Can differences in individual learning explain patterns of technology adoption? Evidence on heterogeneous learning patterns and hybrid rice adoption in Bihar, India.

Jared Gars,Patrick S. Ward

doi:10.1016/j.worlddev.2018.11.014

Abstract

Much empirical research that has shown that an individual's decision to adopt a new technology is the result of learning - both in personal experimentation as well as observing the experimentation of others. Yet even casual observation would suggest significant heterogeneity learning processes, manifesting itself in widely varying patterns of adoption over space and time. In this paper we explore this heterogeneity in the context of early adoption of hybrid rice in rural India. Using specially-designed experiments conducted as part of a primary survey in the field, we are able to identify which of four broad learning heuristics most accurately reflects individuals' information processing strategies. Linking these learning heuristics with observed use of rice hybrids, we demonstrate that pure Bayesian learning is well suited for the tinkering and marginal adjustments that would be required to learn about a technology like hybrid rice, but is also more cognitively taxing, requiring a longer memory and more complex updating processes. Consequently, only about 25 percent of the farmers in our sample can be characterized as pure Bayesian learners. Present-biased learning and relying on first impressions will likely hinder adoption of a technology like hybrid rice, even after controlling for access to credit and a rudimentary proxy for intelligence.

Highlights

In many parts of the developing world, the transition from indigenous agricultural practices to modern technologies is often viewed as a critical step toward achieving broad agricultural development objectives such as food security or self-sufficiency
To foreshadow our results, we find that pure Bayesian learning is well suited for the tinkering and marginal adjustments that would be required to learn about a technology like hybrid rice
In India, as in much of South Asia, the adoption of these modern varieties provided the potential for increased yields, though arguably these new varieties did not reach their full yield potential until they were paired with complementary inputs, such as irrigation and fertilizers

Summary

Introduction

In many parts of the developing world, the transition from indigenous agricultural practices to modern technologies is often viewed as a critical step toward achieving broad agricultural development objectives such as food security or self-sufficiency. In the context of modern seed varieties during the Green Revolution, Foster and Rosenzweig (1995) have demonstrated the importance of learning, both from one’s own experimentation as well as in observing the experimentation of others. This process involves an iterative process of forming and updating beliefs about yield or profit distributions, with new infor-. The technology adoption literature has typically assumed Bayesian learning as it is empirically tractable and theoretically consistent (Conley & Udry, 2010; Foster & Rosenzweig, 1995) This is clearly an unrealistic assumption in almost any real-world context. The literature on learning in non-cooperative games has adapted a variety of models of learning and updating beliefs which may be more relevant in contexts where there are low levels of human capital (Cheung & Friedman, 1997; Camerer & Ho, 1999)

Methods

Results

Conclusion