Sequential conditional correlations: Inference and evaluation

This paper presents a new approach to the modeling of conditional corre- lation matrices within the multivariate GARCH framework. The procedure, which consists in breaking the matrix into the product of a sequence of matri- ces with desirable characteristics, in effect converts a highly dimensional and intractable optimization problem into a series of simple and feasible estima- tions. This in turn allows for richer parameterizations and complex functional forms for the single components. An empirical application involving the condi- tional second moments of 69 selected stocks from the NASDAQ100 shows how the new procedure results in strikingly accurate measures of the conditional correlations.