Asset Pricing Models: A Comparative Exercise Using Neural Networks to the Colombian Stock Market

This study seeks to evaluate the effectiveness that variables like firm size and book-to-market ratio—present in the model of Fama and French—have to capture the average expected return on assets, as compared to macroeconomic fundamentals or the market index. For this purpose, we used an artificial neural network model (ANN), which departs from a structure of non-linear estimation to capture some irregularities that characterize financial markets. We found that the Fama and French model accounts for the conditions of the Colombian stock market better, which suggests the importance of microeconomic risk factors to explain asset returns.