Futures hedging with Markov switching vector error correction FIEGARCH and FIAPARCH

Markov switching vector error correction asymmetric long memory volatility models with fat tailed innovations are proposed. Bivariate two state versions of the models are applied to a futures hedge of the S&P500. Regime switches occur between high and low cost of carry states via changes in the error correction term or basis. Regime identification is therefore dominated by switches in the mean, not volatility. Relative to a number of alternatives, the proposed models provide superior out of sample forecasts of the covariance matrix particularly for horizons greater than 10 days ahead. When hedging, Markov switching with long memory improves the tail risk of hedged returns beyond 10 day horizons, however there is mixed support for models with volatility asymmetries. These findings have important implications for the development of multivariate models and other applications including portfolio management, spread option pricing and arbitrage.