Portfolio risk and performance evaluation using a combination of GARCH and stochastic volatility models with constant and dynamic correlation structures

Sammanfattning

CCC and DCC are used to estimate the conditional correlation between a non-standard mix of GARCH, GARCH-M and a stochastic volatility model, all assuming Gaussian innovations. After demonstrating the theoretical soundness of this combination, a portfolio of ten assets with diverse return dynamics is constructed. Portfolio performance is evaluated using Value at risk, the Sharpe ratio, maximum drawdown and a diversification measure, then compared to the benchmark strategies - Equal Weights and Buy and Hold. The proposed models outperform the benchmarks across the selected performance metrics during validation, although, as expected, forecast accuracy declines with increasing distance into the out-of-sample period and fails to adapt to overall market trend shifts.