No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
A New Approach to Early Warning Systems for Small European Banks
... This article is thus positioned as a comparative empirical paper applying a selected range of descriptive variables and multivariate classification methods to provide as many reliable models as possible to predict the failure of European Union (EU)-27 banks. Similar empirical research has been performed to this point within the frameworks of several publications (see, inter alia, Poghosy and Č ihák, 2009;Betz et al., 2013;Gerhardt and Vennet, 2017;Bräuning et al., 2019), although not necessarily using the same variables and methods as those in this article. ...
... It is further acknowledged that empirical results justifying the superiority of ensemble classifiers and neural networks (NNs) over conventional techniques in bank failure prediction are already available in the literature (see, inter alia, Tanaka et al., 2016;Le, Viviani, 2018;Bräuning et al., 2019;Appiahene et al., 2020;Shrivastava et al., 2020). It can be simultaneously observed that a relatively large share of empirical results in this field originates from the United States (US) (see, inter alia, Gogas et al., 2018;Jing and Fang, 2018;Manthoulis et al., 2020;Petropoulos et al., 2020). ...
... Essential technical aspects can further improve the predictive accuracy of XGB that can outperform gradient boosting and adaboost techniques (Pham and Ho, 2021). Bräuning et al. (2019) developed an early warning system for 3000 small European banks between 2014 and 2016. The C5.0 DT model achieved a 92% AUROC score from the testing sample data, thereby exceeding the 90% performance level of the benchmark logit model. ...
This article provides evidence that machine learning methods are suitable for reliably predicting the failure risk of European Union-27 banks from the experiences of the past decade. It demonstrates that earnings, capital adequacy, and management capability are the strongest predictors of bank failure. Critical and relevant field research is presented in the context of economic uncertainties arising from the COVID-19 pandemic. The results suggest that the developed models possess high predictive power, with the C5.0 decision tree model providing the best performance. The findings have policy implications for bank supervisory authorities, bank executives, risk management professionals, and policymakers working in finance. The models can be used to recognize bank weaknesses in time to take appropriate mitigating actions.
... In detail, Halteh et al. [57] use the Altman Z-score and the standardized profits to identify distressed banks, while Zaki et al. [58] computed percentage changes in annual equity, annual return on average equity (ROAE), and annual net interest margin (NIM) for each year from 2000 to 2008 for UAE banks and classified them as "distressed" if they experienced an annual variation in such index below the median. 4. Expansive definition of distress: Some studies (e.g., Ref. [29,[59][60][61][62]) opt for broader definitions encompassing various distress events beyond mere bank failures or receiverships This approach implies the inclusion of additional events, such as state support or government intervention, removal of management body, breach of minimum capital requirements, merger (or distressed merger) and acquisition, failure to meet obligations. ...
... Ref. [74,76,81]), and decision trees (e.g. Ref. [61,76,80]) have evolved as systems that 'learn' from previous experience in order to improve their problem-solving performance and to deal with non-linearity. The last decade has also seen the diffusion of Ensemble and Hybrid models, which have progressively become more attractive. ...
This paper presents a literature review of recent empirical contributions on bank default prediction. The topic has
always been important in the banking and finance literature, but it gained increasing interest especially after the
2008–2009 global financial crisis. The significant consequences of bankruptcy cases have highlighted the need
for managers and regulators to develop and adopt appropriate early warning systems.
Previous studies are analysed in this review according to three different aspects: definition of default and
financial distress, application of statistical and intelligent techniques, and the selection of variables. The review
also proposes some possible upgrades to promote future research on the topic, that is, pointing out the potential
role of non-financial information and ESG (Environmental, Social and Governance) variables as a determinant in
the default prediction of banking institutions. While the predictive accuracy of macroeconomic information has
been tested in several works, there is only limited evidence of the influence of non-financial performance on a
bank’s probability of default
... Literature around early warning indicators peaked after the 2007-2008 Global Financial Crisis. There was a similar increase in research interest in this area following the financial crisis in Asia (Bräuning et al., 2019). For instance, Surjaningsih, Yumanita and Deriantino (2014) developed EWIs for banking liquidity risk and could detect liquidity imbalances one year before October 2008, reporting 67% predictive power. ...
... Models were listed under six different types and, using machine learning, family artificial neural networks and genetic algorithms were reported. In a more recent study employing a machine learning model to predict bank distress, Bräuning et al. (2019) employed a decision tree model with the Quinlan C5.0 algorithm to identify individual bank distress for small European banks. The study reported low Type I and Type II errors, with the decision tree model outperforming the logit model. ...
