The Logic Of Adaptive Sequential Experimentation In Policy Design

Inspired by the wide adoption of rigorous randomized controlled trials (RCTs) in medical research, economists and other social scientists have increasingly used RCTs in their research. As researchers pick up projects amenable to the RCT methodology, they likely leave out important questions to which RCTs cannot be directly applied. As a result, RCTs have been criticized for the proclivity of addressing trivial questions. As a matter of fact, in medical research RCTs are an integral part of adaptive sequential experiment design—a few steps must be taken to screen out drugs that have toxins and strong side effects before running any RCTs on humans. In this paper, we argue that economists can learn a great deal from the design principles implemented in medical research. We develop a theoretical model to show the logic of adaptive sequential experiment design in the presence of uncertainty over negative effects and discuss how to choose samples in a population to minimize the experiment cost. We also point out the applications of our proposed framework in the economic domain, such as economic reforms and new product design.

Cambodia has been identified as one of the most vulnerable countries to climate change, given the predicted changes in temperature and precipitation, the share of labor in agriculture, and the country’s low adaptive capacity due to widespread poverty. In this study, we use climate data from four general circulation models (GCMs) to evaluate the impact of climate change on agriculture in Cambodia by 2050. We used the Decision Support System for Agrotechnology Transfer crop modeling software to evaluate crop yields, first for the 1950–2000 period (actual climate) and then for the climates given by the four GCMs for 2050. We evaluated crop yields for eight different crops at 2,162 points in Cambodia, using a grid of 10-kilometer squares, for 2000 and 2050. For each crop, we searched for the best cultivar (variety) in each square, rather than assuming the same cultivar to be used in all locations. We also searched for the best planting month in each square. We explored potential gains from changing fertilizer levels and from using irrigation to compensate for rainfall changes. This analysis indicates that when practiced together, using improved cultivars better suited for the changing climate conditions and adjusting planting dates can lessen the impact of climate change on yields, including for both wet- and dry-season rice. In addition, the analysis shows that losses in yield due to climate change can be compensated for—for many crops—by increasing the availability of nitrogen in the soil. To provide context to the modeling analysis, a survey of 45 communes was conducted using focus group discussions to solicit information on agricultural practices. Questions were asked about fertilizer, irrigation, seeds, tillage, and pest management, as well as about natural disasters and how farmers respond. Key results indicate that in response to extreme weather, only 7 to 16 percent of farmers report changing crop variety and only 20 percent of farmers report changing planting dates. Since the modeling results indicate that adaptation to climate change by changing crop variety and planting dates will be critical in order to avoid yield losses over the next 40 years, it is recommended that farmers expand their capacity to adapt in this way. In addition, every commune reported using some type of chemical fertilizer; however, in a typical commune, only 50 percent of the farmers were using any chemical fertilizer. This indicates that there is room to increase the use of chemical fertilizers. Finally, focus group participants were asked to name the top three natural disasters of concern. Drought was the most cited, reported in 44 of the 45 communes (98 percent), while flooding was cited in 67 percent of the communes surveyed. Despite this, in fully 58 percent of the communes, farmers reported taking no action in response to floods. In response to drought, farmers reported switching to other crops in 16 percent of the communes, and changed planting dates in 19 percent, while only 7 percent of the communes reported no adaptation in farming practices. These findings indicate the need for intervention to help farmers deal with floods in particular, and to determine whether strategies for adapting to drought are the best suited to mitigate crop loss.

In developing countries, all too often policies formulated in response to high food prices are inspired by ideology instead of evidence-based policy research. We look at the immediate effects of these shocks faced by households in Uganda on their poverty and well-being. In addition, we look at the economywide impact in the long run when all markets have settled at a new equilibrium. We find that in the short run, poverty has increased substantially. However, in the longer run, we find welfare levels of rural farm households in particular to rise sharply, primarily as a result of increased returns to farm labor and agricultural land coupled with improved market prices for output sold. These results call for policies that aim to protect the most vulnerable against high food prices and extreme volatility in the short run, without eliminating the incentives of steadily rising commodity prices for longer-run structural agricultural development.

Agricultural extension is critical for agricultural growth and food security, but making the extension system effective, demand driven, and responsive to the needs of a diverse set of producers remains a challenge. As part of the institutional reforms in the Democratic Republic of Congo (DRC), the extension system is being reviewed to identify strategies and practical actions to transform the system to better respond to the knowledge needs in a rapidly changing agriculture and food sector. This paper provides an in-depth review of the agricultural extension system of DRC including an analysis of its policies and legal framework, organization, and management; links to critical institutions; and capacity and incentive of different actors in the system. This review involved document analyses, interviews with key informants, and surveys of 107 extension organizations and 162 extension agents in 156 randomly selected villages in western DRC. This review suggests serious funding constraints, human resource management problems, no linkage and coordination within the extension system and with research and education systems, and a majority of underserved communities and farmers. This review also highlights a good opportunity given the huge human resources (more than 11,000 agents) deployed into the sectors and territories as part of the Ministry’s agricultural inspection system. However, their mandate is not clear and they currently focus on data collection (census) and checking what farmers do and limited in extension and advisory work. This review highlights the urgent need for human resource or civil service reform; public-sector commitment and funding; infrastructure rebuilding and skills upgrading in extension, education and training, and research organizations; and a unified policy or strategy with clear vision, mandate, targets, and performance-based incentive system.

