Abstract: Following the introduction of genetically modified bt cotton seeds in 2002, cotton farmers in India overwhelmingly switched to the more profitable bt seed variety but failed to adjust the level of pesticide use leading to lower profit. Since over use of pesticides cannot be explained by financial or learning constraint we consider a decision problem involving learning constraint for a rational producer. The producer needs to choose more than one input to maximize payoff. We assume he faces the choice of learning about each input unconditionally, i.e, learning about best seed, or learning conditionally about an input combination, i.e, mix of seed and pesticide, subject to two different costs of learning. We find the optimal learning strategy is observing averages first followed by cells unless the difference between the two costs are extreme. For very small and large differece in cost the optimal strategy is to observe only cells or averages. We also find that reducing only one cost can reduce the precision of posterior belief which ca propagate through social learning.
For questions, please contact: Srijita Ghosh - firstname.lastname@example.org