Estimation of mean and variance episodes in the price return of the stock exchange of Thailand

Abstract

This paper extends the novel and useful recent work of Inclán and Tiao (1994). Their invention was a procedure, based on CUSUM statistics, that would estimate the location of change points in the variance of a time series. With a Monte Carlo analysis this paper shows that the basic Inclán and Tiao procedure will also estimate changes in the average of a time series. This finding substantially weakens the value of the original Inclán and Tiao procedure and brings into question the veracity of the prior work using this procedure. To resurrect the original Inclán and Tiao procedure this paper introduces an addition to the Inclán and Tiao procedure that will statistically identify whether an estimated change point is a change in average, variance, or both. To ascertain its practical value, the augmented procedure is then applied to the daily return on the index of the Stock Exchange of Thailand (SET).