Forecasting net charge-off rates of banks: What model works best?

James R. Barth; Sumin Han; Sunghoon Joo; Kang Bok Lee; Stevan Maglic; Xuan Shen; James R. Barth; Sumin Han; Sunghoon Joo; Kang Bok Lee; Stevan Maglic; Xuan Shen

doi:10.3934/QFE.2018.3.554

Quantitative Finance and Economics

2018, Volume 2, Issue 3: 554-589. doi: 10.3934/QFE.2018.3.554

Previous Article Next Article

Research article Special Issues

Forecasting net charge-off rates of banks: What model works best?

1.
Lowder Eminent Scholar in Finance, Raymond J. Harbert College of Business, 316 Lowder Hall, Auburn University, Auburn, AL 36849, USA
2.
Assistant Professor in Business Analytics, Raymond J. Harbert College of Business, 413 Lowder Hall, Auburn University, Auburn, AL 36849, USA
3.
Ph.D. Student in Finance, Raymond J. Harbert College of Business, 306 Lowder Hall, Auburn University, Auburn, AL 36849, USA
4.
Assistant Professor in Business Analytics, Raymond J. Harbert College of Business, 424 Lowder Hall, Auburn University, Auburn, AL 36849, USA
5.
Senior Vice President and Head of Quantitative Risk Analytics, Regions Bank, 1900 5th Avenue North, Birmingham, AL 35203, USA
6.
Vice President and Risk Quantitative Analyst, Regions Bank, 1900 5th Avenue North, Birmingham, AL 35203, USA

Received: 23 March 2018 Accepted: 05 June 2018 Published: 27 July 2018
JEL Codes: C38, C53, C54, G17, G21, G32

The purpose of this paper is to focus on the losses of two very big banks, Citigroup (Citi) and Wells Fargo & Company (Wells Fargo), and two very small banks, First Busey Corporation (Busey) and Capital City Bank Group (Capital), over the period 1991–2016. The federal government actually bailed out the two big banks, as measured by total assets, whereas neither of the two small banks required a bail out. Clearly, if one is able to use a variety of predictor variables to forecast accurately the losses of banks of various sizes, in different geographical locations, and operating a variety of business models, this may help identify potential causes of future banking problems and thereby lessen, if not eliminate, the need for future bailouts. This is important for both the banks and the bank regulatory authorities. In particular, those banks expected to suffer significant losses on loans may be in a position to increase their provisioning and thus loan loss allowances. If such banks are unable to take this type of action or other corrective action to address expected losses, regulatory action may become necessary in response to this situation. The motivation for our paper is this very issue: can one obtain accurate forecasts of losses, or the net charge-off rates, of banks? We provide an answer to this question by examining the four banks mentioned using several hundred predictor variables and several different forecast techniques.
- forecasting,
- banking,
- predictive models,
- ridge estimator,
- partial least squares
Citation: James R. Barth, Sumin Han, Sunghoon Joo, Kang Bok Lee, Stevan Maglic, Xuan Shen. Forecasting net charge-off rates of banks: What model works best?[J]. Quantitative Finance and Economics, 2018, 2(3): 554-589. doi: 10.3934/QFE.2018.3.554

Related Papers:

Abstract

The purpose of this paper is to focus on the losses of two very big banks, Citigroup (Citi) and Wells Fargo & Company (Wells Fargo), and two very small banks, First Busey Corporation (Busey) and Capital City Bank Group (Capital), over the period 1991–2016. The federal government actually bailed out the two big banks, as measured by total assets, whereas neither of the two small banks required a bail out. Clearly, if one is able to use a variety of predictor variables to forecast accurately the losses of banks of various sizes, in different geographical locations, and operating a variety of business models, this may help identify potential causes of future banking problems and thereby lessen, if not eliminate, the need for future bailouts. This is important for both the banks and the bank regulatory authorities. In particular, those banks expected to suffer significant losses on loans may be in a position to increase their provisioning and thus loan loss allowances. If such banks are unable to take this type of action or other corrective action to address expected losses, regulatory action may become necessary in response to this situation. The motivation for our paper is this very issue: can one obtain accurate forecasts of losses, or the net charge-off rates, of banks? We provide an answer to this question by examining the four banks mentioned using several hundred predictor variables and several different forecast techniques.

