How to measure and model volatility is an important issue in finance. Recent research uses high frequency intraday data to construct ex post measures of daily volatility. This measure, called realized volatility, permits the modeling of volatility by traditional time-series methods. Barndor?Nielsen and Shephard(2004) have introduced additional volatility instruments called realized power variation and realized bipower variation. We investigate the benefits of these volatility instruments in modeling and forecasting volatility. The first contribution of this paper is to demonstrate that realized power variation can provide dramatic improvements in predicting volatility for foreign exchange and equity markets. Secondly, given the large number of possible models, we consider the benefits of Bayesian model averaging. The model average reduces the risk of choosing an individual model and provides overall strong performance for each volatility series and forecast horizon.
展开▼
