No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Constructing Early Warning Indicators for the Banks Using Machine Learning Models
Abstract
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.
... By employing supervised machine learning models to offer banks with early warnings of liquidity stress using market base indicators, Tarkocin and Donduran (2021) focused on the management of bank liquidity. To train the machine learning model, they used publically accessible data from 2007 to 2021. ...
Early warning indicators (EWIs) of banking crises should ideally be judged on how well they function in relation to the choice issue faced by macroprudential policymakers. However, the effectiveness of EWIs depends upon the strength of the predicting power, stability, and timeliness of the signal. Using a balanced panel of 6 countries’ experience with banking and currency crises in recent times, this paper evaluates the effectiveness of EWIs using Receiver Operating Characteristics. Following the drivers of the banking crisis and currency crisis, the paper evaluates the effectiveness of aggregate credit growth, sectoral deployment of credit along and other macroeconomic indicators generally used as EWI. The paper observes that credit disbursements to non-financial sectors and the central government provides stable signals about systemic risks. Further debt service ratio, interbank rates and total reserves are also found to be useful in predicting these crises. Lastly, the effective EWIs are combined using shrinkage regression methods to evaluate the improvement of signal strength of the combination of EWIs. The predictive power of the combination of EWIs provides better signal strength in predicting the macroprudential crisis.
The aim of the article is to substantiate the functionality and ways to modernize early warning systems in banks as an instrument for anti-crisis management. It is proved that banking crises at the macro and micro levels are identified by: bank runs (banking panic), growth of interest rates on deposits and loans; provision of the State financial assistance to banks to avoid a systemic crisis; reorganization and restructuring in the banking system. It is proposed to consider early warning systems as an aggregate of methods and models that allow to identify potential and real signs of crisis phenomena in banks at the macro and micro levels, and to develop preventive measures of anti-crisis management at the early stages of manifestation. It is noted that the creation of an early warning model is a sequential process that includes three stages: preliminary modeling, modeling, and post-modeling. Such actions are directed towards defining the goals and objectives of the model, evaluating the model, and substantiating the initial representativeness of the model. An approach to systematization of early warning indicators with their division into external and internal ones has been developed. External indicators include: insufficient level of capitalization and profitability, high share of non-performing loans, banking risks, unsatisfactory asset quality, bank runs, imbalance of assets and liabilities in terms of amounts and terms, liquidity problems, and others. Internal indicators are as follows: low level of capitalization, quality of assets of an individual institution, deposit potential, liquidity, profitability, spread, etc. Approaches to improving early warning systems have been developed by modernizing them on the basis of artificial intelligence. An anti-crisis management instrumentarium in banks has been proposed, depending on the type and depth of the crisis. The instruments are systematized into non-market ones, which include emergency measures, restrictions on banking competition and the scope of activity, and market instruments – financial, operational and structural. Prospects for further research are the development of indicators for assessing the depth of the crisis at the macro and micro levels along with determining its level, substantiating strategic and tactical methods of anti-crisis management.
This paper provides a comprehensive state-of-the-art investigation of the recent advances in data science in emerging economic applications. The analysis is performed on the novel data science methods in four individual classes of deep learning models, hybrid deep learning models, hybrid machine learning, and ensemble models. Application domains include a broad and diverse range of economics research from the stock market, marketing, and e-commerce to corporate banking and cryptocurrency. Prisma method, a systematic literature review methodology, is used to ensure the quality of the survey. The findings reveal that the trends follow the advancement of hybrid models, which outperform other learning algorithms. It is further expected that the trends will converge toward the evolution of sophisticated hybrid deep learning models.
This study investigates an early warning indicator for liquidity shortages in the short‐term interbank market. To identify structural breaks and their persistence, an autoregressive two‐state regime switching model is presented. The variability in the LIBOR–OIS spread along with thresholds, which delimit four intensities, reveals regime changes consistent with liquidity crashes. The transition between the states is state dependent, and the posterior estimates for the crisis and noncrisis states are estimated using the Gibbs sampler. We forecast our early warning indicator up to December 2011 and show that the estimates are superior to a random walk with drift. Therefore, the model is an effective early warning indicator of an imminent liquidity shortage impacting the interbank market.
