed out at a roughly 8% reduction on average. While most European countries had recovered to nearly normal by the end of the summer, Maharashtra, IN was still 20% below counterfactual levels. These data will be of continuing interest to provide a timely, high-quality proxy of economic activity as the COVID-19 pandemic progresses.
A growing literature associates poverty with anomalies in decision-making. Fehr, Fink, and Jack investigate this link in a sample of over 3,000 small-scale farmers in Zambia, who were given the opportunity to exchange randomly assigned household items for alternative items of similar value. Analyzing a total of 5,842 trading decisions over a range of items, including cash, the researchers show that exchange asymmetries are sizable and remarkably robust across items and experimental procedures. Using cross-sectional, seasonal, and randomized variation in financial resource availability, they show that exchange asymmetries decrease in magnitude when subjects are more constrained. Consistent with the interpretation that variation in decision stakes drive the results, the researchers also show that trading probabilities increase when the value of the items involved is exogenously increased.
Religious festivals are widespread around the world, and often make up nontrivial shares of household expenditures. What are festivals' economic and social consequences? Montero and Yang study Catholic patron saint day festivals in Mexico, exploiting two features of the setting: (i) festival dates vary across the calendar and were determined in the early history of towns during conquest, and (ii) there is considerable variation in the intra-annual timing of agricultural seasons. The researchers compare municipalities with "poorly-timed" festivals (those that overlap with peak planting and harvest months) to other municipalities, examining differences in long-run economic development and social outcomes. Poorly-timed festivals lead to worse economic development along a range of measures. They also lead to lower agricultural productivity, higher shares of the labor force in agriculture, and higher religiosity. The negative impact of poorly-timed festivals on agricultural productivity impedes the structural transformation out of agriculture and thus overall economic growth. The increase in religiosity may help explain why poorly-timed festivals persist in spite of their negative growth consequences.