Further Analysis of Efficient Portfolios with the USER Data

Abstract

In this study, we show that earnings forecasting and price momentum strategies complement fundamental stock selection strategies such that a composite model can be effectively implemented using both enhanced index-tracking portfolios and traditional mean–variance portfolios. The mean–variance optimization model produces statistically significant asset selection portfolios that dominate less-aggressive enhanced index-tracking portfolio construction models. We show that portfolios that use tracking error in risk optimization techniques produce a superior risk–return trade-off than traditional mean–variance optimization techniques. A portfolio manager should use a data mining corrections test to minimize the probability that the models selected resulted from a near-random process.

TOPICS: VAR and use of alternative risk measures of trading risk, big data/machine learning, portfolio construction