Stationarity Tests of the Market Model for Security Returns

Abstract

This paper shows by simulation that the cusum-of-squares test rejects stationarity of security returns too often when residual returns are nonnormal and/or heteroscedastic. An alternative test, suggested by Farley, Hinich, and McGuire [12], appears to be more powerful and robust than the cusum-of-squares test. The Farley-Hinich test finds only a small number of securities for which the slope or the intercept of the market model are nonstationary, suggesting that the cusum-of-squares test results may reflect the departures of residual returns from the ideal test assumptions.