Imperfect information and learning: Evidence from cotton cultivation in Pakistan

Abstract

Information problems are pervasive in developing economies and can hinder productivity growth. This paper studies how much rural producers in developing countries can learn from their own cultivation experience, i.e. learning by doing, to redress important information gaps about imperfectly known input technologies. First, I build a theoretical model which links learning by doing in one period to improved input choices in the next period, and show that this can be impeded by uncertainty about what is being observed due to noisy cultivation signals and by uncertainty about what to infer about market varieties due to imperfect variety integrity. Second, I apply this framework to cotton farmers in Pakistan, where farmers have imperfect information prior to cultivation about the extent to which their seeds have pest resistant biotechnology. The results suggest that farmers are unable to learn by doing about this aspect of their seeds due to a high degree of noise in cultivation signals. The paper highlights the potential difficulties in parsing out and processing information from cultivation experience alone and therefore of learning by doing by rural producers in a development context.