Testing for Parameter Instability using the R/S Statistic

This paper explores the use of the R/S statistic as a means of checking for parameter instability. The nature and properties of the statistic are described, and its behaviour and power in the context of three situations of structural change are examined. The results suggest that the R/S statistic has an ability to detect shifts in means and changes in the intercept of a regression relationship. Moreover, it is more powerful than the OLS variant of the well known cumulative sum of squares residual test for detecting unknown structural breaks in the cases involving the linear regression models that were examined. It would therefore seem to us that the application of the R/S statistic to the problem of testing for structural instability is worthy of further investigation.