Behavioral investment strategy matters: a statistical arbitrage approach

In this study, we employ a statistical arbitrage approach to demonstrate that momentum investment strategy tend to work better in periods longer than six months, a result different from findings in past literature. Compared with standard parametric tests, the statistical arbitrage method produces more clearly that momentum strategies work only in longer formation and holding periods. Also they yield positive significant returns in an up market, but negative yet insignificant returns in a down market. Disposition and over-confidence effects are important factors contributing to the phenomenon. The over-confidence effect seems to dominate the disposition effect, especially in an up market. Moreover, the over-confidence investment behavior of institutional investors is the main cause for significant momentum returns observed in an up market. In a down market, the institutional investors tend to adopt a contrarian strategy while the individuals are still maintaining momentum behavior within shorter periods. The behavior difference between investor groups explains in part why momentum strategies work differently between up and down market states. Robustness tests confirm that the momentum returns do not come from firm size, overlapping execution periods, market states definition or market frictions