Testing for covariance stationarity of stock market returns : an intervention analysis

Abstract

This thesis adopts the Omran and McKenzie (1999) testing procedure to test for covariance stationarity in the stock's index return series for stock markets in Hong Kong, Singapore, Australia and the United States. The objective of the test is to investigate the role of the structural break on covariance stationarity. The procedure comprises both the Loretan and Phillips (1994) test and intervention analysis. The results suggest that Hang Seng Index (Hong Kong) returns and Strait Time Index (Singapore) returns are covariance stationary if they properly filtered out the unusual periods. Moreover, the effects of structural break are found to be not significant in the return series of the All Ordinance Index (Australia) and the Dow Jones Industrial Average Index (the United States). They can be assumed as covariance stationarity and covariance non-stationarity, respectively.