research paper on banking risk

• Open access
• Published: 03 September 2022

A literature review of risk, regulation, and profitability of banks using a scientometric study

• Shailesh Rastogi 1 ,
• Arpita Sharma 1 ,
• Geetanjali Pinto 2 &
• Venkata Mrudula Bhimavarapu   ORCID: orcid.org/0000-0002-9757-1904 1 , 3  

Future Business Journal volume  8 , Article number:  28 ( 2022 ) Cite this article

14k Accesses

2 Citations

Metrics details

This study presents a systematic literature review of regulation, profitability, and risk in the banking industry and explores the relationship between them. It proposes a policy initiative using a model that offers guidelines to establish the right mix among these variables. This is a systematic literature review study. Firstly, the necessary data are extracted using the relevant keywords from the Scopus database. The initial search results are then narrowed down, and the refined results are stored in a file. This file is finally used for data analysis. Data analysis is done using scientometrics tools, such as Table2net and Sciences cape software, and Gephi to conduct network, citation analysis, and page rank analysis. Additionally, content analysis of the relevant literature is done to construct a theoretical framework. The study identifies the prominent authors, keywords, and journals that researchers can use to understand the publication pattern in banking and the link between bank regulation, performance, and risk. It also finds that concentration banking, market power, large banks, and less competition significantly affect banks’ financial stability, profitability, and risk. Ownership structure and its impact on the performance of banks need to be investigated but have been inadequately explored in this study. This is an organized literature review exploring the relationship between regulation and bank performance. The limitations of the regulations and the importance of concentration banking are part of the findings.

Introduction

Globally, banks are under extreme pressure to enhance their performance and risk management. The financial industry still recalls the ignoble 2008 World Financial Crisis (WFC) as the worst economic disaster after the Great Depression of 1929. The regulatory mechanism before 2008 (mainly Basel II) was strongly criticized for its failure to address banks’ risks [ 47 , 87 ]. Thus, it is essential to investigate the regulation of banks [ 75 ]. This study systematically reviews the relevant literature on banks’ performance and risk management and proposes a probable solution.

Issues of performance and risk management of banks

Banks have always been hailed as engines of economic growth and have been the axis of the development of financial systems [ 70 , 85 ]. A vital parameter of a bank’s financial health is the volume of its non-performing assets (NPAs) on its balance sheet. NPAs are advances that delay in payment of interest or principal beyond a few quarters [ 108 , 118 ]. According to Ghosh [ 51 ], NPAs negatively affect the liquidity and profitability of banks, thus affecting credit growth and leading to financial instability in the economy. Hence, healthy banks translate into a healthy economy.

Despite regulations, such as high capital buffers and liquidity ratio requirements, during the second decade of the twenty-first century, the Indian banking sector still witnessed a substantial increase in NPAs. A recent report by the Indian central bank indicates that the gross NPA ratio reached an all-time peak of 11% in March 2018 and 12.2% in March 2019 [ 49 ]. Basel II has been criticized for several reasons [ 98 ]. Schwerter [ 116 ] and Pakravan [ 98 ] highlighted the systemic risk and gaps in Basel II, which could not address the systemic risk of WFC 2008. Basel III was designed to close the gaps in Basel II. However, Schwerter [ 116 ] criticized Basel III and suggested that more focus should have been on active risk management practices to avoid any impending financial crisis. Basel III was proposed to solve these issues, but it could not [ 3 , 116 ]. Samitas and Polyzos [ 113 ] found that Basel III had made banking challenging since it had reduced liquidity and failed to shield the contagion effect. Therefore, exploring some solutions to establish the right balance between regulation, performance, and risk management of banks is vital.

Keeley [ 67 ] introduced the idea of a balance among banks’ profitability, regulation, and NPA (risk-taking). This study presents the balancing act of profitability, regulation, and NPA (risk-taking) of banks as a probable solution to the issues of bank performance and risk management and calls it a triad . Figure  1 illustrates the concept of a triad. Several authors have discussed the triad in parts [ 32 , 96 , 110 , 112 ]. Triad was empirically tested in different countries by Agoraki et al. [ 1 ]. Though the idea of a triad is quite old, it is relevant in the current scenario. The spirit of the triad strongly and collectively admonishes the Basel Accord and exhibits new and exhaustive measures to take up and solve the issue of performance and risk management in banks [ 16 , 98 ]. The 2008 WFC may have caused an imbalance among profitability, regulation, and risk-taking of banks [ 57 ]. Less regulation , more competition (less profitability ), and incentive to take the risk were the cornerstones of the 2008 WFC [ 56 ]. Achieving a balance among the three elements of a triad is a real challenge for banks’ performance and risk management, which this study addresses.

Triad of Profitability, regulation, and NPA (risk-taking). Note The triad [ 131 ] of profitability, regulation, and NPA (risk-taking) is shown in Fig.  1

Triki et al. [ 130 ] revealed that a bank’s performance is a trade-off between the elements of the triad. Reduction in competition increases the profitability of banks. However, in the long run, reduction in competition leads to either the success or failure of banks. Flexible but well-expressed regulation and less competition add value to a bank’s performance. The current review paper is an attempt to explore the literature on this triad of bank performance, regulation, and risk management. This paper has the following objectives:

To systematically explore the existing literature on the triad: performance, regulation, and risk management of banks; and

To propose a model for effective bank performance and risk management of banks.

Literature is replete with discussion across the world on the triad. However, there is a lack of acceptance of the triad as a solution to the woes of bank performance and risk management. Therefore, the findings of the current papers significantly contribute to this regard. This paper collates all the previous studies on the triad systematically and presents a curated view to facilitate the policy makers and stakeholders to make more informed decisions on the issue of bank performance and risk management. This paper also contributes significantly by proposing a DBS (differential banking system) model to solve the problem of banks (Fig.  7 ). This paper examines studies worldwide and therefore ensures the wider applicability of its findings. Applicability of the DBS model is not only limited to one nation but can also be implemented worldwide. To the best of the authors’ knowledge, this is the first study to systematically evaluate the publication pattern in banking using a blend of scientometrics analysis tools, network analysis tools, and content analysis to understand the link between bank regulation, performance, and risk.

This paper is divided into five sections. “ Data and research methods ” section discusses the research methodology used for the study. The data analysis for this study is presented in two parts. “ Bibliometric and network analysis ” section presents the results obtained using bibliometric and network analysis tools, followed by “ Content Analysis ” section, which presents the content analysis of the selected literature. “ Discussion of the findings ” section discusses the results and explains the study’s conclusion, followed by limitations and scope for further research.

Data and research methods

A literature review is a systematic, reproducible, and explicit way of identifying, evaluating, and synthesizing relevant research produced and published by researchers [ 50 , 100 ]. Analyzing existing literature helps researchers generate new themes and ideas to justify the contribution made to literature. The knowledge obtained through evidence-based research also improves decision-making leading to better practical implementation in the real corporate world [ 100 , 129 ].

As Kumar et al. [ 77 , 78 ] and Rowley and Slack [ 111 ] recommended conducting an SLR, this study also employs a three-step approach to understand the publication pattern in the banking area and establish a link between bank performance, regulation, and risk.

Determining the appropriate keywords for exploring the data

Many databases such as Google Scholar, Web of Science, and Scopus are available to extract the relevant data. The quality of a publication is associated with listing a journal in a database. Scopus is a quality database as it has a wider coverage of data [ 100 , 137 ]. Hence, this study uses the Scopus database to extract the relevant data.

For conducting an SLR, there is a need to determine the most appropriate keywords to be used in the database search engine [ 26 ]. Since this study seeks to explore a link between regulation, performance, and risk management of banks, the keywords used were “risk,” “regulation,” “profitability,” “bank,” and “banking.”

Initial search results and limiting criteria

Using the keywords identified in step 1, the search for relevant literature was conducted in December 2020 in the Scopus database. This resulted in the search of 4525 documents from inception till December 2020. Further, we limited our search to include “article” publications only and included subject areas: “Economics, Econometrics and Finance,” “Business, Management and Accounting,” and “Social sciences” only. This resulted in a final search result of 3457 articles. These results were stored in a.csv file which is then used as an input to conduct the SLR.

Data analysis tools and techniques

This study uses bibliometric and network analysis tools to understand the publication pattern in the area of research [ 13 , 48 , 100 , 122 , 129 , 134 ]. Some sub-analyses of network analysis are keyword word, author, citation, and page rank analysis. Author analysis explains the author’s contribution to literature or research collaboration, national and international [ 59 , 99 ]. Citation analysis focuses on many researchers’ most cited research articles [ 100 , 102 , 131 ].

The.csv file consists of all bibliometric data for 3457 articles. Gephi and other scientometrics tools, such as Table2net and ScienceScape software, were used for the network analysis. This.csv file is directly used as an input for this software to obtain network diagrams for better data visualization [ 77 ]. To ensure the study’s quality, the articles with 50 or more citations (216 in number) are selected for content analysis [ 53 , 102 ]. The contents of these 216 articles are analyzed to develop a conceptual model of banks’ triad of risk, regulation, and profitability. Figure  2 explains the data retrieval process for SLR.

