An Analysis of the Conditional Dependence of the Stock Market Using the Network Model

Abstract

This study used the factor decomposition model combined with the long-run variance decomposition network (LVDN) model to analyze interdependencies in a network. It exercised the two-step procedure as follows. First, the stock index returns were decomposed into the components of the common and idiosyncratic returns. The factor decomposition procedure was again applied to the estimated residual variances from the first step, where the variances were decomposed into the common and idiosyncratic volatility components.

Second, the LVDN was obtained from the vector autoregressive (VAR) representation of the idiosyncratic component of volatility by choosing a Choleski decomposition of the idiosyncratic shocks' covariance matrix. The LVDN estimation result showed that interdependencies between a pair of stock markets in the network increased significantly during the financial crisis period of 2007-2009. From the analysis, this study concluded that the stock markets became more susceptible to the idiosyncratic component of volatility shocks in other stock markets in the network. The value of total interdependence verified this result.

Keywords: network model, generalized dynamic factor model, LVDN, systemic risk, sparse vector autoregression

JEL Classification: C13, C22