This research contributes to bank liquidity risk management by employing supervised machine learning models to provide banks with early warnings of liquidity stress using market-based indicators. Identifying increasing levels of stress as early as possible provides management with a crucial window of time in which to assess and develop a potential response. This study uses publicly available data from 2007 to 2021, covering two severe stress periods: the 2007–2008 global financial crisis and the COVID-19 crisis. The current version of the developed model then applies backtesting using the data from the COVID-19 crisis. The findings of this study show that the ensemble model with the RUSBoost algorithm predicts “red” and “amber” days with a success rate 21% greater than the average of other machine learning models; thus, it can greatly contribute to bank risk management.
... After the 2007-2008 Global Financial Crisis (GFC), literature around early warning indicators peaked and further developed. There was a similar increase in research interest in this area following the financial crisis in Asia (Bräuning et al, 2019). For instance, Surjaningsih, Yumanita, Deriantino (2014) developed EWIs for banking liquidity risk and could detect liquidity imbalances one year before October 2008, with 67% predictive power reported. ...
... In a more recent study employing machine learning model to predict bank distress, Bräuning et al (2019) employed a decision tree model with the Quinlan C5.0 algorithm to identify individual bank distress for small European banks. The study reported low type I and type II errors, with the decision tree model outperforming the logit model (Bräuning, Malikkidou, Scalone, & Scricco, 2019). ...
This work contributes the bank liquidity management by applying supervised machine learning models to provide banks with early warnings of liquidity stress using market base indicators. Identifying increasing levels of stress as early as possible provides management with a crucial window of time in which to assess and develop a potential response. In this study, market stress was transformed into a classification problem. Publicly available data from 2007 to 2021 was used to train the machine learning model; this period covers two severe stress periods, namely the 2007-2008 Global Financial Crisis and the Covid-19 crisis. The St. Louis Fed Financial Stress Index was used to define the level of stress in the market. Each day trading was assigned a red-amber-green (RAG) status to identify the risk level for that day. Machine learning models were then applied to predict the RAG status of each day. Due to a significantly limited number of “red” status days, modelling became more challenging. An ensemble model with a random under-sampling boosting algorithm (RUSBoost) was used to improve predictions of imbalanced data. Initial results show the machine learning model used in this study can predict 80% of “red” risk days. The current version of the developed model has been back-tested using data from the Covid-19 crisis and it was able to trigger “amber” and “red” status as crisis unfolding. These findings show that the ensemble model with the RUSBoost algorithm predicts “red” and “amber” days on average 36% more than other machine learning models and can contribute to bank risk management.
... Empirical research has proposed different approaches to early warning systems (EWS) for banks as lenders, but the studies concentrate on developed economies. In a similar study by Bräuning et al. (2020), the authors argue that the existing literature on EWS is not well-suited for small banks, as it is based on data from large banks and does not take into account the specific characteristics of small banks. Based on a novel decisiontree approach, this study identifies variables from the qualitative banking sector and macroeconomic indicators, to identify small banks that are at risk of financial distress. ...
... The evidence from the majority of the recent works show that the machine learning models mostly outperform benchmark logistic regression methods in out of sample predictions and forecasting (e.g. Bluwstein, Buckmann, Joseph, Kang, Kapadia, & Simsek, 2020;Bräuning, Malikkidou, Scricco, & Scalone, 2019;Casabianca, Catalano, Forni, Giarda, & Passeri, 2016;Holopainen & Sarlin, 2016;Jarmulska, 2020;Peltonen, Sarlin, & Piloiu, 2016;Tölö, 2019). However, Beutel, List, & von Schweinitz (2019) investigated financial crises over the past 45 years in 15 advanced countries (including the US and Japan) and stated that machine learning methods are not superior to conventional models in predicting financial crises. ...
Aim/purpose – This paper investigates the accuracy of leading indicators in the case of the 2001 sovereign default crisis and the 2018 currency turmoil in Argentina.
Design/methodology/approach – In this paper, we conducted early warning signals analysis based on a-priori selected variables. For each of the macroeconomic variables, we computed yearly changes and selected the threshold to minimise the noise-to-signal ratio, i.e. the ratio of percentage of false signals in ‘normal’ times to percentage of good signals in a two-year period preceding each of the crises.
