Predicting the trading behavior of socially connected investors

We find that investors’ future trading decisions are driven by the patterns of their social neighborhood
and the trading activity therein. Moreover, we provide evidence that investors weigh their social connections
differently in terms of information transfer. Methodologically, we tackle the complex, cyclical patterns of
investor social networks by graph neural networks, which allow us to propose a sophisticated way to predict
the behavior of investors with data on their social connections. Our analysis is based on the unique data on
observed social links through director (insider) positions on the same companies as well as links to family
members, together with full investor-level market-wise transaction data.

The data is available online at https://doi.org/10.6084/m9.figshare.20310240.v1