The economic value of flexible dynamic correlation models

This article assesses the ability of flexible dynamic correlation specifications to improve asset allocation decisions. To that end, we use the recently proposed Rotated Dynamic Conditional Correlation (RDCC) model that enables the estimation of models with high degree of parameterization and large number of assets. We also extend the RDCC model to incorporate 'rotated' realized correlation measures which exploit the information content of intra-day data. The empirical evidence, based on ten US equities and three years of out-of-sample forecasting (2007-2009), support the use of flexible diagonal RDCC specifications for portfolio management purposes. However, simpler scalar specifications enhanced with realized correlation measures can produce superior or in some cases similar results. Overall, our findings give evidence in favor of inter-daily flexible RDCC models for asset allocation purposes when the computation of realized correlation measures is practically unfeasible.