Mining the Value of Network Structure on Stock Performance

We propose a novel Fintech application which models investors' co-attention to diversified stocks as networks by utilizing visitors' online correlated searches for a set of stocks. We adopt a sliding-window procedure to capture the dynamics of the co-search networks and develop panel-data models to examine the impact of the network structure on individual stock performance. By arguing the heterogeneous information flows transmitting through the network matter for an ego's future performance, we explicitly divide the alters of an ego stock into two distinct groups based on their historical return trends. Our empirical results suggest an increase of network centrality and closure among historical winners can positively contribute to an ego's future return whereas that among historical losers can exert significant and negative impact. We also find a higher level of centrality is associated with a lower level of future volatility.