Smallholder agriculture continues to be the main source of livelihood for a large portion of Sub-Saharan Africa's population. In recent years, compounding crises and shocks have threatened this livelihood basis. How did smallholder farmers in Sub-Saharan Africa fare throughout this period? What measures did they take to cope with the repeated disruption to their farming activities? We address this knowledge gap using nationally representative, cross-country comparable longitudinal microdata collected from over 16,000 agricultural households in six Sub-Saharan African countries between 2018 and 2024. While overall fertilizer adoption remained relatively stable, we find considerable cross-country heterogeneity and poorer households to be more likely to have discontinued fertilizer use. At the intensive margin, 47% of farmers could not access their desired quantity of fertilizer. On average, these farmers used less than half as much inorganic fertilizer as they desired with affordability being the main constraint. Households adopted a range of coping strategies, some of which may compromise productivity and heighten future vulnerability.

Using phone surveys and administrative data in 36 LMICs, we find survey-based COVID-19 vaccine coverage to systematically exceed administrative figures by 47% on average. This discrepancy is strongest in Sub-Saharan Africa, raising questions about the role of data quality for our understanding of vaccination progress. We isolate the effect of sampling and measurement errors on the estimated vaccination rate in six survey experiments conducted in five Sub-Saharan African countries. Accounting for respondent selection effects reduces misalignment between data sources by 42% on average. Proxy reports miss some vaccinations, but other error sources including strategic misreporting, panel conditioning, or survey mode effects do not affect survey estimates. After adjusting for errors in the survey data, a substantial average gap of 9 percentage points remains with the official figures and seems to relate to flaws in administrative records that we document using data covering all 134 LMICs worldwide. Our results provide novel evidence on the size and sources of measurement error in modern data sources that have seen a surge in use by development economists. Our error-corrected estimates imply that vaccination progress may have been quicker than what African countries were credited for in the public discourse.

Climate change is affecting the frequency and severity of extreme weather events, such as droughts or floods, which result in loss and damage to people, crops and infrastructure. Global data on loss and damage used in research, policy and media primarily come from macrostatistics based on disaster inventories. Here, we propose a different approach, based on survey microdata. We harmonize data from 120,000 agricultural fields in six African countries for a period from 2008 to 2019 and quantify crop production losses related to climate shocks. We find substantial damages which affect around 35% of plots and reduce national crop production by 29% on average. The economic impacts are greater than the global disaster data suggest. The economic losses resulting from droughts and flood alone are US$5.1 billion higher than reported in disaster inventories, affecting between 145 and 170 million people. The difference stems mostly from smaller and less severe but frequent adverse events that go under-reported or undetected in disaster inventories and therefore elude macrostatistics and reporting. The findings have implications for measurement and policies related to loss and damage and disaster risk reduction.

Despite improved availability of COVID-19 vaccines in Sub-Saharan Africa, vaccination campaigns in the region have struggled to pick up pace and trail the rest of the world. Yet, a successful vaccination campaign in Sub-Saharan Africa will be critical to containing COVID-19 globally. Here, we present new descriptive evidence on vaccine hesitancy, uptake, last-mile delivery barriers, and potential strategies to reach those who remain unvaccinated. Our data comes from national high frequency phone surveys in six countries in East and West Africa with a total population of 415 million people. Samples were drawn from nationally representative samples of households interviewed in recent in-person surveys. Our estimates are based on a survey module harmonized across countries and are re-weighted to mitigate potential sample selection biases. We show that vaccine acceptance remains generally high among respondents in Sub-Saharan Africa (between 95.1% and 63.3%) even though hesitancy is non-negligible among those pending vaccination. Many who are willing to get vaccinated are deterred by a lack of easy access to vaccines at the local level. Furthermore, social ties and perceptions as well as intra-household power relations matter for vaccine take-up. Among the unvaccinated population, radio broadcasts have widespread reach and medical professionals are highly trusted. Our findings highlight that creating a positive social norm around COVID-19 vaccination, messaging that leverages trusted and accessible information sources and channels, and more easily accessible vaccination sites at the community level are promising policy options to boost vaccination campaigns in the region and end the pandemic everywhere.

