Abstract: Using data from a randomized intervention among smallholder hog farmers in Sichuan, China, we estimate alternative models of insurance demand under neoclassical, present-biased, and reference-dependent decision rules. The neoclassical specifications include a baseline model, one with liquidity constraints, and one with imperfect trust. The reference-dependent model provides the best in-sample fit. Allowing heterogeneity through mixture specifications shows that approximately 90 percent of farmers are reference dependent, while the remaining population share is best described by a neoclassical rule; present bias, liquidity constraints, and distrust do not add explanatory power once heterogeneity is allowed. We evaluate out-of-sample predictions from our structural estimations with machine learning algorithms and find that models incorporating reference dependence remain competitive. Machine learning, however, can help identify likely adopters.
Abstract: Decades of evidence that risk and shocks inhibit agricultural investment and productivity growth motivated the search for risk management solutions. The search for solutions, particularly for smallholder farmers, has recently prioritized agricultural index insurance. Despite empirical evidence that index insurance can solve this problem, the evidence also shows that index insurance is a tough sell to smallholder farmers for a variety of reasons, including its expense, its complexity and the fact that it ironically exposes the farmer to the risk of substantial cost when the underlying insurance index does not accurately capture true losses. Noting that other financial instruments (lines of credit and commitment savings accounts) can be indexed, this paper uses dynamic modeling methods to explore how these different indexed instruments can complement each other to achieve the goal of smallholder investment and economic well-being. Two main avenues of complementarity emerge from the analysis. First, because the instruments differ in their leverage and cost, they are optimally employed at different points of the lifecycle and in the periods around shocks that dissipate wealth. Second, because the instruments differ in the costs associated with index inaccuracy, they offer dynamic learning complementarities. Skeptical rational learners experiment with the low-risk instruments (indexed savings and loans) and only slowly shift to insurance. Among other results, the standard practice of offering only index insurance may well result in no demand and no impact on smallholder farmer investment and productivity.
Abstract: This study examines the relationship between external incentives and intrinsic social preferences in shaping social norms and cooperative behavior through an experiment in Baringo, Kenya. Participants engaged in public goods games that simulated real-world trade-offs between private work and communal land restoration under a program known as participatory rangeland management (PRM). By varying both the incentive level (none, low, and high) and the framing of the incentive (compensation versus collaboration-encouragement), we assess how these factors influence individual contributions and the social norms surrounding cooperative behavior. We find that contributions and social norms differ by site, and that while financial incentives boost contributions, they tend to crowd out intrinsic social preferences. While we do see effects of framing on contribution levels and norms, it does not impact the extent of crowding out.
Abstract: In this study, we evaluate whether participatory rangeland management (PRM), an intervention aimed at helping pastoralist communities in Kenya to manage their resources including grazing lands, has altered the collective action behavior of individuals in the communities. We evaluate this using public goods games where participants decide how to allocate days between communal and private activities. Participants are randomly selected from villages exposed to PRM and those unexposed to PRM. We find negative and mostly insignificant correlation between PRM and contributions to communal activities. We also do not find any evidence that PRM crowds-out traditional rangeland management systems. Our findings suggest that benefits from community level interventions may be undermined by non-recognition of existing traditional management systems and poor implementation of processes.