Extreme News Events, Long-Memory Volatility, and Time Varying Risk Premia in Stock Market Returns

Abstract

This paper proposes a new class of GARCH-jump in mean models to test the presence of time varying risk premia associated with normal and extreme news events. The model allows for a dynamic jump component with autoregressive jump intensity, long-range dependence in volatility dynamics, and volatility in mean structure separately for normal and extreme news events. The results show significant jump risk premia in five stock market index returns. We also find that ignoring the long-memory feature in volatility dynamics leads to false rejection of time varying risk premia.