Background

COVID-19 vaccination efforts are lagging in Sub-Saharan Africa, as just over 20 percent of the population has been fully vaccinated. COVID-19 vaccine hesitancy is considered important as a prerequisite for widespread vaccine take-up. Here, we study the dynamics of vaccine acceptance, its correlates, and reasons for hesitancy over time, drawing on two years of panel survey data.

Methods

In this observational study, we use multiple rounds of data from national High Frequency Phone Surveys (HFPS) in five countries in East and West Africa (Burkina Faso, Ethiopia, Malawi, Nigeria, and Uganda), covering a period between 2020 and 2022. The surveys are cross-country comparable and draw their samples from nationally representative sampling frames. Based on this data source, the study presents population-weighted means and performs multivariate regression analysis.

Results

COVID-19 vaccine acceptance was high throughout the study period (68% to 98%). However, acceptance levels were lower in 2022 than in 2020 in three countries (Burkina Faso, Malawi, Nigeria), and higher in one country (Uganda). Moreover, individuals are observed to change their stated vaccine attitudes between survey rounds, to a limited extent in some countries (Ethiopia) and more frequently in others (Burkina Faso, Malawi, Nigeria, Uganda). Vaccine hesitancy is higher in richer households, and those residing in urban areas; among women and those better educated. Hesitancy is lower in larger households and among heads of the household. The main reasons for hesitancy are concerns about side effects of the vaccine, its safety and efficacy, as well as assessments of COVID-19 risk, though these reasons fluctuate over time.

Conclusions

Reported COVID-19 vaccine acceptance levels remain far above vaccination rates in the study countries, suggesting that vaccine hesitancy is not the primary obstacle to reaching greater vaccine coverage, which may instead be related to access and delivery barriers as well as supply shortages. Nevertheless, vaccine attitudes appear malleable so that continued efforts are needed to retain high levels of vaccine acceptance.

Objectives To estimate the willingness to accept a COVID-19 vaccine in six sub-Saharan African countries and identify differences in acceptance across countries and population groups.

Design Cross-country comparable, descriptive study based on a longitudinal survey.

Setting Six national surveys from countries representing 38% of the sub-Saharan African population (Burkina Faso, Ethiopia, Malawi, Mali, Nigeria and Uganda).

Participants Respondents of national high-frequency phone surveys, aged 15 years and older, drawn from a nationally representative sample of households.

Main outcome measures Willingness to get vaccinated against COVID-19 if an approved vaccine is provided now and for free, disaggregated by demographic attributes and socioeconomic factors obtained from national household surveys. Correlates of and reasons for vaccine hesitancy.

Results Acceptance rates in the six sub-Saharan African countries studied are generally high, with at least four in five people willing to be vaccinated in all but one country. Vaccine acceptance ranges from nearly universal in Ethiopia (97.9%, 95% CI 97.2% to 98.6%) to below what would likely be required for herd immunity in Mali (64.5%, 95% CI 61.3% to 67.8%). We find little evidence for systematic differences in vaccine hesitancy by sex or age but some clusters of hesitancy in urban areas, among the better educated, and in richer households. Safety concerns about the vaccine in general and its side effects specifically emerge as the primary reservations toward a COVID-19 vaccine across countries.

Conclusions Our findings suggest that inadequate demand is unlikely to represent the key bottleneck to reaching high COVID-19 vaccine coverage in sub-Saharan Africa. To turn intent into effective demand, targeted information, sensitisation and engagement campaigns bolstering confidence in the safety of approved vaccines and reducing concerns about side effects will be crucial to safeguard the swift progression of vaccine rollout in one of the world’s poorest regions.

Every year, disasters account for billions of dollars in crop production losses in low- and middle-income countries and particularly threaten the lives and livelihoods of those depending on agriculture. With climate change accelerating, this burden will likely increase in the future and accurate, micro-level measurement of crop losses will be important to understand disasters’ implications for livelihoods, prevent humanitarian crises, and build future resilience. Survey data present a large, rich, highly disaggregated information source that is trialed and tested to the specifications of smallholder agriculture common in low- and middle-income countries. However, to tap into this potential, a thorough understanding of and robust methodology for measuring disaster crop production losses in survey microdata is essential. This paper exploits plot-level panel data for almost 20,000 plots on 8,000 farms in three Sub-Saharan African countries with information on harvest, input use, and different proxies of losses; household and community-level data; as well data from other sources such as crop cutting and survey experiments, to provide new insights into the reliability of survey-based crop loss estimates and their attribution to disasters. The paper concludes with concrete recommendations for methodology and survey design and identifies key avenues for further research.

