I develop an estimation methodology that allows for three-sided heterogeneity. I implement this on matched panel data. I use machine learning in the classification step of the estimation of a Markovian model of worker mobility and apply it to novel data on police departments.
Using novel data combining the religion of the final pre-colonial ruler with literacy outcomes of Hindus and Muslims in India between 1881 and 1921, we show that religious groups resisted modern education due to foreign occupation.
We examine the high-powered financial incentives for bureaucrats. We digitize the financial disclosures of elite bureaucrats from India and combine this novel data with web-scraped career histories to estimate the private returns to public servants …