New study indicates that some investors may take advantage of their social connections in stock trading based on their joint board memberships
The study Predicting the Trading Behavior of Socially Connected Investors: Graph Neural Network Approach with Implications to Market Surveillance has developed a tool for market regulators to identify investors who may be using inside information spread through social connections in their stock trading. The study has been published in the scientific journal Expert Systems with Applications.
The insider network consists of persons who know each other through board memberships in joint listed companies. Since many investors as board members have positions in several companies, insider contacts form a highly connected network that enables the wide spread of information across company borders.
The research is based on Dr. Kęstutis Baltakys’ project, which has been funded by, among others, the OP Group Research Foundation. The study has been conducted together with Dr. Negar Heidar and Professor Alexandros Iosifidis from Aarhus University, Denmark.
The study has used pseudonymised data, that is, data from which the identification of individuals has been removed.
“We have collected material from data sources that have been public under the Securities Market Act and have been processing and modelling data for this research approach for more than four years. Pseudonymised data is globally very unique. It allows us to determine how equity investors acting as insiders have traded all shares, including those not related to insider regulations,” says Juho Kanniainen, who led the study.
New tool to control the use of insider information spread through relationships
The publication shows that the share trades of investors in the insider network are surprisingly widely predictable with previous stock trading by colleagues they know, even in the case of shares that are not related to insiders. This indicates that insiders may disseminate information or views to colleagues outside the company, who may then transmit it.
“The focus was on investors who are themselves insiders and also connected to other insiders. Our main result is a method based on graph neural networks that allows market regulators to find investors who may be using inside information spread through social connections in their stock trading,” Kanniainen says.
Market regulators and stock exchanges may particularly benefit from the developed system. With the system, it is possible to predict a retail investor’s future trading decisions and use the information obtained to identify so-called suspicious investors.
In general, the application of the system to reduce market abuse can benefit all investors.
“In the future, research on the topic will continue and will focus on how insiders may have shared information related to stock exchange releases on their social network before the release is published. This way it is possible to find out where and how the private information has been spreading in a social network and how it has been utilised,” Kanniainen summarises.