Time variation and asymmetry in systematic risk: evidence from the Finnish stock exchange

Abstract

This paper models time-varying betas in the Finnish Stock Exchange using a bivariate version of the asymmetric GARCH model introduced by Glosten et al. (unpublished working paper, Department of Finance, Colombia University, 1989). An extension of Bollerslev's (Rev. Econ. Stat. 72 (1990) 498) constant correlation model is used for the conditional covariance. This extension allows the conditional covariance to respond asymmetrically to up and down movements of the market, a sort of ‘leverage effect’ in the covariance. There is evidence of substantial time variation in betas. In all instances, time-varying betas follow stationary processes with slow mean reversion toward their long-tem means. There is some evidence that betas are asymmetric in up and down markets. Using metrics such as the mean square error, the mean percent error and the mean absolute error it is confirmed that time-varying betas offer a better description of the short-term dynamics of systematic risk.