Sentiment in Central Banks’ Financial Stability Reports*

Abstract

We use the text of financial stability reports (FSRs) published by central banks to analyze the relation between the sentiment they convey and the financial cycle. We construct a dictionary tailored specifically to a financial stability context, which classifies words as positive or negative based on the sentiment they convey in FSRs. With this dictionary, we construct financial stability sentiment (FSS) indexes for thirty countries between 2005 and 2017. We find that central banks’ financial stability communications are mostly driven by developments in the banking sector. Moreover, the sentiment captured by the FSS index explains movements in financial cycle indicators related to credit, asset prices, systemic risk, and monetary policy rates. Finally, our results show that the sentiment in central banks’ communications is a useful predictor of banking crises—a one percentage point increase in FSS is followed by a twenty-nine percentage point increase in the probability of a crisis.

Financial Stability and the Fed: Evidence from Congressional Hearings

“…To build indicators for financial stability sentiment, we use the dictionary proposed by Correa et al (2017). Their dictionary consists of 391 words, of which 96 are deemed to have a positive and 295 a negative connotation with respect to financial stability.…”

“…We use the Correa et al (2017) dictionary as it is specifically tailored to the context of financial stability. Using alternatives such as Harvard-IV might be problematic, as words in that lists could have very different connotations in the context of financial stability.…”

“…We considered using the word list by Peek et al (2016). However, that list is much shorter than that from Correa et al (2017), and applying it to the Humphrey-Hawkins testimonies generated too little variation to be included in the Taylor rule models. It is important to note that we go beyond a plain 'bag of words' approach and moderate the influence of contextual language by accounting for simple negation patterns of positive terms, as in Loughran and McDonald (2011) and Correa et al (2017).…”

“…paper relies on a broader range of methods for sentiment analysis and topic modeling. In particular, we will rely on a recent sentiment dictionary by Correa et al (2017) that is especially suited for analyses in the context of financial stability.…”

“…Their sample, however, only tracks back to 2000, and for the Fed the relatively short monetary policy press releases are studied. Our paper is also related to work by Correa et al (2017) the monetary policy stance with members of Congress. There are two key dimensions to our findings.…”