Essays in Applied Machine Learning and Economics

Book excerpt: In this dissertation, we contribute to two strands of economic literature- applied economics and game theory. The contribution to applied economics is developing a prediction model for PM2.5, a key indicator of air pollution. The PM2.5 predictions from this model enable an analysis of economic and environmental policies that were previously infeasible due to a lack of PM2.5 measurements, especially in developing regions. We also demonstrate the unsuitability, for Delhi specifically, of the predominant benchmark estimates for PM2.5, in the applied economics literature supplied by van Donkelaar et al. Additionally, we were able to introduce and demonstrably improve upon a frontier technique from the deep learning literature to the applied economics literature. The contribution to the game theory literature takes the form of assessing the optimality of contests as a mechanism in the context of a standard Myersonian mechanism design environment, a previously unexplored setting. We find that despite extensive usage of contests as a mechanism in the real world, it is not without loss in revenue to use optimal contests. This dissertation's primary contribution is developing a modeling pipeline with lower data requirements and better predictive performance than the existing state-of-the-art estimates in the applied economics literature.