Transforming Rural Africa Conference, Spring 2019

Morjaria and Peck study the role of cost uncertainty and learning in new industries in developing economies. With unique historical data from the cut-flower industry in Ethiopia, the researchers document rapid entry and growth in the industry in the early 2000s. They find concurrent cost uncertainty among potential entrants and inter- and intra-firm variation in growth, consistent with learning. The researchers go on to build a multi-agent, dynamic model that brings the production function uncertainty insights of Hausmann and Rodrik (2003) to the framework of Jovanovic (1982). Model simulations predict that the least productive firms shrink and exit each period, that firm production levels become more disbursed over time, and that firm-level growth is most volatile early in the industrial development process. The researchers find corroborative evidence in transaction-level trade data. They conclude with an exercise that uses their model to show the potential for short-term industrial stimulus to generate long-term economic growth in the industry.

Coordination failure lies at the heart of certain development (poverty) traps, and a key policy question is how to mitigate it. A related literature identifies nonbinding communication (cheap talk), as a potential means for increasing coordination. Most of this literature is theoretical and/or based on conventional laboratory experiments. Aflagah, Bernard, and Viceisza test for this effect in a real-life setting, with Senegalese groundnut cooperatives whose members seek (but mostly fail) to jointly sell their outputs, in order to secure better unit prices. The researchers combine results from a lab-in-the field experiment, with those from a real-life randomized intervention. Together, they provide strong support for the researchers theoretical predictions: revealing intentions leads to enhanced coordination in larger groups, where coordination is more difficult to start with. The results also point to large and positive effects of the intervention on farmers' income from the sale of groundnuts. Finally, the researchers find evidence that the randomly selected farmers who participated in the "lab" experiments seem to have learned and in turn, transferred knowledge to their environment: they were more likely to commercialize through their groups in the subsequent real-life intervention. The findings are robust to both survey and administrative data as well as a range of falsification/placebo tests. They suggest that a subtle and relatively low-cost "cheap-talk" mechanism (intervention) could help address some of the coordination-based poverty traps.

Raising smallholder agricultural productivity has the potential to boost GDP and to reduce rural poverty, but the evidence on how to best achieve productivity gains remains mixed. Technologies exist that can increase smallholder yields and profits, but researchers and practitioners grapple with how to induce technology adoption in this population. This task is particularly challenging and pressing in sub-Saharan Africa, where agricultural yields lag all other regions and where poverty is often concentrated in rural areas. Deutschmann, Duru, Siegal, and Tjernström present randomized evidence of a rare scaled-up success story: One Acre Fund's (1AF) small farmer program. Much like anti-poverty "graduation" programs, 1AF's program is designed around the notion that farmers face multiple constraints simultaneously. Participating farmers receive input loans, crop insurance, and information about improved farming practices. Analyzing data from a pre-registered randomized control trial, the researchers show that participation in 1AF's program causes statistically and economically significant increases in yields and profits. They find evidence that relaxing information constraints alone unlikely explains the program's success. Using various approaches to heterogeneity analysis, the researchers find suggestive evidence that more disadvantaged farmers benefit more from the program and that program impacts accumulate over time. Both results are in line with the literature on graduation programs, which typically target the most disadvantaged, provide them with a bundle of different types of support, and continue to work with them as they climb out of poverty.

