Multivariate Range-Based Egarch Models

The dynamic conditional correlation (DCC) and co-range models are two main frameworks widely used to employ range-based univariate volatility. Using the two approaches, we build novel multivariate range-based EGARCH (REGARCH) models: a DCC-REGARCH and co-range REGARCH (CRREGARCH) model, as well as a co-range CARR (CRCARR) model. We compare these models with five extant models over twelve forecast horizons, ranging from one to twelve weeks, and examining currencies and E-mini futures. Out of the eight models, the DCC-REGARCH and CRREGARCH models demonstrate the best performance in out-of-sample forecasts of the variance-covariance matrix, generating the lowest variance and turnover for global minimum-variance (GMV) portfolios in the majority of cases