Tax nudges in the form of reminders and deterrence messages sent to taxpayers have become a common policy instrument and a large literature finds them to be highly cost-effective at raising tax revenue. We conduct a field experiment with over 60,000 firms in Uganda to investigate if the existing evidence misrepresents the true rate of return when conducting tax nudges at scale, that is, across different tax schemes and over time. By merging local and national business tax registries, we create a unique administrative dataset that lets us measure (i) spillover effects on tax obligations payable to other tiers of government not targeted by the nudge; (ii) effects across the firm-size distribution including smaller, semi-informal, and large businesses. By following up on a tax nudge intervention conducted four years prior, we further (iii) investigate long-run dynamics in the effect of tax nudges and (iv) study their effectiveness under repeated exposure without an accompanying shift in enforcement capacity.

Understanding the impact of tax policies on fiscal space and business formalization are key questions for economic development. One such policy are tax amnesties – partial forgiveness of past tax debt and fines in return for voluntary disclosure and compliance. While tax amnesties have been successful in higher income contexts, empirical evidence on their effectiveness in low-income countries is missing. Tax amnesties promise to raise revenue at low enforcement cost and may broaden the tax base. Yet, they may incur costs in terms of foregone revenue, tax morale, and signal weak state capacity – potentially negating financial gains from debt declared under the amnesty. We collaborate with the Uganda Revenue Authority to assess the fiscal impact of a national tax amnesty declared in 2023. Next, we study lasting impacts on attitudes and tax morale through a randomized salience intervention and survey data we collect one year onward. Informed by our survey results, and exploiting an extension of the amnesty until the end of 2024, we study the effectiveness of different amnesty uptake nudges in a large-scale field experiment with 90,000 Ugandan firms.

A fiscal social contract in which citizens pay taxes and the government is accountable to taxpayer demands in the way funds are spent is a fundamental characteristic of many democratic states. However, many LMICs exhibit low tax compliance and dissatisfaction with public services. In these settings, tax policies commonly aim to increase enforcement capacity. We contend that a deterrence-based approach alone has failed to bring about sustainable changes in the tax climate and propose a complementary approach based on increasing political participation. We collaborate with the Kampala Capital City Authority in Uganda and test the effects of an ICT-based civic participation platform integrated into WhatsApp on taxpayer demands for accountability. Leveraging original survey data and cross-linked business tax registries at the national and municipal level, we provide novel evidence on fiscal contracts in low-income countries and the way firms engage with the state.

Although ending hunger all year round is enshrined in Sustainable Development Goal (SDG) 2, canonical data sources do not measure how food security varies throughout the year. In this paper, we show that high-frequency panel data offer a new perspective on the extent and dynamic nature of food insecurity in the world’s poorest region. Using data tracking over 17,000 households in six Sub-Saharan African countries over a four-year period, we show that food security is characterized by substantial changes in the food security of the same households over the course of a year. On average, 70% of households are ever food insecure over a year but only 35% of them are always food insecure.  Households’ food security situation can change rapidly. A quarter of households at least once experience a large deterioration and 62% a large improvement in food security within a six-month period.  Crucially, different households have their peak of food insecurity at different times of the year, suggesting that seasonal patterns only explain part of this variation. An implication of this is that aligning the timing of data collection with the seasonal food insecurity peak would still miss out on 23% of ever-food-insecure households. We separate the chronically from the transitorily food-insecure households and show how this knowledge is useful for more effective and efficient policy targeting. Our results have implications for how we move from 2.3 billion food insecure people to zero and track progress toward this goal.

