This paper investigates the performance of three different approaches to modelling time-variation in conditional asset betas: GARCH models, the extended market model of Schwert and Seguin (1990) and the Kalman Filter algorithm. Using daily UK industry returns, we find the simple market model beta to be as efficient as the more complicated GARCH type models. However, the Kalman Filter algorithm incorporating a random walk parameterisation dominates all other models under the mean-square error criterion. Finally, we provide strong evidence that a combination of the methods under investigation may lead to considerably more powerful estimators of the time-variation in conditional beta.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below