Identification of Information Networks in Stock Markets

Abstract

Existing studies have addressed the significance of social influence and private communication in decision making in stock markets. However, the estimation of investor information networks remains an important and challenging task because the existing network inference methodologies lack the ability to explicitly account for the impact of public information on investor trading decisions. We address this gap by proposing a new framework to estimate private information channels in stock markets. In our approach, the impact of public information on investors' trading events is filtered out from investors' transactions. This allows us to reveal their co-behavior driven by the transfer of private information. Our results show that taking public information into account when inferring investor networks significantly changes their topology and strengthens the relationship between investor's network centrality and returns. Therefore, we believe that our approach leads to a more precise representation of the information network. Furthermore, we find that the association between centrality and returns becomes stronger and both statistically and economically more significant. Moreover, investigating the properties of information networks, we observe that the physical distances between connected investors begin to shrink when network links are validated using harsh thresholds.