Filtering Methods for the Estimation of the Long-Run Risk Asset Pricing Model

Abstract

Previous attempts to estimate the long-run risk (LRR) model revealed serious methodological issues and low estimation precision of the existing econometric approaches. However, this study shows that despite the presence of latent variables asymptotically efficient maximum likelihood (ML) estimation is possible through application of filtering methods. A three-step estimation strategy is suggested that involves ML estimation relying on the Kalman filter and a particle filter, which allows to identify all LRR model parameters. A Monte Carlo study assesses the estimation precision for different sample sizes, an empirical application presents estimation results obtained from U.S. data.