Misspecified Learning in Technology Adoption: Experimental Evidence from Fertilizer Use in China

Study Context

This project shows how agents learn multiple technologies simultaneously. The team documents that when agents face two technologies at the same time, the mistaken beliefs about the first technology will cause them to mislearn the second technology. The team is studying fertilizer application in China. Farmers learn about the returns to two types of fertilizers, nitrogen vs phosphorus and potassium fertilizers. We find convincing evidence that Chinese farmers overuse nitrogen fertilizers but underuse phosphorus and potassium fertilizers. Such simultaneous underuse and overuse of fertilizers are especially puzzling, because farmers have been working with these technologies over decades, which should be a sufficient time frame for them to learn the correct usage, according to classic learning-by-doing models. This learning failure issue serves as the starting point of the paper, which also has significant implications for most developing countries and developed countries.

Study Design

Using a model and two-phase randomized control trials among 1,200 farmers in 200 villages, Ao and team answer four questions: 1) Can farmers learn the correct usage of different fertilizers? 2) How large is the efficiency loss from the misuse of different fertilizers. 3) What theory can explain the simultaneous overuse and underuse of different fertilizers. 4) Can theory-guided intervention resolve such learning failure issue? Using administrative data on soil analysis of millions of local plots, we linked farmers to the nearest testing plots using GPS tracking and generated customized fertilizer recommendations to farmers. In the second field experiment, the team cross-randomized whether farmers receive leaf color charts or not, which can help them better calibrate the optimal level of greenness.

Results and Policy Lessons

The first-phase recommendations are followed by farmers and lead to reduced nitrogen application and increased phosphorus and
potassium use. As a result, average yields and revenues are 6% higher compared to the control group, while total fertilizer costs
remain unchanged. We also find a significant reduction in greenhouse gas (N2O) emissions linked to nitrogen overuse. If we scale up
this intervention to all rice farmers in China, greenhouse gas emissions would reduce by 37 million tons per year, corresponding to 0.4% of China’s annual CO2 emissions and farmer’s revenues would increase by 30 billion Chinese Yuan. The second-phase experiment reduced nitrogen fertilizer application by 11% immediately. It also encouraged a small proportion of farmers to start to use phosphorus and potassium fertilizers for the first time ever, compared to 3% in the control group.