Data retrieval process for SLR. Note Stepwise SLR process and corresponding results obtained

Bibliometric and network analysis

Figure  3 [ 58 ] depicts the total number of studies that have been published on “risk,” “regulation,” “profitability,” “bank,” and “banking.” Figure  3 also depicts the pattern of the quality of the publications from the beginning till 2020. It undoubtedly shows an increasing trend in the number of articles published in the area of the triad: “risk” regulation” and “profitability.” Moreover, out of the 3457 articles published in the said area, 2098 were published recently in the last five years and contribute to 61% of total publications in this area.

Articles published from 1976 till 2020 . Note The graph shows the number of documents published from 1976 till 2020 obtained from the Scopus database

Source of publications

A total of 160 journals have contributed to the publication of 3457 articles extracted from Scopus on the triad of risk, regulation, and profitability. Table 1 shows the top 10 sources of the publications based on the citation measure. Table 1 considers two sets of data. One data set is the universe of 3457 articles, and another is the set of 216 articles used for content analysis along with their corresponding citations. The global citations are considered for the study from the Scopus dataset, and the local citations are considered for the articles in the nodes [ 53 , 135 ]. The top 10 journals with 50 or more citations resulted in 96 articles. This is almost 45% of the literature used for content analysis ( n  = 216). Table 1 also shows that the Journal of Banking and Finance is the most prominent in terms of the number of publications and citations. It has 46 articles published, which is about 21% of the literature used for content analysis. Table 1 also shows these core journals’ SCImago Journal Rank indicator and H index. SCImago Journal Rank indicator reflects the impact and prestige of the Journal. This indicator is calculated as the previous three years’ weighted average of the number of citations in the Journal since the year that the article was published. The h index is the number of articles (h) published in a journal and received at least h. The number explains the scientific impact and the scientific productivity of the Journal. Table 1 also explains the time span of the journals covering articles in the area of the triad of risk, regulation, and profitability [ 7 ].

Figure  4 depicts the network analysis, where the connections between the authors and source title (journals) are made. The network has 674 nodes and 911 edges. The network between the author and Journal is classified into 36 modularities. Sections of the graph with dense connections indicate high modularity. A modularity algorithm is a design that measures how strong the divided networks are grouped into modules; this means how well the nodes are connected through a denser route relative to other networks.

Network analysis between authors and journals. Note A node size explains the more linked authors to a journal

The size of the nodes is based on the rank of the degree. The degree explains the number of connections or edges linked to a node. In the current graph, a node represents the name of the Journal and authors; they are connected through the edges. Therefore, the more the authors are associated with the Journal, the higher the degree. The algorithm used for the layout is Yifan Hu’s.

Many authors are associated with the Journal of Banking and Finance, Journal of Accounting and Economics, Journal of Financial Economics, Journal of Financial Services Research, and Journal of Business Ethics. Therefore, they are the most relevant journals on banks’ risk, regulation, and profitability.

Location and affiliation analysis

Affiliation analysis helps to identify the top contributing countries and universities. Figure  5 shows the countries across the globe where articles have been published in the triad. The size of the circle in the map indicates the number of articles published in that country. Table 2 provides the details of the top contributing organizations.

Location of articles published on Triad of profitability, regulation, and risk

Figure  5 shows that the most significant number of articles is published in the USA, followed by the UK. Malaysia and China have also contributed many articles in this area. Table 2 shows that the top contributing universities are also from Malaysia, the UK, and the USA.

Key author analysis

Table 3 shows the number of articles written by the authors out of the 3457 articles. The table also shows the top 10 authors of bank risk, regulation, and profitability.

Fadzlan Sufian, affiliated with the Universiti Islam Malaysia, has the maximum number, with 33 articles. Philip Molyneux and M. Kabir Hassan are from the University of Sharjah and the University of New Orleans, respectively; they contributed significantly, with 20 and 18 articles, respectively.

However, when the quality of the article is selected based on 50 or more citations, Fadzlan Sufian has only 3 articles with more than 50 citations. At the same time, Philip Molyneux and Allen Berger contributed more quality articles, with 8 and 11 articles, respectively.

Keyword analysis

Table 4 shows the keyword analysis (times they appeared in the articles). The top 10 keywords are listed in Table 4 . Banking and banks appeared 324 and 194 times, respectively, which forms the scope of this study, covering articles from the beginning till 2020. The keyword analysis helps to determine the factors affecting banks, such as profitability (244), efficiency (129), performance (107, corporate governance (153), risk (90), and regulation (89).

The keywords also show that efficiency through data envelopment analysis is a determinant of the performance of banks. The other significant determinants that appeared as keywords are credit risk (73), competition (70), financial stability (69), ownership structure (57), capital (56), corporate social responsibility (56), liquidity (46), diversification (45), sustainability (44), credit provision (41), economic growth (41), capital structure (39), microfinance (39), Basel III (37), non-performing assets (37), cost efficiency (30), lending behavior (30), interest rate (29), mergers and acquisition (28), capital adequacy (26), developing countries (23), net interest margin (23), board of directors (21), disclosure (21), leverage (21), productivity (20), innovation (18), firm size (16), and firm value (16).

Keyword analysis also shows the theories of banking and their determinants. Some of the theories are agency theory (23), information asymmetry (21), moral hazard (17), and market efficiency (16), which can be used by researchers when building a theory. The analysis also helps to determine the methodology that was used in the published articles; some of them are data envelopment analysis (89), which measures technical efficiency, panel data analysis (61), DEA (32), Z scores (27), regression analysis (23), stochastic frontier analysis (20), event study (15), and literature review (15). The count for literature review is only 15, which confirms that very few studies have conducted an SLR on bank risk, regulation, and profitability.

Citation analysis

One of the parameters used in judging the quality of the article is its “citation.” Table 5 shows the top 10 published articles with the highest number of citations. Ding and Cronin [ 44 ] indicated that the popularity of an article depends on the number of times it has been cited.

Tahamtan et al. [ 126 ] explained that the journal’s quality also affects its published articles’ citations. A quality journal will have a high impact factor and, therefore, more citations. The citation analysis helps researchers to identify seminal articles. The title of an article with 5900 citations is “A survey of corporate governance.”

Page Rank analysis

Goyal and Kumar [ 53 ] explain that the citation analysis indicates the ‘popularity’ and ‘prestige’ of the published research article. Apart from the citation analysis, one more analysis is essential: Page rank analysis. PageRank is given by Page et al. [ 97 ]. The impact of an article can be measured with one indicator called PageRank [ 135 ]. Page rank analysis indicates how many times an article is cited by other highly cited articles. The method helps analyze the web pages, which get the priority during any search done on google. The analysis helps in understanding the citation networks. Equation  1 explains the page rank (PR) of a published paper, N refers to the number of articles.

T 1,… T n indicates the paper, which refers paper P . C ( Ti ) indicates the number of citations. The damping factor is denoted by a “ d ” which varies in the range of 0 and 1. The page rank of all the papers is equal to 1. Table 6 shows the top papers based on page rank. Tables 5 and 6 together show a contrast in the top ranked articles based on citations and page rank, respectively. Only one article “A survey of corporate governance” falls under the prestigious articles based on the page rank.

Content analysis

Content Analysis is a research technique for conducting qualitative and quantitative analyses [ 124 ]. The content analysis is a helpful technique that provides the required information in classifying the articles depending on their nature (empirical or conceptual) [ 76 ]. By adopting the content analysis method [ 53 , 102 ], the selected articles are examined to determine their content. The classification of available content from the selected set of sample articles that are categorized under different subheads. The themes identified in the relationship between banking regulation, risk, and profitability are as follows.

Regulation and profitability of banks

The performance indicators of the banking industry have always been a topic of interest to researchers and practitioners. This area of research has assumed a special interest after the 2008 WFC [ 25 , 51 , 86 , 114 , 127 , 132 ]. According to research, the causes of poor performance and risk management are lousy banking practices, ineffective monitoring, inadequate supervision, and weak regulatory mechanisms [ 94 ]. Increased competition, deregulation, and complex financial instruments have made banks, including Indian banks, more vulnerable to risks [ 18 , 93 , 119 , 123 ]. Hence, it is essential to investigate the present regulatory machinery for the performance of banks.

There are two schools of thought on regulation and its possible impact on profitability. The first asserts that regulation does not affect profitability. The second asserts that regulation adds significant value to banks’ profitability and other performance indicators. This supports the concept that Delis et al. [ 41 ] advocated that the capital adequacy requirement and supervisory power do not affect productivity or profitability unless there is a financial crisis. Laeven and Majnoni [ 81 ] insisted that provision for loan loss should be part of capital requirements. This will significantly improve active risk management practices and ensure banks’ profitability.

Lee and Hsieh [ 83 ] proposed ambiguous findings that do not support either school of thought. According to Nguyen and Nghiem [ 95 ], while regulation is beneficial, it has a negative impact on bank profitability. As a result, when proposing regulations, it is critical to consider bank performance and risk management. According to Erfani and Vasigh [ 46 ], Islamic banks maintained their efficiency between 2006 and 2013, while most commercial banks lost, furthermore claimed that the financial crisis had no significant impact on Islamic bank profitability.

Regulation and NPA (risk-taking of banks)

The regulatory mechanism of banks in any country must address the following issues: capital adequacy ratio, prudent provisioning, concentration banking, the ownership structure of banks, market discipline, regulatory devices, presence of foreign capital, bank competition, official supervisory power, independence of supervisory bodies, private monitoring, and NPAs [ 25 ].

