Market Risk Beta Estimation using Adaptive Kalman Filter

Market risk of an asset or portfolio is recognized through beta in Capital Asset Pricing Model (CAPM). Traditional estimation techniques emerge poor results when beta in CAPM assumed to be dynamic and follows auto regressive model. Kalman Filter (KF) can optimally estimate dynamic beta where measurement noise covariance and state noise covariance are assumed to be known in a state-space framework. This paper applied Adaptive Kalman Filter (AKF) for beta estimation when the above covariances are not known and estimated dynamically. The technique is first characterized through simulation study and then applied to empirical data from Indian security market. A modification of the used AKF is also proposed to take care of the problems of AKF implementation on beta estimation and simulations show that modified method improves the performance of the filter measured by RMSE.