Evolution of Asset Pricing Models: From Traditional Methods to Machine Learning Applications

Abstract

This paper examines the development of asset pricing models following chronological order. Beginning with traditional methods, this paper introduces the historical background of the Capital Asset Pricing Model (CAPM) as well as the strengths and weaknesses of the application of this model in the real world. To enhance the predictiveness of CAPM and improve this over-simplified model, researchers have come up with the FAMA-French Model with multiple factors. These models incorporate more factors when predicting returns of assets, but they are still based on linear relationships. With the growing volatility of the financial market, the financial industry requires more complicated models to capture more nuanced trends. With such a need, the researchers apply the most advanced machine learning technique to the asset pricing process, using models such as neural networks, random forest, and gradient boosting to enhance the predictiveness of asset pricing models. However, these advanced methods also bring issues such as overfitting and interpretability to investors. By tracing the development of asset pricing methodologies, this paper offers insights into the ongoing refinement of financial models and points out potential future research direction.