Kanoujiya et al. [ 64 ] revealed through empirical evidence that Indian bank regulations lack a proper understanding of what banks require and propose reforming and transforming regulation in Indian banks so that responsive governance and regulation can occur to make banks safer, supported by Rastogi et al. [ 105 ]. The positive impact of regulation on NPAs is widely discussed in the literature. [ 94 ] argue that regulation has multiple effects on banks, including reducing NPAs. The influence is more powerful if the country’s banking system is fragile. Regulation, particularly capital regulation, is extremely effective in reducing risk-taking in banks [ 103 ].

Rastogi and Kanoujiya [ 106 ] discovered evidence that disclosure regulations do not affect the profitability of Indian banks, supported by Karyani et al. [ 65 ] for the banks located in Asia. Furthermore, Rastogi and Kanoujiya [ 106 ] explain that disclosure is a difficult task as a regulatory requirement. It is less sustainable due to the nature of the imposed regulations in banks and may thus be perceived as a burden and may be overcome by realizing the benefits associated with disclosure regulation [ 31 , 54 , 101 ]. Zheng et al. [ 138 ] empirically discovered that regulation has no impact on the banks’ profitability in Bangladesh.

Governments enforce banking regulations to achieve a stable and efficient financial system [ 20 , 94 ]. The existing literature is inconclusive on the effects of regulatory compliance on banks’ risks or the reduction of NPAs [ 10 , 11 ]. Boudriga et al. [ 25 ] concluded that the regulatory mechanism plays an insignificant role in reducing NPAs. This is especially true in weak institutions, which are susceptible to corruption. Gonzalez [ 52 ] reported that firm regulations have a positive relationship with banks’ risk-taking, increasing the probability of NPAs. However, Boudriga et al. [ 25 ], Samitas and Polyzos [ 113 ], and Allen et al. [ 3 ] strongly oppose the use of regulation as a tool to reduce banks’ risk-taking.

Kwan and Laderman [ 79 ] proposed three levels in regulating banks, which are lax, liberal, and strict. The liberal regulatory framework leads to more diversification in banks. By contrast, the strict regulatory framework forces the banks to take inappropriate risks to compensate for the loss of business; this is a global problem [ 73 ].

Capital regulation reduces banks’ risk-taking [ 103 , 110 ]. Capital regulation leads to cost escalation, but the benefits outweigh the cost [ 103 ]. The trade-off is worth striking. Altman Z score is used to predict banks’ bankruptcy, and it found that the regulation increased the Altman’s Z-score [ 4 , 46 , 63 , 68 , 72 , 120 ]. Jin et al. [ 62 ] report a negative relationship between regulation and banks’ risk-taking. Capital requirements empowered regulators, and competition significantly reduced banks’ risk-taking [ 1 , 122 ]. Capital regulation has a limited impact on banks’ risk-taking [ 90 , 103 ].

Maji and De [ 90 ] suggested that human capital is more effective in managing banks’ credit risks. Besanko and Kanatas [ 21 ] highlighted that regulation on capital requirements might not mitigate risks in all scenarios, especially when recapitalization has been enforced. Klomp and De Haan [ 72 ] proposed that capital requirements and supervision substantially reduce banks’ risks.

A third-party audit may impart more legitimacy to the banking system [ 23 ]. The absence of third-party intervention is conspicuous, and this may raise a doubt about the reliability and effectiveness of the impact of regulation on bank’s risk-taking.

NPA (risk-taking) in banks and profitability

Profitability affects NPAs, and NPAs, in turn, affect profitability. According to the bad management hypothesis [ 17 ], higher profits would negatively affect NPAs. By contrast, higher profits may lead management to resort to a liberal credit policy (high earnings), which may eventually lead to higher NPAs [ 104 ].

Balasubramaniam [ 8 ] demonstrated that NPA has double negative effects on banks. NPAs increase stressed assets, reducing banks’ productive assets [ 92 , 117 , 136 ]. This phenomenon is relatively underexplored and therefore renders itself for future research.

Triad and the performance of banks

Regulation and triad.

Regulations and their impact on banks have been a matter of debate for a long time. Barth et al. [ 12 ] demonstrated that countries with a central bank as the sole regulatory body are prone to high NPAs. Although countries with multiple regulatory bodies have high liquidity risks, they have low capital requirements [ 40 ]. Barth et al. [ 12 ] supported the following steps to rationalize the existing regulatory mechanism on banks: (1) mandatory information [ 22 ], (2) empowered management of banks, and (3) increased incentive for private agents to exert corporate control. They show that profitability has an inverse relationship with banks’ risk-taking [ 114 ]. Therefore, standard regulatory practices, such as capital requirements, are not beneficial. However, small domestic banks benefit from capital restrictions.

DeYoung and Jang [ 43 ] showed that Basel III-based policies of liquidity convergence ratio (LCR) and net stable funding ratio (NSFR) are not fully executed across the globe, including the US. Dahir et al. [ 39 ] found that a decrease in liquidity and funding increases banks’ risk-taking, making banks vulnerable and reducing stability. Therefore, any regulation on liquidity risk is more likely to create problems for banks.

Concentration banking and triad

Kiran and Jones [ 71 ] asserted that large banks are marginally affected by NPAs, whereas small banks are significantly affected by high NPAs. They added a new dimension to NPAs and their impact on profitability: concentration banking or banks’ market power. Market power leads to less cost and more profitability, which can easily counter the adverse impact of NPAs on profitability [ 6 , 15 ].

The connection between the huge volume of research on the performance of banks and competition is the underlying concept of market power. Competition reduces market power, whereas concentration banking increases market power [ 25 ]. Concentration banking reduces competition, increases market power, rationalizes the banks’ risk-taking, and ensures profitability.

Tabak et al. [ 125 ] advocated that market power incentivizes banks to become risk-averse, leading to lower costs and high profits. They explained that an increase in market power reduces the risk-taking requirement of banks. Reducing banks’ risks due to market power significantly increases when capital regulation is executed objectively. Ariss [ 6 ] suggested that increased market power decreases competition, and thus, NPAs reduce, leading to increased banks’ stability.

Competition, the performance of banks, and triad

Boyd and De Nicolo [ 27 ] supported that competition and concentration banking are inversely related, whereas competition increases risk, and concentration banking decreases risk. A mere shift toward concentration banking can lead to risk rationalization. This finding has significant policy implications. Risk reduction can also be achieved through stringent regulations. Bolt and Tieman [ 24 ] explained that stringent regulation coupled with intense competition does more harm than good, especially concerning banks’ risk-taking.

Market deregulation, as well as intensifying competition, would reduce the market power of large banks. Thus, the entire banking system might take inappropriate and irrational risks [ 112 ]. Maji and Hazarika [ 91 ] added more confusion to the existing policy by proposing that, often, there is no relationship between capital regulation and banks’ risk-taking. However, some cases have reported a positive relationship. This implies that banks’ risk-taking is neutral to regulation or leads to increased risk. Furthermore, Maji and Hazarika [ 91 ] revealed that competition reduces banks’ risk-taking, contrary to popular belief.

Claessens and Laeven [ 36 ] posited that concentration banking influences competition. However, this competition exists only within the restricted circle of banks, which are part of concentration banking. Kasman and Kasman [ 66 ] found that low concentration banking increases banks’ stability. However, they were silent on the impact of low concentration banking on banks’ risk-taking. Baselga-Pascual et al. [ 14 ] endorsed the earlier findings that concentration banking reduces banks’ risk-taking.

Concentration banking and competition are inversely related because of the inherent design of concentration banking. Market power increases when only a few large banks are operating; thus, reduced competition is an obvious outcome. Barra and Zotti [ 9 ] supported the idea that market power, coupled with competition between the given players, injects financial stability into banks. Market power and concentration banking affect each other. Therefore, concentration banking with a moderate level of regulation, instead of indiscriminate regulation, would serve the purpose better. Baselga-Pascual et al. [ 14 ] also showed that concentration banking addresses banks’ risk-taking.

Schaeck et al. [ 115 ], in a landmark study, presented that concentration banking and competition reduce banks’ risk-taking. However, they did not address the relationship between concentration banking and competition, which are usually inversely related. This could be a subject for future research. Research on the relationship between concentration banking and competition is scant, identified as a research gap (“ Research Implications of the study ” section).

Transparency, corporate governance, and triad

One of the big problems with NPAs is the lack of transparency in both the regulatory bodies and banks [ 25 ]. Boudriga et al. [ 25 ] preferred to view NPAs as a governance issue and thus, recommended viewing it from a governance perspective. Ahmad and Ariff [ 2 ] concluded that regulatory capital and top-management quality determine banks’ credit risk. Furthermore, they asserted that credit risk in emerging economies is higher than that of developed economies.

Bad management practices and moral vulnerabilities are the key determinants of insolvency risks of Indian banks [ 95 ]. Banks are an integral part of the economy and engines of social growth. Therefore, banks enjoy liberal insolvency protection in India, especially public sector banks, which is a critical issue. Such a benevolent insolvency cover encourages a bank to be indifferent to its capital requirements. This indifference takes its toll on insolvency risk and profit efficiency. Insolvency protection makes the bank operationally inefficient and complacent.

