Machine Learning Classification Methods and Portfolio Allocation : An Examination of Market Efficiency

We design a novel empirical framework to examine market efficiency through out-of-sample(OOS) predictability. We frame the classic empirical asset pricing problem as a machine learningclassification problem. We construct classification models to predict return states. The prediction- based portfolios beat the market in time series and cross-sections with significant economic gains.We directly measure prediction accuracies. For each model, we introduce a novel applicationof binomial test to test the accuracy of 3.34 million return state predictions. Our models cangenerate information about the relation between future returns and historical information. Theestablishment of predictability questions the correctness of prices