Demand heterogeneity often makes it profitable for firms to price and promote goods and services differently in different market segments. When private consumption brings public benefits, this same heterogeneity can be used to target public subsidies. We explore the design of public–private targeting and segmentation strategies in the case of a resource-conserving agricultural technology in India. To understand farmers’ heterogeneous demand for laser land leveling (LLL), we conducted an experimental auction for LLL services with an integrated randomized controlled trial to estimate the private benefits of the technology. We use graphical and econometric approaches to characterize farmer demand for LLL. We then add detailed cost data from LLL providers to simulate and evaluate several potential targeted delivery strategies based on measures of (1) the cost-effectiveness of expanding LLL dissemination, (2) water savings, and (3) market surplus in a welfare framework. These simulations demonstrate inherent tradeoffs between increasing the amount of land that is leveled and expanding the number of farmers who adopt the technology, and between adoption and water savings. While segmenting and targeting are popular elements of many public–private partnerships to develop and disseminate agricultural technologies, formulating and implementing effective delivery strategies requires a rich understanding of costs, benefits, and demand. Our experimental approach generates such an understanding and may be relevant in other contexts.

The Nigerian Comprehensive Food Security and Vulnerability Analysis (CFSVA) provides an in-depth assessment of the food security situation within Nigeria. This is very important as it equips policymakers with timely and relevant information that will aid the targeting of interventions. Some of the most pertinent findings of the study are listed below: • Food insecurity and poverty are intricately linked. Some 29 percent of households in the poorest wealth quintiles have unacceptable diets (9 percent poor and 20 percent borderline) compared with 15 percent in the wealthiest (2 percent poor and 13 percent borderline). • The poorest livelihoods are found in agriculture. Seventy-seven percent of subsistence farmers are found in the two poorest wealth quintiles, as are 70 percent of mixed or cash crop farmers. • The general state of water and sanitation facilities available to households in all wealth categories is very poor, with consequent health implications. Forty-five percent of respondents do not have access to decent toilets, and 85 percent have no proper means of refuse disposal. • The vulnerable and food insecure are mostly found in rural areas and the North West and North East regions of Nigeria. • Most households in all regions and at all wealth levels purchase food, but rural households and poorer households (by wealth and livelihood) also rely heavily on own food production. Households in the poorest quintiles in both rural and urban areas rely on own production (32 percent rural and 24 percent urban). Wealthier urban households rely mostly on purchases, whereas own production is common at varying levels across all wealth levels for rural households. • Nigerians generally consume a starchy diet, but wealthier households can afford more nutrient-rich foods (including animal-based proteins) than poorer households. For instance, the wealthiest households consume meat, fish, and eggs an average of four days a week compared with only two days for the poorest households. • Most households protect vulnerable household members in terms of food allocations (women and children), but that may not hold in the poorest households where some difficult allocation decisions may have to be made. • Poorer households are more likely to engage in extreme coping strategies (like going a whole day without food) to deal with food shortages.

Uganda has made notable progress in reducing micronutrient deficiencies in recent years, but the prevalence of vitamin A deficiency (VAD) and anemia among children under 5 remain unacceptably high. According to World Health Organization criteria, VAD remains a public health problem in Uganda, and anemia is a severe problem. In this paper we explore the potential contribution to reducing both of these deficiencies using a genetically modified, high–provitamin A and high-iron banana (HPVAHIB) that is currently being developed. We present an ex ante analysis of the costs and nutritional benefits of HPVAHIB. Using the Ugandan National Household Survey of 2005/06, we analyzed the production and consumption patterns of highland cooking banana (nakinyika) and sweet banana (sukalindizi). Informed by the empirical findings, we developed geographically differentiated adoption, production, consumption, and diffusion patterns for several types of HPVAHIB. Based on households’ reported quantities of each type of banana currently consumed, we estimated the number of people consuming each banana and the quantities they consume, and then simulated the additional intakes of vitamin A and iron and estimated the number of disability-adjusted life years (DALYs) saved attributable to each. Combining the health impacts with the estimated costs of the project, three impact measures of the HPVAHIB are developed: the cost per DALY saved, the benefit–cost ratio, and the internal rate of return. Eighteen scenarios are estimated. The base scenario, which includes only the biofortification of cooking banana with provitamin A at a level equal to 400 percent its intrinsic provitamin A content, estimates that the net present cost per DALY saved of HPVAHIB is US$62, its benefit–cost ratio is 16, and its internal rate of return is 31 percent. According to criteria established by the World Health Organization and the World Bank, the HPVAHIB project is a “very cost-effective” health intervention.