Class imbalance is a problem that is common to many application domains. When examples of one class in a training data set vastly outnumber examples of the other class(es), traditional data mining algorithms tend to create suboptimal classification models. Several techniques have been used to alleviate the problem of class imbalance, including data sampling and boosting. In this paper, we present a new hybrid sampling/boosting algorithm, called RUSBoost, for learning from skewed training data. This algorithm provides a simpler and faster alternative to SMOTEBoost, which is another algorithm that combines boosting and data sampling. This paper evaluates the performances of RUSBoost and SMOTEBoost, as well as their individual components (random undersampling, synthetic minority oversampling technique, and AdaBoost). We conduct experiments using 15 data sets from various application domains, four base learners, and four evaluation metrics. RUSBoost and SMOTEBoost both outperform the other procedures, and RUSBoost performs comparably to (and often better than) SMOTEBoost while being a simpler and faster technique. Given these experimental results, we highly recommend RUSBoost as an attractive alternative for improving the classification performance of learners built using imbalanced data.
This paper studies the factors associated with the emergence of systemic banking crises in a large sample of developed and developing countries in 1980-94 using a multivariate logit econometric model. The results suggest that crises tend to erupt when the macroeconomic environment is weak, particularly when growth is low and inflation is high. Also, high real interest rates are clearly associated with systemic banking sector problems, and there is some evidence that vulnerability to balance of payments crises has played a role. Countries with an explicit deposit insurance scheme were particularly at risk, as were countries with weak law enforcement. Copyright 1998, International Monetary Fund
This paper presents a review of alternative methodologies for early detection of banking distress. The methodologies proposed are aimed to the early identification of financial distress for countries without an important recent history of bank failure, but facing an unstable international environment. We evaluate several indicators and methodologies to measure financial distress such as qualitative indicators, the signal extraction approach, limited dependent estimation and finally duration models.
The paper tests the effectiveness of financial soundness indicators (FSIs) as harbingers of banking crises, using multivariate logit models to see whether FSIs, broad macroeconomic indicators, and institutional indicators can indeed predict crisis occurrences. The analysis draws upon a data set of homogeneous indicators comparable across countries over the period 2005 to 2012, leveraging the IMF’s FSI database. Results indicate significant correlation between some FSIs and the occurrence of systemic banking crises, and suggest that some indicators are precursors to the occurrence of banking crises.
This paper introduces a coincident indicator of systemic liquidity risk in the Italian
financial markets. In order to take account of the systemic dimension of liquidity stress,
standard portfolio theory is used. Three sub-indices, that reflect liquidity stress in specific
market segments, are aggregated in the systemic liquidity risk indicator in the same way as
individual risks are aggregated in order to quantify overall portfolio risk. The aggregation
takes account of the time-varying cross-correlations between the sub-indices, using a
multivariate GARCH approach. This is able to capture abrupt changes in the correlations and
makes it possible for the indicator to identify systemic liquidity events precisely. We
evaluate the indicator on its ability to match the results of a survey conducted among
financial market experts to determine the most liquidity stressful events for the Italian
financial markets. The results show that the systemic liquidity risk indicator accurately
identifies events characterized by high systemic risk, while not exaggerating the level of
stress during calm periods.
The paper develops an early-warning model for predicting vulnerabilities leading to distress in European banks using both bank and country-level data. As outright bank failures have been rare in Europe, we introduce a novel dataset that complements bankruptcies and defaults with state interventions and mergers in distress. The signals of the early-warning model are calibrated not only according to the policymaker’s preferences between type I and II errors, but also to take into account the potential systemic relevance of each individual financial institution. The key findings of the paper are that complementing bank-specific vulnerabilities with indicators for macro-financial imbalances improves model performance and yields useful out-of-sample predictions of bank distress during the current financial crisis.
Machine Learning in UK Financial Services
- Boe
BOE. (2019, October). Machine Learning in UK Financial Services. Retrieved from
https://www.bankofengland.co.uk/-/media/boe/files/report/2019/machine-learning-inuk-financial-services.pdf
The St. Louis Fed's Financial Stress Index, Version 2.0
- K Kliesen
- M Mccracken
Kliesen, K., & McCracken, M. (2020, March 26). The St. Louis Fed's Financial Stress Index,
Version 2.0. Retrieved from https://fredblog.stlouisfed.org/2020/03/the-st-louis-fedsfinancial-stress-index-version-2-0/
RUSBoost: A Hybrid Approach to Alleviating Class Imbalance
- C Seiffert
- T M Khoshgoftaar
- J V Hulse
- A Napolitano
Seiffert, C., Khoshgoftaar, T. M., Hulse, J. V., & Napolitano, A. (2008). RUSBoost: A Hybrid
Approach to Alleviating Class Imbalance. 19th International Conference on Pattern
Recognition, (pp. 1-4).