Correlation Persistence in Financial Markets: A Network Theory Approach

I propose a persuasion model to show the correlation of stock price can substantially vary over time in a general network economy in which investors update their own information following by the Bayesian rules and evolving in the social network effect. To describe the mutual interconnections and model the financial network, an empirical strategy, Granger-Causality network, is proposed based on the multivariate vector autoregression (MVAR) models. The empirical study is conducted by examining the unique dataset that consists of all brokers' daily trading information for a decade in Taiwan. The empirical results show the density of brokers' financial network can be positively correlated to the market realized volatility. Moreover, the density can also be viewed as an indicator to the systemic risk of market. The empirical evidence shed new light not only on providing an explanation to the phenomenon of correlation persistence but also measuring the systemic risk of the market from the perspective of network.