Searching for periods of volatility: A study of the behavior of volatility in Thai stocks

Abstract

This paper improves the precision of the useful new procedure of Inclán and Tiao (1994) that estimates variance shift points in a time series. It accomplishes this by incorporating the evidence of Bos and Fetherston (1992) that the linear Brown, Durbin, and Evans ( Brown et al., 1975) critical CUSUM of squares boundaries [used by Inclán and Tiao] produce an understatement of instability at the data end points. This is solved by Tanizaki (1995) which, like Bos and Fetherston (1992, 1995), uses the fact that the CUSUM of squares statistic follows a beta distribution. This study uses the Inclán and Tiao procedure with the nonlinear Tanizaki CUSUM of squares boundaries to research volatility in Thai stock returns. The paper's empirical results show that, on any trading day, there is a 1.16% chance that a Thai stock will experience a shift in volatility. The results also show that this incidence is not random, and, hence, it is possible to predict the incidence of shifts. Though the results here cannot answer the question of how to do this, we suspect that movements in average return have a role to play. We propose that the culprit may be changes in the average return, and therefore that the estimated volatility shift points may be spurious.