"Explicit Modeling of Time Variation of Diebold and Yilmaz Connectedness Measures"

Assessment of the Impact of Multivariate Heteroscedasticity on the Dynamics of the Diebold and Yilmaz Measures of Connectedness. The Realized Volatility Case: This paper introduces multivariate heteroscedastic effects in the dynamics of measures of connectedness proposed and developed by Diebold and Yilmaz (Diebold and Yilmaz [8], Diebold and Yilmaz [9], Diebold and Yılmaz [10], Diebold and Yilmaz [11]). The measures are based on forecast error variance decompositions. We consider variability in the covariance matrix of the error vector by using the  Dynamic Conditional Correlation (DCC) model of Engle [12]. We develop this framework both for the Orthogonal Variance Decomposition (OVD) and Generalized Variance Decomposition (GVD) measures of connectedness. We apply this new methodology to measure the variability in the links between a set of weekly time series of realized variances for returns of equity indices. We confirm the presence of heteroscedastic effects and compare our results to a rolling-window methodology.

Generalized Diebold-Yilmaz Connectedness Measures for MS-VARs: The recent financial crisis has shown the world how important is to consider the dynamic nature of the strength of the relationships between financial variables. One way to assess both the magnitude and time variation of such strength is through the Diebold and Yilmaz connectedness framework, which is based on forecast error variance decompositions of VAR systems. Forecast error variance decompositions can be identified either through Cholesky decomposition, by assuming a priorWold (or causal) Order, or via generalized variance decompositions, under some restrictions about the distribution of the system errors. The aim of this paper is to extend the Diebold and Yilmaz framework by relaxing such restrictions and proposing new formulae for the generalized variance decomposition of Markov switching vector autoregressions.