改善資訊交易機率的估計

This project proposes to improve the estimation methods of Easley et. al.’s (1996) probability of informed trading (PIN) and of Duarte and Young’s (2009) adjusted probability of informed trading (AdjPIN). Traditionally, it is assumed that the probability of the occurrence of a good or bad signal is the same in the sample period, so that the maximum likelihood method is employed to estimate the probability. Such an assumption is not entirely reasonable given the fact that the sample period often lasts a quarter or even a year. This project therefore assumes that the probability of a good signal for a security is positively correlated with its rate of return during trading hours, and uses the return to construct the proxy for the probability and to re-write the maximum likelihood function. This project will compare the traditional and the new methods in terms of the convergent rates of the parameters, the precision of the estimations, and the time spent on estimations.