Look inside. Predicting stock prices by analysing an enterprise intranet social network and using word co-occurrence networks

TL;DR

This paper introduces a novel approach to stock price prediction by analyzing employee communication on an enterprise intranet, using social network metrics and word co-occurrence networks to identify significant predictors.

Contribution

It presents new metrics derived from intranet communication and word networks that enhance stock price forecasting models and extends the application of word co-occurrence networks.

Findings

Lower sentiment predicts higher stock prices.

Higher betweenness centrality of the company brand correlates with stock increases.

Denser word co-occurrence networks are associated with rising stock prices.

Abstract

This study looks into employees' communication, offering novel metrics which can help to predict a company's stock price. We studied the intranet forum of a large Italian company, exploring the interactions and the use of language of about 8,000 employees. We built a network linking words included in the general discourse. In this network, we focused on the position of the node representing the company brand. We found that a lower sentiment, a higher betweenness centrality of the company brand, a denser word co-occurrence network and more equally distributed centrality scores of employees (lower group betweenness centrality) are all significant predictors of higher stock prices. Our findings offers new metrics that can be helpful for scholars, company managers and professional investors and could be integrated into existing forecasting models to improve their accuracy. Lastly, we contribute to the research on word co-occurrence networks by extending their field of application.