553 - SenSR: A sentiment-based systemic risk indicator

DNB Working Papers
Date 10 April 2017
The media influence our perception of reality and, since we act on those perceptions, reality is in turn affected by the media. News is a rich source of information, but, in addition, the sentiment (i.e., the tone of financial news) tells us how others perceive the financial system and how that perception changes.
In this paper we propose a new indicator of the systemic risk in the global financial system. We call it SenSR : Sentiment-based Systemic Risk indicator. This measure is constructed by dynamically aggregating the sentiment in news about systemically important financial institutions (SIFIs).
We test the SenSR for its ability to indicate or even forecast systemic stress in the financial system. We compare its performance to other well-known systemic risk indicators, as well as with macroeconomic fundamentals. We find that SenSR anticipates other systemic risk measures such as SRISK or VIX in signaling stressed times. In particular, it leads other systemic risk measures and macroeconomic indicators by as long as 12 weeks.
 
Keywords: systemic risk, sentiment analysis, Granger causality.
JEL classifications: G01, G18, C58, G17.