Foreign equity and corporate governance practices help manage the adverse impact of banks’ risk-taking to ensure the profitability and stability of banks [ 33 , 34 ]. Eastburn and Sharland [ 45 ] advocated that sound management and a risk management system that can anticipate any impending risk are essential. A pragmatic risk mechanism should replace the existing conceptual risk management system.

Lo [ 87 ] found and advocated that the existing legislation and regulations are outdated. He insisted on a new perspective and asserted that giving equal importance to behavioral aspects and the rational expectations of customers of banks is vital. Buston [ 29 ] critiqued the balance sheet risk management practices prevailing globally. He proposed active risk management practices that provided risk protection measures to contain banks’ liquidity and solvency risks.

Klomp and De Haan [ 72 ] championed the cause of giving more autonomy to central banks of countries to provide stability in the banking system. Louzis et al. [ 88 ] showed that macroeconomic variables and the quality of bank management determine banks’ level of NPAs. Regulatory authorities are striving hard to make regulatory frameworks more structured and stringent. However, the recent increase in loan defaults (NPAs), scams, frauds, and cyber-attacks raise concerns about the effectiveness [ 19 ] of the existing banking regulations in India as well as globally.

Discussion of the findings

The findings of this study are based on the bibliometric and content analysis of the sample published articles.

The bibliometric study concludes that there is a growing demand for researchers and good quality research

The keyword analysis suggests that risk regulation, competition, profitability, and performance are key elements in understanding the banking system. The main authors, keywords, and journals are grouped in a Sankey diagram in Fig.  6 . Researchers can use the following information to understand the publication pattern on banking and its determinants.

Sankey Diagram of main authors, keywords, and journals. Note Authors contribution using scientometrics tools

Research Implications of the study

The study also concludes that a balance among the three components of triad is the solution to the challenges of banks worldwide, including India. We propose the following recommendations and implications for banks:

This study found that “the lesser the better,” that is, less regulation enhances the performance and risk management of banks. However, less regulation does not imply the absence of regulation. Less regulation means the following:

Flexible but full enforcement of the regulations

Customization, instead of a one-size-fits-all regulatory system rooted in a nation’s indigenous requirements, is needed. Basel or generic regulation can never achieve what a customized compliance system can.

A third-party audit, which is above the country's central bank, should be mandatory, and this would ensure that all three aspects of audit (policy formulation, execution, and audit) are handled by different entities.

Competition

This study asserts that the existing literature is replete with poor performance and risk management due to excessive competition. Banking is an industry of a different genre, and it would be unfair to compare it with the fast-moving consumer goods (FMCG) or telecommunication industry, where competition injects efficiency into the system, leading to customer empowerment and satisfaction. By contrast, competition is a deterrent to the basic tenets of safe banking. Concentration banking is more effective in handling the multi-pronged balance between the elements of the triad. Concentration banking reduces competition to lower and manageable levels, reduces banks’ risk-taking, and enhances profitability.

No incentive to take risks

It is found that unless banks’ risk-taking is discouraged, the problem of high NPA (risk-taking) cannot be addressed. Concentration banking is a disincentive to risk-taking and can be a game-changer in handling banks’ performance and risk management.

Research on the risk and performance of banks reveals that the existing regulatory and policy arrangement is not a sustainable proposition, especially for a country where half of the people are unbanked [ 37 ]. Further, the triad presented by Keeley [ 67 ] is a formidable real challenge to bankers. The balance among profitability, risk-taking, and regulation is very subtle and becomes harder to strike, just as the banks globally have tried hard to achieve it. A pragmatic intervention is needed; hence, this study proposes a change in the banking structure by having two types of banks functioning simultaneously to solve the problems of risk and performance of banks. The proposed two-tier banking system explained in Fig.  7 can be a great solution. This arrangement will help achieve the much-needed balance among the elements of triad as presented by Keeley [ 67 ].

Conceptual Framework. Note Fig.  7 describes the conceptual framework of the study

The first set of banks could be conventional in terms of their structure and should primarily be large-sized. The number of such banks should be moderate. There is a logic in having only a few such banks to restrict competition; thus, reasonable market power could be assigned to them [ 55 ]. However, a reduction in competition cannot be over-assumed, and banks cannot become complacent. As customary, lending would be the main source of revenue and income for these banks (fund based activities) [ 82 ]. The proposed two-tier system can be successful only when regulation especially for risk is objectively executed [ 29 ]. The second set of banks could be smaller in size and more in number. Since they are more in number, they would encounter intense competition for survival and for generating more business. Small is beautiful, and thus, this set of banks would be more agile and adaptable and consequently more efficient and profitable. The main source of revenue for this set of banks would not be loans and advances. However, non-funding and non-interest-bearing activities would be the major revenue source. Unlike their traditional and large-sized counterparts, since these banks are smaller in size, they are less likely to face risk-taking and NPAs [ 74 ].

Sarmiento and Galán [ 114 ] presented the concerns of large and small banks and their relative ability and appetite for risk-taking. High risk could threaten the existence of small-sized banks; thus, they need robust risk shielding. Small size makes them prone to failure, and they cannot convert their risk into profitability. However, large banks benefit from their size and are thus less vulnerable and can convert risk into profitable opportunities.

India has experimented with this Differential Banking System (DBS) (two-tier system) only at the policy planning level. The execution is impending, and it highly depends on the political will, which does not appear to be strong now. The current agenda behind the DBS model is not to ensure the long-term sustainability of banks. However, it is currently being directed to support the agenda of financial inclusion by extending the formal credit system to the unbanked masses [ 107 ]. A shift in goal is needed to employ the DBS as a strategic decision, but not merely a tool for financial inclusion. Thus, the proposed two-tier banking system (DBS) can solve the issue of profitability through proper regulation and less risk-taking.

The findings of Triki et al. [ 130 ] support the proposed DBS model, in this study. Triki et al. [ 130 ] advocated that different component of regulations affect banks based on their size, risk-taking, and concentration banking (or market power). Large size, more concentration banking with high market power, and high risk-taking coupled with stringent regulation make the most efficient banks in African countries. Sharifi et al. [ 119 ] confirmed that size advantage offers better risk management to large banks than small banks. The banks should modify and work according to the economic environment in the country [ 69 ], and therefore, the proposed model could help in solving the current economic problems.

This is a fact that DBS is running across the world, including in India [ 60 ] and other countries [ 133 ]. India experimented with DBS in the form of not only regional rural banks (RRBs) but payments banks [ 109 ] and small finance banks as well [ 61 ]. However, the purpose of all the existing DBS models, whether RRBs [ 60 ], payment banks, or small finance banks, is financial inclusion, not bank performance and risk management. Hence, they are unable to sustain and are failing because their model is only social instead of a much-needed dual business-cum-social model. The two-tier model of DBS proposed in the current paper can help serve the dual purpose. It may not only be able to ensure bank performance and risk management but also serve the purpose of inclusive growth of the economy.

Conclusion of the study

The study’s conclusions have some significant ramifications. This study can assist researchers in determining their study plan on the current topic by using a scientific approach. Citation analysis has aided in the objective identification of essential papers and scholars. More collaboration between authors from various countries/universities may help countries/universities better understand risk regulation, competition, profitability, and performance, which are critical elements in understanding the banking system. The regulatory mechanism in place prior to 2008 failed to address the risk associated with banks [ 47 , 87 ]. There arises a necessity and motivates authors to investigate the current topic. The present study systematically explores the existing literature on banks’ triad: performance, regulation, and risk management and proposes a probable solution.

To conclude the bibliometric results obtained from the current study, from the number of articles published from 1976 to 2020, it is evident that most of the articles were published from the year 2010, and the highest number of articles were published in the last five years, i.e., is from 2015. The authors discovered that researchers evaluate articles based on the scope of critical journals within the subject area based on the detailed review. Most risk, regulation, and profitability articles are published in peer-reviewed journals like; “Journal of Banking and Finance,” “Journal of Accounting and Economics,” and “Journal of Financial Economics.” The rest of the journals are presented in Table 1 . From the affiliation statistics, it is clear that most of the research conducted was affiliated with developed countries such as Malaysia, the USA, and the UK. The researchers perform content analysis and Citation analysis to access the type of content where the research on the current field of knowledge is focused, and citation analysis helps the academicians understand the highest cited articles that have more impact in the current research area.

Practical implications of the study

The current study is unique in that it is the first to systematically evaluate the publication pattern in banking using a combination of scientometrics analysis tools, network analysis tools, and content analysis to understand the relationship between bank regulation, performance, and risk. The study’s practical implications are that analyzing existing literature helps researchers generate new themes and ideas to justify their contribution to literature. Evidence-based research knowledge also improves decision-making, resulting in better practical implementation in the real corporate world [ 100 , 129 ].

Limitations and scope for future research

The current study only considers a single database Scopus to conduct the study, and this is one of the limitations of the study spanning around the multiple databases can provide diverse results. The proposed DBS model is a conceptual framework that requires empirical testing, which is a limitation of this study. As a result, empirical testing of the proposed DBS model could be a future research topic.

Availability of data and materials

SCOPUS database.