Low labor productivity among farming households have been associated with farmers' exposures to chemical and musculoskeletal health risks. The extent of these health risks exposures among farmers in Nigeria is yet to be addressed by research output. There is paucity of empirical evidence on the effectiveness of appropriate intervention at reducing exposure to these health risks. Olowogbon, Babatunde, and Asiedu used a randomized control trial approach to assign 480 cassava farmers from 24 farming communities to the study. However, 200 farmers were randomly assigned to receive the treatment. The intervention component includes one-time village level agricultural health training and a three-month farm safety mobile text messaging follow up. A peer developed module covering safe ergonomic practices and safe use of agrochemicals was used for the training. The intervention effects were evaluated in short-term (6 months) post treatment. Structured questionnaire, interviews and random farm visit were used for data collection. Exposure to health risks is measured by recurrent self-reported symptoms. Descriptive statistics, ordinary least square regression, and difference-in-difference estimator were used for data analysis. Findings showed that during chemical application, 90% of cassava farmers reported exposure to chemical health risks and at least 40% reported exposure to musculoskeletal disorders at other stages of cassava production, farmers' sickness absence is influenced by age, educational level, daily duration of chemical spray, care time and number of ergonomic exposure (p less than 0.05), every one day increase in sickness absence decreases labor productivity of cassava farmers' by 3% (p less than 0.01), the agricultural health intervention reduced sickness absence in the season by 1.9 out of 6.5 days (29%) with significant improvement in farmers' agricultural health knowledge and attitude (p less than 0.01). The study concluded that cassava farmers were engaged in unsafe farm practices exposing them to some health risks which negatively affect their wellbeing. Although, evidences from the study supports that the agricultural health training intervention enhanced farm safety knowledge, attitude and reduced sickness absence in short term, additional research is needed to establish the long-term intervention effects and explore issues of cost effectiveness. Furthermore, there is a need for inclusive agricultural health policy addressing agricultural health information, agricultural health surveillance and agricultural health services for the teeming Nigerian farming population. By casualty, this will enhance farmers' health, shared agricultural productivity and well-being.

Caeiro analyzes the role of social networks in the diffusion of cultivation techniques introduced by an agricultural project in Guinea-Bissau. Caeiro takes advantage of this intervention to study the diffusion of knowledge and adoption of cultivation techniques from project participants to the wider community. In order to test for social learning, Caeiro exploits a detailed dataset which includes village census and social network data across different social network dimensions. More precisely, Caeiro makes use of a village photo directory in order to obtain a comprehensive and fully mapped social network dataset. Caeiro finds evidence that agricultural information spreads across networks from project participants to non-participants. Specific network groups such as the ‘regular chatting’ and ‘agricultural advice’ networks appear to be more relevant for that process. The dataset also allows Caeiro to disentangle effects stemming from strong and weak network links, and weak links are found to be equally important in the diffusion of agricultural knowledge. Despite positive diffusion effects in knowledge, Caeiro has not found evidence of network effects in adoption behavior.

Aggarwal, Giera, Jeong, Robinson, and Spearot quantify poor market access in rural Africa, and the extent to which it constrains agricultural productivity. They collect granular data on farmer input and sales decisions, input and output prices, and travel costs in all 1,183 villages in two regions of Tanzania. The researchers find that a village in the 90th percentile of the travel-cost adjusted price distribution faces input and output prices 40-60% higher than a village at the 10th percentile. In reduced form, an additional standard deviation of travel time is associated with 20-25% lower input adoption and output sales. The researchers develop and quantify a spatial model of input adoption and conservatively estimate that farmers behave as if they face travel costs of 6% ad-valorem per kilometer of travel, which is equivalent to 40% when traveling to the closest retailer. Holding exogenous local factors fixed, the researchers estimate that reducing travel costs by 50% (approximately the effect of paving rural roads) doubles adoption and reduces the adoption-remoteness gradient by 18%.

Low agricultural productivity and severe levels of land degradation, coupled with high population growth rates, result in yield gaps and continue to threaten both food security and ecosystems in many African countries. The concept of integrated soil fertility management (ISFM) provides a promising approach to tackle soil degradation and increase productivity among smallholders in Africa. The core of ISFM is the integrated application of organic and inorganic soil amendments, accompanied by a general improvement of agronomic techniques. Yet, ISFM knowledge and its uptake among farmers are still low. At the same time, “bottom-up” extension models are gaining increased importance, as well as non-traditional forms of agricultural education. Hörner, Bouguen, Frolich, and Wollni assess the effects of a decentralized, participatory extension program and an additional video intervention on the adoption of five ISFM practices – compost, blended fertilizer, improved seeds, line seeding and lime – among small-scale farmers in Ethiopia using a randomized control trial. The sample consists of over 2,400 households spread over three Ethiopian highland regions. The researchers find that the treatments induce ISFM adoption, in particular of compost, line seeding and lime, as well as gains in knowledge. They further find evidence for diffusion effects from farmers who actively participate in the extension activities to their non-active peers in treatment communities regarding increased adoption of ISFM practices at the household level. However, when it comes to integrated use of the practices at the plot level – which is particularly important given the synergistic effects of the technologies – the extension treatment does not affect non-active farmers in treated communities. Yet, the results imply that these “left out” farmers can be reached with the video intervention, successfully encouraging integrated adoption among them.

