Autoregressive conditional beta

The capital asset pricing model provides various predictions about equilibrium expected returns on risky assets. One key prediction is that the risk premium on a risky asset is proportional to the nondiversifiable market risk measured by the asset's beta coefficient. This paper proposes a new method for estimating and drawing inferences from a time varying capital asset pricing model. The proposed method, which can be considered a vector autoregressive model for multiple beta coefficients, is different from existing time-varying capital asset pricing models in that the effects of an exogenous variable on an asset's beta coefficient can be unambiguously determined and the codependence between the beta coefficients of individual assets can be measured and estimated.