Abbreviations

Systematic literature review

World Financial Crisis

Non-performing assets

Differential banking system

SCImago Journal Rank Indicator

Liquidity convergence ratio

Net stable funding ratio

Fast moving consumer goods

Regional rural banks

Agoraki M-EK, Delis MD, Pasiouras F (2011) Regulations, competition and bank risk-taking in transition countries. J Financ Stab 7(1):38–48

Google Scholar  

Ahmad NH, Ariff M (2007) Multi-country study of bank credit risk determinants. Int J Bank Financ 5(1):35–62

Allen B, Chan KK, Milne A, Thomas S (2012) Basel III: Is the cure worse than the disease? Int Rev Financ Anal 25:159–166

Altman EI (2018) A fifty-year retrospective on credit risk models, the Altman Z-score family of models, and their applications to financial markets and managerial strategies. J Credit Risk 14(4):1–34

Alvarez F, Jermann UJ (2000) Efficiency, equilibrium, and asset pricing with risk of default. Econometrica 68(4):775–797

Ariss RT (2010) On the implications of market power in banking: evidence from developing countries. J Bank Financ 34(4):765–775

Aznar-Sánchez JA, Piquer-Rodríguez M, Velasco-Muñoz JF, Manzano-Agugliaro F (2019) Worldwide research trends on sustainable land use in agriculture. Land Use Policy 87:104069

Balasubramaniam C (2012) Non-performing assets and profitability of commercial banks in India: assessment and emerging issues. Nat Mon Refereed J Res Commer Manag 1(1):41–52

Barra C, Zotti R (2017) On the relationship between bank market concentration and stability of financial institutions: evidence from the Italian banking sector, MPRA working Paper No 79900. Last Accessed on Jan 2021 https://mpra.ub.uni-muenchen.de/79900/1/MPRA_paper_79900.pdf

Barth JR, Caprio G, Levine R (2004) Bank regulation and supervision: what works best? J Financ Intermed 2(13):205–248

Barth JR, Caprio G, Levine R (2008) Bank regulations are changing: For better or worse? Comp Econ Stud 50(4):537–563

Barth JR, Dopico LG, Nolle DE, Wilcox JA (2002) Bank safety and soundness and the structure of bank supervision: a cross-country analysis. Int Rev Financ 3(3–4):163–188

Bartolini M, Bottani E, Grosse EH (2019) Green warehousing: systematic literature review and bibliometric analysis. J Clean Prod 226:242–258

Baselga-Pascual L, Trujillo-Ponce A, Cardone-Riportella C (2015) Factors influencing bank risk in Europe: evidence from the financial crisis. N Am J Econ Financ 34(1):138–166

Beck T, Demirgüç-Kunt A, Levine R (2006) Bank concentration, competition, and crises: first results. J Bank Financ 30(5):1581–1603

Berger AN, Demsetz RS, Strahan PE (1999) The consolidation of the financial services industry: causes, consequences, and implications for the future. J Bank Financ 23(2–4):135–194

Berger AN, Deyoung R (1997) Problem loans and cost efficiency in commercial banks. J Bank Financ 21(6):849–870

Berger AN, Udell GF (1998) The economics of small business finance: the roles of private equity and debt markets in the financial growth cycle. J Bank Financ 22(6–8):613–673

Berger AN, Udell GF (2002) Small business credit availability and relationship lending: the importance of bank organisational structure. Econ J 112(477):F32–F53

Berger AN, Udell GF (2006) A more complete conceptual framework for SME finance. J Bank Financ 30(11):2945–2966

Besanko D, Kanatas G (1996) The regulation of bank capital: Do capital standards promote bank safety? J Financ Intermed 5(2):160–183

Beyer A, Cohen DA, Lys TZ, Walther BR (2010) The financial reporting environment: review of the recent literature. J Acc Econ 50(2–3):296–343

Bikker JA (2010) Measuring performance of banks: an assessment. J Appl Bus Econ 11(4):141–159

Bolt W, Tieman AF (2004) Banking competition, risk and regulation. Scand J Econ 106(4):783–804

Boudriga A, BoulilaTaktak N, Jellouli S (2009) Banking supervision and non-performing loans: a cross-country analysis. J Financ Econ Policy 1(4):286–318

Bouzon M, Miguel PAC, Rodriguez CMT (2014) Managing end of life products: a review of the literature on reverse logistics in Brazil. Manag Environ Qual Int J 25(5):564–584. https://doi.org/10.1108/MEQ-04-2013-0027

Article   Google Scholar  

Boyd JH, De Nicolo G (2005) The theory of bank risk taking and competition revisited. J Financ 60(3):1329–1343

Brealey RA, Myers SC, Allen F, Mohanty P (2012) Principles of corporate finance. Tata McGraw-Hill Education

Buston CS (2016) Active risk management and banking stability. J Bank Financ 72:S203–S215

Casu B, Girardone C (2006) Bank competition, concentration and efficiency in the single European market. Manch Sch 74(4):441–468

Charumathi B, Ramesh L (2020) Impact of voluntary disclosure on valuation of firms: evidence from Indian companies. Vision 24(2):194–203

Chen X (2007) Banking deregulation and credit risk: evidence from the EU. J Financ Stab 2(4):356–390

Chen H-J, Lin K-T (2016) How do banks make the trade-offs among risks? The role of corporate governance. J Bank Financ 72(1):S39–S69

Chen M, Wu J, Jeon BN, Wang R (2017) Do foreign banks take more risk? Evidence from emerging economies. J Bank Financ 82(1):20–39

Claessens S, Laeven L (2003) Financial development, property rights, and growth. J Financ 58(6):2401–2436. https://doi.org/10.1046/j.1540-6261.2003.00610.x

Claessens S, Laeven L (2004) What drives bank competition? Some international evidence. J Money Credit Bank 36(3):563–583

Cnaan RA, Moodithaya M, Handy F (2012) Financial inclusion: lessons from rural South India. J Soc Policy 41(1):183–205

Core JE, Holthausen RW, Larcker DF (1999) Corporate governance, chief executive officer compensation, and firm performance. J Financ Econ 51(3):371–406

Dahir AM, Mahat FB, Ali NAB (2018) Funding liquidity risk and bank risk-taking in BRICS countries: an application of system GMM approach. Int J Emerg Mark 13(1):231–248

Dechow P, Ge W, Schrand C (2010) Understanding earnings quality: a review of the proxies, their determinants, and their consequences. J Acc Econ 50(2–3):344–401

Delis MD, Molyneux P, Pasiouras F (2011) Regulations and productivity growth in banking: evidence from transition economies. J Money Credit Bank 43(4):735–764

Demirguc-Kunt A, Laeven L, Levine R (2003) Regulations, market structure, institutions, and the cost of financial intermediation (No. w9890). National Bureau of Economic Research.

Deyoung R, Jang KY (2016) Do banks actively manage their liquidity? J Bank Financ 66:143–161

Ding Y, Cronin B (2011) Popularand/orprestigious? Measures of scholarly esteem. Inf Process Manag 47(1):80–96

Eastburn RW, Sharland A (2017) Risk management and managerial mindset. J Risk Financ 18(1):21–47

Erfani GR, Vasigh B (2018) The impact of the global financial crisis on profitability of the banking industry: a comparative analysis. Economies 6(4):66

Erkens DH, Hung M, Matos P (2012) Corporate governance in the 2007–2008 financial crisis: evidence from financial institutions worldwide. J Corp Finan 18(2):389–411

Fahimnia B, Sarkis J, Davarzani H (2015) Green supply chain management: a review and bibliometric analysis. Int J Prod Econ 162:101–114

Financial Stability Report (2019) Financial stability report (20), December 2019. https://www.rbi.org.in/Scripts/PublicationReportDetails.aspx?UrlPage=&ID=946 Accesses on March 2020

Fink A (2005) Conducting Research Literature Reviews:From the Internet to Paper, 2nd edn. SAGE Publications

Ghosh A (2015) Banking-industry specific and regional economic determinants of non-performing loans: evidence from US states. J Financ Stab 20:93–104. https://doi.org/10.1016/j.jfs.2015.08.004

Gonzalez F (2005) Bank regulation and risk-taking incentives: an international comparison of bank risk. J Bank Financ 29(5):1153–1184

Goyal K, Kumar S (2021) Financial literacy: a systematic review and bibliometric analysis. Int J Consum Stud 45(1):80–105

Grassa R, Moumen N, Hussainey K (2020) Do ownership structures affect risk disclosure in Islamic banks? International evidence. J Financ Rep Acc 19(3):369–391

Haque F, Shahid R (2016) Ownership, risk-taking and performance of banks in emerging economies: evidence from India. J Financ Econ Policy 8(3):282–297

Hellmann TF, Murdock KC, Stiglitz JE (2000) Liberalization, moral hazard in banking, and prudential regulation: Are capital requirements enough? Am Econ Rev 90(1):147–165

Hirshleifer D (2001) Investor psychology and asset pricing. J Financ 56(4):1533–1597

Huang J, You JX, Liu HC, Song MS (2020) Failure mode and effect analysis improvement: a systematic literature review and future research agenda. Reliab Eng Syst Saf 199:106885

Ibáñez Zapata A (2017) Bibliometric analysis of the regulatory compliance function within the banking sector (Doctoral dissertation). Last Accessed on Jan 2021 https://riunet.upv.es/bitstream/handle/10251/85952/Bibliometric%20analysis_AIZ_v4.pdf?sequence=1

Ibrahim MS (2010) Performance evaluation of regional rural banks in India. Int Bus Res 3(4):203–211

Jayadev M, Singh H, Kumar P (2017) Small finance banks: challenges. IIMB Manag Rev 29(4):311–325

