One of the most robust anomalies is the price momentum effect identified by Jegadeesh and Titman (1993) , which is followed by a long-term reversal pattern, initially documented by De Bondt and Thaler (1985, 1987). In recent years, tremendous empirical evidence indicates that investors may overreact to news with salient features, while at the same time underreact to non-salient news; consequently stock returns are overstated in the former case and understated in the latter case. In this project, I propose a method that decomposes observed stock returns into two components: robust returns and residual (or excess) returns, where the latter reflects investors misreaction or correction to previous misreaction. I assume that the cross-section of true stock returns for each month has a normal distribution. A stock's true return is recovered by plugging the percentile of the observed return back to the inverse normal distribution, with mean and variance being their sample counterparts, resulting in a number which is referred to as the "robust return." I then use the robust returns to form robust momentum strategies as in Jegadeesh and Titman (1993). As the robust returns are presumably less affected by the presence of extreme observations, we expect the performance persistence of robust momentum to be stronger.Similarly, the difference between the observed and robust returns is defined as the residual return. We expect the residual momentum constructed on the basis of residual returns to experience quicker reverse with respect to either traditional momentum or the robust momentum. Indeed, the preliminary results support my conjecture. The robust momentum persists, and experiences no return reversals. By contrast, the residual momentum experiences reversals immediately following formation. I believe the project shall provide important insight into behavioral finance literature. This 3-year project shall explore the profitability of proposed strategies from both rational and behavioral perspectives.
|Effective start/end date||1/08/18 → 31/07/19|
- nonparametric statistic
- trading strategies
- investor sentiment
- asset-pricing model.
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.