References

[1]	Adrian T, Ashcraft AB (2016) Shadow banking: A review of the literature. Staff Rep 6: 282–315.
[2]	Barth J, Joo S, Kim H, et al. (2018) Forecasting net charge-off rates of banks: A PLS approach. Unpublished Manuscript.
[3]	Barth JR, Miller SM (2017) A primer on the evolution and complexity of bank regulatory capital standards. Unpublished Manuscript.
[4]	Bastos JA (2010) Forecasting bank loans loss-given-default. J Banking Finance 34: 2510–2517. doi: 10.1016/j.jbankfin.2010.04.011
[5]	Bernoth K, Pick A (2011) Forecasting the fragility of the banking and insurance sectors. J Banking Finance 35: 807–818. doi: 10.1016/j.jbankfin.2010.10.024
[6]	Covas FB, Rump B, Zakrajšek E (2014) Stress-testing US bank holding companies: A dynamic panel quantile regression approach. Int J Forecasting 30: 691–713. doi: 10.1016/j.ijforecast.2013.11.003
[7]	Crook J, Banasik J (2012) Forecasting and explaining aggregate consumer credit delinquency behaviour. Int J Forecasting 28: 145–160. doi: 10.1016/j.ijforecast.2010.12.002
[8]	Drehmann M, Juselius M (2014) Evaluating early warning indicators of banking crises: Satisfying policy requirements. Int J Forecasting 30: 759–780. doi: 10.1016/j.ijforecast.2013.10.002
[9]	Fitzpatrick BD, Reichmeier J, Dowell J (2017) Back to the future: The Landscape of the Financial Services Industry 2020 and Beyond. J Adv Econ Finance 2: 40–53.
[10]	Geladi P, Kowalski BR (1986) Partial least-squares regression: A tutorial. Anal Chim Acta 185: 1–17. doi: 10.1016/0003-2670(86)80028-9
[11]	Guerrieri L, Welch M (2012) Can macro variables used in stress testing forecast the performance of banks? Unpublished Manuscript.
[12]	Hirtle B, Kovner A, Vickery J, et al. (2016) Assessing financial stability: The capital and loss assessment under stress scenarios (CLASS) model. J Banking Finance 69: S35–S55. doi: 10.1016/j.jbankfin.2015.09.021
[13]	Hyndman RJ, Koehler AB (2006) Another look at measures of forecast accuracy. Int J Forecasting 22: 679–688. doi: 10.1016/j.ijforecast.2006.03.001
[14]	Jakšič M, Marinč M (2017) Relationship banking and information technology: The role of artificial intelligence and FinTech. Risk Manage 2017: 1–18.
[15]	Kupiec P (2018) Inside the black box: The accuracy of alternative stress test models. Unpublished Manuscript.
[16]	Luttrell D, Atkinson T, Rosenblum H (2013) Assessing the costs and consequences of the 2007–2009 financial crisis and its aftermath. Econ Lett 8: 1–4.
[17]	Pesaran MH (2006) Estimation and inference in large heterogeneous panels with a multifactor error structure. Econometrica 74: 967–1012. doi: 10.1111/j.1468-0262.2006.00692.x
[18]	Roy AD (1952) Safety first and the holding of assets. Econometrica 20: 431–449. doi: 10.2307/1907413
[19]	Tibshirani JR (1996) Regression shrinkage and selection via the Lasso. J R Stat Soc 58: 267–288.
[20]	Zou H, Hastie T (2010) Regularization and variable selection via the elastic net. J R Stat Soc 67: 301–320.

Reader Comments

Your name:*

Email:*
© 2018 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)