Jin JY, Kanagaretnam K, Lobo GJ, Mathieu R (2013) Impact of FDICIA internal controls on bank risk taking. J Bank Financ 37(2):614–624

Joshi MK (2020) Financial performance analysis of select Indian Public Sector Banks using Altman’s Z-Score model. SMART J Bus Manag Stud 16(2):74–87

Kanoujiya J, Bhimavarapu VM, Rastogi S (2021) Banks in India: a balancing act between profitability, regulation and NPA. Vision, 09722629211034417

Karyani E, Dewo SA, Santoso W, Frensidy B (2020) Risk governance and bank profitability in ASEAN-5: a comparative and empirical study. Int J Emerg Mark 15(5):949–969

Kasman S, Kasman A (2015) Bank competition, concentration and financial stability in the Turkish banking industry. Econ Syst 39(3):502–517

Keeley MC (1990) Deposit insurance, risk, and market power in banking. Am Econ Rev 1:1183–1200

Khaddafi M, Heikal M, Nandari A (2017) Analysis Z-score to predict bankruptcy in banks listed in indonesia stock exchange. Int J Econ Financ Issues 7(3):326–330

Khanna T, Yafeh Y (2007) Business groups in emerging markets: Paragons or parasites? J Econ Lit 45(2):331–372

King RG, Levine R (1993) Finance and growth: schumpeter might be right. Q J Econ 108(3):717–737

Kiran KP, Jones TM (2016) Effect of non performing assets on the profitability of banks–a selective study. Int J Bus Gen Manag 5(2):53–60

Klomp J, De Haan J (2015) Banking risk and regulation: Does one size fit all? J Bank Financ 36(12):3197–3212

Koehn M, Santomero AM (1980) Regulation of bank capital and portfolio risk. J Financ 35(5):1235–1244

Köhler M (2015) Which banks are more risky? The impact of business models on bank stability. J Financ Stab 16(1):195–212

Kothari SP (2001) Capital markets research in accounting. J Account Econ 31(1–3):105–231

Kumar S, Goyal N (2015) Behavioural biases in investment decision making – a systematic literature review. Qual Res Financ Mark 7(1):88–108

Kumar S, Kamble S, Roy MH (2020) Twenty-five years of Benchmarking: an International Journal (BIJ): a bibliometric overview. Benchmarking Int J 27(2):760–780. https://doi.org/10.1108/BIJ-07-2019-0314

Kumar S, Sureka R, Colombage S (2020) Capital structure of SMEs: a systematic literature review and bibliometric analysis. Manag Rev Q 70(4):535–565. https://doi.org/10.1007/s11301-019-00175-4

Kwan SH, Laderman ES (1999) On the portfolio effects of financial convergence-a review of the literature. Econ Rev 2:18–31

Lado AA, Boyd NG, Hanlon SC (1997) Competition, cooperation, and the search for economic rents: a syncretic model. Acad Manag Rev 22(1):110–141

Laeven L, Majnoni G (2003) Loan loss provisioning and economic slowdowns: Too much, too late? J Financ Intermed 12(2):178–197

Laeven L, Ratnovski L, Tong H (2016) Bank size, capital, and systemic risk: Some international evidence. J Bank Finance 69(1):S25–S34

Lee C-C, Hsieh M-F (2013) The impact of bank capital on profitability and risk in Asian banking. J Int Money Financ 32(1):251–281

Leech D, Leahy J (1991) Ownership structure, control type classifications and the performance of large British companies. Econ J 101(409):1418–1437

Levine R (1997) Financial development and economic growth: views and agenda. J Econ Lit 35(2):688–726

Lim CY, Woods M, Humphrey C, Seow JL (2017) The paradoxes of risk management in the banking sector. Br Acc Rev 49(1):75–90

Lo AW (2009) Regulatory reform in the wake of the financial crisis of 2007–2008. J Financ Econ Policy 1(1):4–43

Louzis DP, Vouldis AT, Metaxas VL (2012) Macroeconomic and bank-specific determinants of non-performing loans in Greece: a comparative study of mortgage, business and consumer loan portfolios. J Bank Financ 36(4):1012–1027

Maddaloni A, Peydró J-L (2011) Bank risk-taking, securitization, supervision, and low interest rates: evidence from the Euro-area and the U.S. lending standards. Rev Financ Stud 24(6):2121–2165. https://doi.org/10.1093/rfs/hhr015

Maji SG, De UK (2015) Regulatory capital and risk of Indian banks: a simultaneous equation approach. J Financ Econ Policy 7(2):140–156

Maji SG, Hazarika P (2018) Capital regulation, competition and risk-taking behavior of Indian banks in a simultaneous approach. Manag Financ 44(4):459–477

Messai AS, Jouini F (2013) Micro and macro determinants of non-performing loans. Int J Econ Financ Issues 3(4):852–860

Mitra S, Karathanasopoulos A, Sermpinis G, Dunis C, Hood J (2015) Operational risk: emerging markets, sectors and measurement. Eur J Oper Res 241(1):122–132

Mohsni S, Otchere I (2018) Does regulatory regime matter for bank risk-taking? A comparative analysis of US and Canada, d/Seas Working Papers-ISSN 2611-0172 1(1):28–28

Nguyen TPT, Nghiem SH (2015) The interrelationships among default risk, capital ratio and efficiency: evidence from Indian banks. Manag Financ 41(5):507–525

Niinimäki J-P (2004) The effects of competition on banks’ risk taking. J Econ 81(3):199–222

Page L, Brin S, Motwani R, Winograd T (1999) The PageRank citation ranking: bringing order to the web. Stanford InfoLab

Pakravan K (2014) Bank capital: the case against Basel. J Financ Regul Compl 22(3):208–218

Palacios-Callender M, Roberts SA, Roth-Berghofer T (2016) Evaluating patterns of national and international collaboration in Cuban science using bibliometric tools. J Doc 72(2):362–390. https://doi.org/10.1108/JD-11-2014-0164

Pinto G, Rastogi S, Kadam S, Sharma A (2019) Bibliometric study on dividend policy. Qual Res Financ Mark 12(1):72–95

Polizzi S, Scannella E (2020) An empirical investigation into market risk disclosure: Is there room to improve for Italian banks? J Financ Regul Compl 28(3):465–483

Prasad P, Narayanasamy S, Paul S, Chattopadhyay S, Saravanan P (2019) Review of literature on working capital management and future research agenda. J Econ Surv 33(3):827–861

Rahman MM, Zheng C, Ashraf BN, Rahman MM (2018) Capital requirements, the cost of financial intermediation and bank risk-taking: empirical evidence from Bangladesh. Res Int Bus Financ 44(1):488–503

Rajan RG (1994) Why bank credit policies fluctuate: a theory and some evidence. Q J Econ 109(2):399–441

Rastogi S, Gupte R, Meenakshi R (2021) A holistic perspective on bank performance using regulation, profitability, and risk-taking with a view on ownership concentration. J Risk Financ Manag 14(3):111

Rastogi S, Kanoujiya J (2022) Does transparency and disclosure (T&D) improve the performance of banks in India? Int J Product Perform Manag. https://doi.org/10.1108/IJPPM-10-2021-0613

Rastogi S, Ragabiruntha E (2018) Financial inclusion and socioeconomic development: gaps and solution. Int J Soc Econ 45(7):1122–1140

RBI (2001) Prudential Norms on income recognition, asset classification, and provisioning -pertaining to advances. Accessed on Apr 2020. https://rbidocs.rbi.org.in/rdocs/notification/PDFs/23068.pdf

Reddy S (2018) Announcement of payment banks and stock performance of commercial banks in India. J Internet Bank Commer 23(1):1–12

Repullo R (2004) Capital requirements, market power, and risk-taking in banking. J Financ Intermed 13(2):156–182

Rowley J, Slack F (2004) Conducting a literature review. Manag Res News 27(6):31–39. https://doi.org/10.1108/01409170410784185

Salas V, Saurina J (2003) Deregulation, market power and risk behaviour in Spanish banks. Eur Econ Rev 47(6):1061–1075

Samitas A, Polyzos S (2015) To Basel or not to Basel? Banking crises and contagion. Journal of Financial Regulation and Compliance 23(3):298–318

Sarmiento M, Galán JE (2017) The influence of risk-taking on bank efficiency: evidence from Colombia. Emerg Mark Rev 32:52–73. https://doi.org/10.1016/j.ememar.2017.05.007

Schaeck K, Cihak M, Wolfe S (2009) Are competitive banking systems more stable? J Money Credit Bank 41(4):711–734

Schwerter S (2011) Basel III’s ability to mitigate systemic risk. J Financ Regul Compl 19(4):337–354

Sen S, Sen RL (2014) Impact of NPAs on bank profitability: an empirical study. In: Ray N, Chakraborty K (eds) Handbook of research on strategic business infrastructure development and contemporary issues in finance. IGI Global, pp 124–134. https://doi.org/10.4018/978-1-4666-5154-8.ch010

Chapter   Google Scholar  

Shajahan K (1998) Non-performing assets of banks: Have they really declined? And on whose account? Econ Pol Wkly 33(12):671–674

Sharifi S, Haldar A, Rao SN (2016) Relationship between operational risk management, size, and ownership of Indian banks. Manag Financ 42(10):930–942

