Essays on Economic Outcomes of Training-Based Knowledge Interventions

Abstract

This dissertation studies the economic impacts of training-based knowledge interventions in low-income settings, with a focus on financial literacy and agricultural technology adoption. Across three chapters, it examines how individuals respond to information delivered through training programs and whether such interventions translate into sustained economic gains. A central theme across the chapters is that although low-cost knowledge interventions can lead to meaningful behavioral change, their effectiveness in driving more complex outcomes, such as technology adoption, depends crucially on contextual factors. The first chapter evaluates the impact of a digital financial literacy training program targeted at refugee youth in Uganda. As humanitarian assistance increasingly shifts from in-kind transfers to cash-based and digitally delivered programs, complementary financial capabilities have become critical. Exploiting the staggered geographic rollout of the program, I implement a strategy that closely emulates a natural experiment. Using reduced-form econometric analyses, robust to various specifications, I find that participation in the training program is associated with significant positive effects on financial knowledge and financial behavior among young refugees. In addition, the program enhances participants' confidence in navigating financial systems and integrating with host communities, suggesting broader social benefits beyond purely economic outcomes. The second chapter examines a bundled agricultural training program for groundnut farmers in rural Bangladesh, which combines Integrated Pest Management (IPM), Good Agricultural Practices (GAP), and fertilizer guidance. Using panel data and reduced-form estimation, I find that adoption is selective and cost-ordered. Farmers adjust fertilizer use in line with recommendations but largely avoid newly introduced IPM technologies, while responses to GAP are limited. As a result, the intervention generates no measurable gains in yields or profits, since the highest-impact components remain unadopted. To interpret these patterns, I develop a simple conceptual framework based on a signal extraction problem: when training provides information at the bundle level, farmers cannot infer component-specific returns and instead rely on observable costs when making adoption decisions. The findings highlight a key limitation of bundled interventions, where gains in scalability and administrative convenience may come at the cost of selective adoption of lower-impact practices. Building on the findings from the second chapter, the intervention was refined. The third chapter studies the adoption of multi-stage IPM technologies using evidence from a cluster randomized controlled trial that combines an improved one-day training with season-long extension support. This chapter addresses the limited evidence on the profitability of IPM technologies while also examining how farmers respond to multiple components introduced across stages. I analyze stage-wise adoption decisions using a generated regressor framework to account for the sequential nature of technology uptake. Using reduced-form econometric analyses, I also estimate the impacts of the program on farmer outcomes. The results show that the intervention increases adoption of several recommended practices and leads to substantial reductions in chemical input use, indicating movement toward more environmentally sustainable production. However, these changes do not translate into significant gains in yields or profits in the short run. I further show that early-stage adoption decisions are shaped by farmer characteristics such as landholdings and soil conditions, while later-stage adoption is largely driven by path dependence and continuation of prior choices. Overall, the chapters show that training programs can shift knowledge and behavior, but translating these into meaningful economic gains is limited by learning costs, financial barriers, and complexity. The findings highlight the importance of accounting for these constraints in program design and suggest that simpler interventions can improve effectiveness.