Nonlinear trend prediction and active investment strategies in intertemporal manipulation and speculation markets
by Liqun Xu
THESIS
1997
Ph.D. Industrial Engineering and Engineering Management
xvi, 200 leaves : ill. ; 30 cm
Abstract
This thesis first models and analyzes a three-period intertemporal market of manipulation and speculation on themes where manipulators learn themes eariier while active speculators and passive speculators have heterogeneous information on the themes. We show that in a rational expectation framework, equilibrium exists when asymmetrically informed risk-averse traders maximize their expected utility over projected wealth. Our model suggests that the manipuiation and speculation activities on themes gradually increase the information content of price before themes become public, and the existence of consistent patterns strongly implies the possibility of prediction. The thesis then presents a novel method of trend prediction in intertemporal manipulation speculation periods. Incorporating...[ Read more ]
This thesis first models and analyzes a three-period intertemporal market of manipulation and speculation on themes where manipulators learn themes eariier while active speculators and passive speculators have heterogeneous information on the themes. We show that in a rational expectation framework, equilibrium exists when asymmetrically informed risk-averse traders maximize their expected utility over projected wealth. Our model suggests that the manipuiation and speculation activities on themes gradually increase the information content of price before themes become public, and the existence of consistent patterns strongly implies the possibility of prediction. The thesis then presents a novel method of trend prediction in intertemporal manipulation speculation periods. Incorporating a feedforward neural network, the prediction system is trained to predict the short-term trends of price movement. The performance of the out-of-sample forecasting using real trading data from the Hong Kong stock market is several times better than chance. Even after considering transaction costs, interest rate and risk adjustment, our dynamic strategies based on trend prediction generate much higher return than the passive strategy of buy-and-hold on index. These results strongly support the intertemporal predictability implied by the three-period model of market manipulation and speculation.
Post a Comment