Delesalle studies the impact of education on labor market participation and on household consumption in a rural environment. To do so, Delesalle uses the Universal Primary Education (UPE) program, whose intensity was varying across locations and over time. This program proved to be efficient at reducing inequalities of access to education and at providing basic agricultural skills. Based on a difference-in difference approach, and exploiting these two exogenous variations to instrument education and Delesalle finds that education raises household consumption, especially in agriculture. Delesalle also provides evidence that education increases the probability of working in agriculture at the expense of non-agricultural self-employed activities. The results illustrate the particularity of the program and suggest that returns to education in agriculture are positive, provided that the curriculum at school is suitable agriculture.

In recent decades contract farming has emerged as a popular mechanism to encourage vertical coordination in developing country agriculture. The goal of such coordination is to better integrate smallholder farmers into the modern agricultural food system, fostering rural transformation. Arouna, Michler, and Lokossou use panel data from a randomized control trial to quantify the impact of different contract attributes on rural transformation and welfare of smallholder rice farmers in Benin. The researchers vary the terms of contract, with some farmers being offered a contract that only guarantees a price, while other contracts add extension training or input loans. While all three types of contracts had positive and significant effects, the researchers find that contracts which only included an agreement on price had nearly as large of an impact as did contracts with additional attributes. This suggests that once price uncertainty is resolved, farmers are able to address other constraints on their own.

The Green Revolution bolstered agricultural yields and rural well-being in Asia and Latin America, but bypassed sub-Saharan Africa. Carter, Laajaj, and Yang study the first randomized controlled trial of a government-implemented input subsidy program (ISP) in Africa. A temporary subsidy for Mozambican maize farmers stimulates Green Revolution technology adoption, and effects persist in later unsubsidized years. Social networks of subsidized farmers benefit from spillovers, experiencing increases in technology adoption, yields, and expected returns to the technologies. Spillovers account for the vast majority of subsidy-induced gains. ISPs alleviate informational market failures, stimulating learning about new technologies by subsidy recipients and their social networks.

The mechanism(s) that generate measurement error matter to inference. Survey measurement error is typically thought to represent simple misreporting correctable through improved measurement. But errors might also or alternatively reflect respondent misperceptions that materially affect the respondent decisions under study. Abay, Bevis, and Barrett show analytically that these alternate data generating processes imply different appropriate regression specifications and have distinct effects on the bias in parameter estimates. The researchers introduce a simple empirical technique to generate unbiased estimates under more general conditions and to apportion measurement error between misreporting and misperceptions in measurement error when one has both self-reported and objectively-measured observations of the same explanatory variable. They then apply these techniques to the longstanding question of agricultural intensification: do farmers increase input application rates per unit area as the size of the plots they cultivate decreases? Using nationally representative data from four sub-Saharan African countries, the researchers find strong evidence that measurement error in plot size reflects a mixture of farmer misreporting and misperceptions. The results matter to inference around the intensification hypothesis and call into question whether more objective, precise measures are always preferable when estimating behavioral parameters.

Are roads in Africa connecting the right places to promote beneficial trade? Graff assesses the efficiency of transport networks for every country in Africa. Using rich data from satellites and online routing services, Graff simulates optimal trade flows over a comprehensive grid of more than 70,000 links covering the entire continent. Graff employs a recently established framework from the optimal transport in economics literature to maximize over the space of networks and find the optimal road system for every African state. Where would the social planner ideally build new roads and which roads are superfluous in promoting trade? These simulations predict that the entire continent would gain more than 1.1% of total welfare from better organising its national road systems. Comparing current and optimal networks, Graff then constructs a novel dataset of local network inefficiency for more than 10,000 African grid cells. Graff analyses roots of the substantial imbalances present in this dataset and finds that colonial infrastructure projects from more than a century ago still persist in significantly skewing trade networks towards a sub-optimal equilibrium. Areas close to former colonial railroads have about 1.7% too much welfare given their position in the network. Evidence is also found for regional favouritism, as the birthplaces of African leaders are overequipped with unnecessary roads. Lastly, Graff uncovers a descriptive relationship whereby large transport infrastructure projects from The World Bank are not allocated to regions most in need of additional roads.