Valerio Potì (University College Dublin)

COVID Narrative Risk: A Computational Linguistic Approach to the Econometric Identification of Narrative Risk During the COVID-19 Pandemic

Abstract

In this paper, we study the role in financial markets of narratives related to the ongoing COVID-19 pandemic. The pandemic represents a natural setting for the development of narratives that may effects or be affected by financial markets. We thus treat the pandemic as a natural experiment on the relation between prevailing narratives and financial markets. We adopt the SIR model and natural language processing on financial newspaper news and Twitter tweets that deal at the same time with financial market topics and COVID-19 to study the dynamics and determinants of coronavirus narrative epidemics. Our aim is to establish whether there an “infodemic” develops, and whether the prevailing narrative, whether resemnbling an infodemic or otherwise, drives or is driven by financial markets developments, controlling for developments regarding the COVID-19 pandemic. We find associations between narratives about the epidemic, stock market dynamics (both regarding returns and volatility) and government responses to COVID-19. We also find that the narrative spread does resemble an infodemic, since it is well described by the SIR epidemic model. Our estimates of the shape of the narrative infodemic curves in different countries depend on whether the COVID-19 outbreak occurred early and on its severity. Negative tones and tones communicating uncertainty are prevalent in the growing stage of the infodemic. We find preliminary evidence of a causality relation between the negative and uncertainty tones of coronavirus tweets and both market return and volatility change.