| Summary: | 碩士 === 國立交通大學 === 管理科學系 === 91 === The performance of portfolio-based momentum strategies on the U.S. market is investigated in this paper. Three main conclusive remarks are drawn. Firstly, contrary to the finding of individual stock-based trading strategies, most portfolio-based momentum strategies are found to perform well in the short- and long- term. Further, momentum in returns seems a pervasive phenomenon in our robust tests. Also, we document different grouping methods affect the profitability of momentum strategies. Several possible explanations are proposed for the profitability of executing portfolio-based momentum strategies in short- and long- term, but not so when executing individual stock-based strategies. Secondly, it is found that time-series predictability plays an important role in portfolio returns. Furthermore, the source of momentum profits depends on the status of the sign of covariance matrix. If autocorrelations and cross-autocorrelation are positive, the source of momentum profit is mainly due to the autocorrelation. On the other hand, if autocorrelation and cross-autocorrelation are negative, the source of momentum profit is mainly due to the cross-autocorrelation and the source of contrarian profit is mainly due to the autocorrelation. Moreover, the sign of covariance matrix is driven by the serial correlation of the market factor, and thus confirm the inference advocated by Chen and Hong (2002). Thirdly, the abnormal returns are still survival in the asset-pricing models. It is concluded that risk-adjusted returns make the momentum effect appear more at odds with the joint hypothesis of market efficiency.
|
|---|
