CMStatistics 2016: Start Registration
View Submission - CMStatistics
B0596
Title: Causality networks: Estimation and combination Authors:  Giovanni Bonaccolto - University of Padova (Italy)
Massimiliano Caporin - University of Padova (Italy)
Roberto Panzica - Goethe University House of finance (Italy) [presenting]
Abstract: Granger causality test is commonly used in finance for estimating the causality relationship-based network among financial assets. Nevertheless, for investigating the comovement of assets, causality in the sense of Granger might not be the most appropriate tool as it focuses on mean relationships. The econometrics literature includes generalizations of the Granger causality concept for investigating the causality across quantiles of two distributions. However, this approach has never been considered to estimate a financial network. Moreover, the relationship in term of contribution of these two forms of causality within a more general equilibrium model is unclear. Previous works are extended by introducing a combination various causality networks within a standard multifactor model. The various networks will be considered as part of a more general multiplex network. Our advantage is to provide a measure of the relevance of each layer within the assets interconnectedness. In the empirical analysis we estimate the layers of the multiplex network by using the standard Granger causality and quantile Granger causality tests, using both parametric and non-parametric methods. We point at showing both the relevance of quantile-based causality in network estimation and the role played by alternative networks in a multifactor model accounting for asset interconnections.