Demonstration activities are often used to diffuse new technologies in the field. Implicit in this choice is that social learning is the prevalent learning mechanism leading to adoption. However, previous research suggests that learning from experience may be more powerful than learning from others. We design a field experiment in Bangladesh to estimate the relative importance of learning from self vs. others in the adoption of new crops.
We randomly assign 102 villages to one of three modalities: a traditional demonstration model and two alternative models in which the resources to demonstrate are divided among progressively larger numbers of farmers, who cultivate smaller experimental plots on their own farms.
We find that splitting the package across as many farmers as want to experiment with the new crops leads to significantly more adoption in the two years following the demonstration season. To identify the mechanisms behind this differential increase, we exploit experimental variations in peer and self-learning and calibrate a target-input model of technology adoption to our experiment. Both exercises lead us to conclude that the learning effects of self-experimentation are greater than peer learning effects within these farmer groups.