Findings – The predictive power of indicators differs significantly in various crisis episodes. For the 2001 crisis, the decline in value of bank deposits was the best leading indicator based on the noise-to-signal ratio. For the 2018 currency crisis, the lowest noise-to-signal ratio was observed for the lending-deposit rate ratio.
Research implications/limitations – The survey is limited mostly by the data availability and their quality.
Originality/value/contribution – This paper gives a complex review of the major early warning indicators in the context of the most recent history of Argentina’s economy. It applies a set of classical leading indicators to two modern cases of financial crises. The paper proposes an original ‘knocking the window’ approach to the presentation of traditional warning concepts in the context of current economic events.
This paper is devoted to modeling the probability of default of Russian banks in 2015–2020. There are relatively few studies on defaults of Russian banks after 2015, and our work intends to partly fill this gap. The purpose of this research is to determine the main variables which significantly impact the risk of default of Russian banks. The work seeks to identify additional factors associated with an increased risk of bank defaults during a relatively stable period of development of the Russian economy (2015–2020) without external shocks, such as COVID‑19 or international sanctions. We apply an integrated approach to modeling the risk of bank defaults. Empirical methodology is represented by logit and probit models, as well as Cox regression. The set of potential predictors for bank defaults include the variables, characterizing various aspects of credit institutions functioning (in accordance with the CAMELS system), as well as macroeconomic variables. The most significant predictors of default turn out to be the capital adequacy ratio N1, bank net assets, the ratio of total loans to assets and the size of secured loan portfolio. In general, the results we obtain are consistent with the CAMELS system of indicators assessing the sustainability of commercial banks, while the impact of macroeconomic indicators tends to be insignificant. The results of the study could be of interest to the regulator both for the purposes of ongoing monitoring of financial stability as well as for default risk prevention; to credit institutions which elaborate internal systems for monitoring their financial soundness; and to financial market participants to select the most stable companies in terms of investment and allocation of funds. Further directions of research are related to the inclusion of a crisis period into the analysis and comparing the set of significant predictors for bank defaults during a crisis and a stable period of economic development, as well as the use of alternative methods, in particular, machine learning algorithms.
This paper aims to determine the ‘new normal’ for banking stability in terms of capital adequacy, reviewing the incidence of banking stress episodes by lagged solvency ratios, based on the experience at the European level after the global financial crisis. We provide rating ladders for both risk‐weighted solvency ratios and a simple gearing (leverage) ratio for time horizons of up to 3 years using well‐known credit risk scoring procedures. Our findings empirically confirm that the recent dual metric structure of the capital adequacy framework is conducive to enhancing the accuracy of banking stability assessment. Specifically, our empirical analysis suggests that both tier 1 capital ratio and leverage ratio generally remain statistically significant in multivariate combinations for crisis probability measurement purposes. Robustness checks with well‐established macrofinancial indicators as control variables suggest that this tandem is hardly replaceable in multivariate early warning systems by combinations of macroimbalance and financial soundness indicators traditionally employed as leading factors of banking crises. Moreover, the pandemic period provides meaningful evidence that robust capital positions, in line with our estimate, have so far been ‘part of the solution’ for dealing with systemic events.
The classification of assets based on their liquidity behaviour under stress is a crucial element of bank liquidity stress testing. It is also important to define how financial institutions should fund these assets within the current business model whilst avoiding excessive liquidity risk. This study aims to revisit the liquidity coverage ratio (LCR) assumptions for common equity shares using new data attributes and supervised machine learning models. This research contributes to the literature by providing fresh insight into which characteristics impact share behaviour under liquidity stress. Empirical results suggest sector, share beta, industry, and market capitalisation of the share are contributing factors which help predict shares’ liquidity behaviour under stress. This study also finds that
the financial, basic materials and energy sectors are more volatile and less liquid under market stress; shares with lower beta show more liquid characteristics, and higher market cap stocks show more liquid behaviour.
The interest in banks' bankruptcy prediction has rapidly increased especially after the 2008-2009 global financial crisis. The relevant consequences of bankruptcy cases have indeed highlighted the necessity for managers and regulators to develop and adopt appropriate early warning systems. The purpose of this paper is therefore to conduct a literature review of recent empirical contributions on bank's default prediction by analysing three underlying aspects: definition of default and financial distress, application of statistical and intelligent techniques, variables selection. The review also proposes some possible upgrades to promote future research on the topic, i.e. pointing out the potential role of non-financial information as good default predictors.
ResearchGate has not been able to resolve any references for this publication.