| Summary: | 碩士 === 逢甲大學 === 統計學系統計與精算碩士班 === 102 === In this paper, we develop a nonlinear quantile CAPM with heteroskedasticity, nonlinear market betas, nonlinear lagged abnormal volume factor, and nonlinear volatility dynamics. It’s widely reported that volume is related to return and such volume-return relationship provides insight into financial market structures and reflects investors’ preferences. Hence, we employ HP-filter to separate the log-volume time series into a stochastic growth trend of volume and the abnormal volume time series. We add the lagged abnormal volume factor in CAPM to capture irrational behavior, to provide predicting information, and to enhance the explanatory power of CAPM. Quantile regression is employed to examine the dependence of lagged volume on return which is uncovered by mean regression. To efficiently estimate the coefficients, Bayesian MCMC methods with asymmetric Laplace distribution are utilized. We analyze six Dow Jones Industrial stocks to demonstrate our proposed models. The results exhibit significantly negative effects of abnormal volume on stock excess return under low quantile levels while there are significantly positive effects under high quantile levels. Each Market beta varies with different quantile levels, representing fluctuations of systematic risk in the stock market. We observe that the coefficients of lag-one stock excess return and abnormal volume are asymmetric between lower regime and upper regime under extreme quantile levels. This work confirms that extreme trading volume contains information about the future evolution of stock prices. More importantly, considering abnormal volume factor could enhance the explanatory power of CAPM and provide considerations in behavioral finance. Adopting these findings, fund managers and investors could have more flexible strategies than using the traditional CAPM.
|
|---|
