Abstract We examine the long-term relationship between signals derived from nine years of unstructured social media microblog text data and financial market developments in five major economic regions. Employing statistical language modeling techniques we construct directional sentiment metrics and link these to aggregate stock index returns. To address the noise in finance-related Twitter messages we identify expert users whose tweets predominantly focus on finance topics. We document that expert users are the main drivers behind an interdependence between Twitter sentiment and financial markets. The direct prediction value of expert sentiment metrics for stock index returns, however, is found to be elusive and short-lived. Yet, we detect significant predictive gains over benchmark models in times of negative market returns. In consequence, the relation between expert sentiment metrics and stock indices is sufficient to devise hypothetically profitable cross-sectional as well as time series momentum investment strategies for futures based on Twitter signals that survive basic transaction cost assumptions. In this context, our results show that expert sentiment signals can yield higher risk-adjusted returns than classical price-based signals.
Paper
Full text
Buzzwords build momentum: Global financial Twitter sentiment and the aggregate stock market
Semantic Scholar · Economics · 2019
Abstract
Abstract We examine the long-term relationship between signals derived from nine years of unstructured social media microblog text data and financial market developments in five major economic regions. Employing statistical language modeling techniques we construct directional sentiment metrics and link these to aggregate stock index returns. To address the noise in finance-related Twitter messages we identify expert users whose tweets predominantly focus on finance topics. We document that expert users are the main drivers behind an interdependence between Twitter sentiment and financial markets. The direct prediction value of expert sentiment metrics for stock index returns, however, is found to be elusive and short-lived. Yet, we detect significant predictive gains over benchmark models in times of negative market returns. In consequence, the relation between expert sentiment metrics and stock indices is sufficient to devise hypothetically profitable cross-sectional as well as time series momentum investment strategies for futures based on Twitter signals that survive basic transaction cost assumptions. In this context, our results show that expert sentiment signals can yield higher risk-adjusted returns than classical price-based signals.