The last decade has seen a concurrent, large increase in forced displacement and in shocks such as conflict, natural disasters, and public health emergencies in low- and middle-income countries. Crises cause displacement so FDPs often geographically cluster around areas susceptible to adverse events and shocks. Crises also hurt those the most that are the most vulnerable to start with. Despite these established facts, data limitations have meant that there is a lack of evidence quantifying differences in vulnerability between forcibly displaced households and their non-displaced counterparts - with important implications for what kind of and for whom policy is made. In this paper, we assemble a longitudinal dataset tracking the vulnerability of over 18,000 forcibly displaced households and 34,000 non-displaced households between April 2020 and March 2023 across 11 low- and middle-income countries. We find that displaced households are almost twice as likely to be severely food insecure, a third less likely to be in employment, and 9% less likely to receive medical care when needed. Part of this vulnerability differential relates to FDPs living in areas with harsher conditions and that are more prone to shocks. We find that FDPs live in areas that are on average 0.7°C hotter, receive 0.65 mm less rain per month, and that record about one additional violent conflict event per 100,000 people and 1.2 additional conflict fatalities in a 6-month period. Yet, FDPs are also substantially more vulnerable than the local host population inhabiting the same areas. This suggests that geography alone explains all but one part of the vulnerability differential which arises even though FDPs are almost three times more likely to receive a form of social assistance. We find that targeting assistance to the broad group of FDPs within a country would do reasonably well in targeting vulnerabilities without the need for data-intensive (proxy) means testing. In an age of compounding crises, mass displacement, and shrinking aid budgets, our findings demonstrate the urgency of more, not less, development funding to be allocated to address the plight of forcibly displaced populations.

Abstract coming soon.

Economists commonly rely on survey data to measure attitudes and behavior that are not directly observable, yet we lack a thorough understanding of two key questions: how survey design shapes respondent behavior and how respondent behavior mediates the reliability of inference. While previous work has documented that survey responses can be sensitive to design choices, we quantify the size and sources of resulting measurement uncertainty by surveying 3,400 business owners about their tax compliance, validating their responses against administrative tax records, and experimentally varying survey design features such as question format, wording, or complexity. We show that measurement uncertainty is quantitatively large but usually neglected when drawing inferences. This results in a fundamentally flawed understanding of the accuracy of survey data and popular methods to measure sensitive behaviors, such as list experiments. To address this, we propose a Bayesian model aggregation approach that provides a more faithful representation of true inferential uncertainty in survey estimates and reduces the risk of misguided policy decisions. Our findings suggest that recent disagreements in empirical economics could stem from unrecognized measurement uncertainty rather than true differences in underlying phenomena.

In a large-scale survey experiment in rural Nigeria, we study the role of survey mode — in-person vs. over the phone — for survey measurement and data quality. Our experimental design isolates mode effects from other common sources of errors in surveys and covers twenty outcome measures across topics such as health, labor, shocks, wellbeing, and food security. We consistently find mode effects across outcomes, with phone responses differing from in-person responses by up to 63%, and a median difference of 17–18%. We document that these effects are large relative to other errors in phone surveys such as undercoverage of households without phones. Using a within-subject design, we measure the full, respondent-level distribution of mode effects and find them to vary much more than averages let on. Respondents with higher education levels are less prone to mode effects whereas mode effects sharply increase in prevalence as respondents face more answer options. We present evidence that mode effects are driven by lower cognitive engagement of respondents over the phone. Given the growing reliance on phone surveys in LMICs, our findings provide novel guidance for survey design and advance the research frontier on how survey mode influences measurement.

Over the last two decades, field experiments and rigorous policy evaluation have become a core part of the work applied economists do. However, there exists a systematic gap in the time horizon and frequency at which economists study policy interventions: Much work focuses on testing novel interventions and evaluating their impacts at one or two follow-up dates over a relatively short time horizon (a few months or years). Conversely, much less research is evaluated at higher frequency and over the long run (10 years or more) resulting in an evidence gap on the dynamic impacts of policy.

This evidence gap in knowledge creation is problematic from a social impact perspective for two main reasons. First, it limits our ability to ’pick winners’ for economic development - those interventions that are transformative in the long run, over and above short run success in the researcher’s sandbox. Second, a focus on novel interventions and short run impacts rather than tracing longer run impacts of existing interventions creates systematic biases and blind spots in what policies are considered effective. Similarly, a lack of high-frequency follow up data limits our understanding of when, why, and how dynamic effects of policy interventions emerge. This leaves us with an incomplete understanding of policies’ full impact. Both issues fundamentally affect the degree to which policy evaluation is informative for real-world, impactful policy making.

