Machine Learning Asset Pricing Factors in an Emerging Stock Market

Abstract

We analyze risk factors in the emerging stock market of Romania using Machine Learning classification models and find novel evidence showing that Liquidity, Conditional Skewness, and Volatility display predictive power over long-term stock returns. In our sample, a portfolio formed using information derived from all three factors earns an average excess return of 12% per year, a result with t≥3 significance. This is adjusted for the risk premiums associated with the Market, Size, and Value factors, which are reconfirmed as being significant in our analysis. Momentum is shown to have no influence, adding to existing evidence pointing toward the same conclusion for markets from the Central and Eastern European region. Besides uncovering potential novel risk factors, our paper shows that Machine Learning models are a useful tool for studying asset pricing in small, emerging stock markets.