Bayesian and mixed estimators of time varying betas

Based on the random coefficient model, Vasicek's static Bayesian beta coefficient adjustment model is extended to a dynamic model. It is shown that the time-varying security beta model can be used to identify and resolve the existence of nonstationary (weak stationary) beta coefficient over time. The implication of a random beta coefficient on th standard Bayesian adjustment is also explored. The usefulness of employing time-varying security beta estimates in forecasting the future beta in terms of Box and Jenkins' ARIMA model is also empirically demonstrated.