SBR-9422942 Abhijit Banerjee During the last fifty years, there have been many empirical studies about the specific process by which ideas, innovations, fads or new behaviors spread through a society. By contrast, the theoretical literature on social diffusion is much smaller and more recent and follows two approaches: the IO approach that emphasizes network externalities, and the approach that emphasizes social learning: that is, the phenomenon of economic agents learning from each other's actions. The research to date has shown that change can be both inefficient and abrupt and that the outcome can be dependent on the choices made by a small number of poorly informed early decision-makers. The latter property is sometimes called "herding" behavior. This has made social learning models very attractive for explaining the volatility of collective behavior. Given this interest, it is important to understand the underpinnings of the herding results. The current research does this by examining a model of social learning which is a natural alternative to the herd behavior model. The key assumption dropped in the new research is that everyone knows all the choices made in the past. If, instead, every agent gets to sample only a small number of previous decisions, then cascading to an inefficient outcome may no longer occur. The research goes on to examine the effect of forgetting and reporting and perception biases on the social diffusion process. This helps to get a clearer idea of the environments in which herding occurs and those in which it does not. The second project looks at the relationship between a traditional rural sector and an industrial urban sector. The previous literature assumed that the rural peasants were not economically rational and that urban wages were not flexible downward. Because of this, the literature fell into disrepute. The purpose of the research here is to revive the distinction between urban and rural sectors without assuming irrationality of persi stent disequilibrium. This is accomplished by finding a compensating differential that accounts for the difference in wages in the two sectors: that differential is a lower cost of borrowing in the rural sector due to better monitoring. The model is capable of answering many questions about the pattern or income distribution and the evolution of the rural-urban wage differential.
