DescriptionThis dissertation studies the cross-section of asset returns. That is, why do certain assets receive higher average returns than others and what factors drive these differences in returns. Chapter 1 presents with a brief introduction, discussing the rise of machine learning in the context of asset pricing.Chapter 2 presents a novel comparison of econometric and machine learning asset pricing models. Previous comparisons of econometric and machine learning based asset pricing models have focused on statistical measures such as the R2. In this chapter I compare popular machine learning techniques with traditional factor models using the level of mispricing, as measured by the model’s alpha, as the primary evaluation metric. In making this comparison I highlight where machine learning models cannot be implemented in the traditional asset pricing framework. For comparison to previous papers in the literature I also compare models based on forecast accuracy as measured by the out-of-sample R2. Using the 30 industry portfolios as test assets traditional factor models achieve smaller average levels of mispricing and more accurate forecasts of asset returns. The benefits of deep learning recently documented in the literature appear the result of the empirical implementation of specific and not a universal truth. Chapter 3 introduces a factor based on an estimated probability of bankruptcy, making an explicit connection between risk and return. Using an underlying model of corporate bankruptcy built as a sequence of two random forests, I demonstrate my bankruptcy risk measure not only predicts bankruptcy more accurately than existing models, but the factor created using this measure earns statistically significant monthly returns of 0.23%, 0.15%, 1.97% and 1.04% in equity, bond, option and credit default swap markets, respectively. In markets with existing common factors I find statistically significant alpha with respect to these factors indicating they contain information about these markets that is orthogonal to these existing common factors. Chapter 4 investigates the equity cross-sections of real estate investment trusts (REITs) and states. A nine-factor asset pricing model which critically relies on the bankruptcy risk factor of Neumann (2021) produces REIT portfolios which outperform the REIT market in terms of Sharpe ratio and the S&P 500 index in terms of absolute returns. This chapter also develops a cross-sectional neural network for municipal bond yields, filling a need to extrapolate yields on these illiquid assets. Additionally, counterfactual municipal bond yields generated by this neural network produce portfolios which outperform the S&P 500 index in terms of absolute returns.