I use night lights data to construct a worldwide panel of city shapes and sizes. Existing methods define a city as contiguous pixels with high light intensity, using an arbitray intensity threshold to define the boundary. I develop a data-driven method to get the optimal city-specific light intensity threshold, using the gradient and the hessian of the light intensity field. This has two main benefits. First, it gives more accurate city boundaries across a wide range of city sizes allowing analysis over large regions. Second, it controls for the non-calibrated variation in sensitivity of satellite sensors across time, which allows for construction of a panel of city shapes, whereas previous approach could only construct cross-sections. Having a accurate measure of city boundaries is critical for research on topics such as segregation. Hence, this method provides a data foundation for large scale panel data analysis on such topics in urban and spatial economics in low-income countries, where standard measures of city boundaries—such as commuting zones—are not available.

In this project I explore the long-term impact of colonial institutions on contemporary ethnic relations by estimating the impact of British rule on residential segregation in India. I compare the levels of segregation in regions that were directly ruled by British versus those that were ruled be them indirectly through native kings who had high levels of autonomy. For causal identification, I use random deaths of kings without heirs (Iyer, 2010) and European Wars (Mukherjee, 2017) as instruments for direct rule.