Sharma A, Theresa L, Mhatre J, Sajid M (2019) Application of altman Z-Score to RBI defaulters: Indian case. Asian J Res Bus Econ Manag 9(4):1–11

Shehzad CT, De Haan J (2015) Supervisory powers and bank risk taking. J Int Finan Markets Inst Money 39(1):15–24

Shen L, Xiong B, Hu J (2017) Research status, hotspotsandtrends forinformation behavior in China using bibliometric and co-word analysis. J Doc 73(4):618–633

Shleifer A, Vishny RW (1997) A survey of corporate governance. J Financ 52(2):737–783

Singh HP, Kumar S (2014) Working capital management: a literature review and research agenda. Qual Res Financ Mark 6(2):173–197

Tabak BM, Fazio DM, Cajueiro DO (2013) Systemically important banks and financial stability: the case of Latin America. J Bank Financ 37(10):3855–3866

Tahamtan I, SafipourAfshar A, Ahamdzadeh K (2016) Factors affecting number of citations: a comprehensive review of the literature. Scientometrics 107(3):1195–1225

Thakor AV (2018) Post-crisis regulatory reform in banking: Address insolvency risk, not illiquidity! J Financ Stab 37(1):107–111

Thomsen S, Pedersen T (2000) Ownership structure and economic performance in the largest European companies. Strategic Manag J 21(6):689–705

Tranfield D, Denyer D, Smart P (2003) Towards a methodology for developing evidence-informed management knowledge by means of systematic review. Br J Manag 14(3):207–222

Triki T, Kouki I, Dhaou MB, Calice P (2017) Bank regulation and efficiency: What works for Africa? Res Int Bus Financ 39(1):183–205

Tsay M, Shu Z (2011) Journal bibliometric analysis: a case study on the journal of documentation. J Doc 67(5):806–822

Vento GA, La Ganga P (2009) Bank liquidity risk management and supervision: which lessons from recent market turmoil. J Money Invest Bank 10(10):78–125

Wahid ANM (1994) The grameen bank and poverty alleviation in Bangladesh: theory, evidence and limitations. Am J Econ Sociol 53(1):1–15

Xiao Y, Watson M (2019) Guidance on conducting a systematic literature review. J Plan Educ Res 39(1):93–112

Xu X, Chen X, Jia F, Brown S, Gong Y, Xu Y (2018) Supply chain finance: a systematic literature review and bibliometric analysis. Int J Prod Econ 204:160–173

Yadav M (2011) Impact of non performing assets on profitability and productivity of public sector banks in India. AFBE J 4(1):232–239

Yong-Hak J (2013), Web of Science, Thomson Reuters

Zheng C, Rahman MM, Begum M, Ashraf BN (2017) Capital regulation, the cost of financial intermediation and bank profitability: evidence from Bangladesh. J Risk Financ Manag 10(2):9

Download references

Acknowledgements

Not Applicable.

Author information

Authors and affiliations.

Symbiosis Institute of Business Management, Symbiosis International (Deemed University), Pune, India

Shailesh Rastogi, Arpita Sharma & Venkata Mrudula Bhimavarapu

SIES School of Business Studies, Navi Mumbai, India

Geetanjali Pinto

School of Commerce and Management, D Y Patil International University, Akurdi, Pune, India

Venkata Mrudula Bhimavarapu

You can also search for this author in PubMed   Google Scholar

Contributions

‘SR’ performed Abstract, Introduction, and Data methodology sections and was the major contributor; ‘AS’ performed Bibliometric and Network analysis and conceptual framework; ‘GP’ performed citation analysis and discussion section; ‘VMB’ collated data from the database and concluded the article. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Venkata Mrudula Bhimavarapu .

Ethics declarations

Ethics approval and consent to participate, consent for publication, competing interests.

The authors declare that they have no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Rastogi, S., Sharma, A., Pinto, G. et al. A literature review of risk, regulation, and profitability of banks using a scientometric study. Futur Bus J 8 , 28 (2022). https://doi.org/10.1186/s43093-022-00146-4

Download citation

Received : 11 March 2022

Accepted : 16 August 2022

Published : 03 September 2022

DOI : https://doi.org/10.1186/s43093-022-00146-4

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Bank performance
• Profitability
• Bibliometric analysis
• Scientometric analysis

Information

• Author Services

Initiatives

You are accessing a machine-readable page. In order to be human-readable, please install an RSS reader.

All articles published by MDPI are made immediately available worldwide under an open access license. No special permission is required to reuse all or part of the article published by MDPI, including figures and tables. For articles published under an open access Creative Common CC BY license, any part of the article may be reused without permission provided that the original article is clearly cited. For more information, please refer to https://www.mdpi.com/openaccess .

Feature papers represent the most advanced research with significant potential for high impact in the field. A Feature Paper should be a substantial original Article that involves several techniques or approaches, provides an outlook for future research directions and describes possible research applications.

Feature papers are submitted upon individual invitation or recommendation by the scientific editors and must receive positive feedback from the reviewers.

Editor’s Choice articles are based on recommendations by the scientific editors of MDPI journals from around the world. Editors select a small number of articles recently published in the journal that they believe will be particularly interesting to readers, or important in the respective research area. The aim is to provide a snapshot of some of the most exciting work published in the various research areas of the journal.

Original Submission Date Received: .

• Active Journals
• Find a Journal
• Proceedings Series
• For Authors
• For Reviewers
• For Editors
• For Librarians
• For Publishers
• For Societies
• For Conference Organizers
• Open Access Policy
• Institutional Open Access Program
• Special Issues Guidelines
• Editorial Process
• Research and Publication Ethics
• Article Processing Charges
• Testimonials
• Preprints.org
• SciProfiles
• Encyclopedia

Article Menu

• Subscribe SciFeed
• Recommended Articles
• Google Scholar
• on Google Scholar
• Table of Contents

Find support for a specific problem in the support section of our website.

Please let us know what you think of our products and services.

Visit our dedicated information section to learn more about MDPI.

JSmol Viewer

Machine learning in banking risk management: a literature review.

1. Introduction

2. theoretical background, 2.1. risk management at banks, 2.2. machine learning, 3. materials and methods, 3.1. credit risk, 3.2. market risk, 3.3. liquidity risk, 3.4. operational risk, 4. discussion, 5. conclusions, author contributions, conflicts of interest.

