Excess Returns and Risk Premia: Using Multi-Factor Models for Understanding Excess Returns

Abstract

This paper utilizes linear regression (parametric) and random forest (non-parametric) models to develop predictions of excess returns on publicly listed US and European firms, by using a variety of ESG, liquidity, financial, and market variables, that are firm-specific. The findings show that market variables are generally strong predictors of excess returns, with firm financial health also being a strong indicator of excess returns, and liquidity and ESG variables only having situational predictive capabilities. Financial health is found to be a more important indicator of excess returns than financial growth, indicating that investors are loss-averse, more so than risk-averse, consistent with behavioral economic theory. Explanatory variables correlation to excess returns are found to be nonlinear and conditional, as different variables are only statistically significant to the prediction of excess returns depending on firm characteristics. It is suggested that parametric models are useful in understanding the correlations between excess returns and explanatory variables, while non-parametric models are more accurate in predictions, and therefore the models should be used in combination.