Media Briefings

TEXT ANALYSIS LAYS BARE ANIMAL SPIRITS

  • Published Date: April 2017

Broker reports and financial news contain useful information in shifting sentiment

Text analysis of broker and news reports can predict swings in ‘animal spirits’ in financial markets, according to new research by Professor David Tuckett and colleagues, to be presented at the Royal Economic Society's annual conference at the University of Bristol in April 2017. Their study provides valuable insights into the role played by narratives and emotions in driving developments in the financial system.

Using sets of predefined word lists to analyse sentiment in internal Bank of England daily commentary on market news and events, broker research reports and Reuters news articles in the UK, the authors find that all three sources showed high levels of excitement relative to anxiety leading up to the financial crash of 2007. The sentiments in all these sources were highly correlated with each other, and could also be used to predict swings in existing indicators of market sentiment, such as the ViX and MSCI.

The authors, part of a collaboration between the Bank of England and the University College London Centre for Decision-Making Uncertainty, conclude: ‘Measures of narrative consensus and sentiment shift have the potential to become an important part of the overall toolkit available to central banks to warn of impending financial system distress and potential systemic risk.’

More…

The years preceding the global financial crisis were characterised by widespread reports of exuberance in the financial sector. As has often occurred throughout history, consensus apparently emerged over a new paradigm, under which the greater efficiency of markets and distribution of risk around the system was thought to justify new approaches to capital ratios and very strong positive sentiment.

But when the crash came during 2007 and 2008, it became very obvious that sentiment reversed rapidly with fear and anxiety pervading the financial and economic system.

To date, emotion and narrative have played a marginal role in economic analysis. This study, the outcome of a collaboration between colleagues at the Bank of England and the University College London Centre for Decision-Making Uncertainty, describes a novel automated set of analytic tools, applied to large amounts of financial market text-based data, to assess if there is any evidence as to how narratives and emotions play a role in driving developments in the financial system.

The authors describe measures of narrative sentiment that use pre-defined word lists representing two specific emotion groups. The emotion groups and the words representing them have been developed through the lens of a social-psychological theory of ‘conviction narratives’ (CNT) – one way of empirically formulating the idea of ‘animal spirits’ described by Keynes.

CNT focuses attention on the specific emotional elements of narratives that evoke attraction or approach to an object of investment (broadly conceived), versus emotions that evoke repulsion or avoidance of that object. This emphasis on approach and avoidance specifies often-vague discussions of positive/negative sentiment. In more ordinary language, the focus is on excitement about the potential gains from an action relative to anxiety about the potential losses.

With this in mind, the researchers analyse three unstructured text-based data sources of potential interest: internal Bank of England daily commentary on market news and events; broker research reports; and Reuters’ news articles in the UK.

They report first, that changes in the emotional content in market narratives are highly correlated across data sources (Figure 1).

The changes through time show clearly the formation (and subsequent collapse) of very high levels of sentiment – high excitement relative to anxiety – leading up to the global financial crisis.

Second, they find that the shifts have predictive power for other commonly used measures of sentiment and volatility the ViX (Figure 2) and the Michigan Consumer Sentiment Index (MSCI) (Figure 3). In the MCSI case, the new series outperforms consensus economic predictions.

Third, they find that a new related methodology, designed to capture changes in the diversity of market narratives (topic entropy), captures the relative homogenisation of beliefs around a new financial paradigm prior to the crisis (Figure 4).

The study concludes that there is much more scope to attend to narrative and emotion, which could add weight to economic analysis. The measures of narrative consensus and sentiment shift have the potential to become an important part of the overall toolkit available to central banks to warn of impending financial system distress and potential systemic risk.

ENDS


News and narratives in financial systems:
Exploiting big data for systemic risk assessment

Rickard Nyman, David Gregory, Sujit Kapadia, Paul Ormerod, David Tuckett and Robert Smith

The views expressed in this study are those of the authors and should not be thought to represent those of the Bank of England, Monetary Policy Committee members or Financial Policy Committee members

Contact:
David Tuckett
d.tuckett@ucl.ac.uk


Figure 1 : Relative sentiment of MCDAILY (black), RTRS (green) and BROKER (red). The y-axis displays the normalised values with 0 mean and standard deviation 1

Figure 2 Relative sentiment of MCDAILY, inverted for convenience, (black) compared to the VIX (yellow). The y-axis displays the normalized values with 0 mean and standard deviation 1

Figure 3 Change in MCI compared to forecasts of the change made using BROKER and consensus economist forecasts


Figure 4 Relative sentiment (black) and entropy (yellow) in Reuters’ London news. The y-axis displays the normalized values with 0 mean and standard deviation 1