CMStatistics 2018: Start Registration
View Submission - CFE
Title: Application of multivariate Hawkes graphs to uncover Granger causality of financial news Authors:  Anastasija Tetereva - University St Gallen (Switzerland) [presenting]
Abstract: A considerable amount of current research in financial business addresses the influence of news and social media on stock returns and volatility. Although news data are used in many applications, the mutual relationship among public announcements remains unclear. Moreover, the majority of studies are conducted using aggregated data, which are less effective in detecting causal links than observations of higher frequency. Evidence of self and mutual triggering of news announcements in the financial sector is provided. It is proposed that the news arrival times be modelled as a multivariate Hawkes process to test the Granger causality of company-specific news and to detect the most influential companies. Based on this information, a novel method of constructing a composite news intensity index (NII) is presented. The NII demonstrates the ability to timeously describe the uncertainty in financial markets. The proposed measure Granger causes VIX at 6-month lag and can therefore be used to diagnose the health of a financial system.