Performance evaluation of modified adaptive Kalman filters, least means square and recursive least square methods for market risk beta and VaR estimation

Adaptive Kalman Filters (AKFs) are well known for their navigational applications. This work bridges the gap in the evolution of AKFs to handle parameter inconsistency problems with adaptive noise covariances. The focus is to apply proposed techniques for beta and VaR estimation of assets. The empirical performance of the proposed filters are compared with the standard least square family and KF with respect to VaR backtesting, expected shortfall analysis and in-sample forecasting performance analysis using Indian market data. Results show that the Modified AKFs are performing at par with the bench mark even with these adaptive noise covariance assumptions.