Time-varying risk premia traditionally have been associated with the empirical fact that conditional second moments are time-varying. This paper additionally examines another possible source for time-varying risk premia, namely the market price of risk (lambda). For utility functions that do not imply constant risk aversion measures, the market price of risk will in general change over time. We provide empirical evidence for the German stock market in a bivariate GARCH-M framework using alternative specifications for lambda. The results indicate that a model with lambda being a function of typical volatility measures performs best for most series. To facilitate the interpretation of the results, we plot impulse response functions of the risk premia.
impulse response analysis Market price of risk Multivariate GARCH-Models CAPM