Analysing financial stability reports as crisis predictors with the use of text-mining

Central bank analyses and external communication play an important role in maintaining its credibility, as well as effectiveness of monetary and macroprudential policies. A financial stability report is one of the main channels of communication between central banks and the financial market. The aim of the study is to verify whether the linguistic content of those reports contains early warning signals of an upcoming financial crisis. We carefully select 848 words related to Early Warning Indicators and examine whether their appearance in the 604 financial stability reports published by 18 central banks could have indicated an impending global financial crisis (2007+). We use the novel approach of joint application of text-mining and the concept of receiver operating characteristic curve to compare the frequency of selected words before and after the global financial crisis. According to the results, the linguistic content of financial stability reports does not emit any early warning signals (except for the single case of the Central Bank of Iceland). On the one hand, this may indicate potential weaknesses in the quality of analyses in those reports, but on the other hand, it may show a central bank's deliberate avoidance of actions prompting negative effects of a ‘self-fulfilling prophecy’.