Applications of a Multivariate Hawkes Process to Joint Modeling of Sentiment and Market Return Events

To investigate the complex interactions between market events and investor sentiment, we employ a multivariate Hawkes process to evaluate dynamic effects among four types of distinct events: positive returns, negative returns, positive sentiment and negative sentiment. Using both intraday S&P 500 return data and Thomson Reuters News sentiment data from 2008 to 2014, we find: a) self-excitation is strong for all four types of events at 15 minutes time scale; b) there is a significant mutual-excitation between positive returns and positive sentiment, and negative returns and negative sentiment; c) decay of return events is almost twice as fast as sentiment events, which means market prices move faster than investor sentiment changes; d) positive sentiment shocks tend to generate negative price jumps; and e) the cross-excitation between positive and negative sentiments is stronger than their self-excitation. These findings provide further understanding of investor sentiment and its intricate interactions with market returns