Detecting Risks in the Banking System by Sentiment Analysis

Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 591–600, Lisbon, Portugal, 17-21 September 2015. c⃝2015 Association for Computational Linguistics. Detecting Risks in the Banking System by Sentiment Analysis Clemens Nopp TU Wien clemens.nopp@alumni.tuwien.ac.at Allan Hanbury Institute of Software Technology and Interactive Systems TU Wien hanbury@ifs.tuwien.ac.at Abstract In November 2014, the European Cen- tral Bank (ECB) started to directly su- pervise the largest banks in the Euro- zone via the Single Supervisory Mecha- nism (SSM). While supervisory risk as- sessments are usually based on quantita- tive data and surveys, this work explores whether sentiment analysis is capable of measuring a bank’s attitude and opinions towards risk by analyzing text data. For realizing this study, a collection consisting of more than 500 CEO letters and outlook sections extracted from bank annual re- ports is built up. Based on these data, two distinct experiments are conducted. The evaluations find promising opportunities, but also limitations for risk sentiment anal- ysis in banking supervision. At the level of individual banks, predictions are relatively inaccurate. In contrast, the analysis of ag- gregated figures revealed strong and sig- nificant correlations between uncertainty or negativity in textual disclosures and the quantitative risk indicator’s future evolu- tion. Risk sentiment analysis should there- fore rather be used for macroprudential analyses than for assessments of individ- ual banks. 1 Introduction From 2007 on, a global crisis struck the financial markets and led to a severe slow-down of the real economy. It was triggered by the collapsing US subprime mortgage sector, where loans had been issued to borrowers with poor credit ratings. Due to the tight interconnectedness of the financial sys- tem, problems quickly propagated in the global banking system. Governments had to bail out im- portant institutions like Northern Rock, but such solutions could not be provided for every troubled bank. In September 2008, the large investment bank Lehman Brothers had to file bankruptcy. In the aftermath of this event, further banks had to be rescued in order to stabilize the financial system. This deep financial crisis highlighted the necessity of better financial regulation as well as more ef- fective financial supervision in the future (Hodson and Quaglia, 2009). As a reaction to the crisis and its severe eco- nomic consequences, EU institutions decided to build up an European Banking Union (EBU). The EBU consists of three pillars, one of them being a new system of financial supervision, the Sin- gle Supervisory Mechanism (SSM). Its goal is to “promote long-term safety and soundness of credit institutions and the stability of the financial sys- tem within the Union and each Member State [...]” (Council of the EU, 2013, p. 72). For supervising over 120 of the largest banks in the Eurozone, the SSM utilizes a range of in- formation sources in order to detect vulnerabilities and risks. The sources include mainly backward- looking quantitative Key Risk Indicators (KRIs), which are complemented with surveys in order to include forward-looking information as well (Eu- ropean Banking Authority, 2014). However, …

Content truncated to 500 words