Essays on Probabilistic Machine Learning for Economics

Abstract

This thesis consists of three essays that explore the use of probabilistic machine learning techniques in combination with information-theoretic concepts to answer economic questions. Over the past years, economists have started applying machine learning methods to a wide range of topics. Probabilistic methods in the context of unsupervised learning represent one particular modelling approach at the intersection of computer science and statistics. While widely used in applied statistics, these models, however, do not necessarily provide relevant and interpretable outputs from an economist's perspective. In this thesis, I appeal to information-theoretic methods to summarise the probabilistic information inferred from such models and construct economically meaningful measures.