In this paper, we propose a simple mechanism to correct this market failure in knowledge creation: by reducing the costs of long-run and high-frequency policy evaluation, generating such evidence can become optimal from a social welfare and private researcher’s perspective. A promising cost-savings margin is data collection mode: Defaulting to in-person surveys might be an inefficient choice when outcomes can be collected with sufficient precision at relatively cheaper (marginal) costs using an alternative data collection mode. One such approach could be phone surveys which have grown in popularity in recent years. Under what conditions researchers can produce sufficiently high-powered long-run and high-frequency evidence at favorable cost by choosing survey mode optimally is the subject of this project.

This Working Paper presents an analysis into the adequacy of Pakistan’s social protection response to the COVID-19 pandemic and highlights pertinent policy takeaways for turning the health crisis into an opportunity for a strengthened social protection system in Pakistan. It focuses on Ehsaas Emergency Cash (EEC) as the country’s flagship initiative to attenuate the economic fallout for the most vulnerable people during and after the crisis. The report starts with a brief analysis into the economic context of the crisis in Pakistan and describes the country’s social protection landscape. Furthermore, the paper distinguishes adequacy considerations for emergency measures put in place from those for regular social protection. Subsequently, we introduce EEC’s programme design in light of the crisis and later discuss its adequacy regarding benefit values, coverage, and selection and payment mechanisms. Throughout this analysis, the paper shines a light on the merits of the existing social protection response and highlights important benefits as well as potential shortcomings of the measures taken. Drawing on some of EEC’s most significant limitations as previously identified, we then shift our focus to estimates of the costs of a number of stylised changes to the programme’s design. The paper concludes with a look ahead into challenges for social protection in Pakistan and summarises some of its most pertinent findings into brief, policy-oriented recommendations. We find that EEC has managed to expand social protection coverage substantially to cover those who have been heavily affected by the health crisis but lack regular access to social protection. Furthermore, EEC has revealed important opportunities for future resiliencebuilding through strong social protection systems in Pakistan. At the same time, the paper cautions that the economic ramifications stemming from the crisis seem certain to take a lasting toll on (informal) livelihoods in the country for months to come. A continued response and important possible amendments to it as identified by our adequacy analysis, as well as the inclusion of lessons learned from the crisis response, could lay the basis for building back better and turning the COVID-19 crisis into an opportunity for social protection in Pakistan. 

Do country-level outcomes relate to latent preference types within the population? This paper uses six different preference measures obtained from a global sample of 80,000 individuals to construct a set of unique preference types. It then asks whether the prevalence of these types correlates with outcomes in the domains of income, entrepreneurship, conflict, democratization, and gender equality. To test this hypothesis, it draws on and contributes to three distinct streams of literature: Firstly, it draws on the psychological literature on personality traits and types and transfers this concept to economic preferences to explore the possibility of a preference typology. In a second step, it expands on previous work on global preference heterogeneity by adding a person-centered perspective. This accounts for complex relationships between different preference measures, lets them cluster into characteristic types and allows us to determine individuals’ particular type membership. Exploiting a recently published, comprehensive dataset, the paper identifies four robust “pure” preference types. Furthermore, it discovers considerable global variation in their prevalence. To uncover this latent grouping structure of the data, the paper applies latent profile analysis, a type of mixture model related to unsupervised machine learning, and demonstrates its use for economic research. Therefore, this paper lastly makes a contribution on a methodological level. It provides a framework for a rigorous approach to typologies and highlights its applicability to a multitude of questions economists face. In this regard, it outlines how employing a similar approach to divide a population into latent groups can present a stepping-stone for new insights into development research. 

Including

• Socio-economic impacts of COVID-19, policy responses and the missing middle in South Asia

• Overview of non-contributory social protection programmes in South Asia from a child and equity perspective

• Social assistance programmes in South Asia: an analysis of socio-economic impacts