Long and short run dynamics in realized covariance matrices: a robust MIDAS approach

A recent stream of the econometric literature is devoted to modelize unobservable short and long–run components in volatility and time–varying correlations of financial assets. In such models two typical problems are the sensitivity of the estimation results to the order in which the assets enter the model and the trade-off between the flexibility of the model and its parsimony. We propose a new class of additive component models belonging to the MIDAS family, that overcomes some drawback related to the use of the Cholesky decomposition to the covariance matrix, avoiding the effect of the order of the series on the estimation process. Moreover, we deal with the curse of dimensionality problem by adopting the Hadamard exponential function which allows asset-pair-specific and time-varying parameters.We verify the advantage of the proposed models by comparing them with some benchmarks, both in terms of in–sample and out–of–sample performance, through some statistical and economic loss functions.