Essays on development economics and industrial organization

Abstract

This dissertation studies two disparate topics in development economics and industrial organization respectively: (a) the role of financial intermediation in promoting economic growth in developing countries and (b) the effects of learning on agents' search behavior. The first essay investigates the effects of commercial banking on economic growth. The tendency of banks to locate in profitable areas experiencing higher growth typically complicates the identification of banking effects. I exploit a previously unstudied reform of bank branching policy in India between 2005-06 that led to a large expansion in private bank credit to financially underserved areas. Using iterations of a regression discontinuity design, I trace the exogenous expansion of banking services through time at the district level. I show this expansion produced positive effects in agriculture and manufacturing. I confirm greater gains in local GDP growth using remote sensing data to overcome the lack of official GDP statistics at the district level. These results offer evidence of a causal impact of financial system expansion on economic development. The second essay examines how the geographic reach of a bank's network of branches can affect its ability to spread risks across spatially separated regions. I investigate the causal impact of the spatial expansion of Indian banks resulting from the bank branching policy reform on smoothing the consumption of households with respect to local weather and agricultural productivity shocks. The third essay (coauthored with Sergei Koulayev) extends a model of sequential search for differentiated goods by relaxing the standard assumption of rational expectations. Agents are likely to refine their imperfect knowledge of product markets while searching. By introducing Bayesian learning into agents' beliefs, the model better replicates important aspects of search behavior. Using data from a popular internet hotel search site, we estimate lower median search costs in the model with Bayesian learning. Considering a counterfactual in which the first page of search results present the most popular hotel options, we estimate an increase in the number of successful searches.