UC San Diego

The three chapters of this dissertation examine the spatial organization of economic activities on a micro scale in a growing historical city, and on a national scale in contemporary India. The first chapter focuses on firms, investigating the question why small family firms were so prevalent in historical cities. I test whether this phenomenon is caused by a lack of technological capability to move goods and people. I exploit the natural experiment that Boston quickly electrified its previous horsedrawn streetcar system between 1889 and 1896. Analyzing new data transcribed from Boston business records from 1885 to 1905, I find that rail-connected locations experienced a 5.3-percentage point relative drop in the share of sole proprietorship establishments after the streetcar electrification, indicating that improved market access leads to an increase in average firm size. The second chapter focuses on individuals, investigating the question whether immigrants stay in ethnic enclaves due to a lack of information about outside communities within the same city. Based on individual records linked between city directories and the decennial census in Boston from 1885 to 1900, I track within-city migrations of immigrants in response to the same transport upgrade event in Boston. I find that immigrant enclave residents who worked within 25 meters of the streetcar rails in 1885 were much more likely to move to a less segregated neighborhood in 1900, compared to enclave residents who worked between 25 and 50 meters away from the rails. Evidence suggests interactions in workplace had an impact on the choices of the residential locations fifteen years after. In the third chapter, my coauthors and I perform a large scale classification of satellite imagery into “built-up” or “not built-up” areas to measure the urbanization process in India, using a reliable and comprehensive ground-truth data set we construct and a cloud-based computational platform - Google Earth Engine (GEE). Our methodology yields a classification accuracy rate between 70% and 85%, and can easily be applied to other countries and regions.