• Ala’raj, Maher, and Maysam F. Abbod. 2016a. A New Hybrid Ensemble Credit Scoring Model Based on Classifiers Consensus System Approach. Expert Systems with Applications 64: 36–55. [ Google Scholar ] [ CrossRef ]
• Ala’Raj, Maher, and Maysam F. Abbod. 2016b. Classifiers Consensus System Approach for Credit Scoring. Knowledge-Based Systems 104: 89–105. [ Google Scholar ] [ CrossRef ]
• Apostolik, Richard, Christopher Donohue, Peter Went, and Global Association of Risk Professionals. 2009. Foundations of Banking Risk: An Overview of Banking, Banking Risks, and Risk-Based Banking Regulation . New York: John Wiley. [ Google Scholar ]
• Arezzo, Maria, and Giuseppina Guagnano. 2018. Response-Based Sampling for Binary Choice Models with Sample Selection. Econometrics 6: 12. [ Google Scholar ] [ CrossRef ]
• Awad, Mariette, and Rahul Khanna. 2015. Machine Learning in Action: Examples. Efficient Learning Machines . [ Google Scholar ] [ CrossRef ]
• Aziz, Saqib, and Michael M. Dowling. 2018. AI and Machine Learning for Risk Management. SSRN Electronic Journal . [ Google Scholar ] [ CrossRef ]
• Bacham, Dinesh, and Janet Zhao. 2017. Machine Learning: Challenges and Opportunities in Credit Risk Modeling. Available online: https://www.moodysanalytics.com/risk-perspectives-magazine/managing-disruption/spotlight/machine-learning-challenges-lessons-and-opportunities-in-credit-risk-modeling (accessed on 2 April 2018).
• Barboza, Flavio, Herbert Kimura, and Edward Altman. 2017. Machine learning models and bankruptcy prediction. Expert Systems with Applications 83: 405–17. [ Google Scholar ] [ CrossRef ]
• Basel Committee on Banking Supervision. 2005a. Guidance on Paragraph 468 of the Framework Document . Basel: Bank for International Settlements. [ Google Scholar ]
• Basel Committee on Banking Supervision. 2005b. An Explanatory Note on the Basel II IRB Risk Weight Functions . Basel: Bank for International Settlements. [ Google Scholar ]
• Basel Committee on Banking Supervision. 2006. Minimum Capital Requirements for Market Risk . Basel: Bank for International Settlements. [ Google Scholar ]
• Basel Committee on Banking Supervision. 2008. Principles for Sound Liquidity Risk Management and Supervision . Basel: Bank for International Settlements. [ Google Scholar ]
• Basel Committee on Banking Supervision. 2011. Principles for the Sound Management of Operational Risk . Basel: Bank for International Settlements, pp. 1–27. [ Google Scholar ]
• Bastos, João A. 2014. Ensemble Predictions of Recovery Rates. Journal of Financial Services Research 46: 177–93. [ Google Scholar ] [ CrossRef ]
• Bauguess, Scott W. 2015. The Hope and Limitations of Machine Learning in Market Risk Assessment . Washington, DC: U.S. Securities and Exchange Commission. [ Google Scholar ]
• Bellotti, Tony, and Jonathan Crook. 2009. Support Vector Machines for Credit Scoring and Discovery of Significant Features. Expert Systems with Applications . [ Google Scholar ] [ CrossRef ]
• Blom, Tineke. 2015. Top down Stress Testing: An Application of Adaptive Lasso to Forecasting Credit Loss Rates. Master’s Thesis, Faculty of Science, Hongkong, China. [ Google Scholar ]
• Brown, Iain, and Christophe Mues. 2012. An experimental comparison of classification algorithms for imbalanced credit scoring data sets. Expert Systems with Applications 39: 3446–53. [ Google Scholar ] [ CrossRef ]
• Cao, Jie, Hongke Lu, Weiwei Wang, and Jian Wang. 2013. A Loan Default Discrimination Model Using Cost-Sensitive Support Vector Machine Improved by PSO. Information Technology and Management 14: 193–204. [ Google Scholar ] [ CrossRef ]
• Chan-Lau, Jorge. 2017. Lasso Regressions and Forecasting Models in Applied Stress Testing. IMF Working Papers 17: 1. [ Google Scholar ] [ CrossRef ]
• Chen, Ning, Bernardete Ribeiro, and An Chen. 2016. Financial Credit Risk Assessment: A Recent Review. Artificial Intelligence Review 45: 1–23. [ Google Scholar ] [ CrossRef ]
• Dal Pozzolo, Andrea. 2015. Adaptive Machine Learning for Credit Card Fraud Detection. Unpublished doctoral dissertation, Université libre de Bruxelles, Faculté des Sciences—Informatique, Bruxelles. [ Google Scholar ]
• Deloitte University Press. 2017. Global Risk Management Survey , 10th ed. Deloitte University Press: Available online: https://www2.deloitte.com/tr/en/pages/risk/articles/global-risk-management-survey-10th-ed.html (accessed on 4 October 2018).
• Financial Stability Board. 2017. Artificial Intelligence and Machine Learning in Financial Services. Market Developments and Financial Stability Implications. Financial Stability Board . November 1. Available online: http://www.fsb.org/2017/11/artificial-intelligence-and-machine-learning-in-financial-service/ (accessed on 2 July 2018).
• Galindo, Jorge, and Pablo Tamayo. 2000. Credit Risk Assessment Using Statistical and Machine Learning: Basic Methodology and Risk Modeling Applications. Computational Economics 15: 107–43. [ Google Scholar ] [ CrossRef ]
• Gotoh, Jun-ya, Akiko Takeda, and Rei Yamamoto. 2014. Interaction between financial risk measures and machine learning methods. Computational Management Science 11: 365–402. [ Google Scholar ] [ CrossRef ]
• Greene, William H. 1992. A Statistical Model for Credit Scoring. NYU Working Paper No. EC-92-29. Available online: https://ssrn.com/abstract=1867088 (accessed on 8 April 1992).
• Guegan, Dominique, Peter Addo, and Bertrand Hassani. 2018. Credit risk analysis using machine and deep learning models. Risks 6: 38. [ Google Scholar ]
• Hamori, Shigeyuki, Minami Kawai, Takahiro Kume, Yuji Murakami, and Chikara Watanabe. 2018. Ensemble Learning or Deep Learning? Application to Default Risk Analysis. Journal of Risk and Financial Management 11: 12. [ Google Scholar ] [ CrossRef ]
• Hand, David J., and William E. Henley. 1997. Statistical Classification Methods in Consumer Credit Scoring: A Review. Journal of the Royal Statistical Society Series A: Statistics in Society . [ Google Scholar ] [ CrossRef ]
• Harris, Terry. 2013. Quantitative credit risk assessment using support vector machines: Broad versus Narrow default definitions. Expert Systems with Applications 40: 4404–13. [ Google Scholar ] [ CrossRef ]
• Helbekkmo, Hans, Alok Kshirsagar, Andreas Schlosser, Francesco Selandari, Uwe Stegemann, and Joyce Vorholt. 2013. Enterprise Risk Management—Shaping the Risk Revolution . New York: McKinsey & Co., Available online: www.rmahq.org (accessed on 18 June 2018).
• Huang, Cheng Lung, Mu Chen Chen, and Chieh Jen Wang. 2007. Credit Scoring with a Data Mining Approach Based on Support Vector Machines. Expert Systems with Applications 33: 847–56. [ Google Scholar ] [ CrossRef ]
• Hull, John. 2012. Risk Management and Financial Institutions . New York: John Wiley and Sons, vol. 733. [ Google Scholar ]
• Islam, Tushith, Christos Vasilopoulos, and Erik Pruyt. 2013. Stress—Testing Banks under Deep Uncertainty. Paper presented at the 31st International Conference of the System Dynamics Society, Cambridge, MA, USA, July 21–25; Available online: http://repository.tudelft.nl/islandora/object/uuid:c162de43-4235-4d29-8eed-3246df87e119?collection=education (accessed on 17 July 2018).
• Jacobs, Michael, Jr. 2018. The validation of machine-learning models for the stress testing of credit risk. Journal of Risk Management in Financial Institutions 11: 218–43. [ Google Scholar ]
• Jorion, Philippe. 2007. Value at Risk: The New Benchmark for Managing Financial Risk . New York: McGraw-Hill. [ Google Scholar ]
• Kanevski, Mikhail F., and Vadim Timonin. 2010. Machine learning analysis and modeling of interest rate curves. Paper presented at the 18th European Symposium on Artificial Neural Networks ESANN, Bruges, Belgium, April 28–30; Available online: https://www.elen.ucl.ac.be/Proceedings/esann/esannpdf/es2010-17.pdf (accessed on 18 June 2018).
• Kannan, Somasundaram, and K. Somasundaram. 2017. Autoregressive-Based Outlier Algorithm to Detect Money Laundering Activities. Journal of Money Laundering Control 20: 190–202. [ Google Scholar ] [ CrossRef ]
• Keramati, Abbas, and Niloofar Yousefi. 2011. A proposed classification of data mining techniques in credit scoring. Paper presented at the 2011 International Conference of Industrial Engineering and Operations Management, Kuala Lumpur, Malaysia, January 22–24. [ Google Scholar ]
• Khandani, Amir E., Adlar J. Kim, and Andrew W. Lo. 2010. Consumer credit-risk models via machine-learning algorithms. Journal of Banking & Finance 34: 2767–87. [ Google Scholar ]
• Khrestina, Marina Pavlovna, Dmitry Ivanovich Dorofeev, Polina Andreevna Kachurina, Timur Rinatovich Usubaliev, and Aleksey Sergeevich Dobrotvorskiy. 2017. Development of Algorithms for Searching, Analyzing and Detecting Fraudulent Activities in the Financial Sphere. European Research Studies Journal 20: 484–98. [ Google Scholar ]
• Lai, Kin Keung, Lean Yu, Ligang Zhou, and Shouyang Wang. 2006. Credit risk evaluation with least square support vector machine. In International Conference on Rough Sets and Knowledge Technology . Berlin/Heidelberg: Springer, pp. 490–95. [ Google Scholar ]
• Lessmann, Stefan, Bart Baesens, Hsin Vonn Seow, and Lyn C. Thomas. 2015. Benchmarking State-of-the-Art Classification Algorithms for Credit Scoring: An Update of Research. European Journal of Operational Research 247: 124–36. [ Google Scholar ] [ CrossRef ]
• Mahdavi-Damghani, Babak, and Stephen Roberts. 2017. A Proposed Risk Modeling Shift from the Approach of Stochastic Differential Equation towards Machine Learning Clustering: Illustration with the Concepts of Anticipative and Responsible VaR. SSRN Electronic Journal . [ Google Scholar ] [ CrossRef ]
• Mainelli, Michael, and Mark Yeandle. 2006. Best execution compliance: New techniques for managing compliance risk. The Journal of Risk Finance 7: 301–12. [ Google Scholar ] [ CrossRef ]
• Malhotra, Rashmi, and D. K. Malhotra. 2003. Evaluating Consumer Loans Using Neural Networks. Omega 31: 83–96. [ Google Scholar ] [ CrossRef ]
• MetricStream. 2018. The Chief Risk Officer’s Role in 2018 and Beyond Managing the Challenges and Opportunities of a Digital Era New Roles of the CRO. Available online: https://www.metricstream.com/insights/chief-risk-officer-role-2018.htm (accessed on 23 June 2018).
• Monfared, Soheil Almasi, and David Enke. 2014. Volatility Forecasting Using a Hybrid GJR-GARCH Neural Network Model. Procedia Computer Science 36: 246–53. [ Google Scholar ] [ CrossRef ]
• Ngai, Eric W. T., Yong Hu, Yiu Hing Wong, Yijun Chen, and Xin Sun. 2011. The Application of Data Mining Techniques in Financial Fraud Detection: A Classification Framework and an Academic Review of Literature. Decision Support Systems 50: 569. [ Google Scholar ] [ CrossRef ]
• Oliver Wyman. 2017. Next Generation Risk Management. Available online: https://www.oliverwyman.com/content/dam/oliver-wyman/v2/publications/2017/aug/Next_Generation_Risk_Management_Targeting_A-Technology_Dividend.pdf (accessed on 1 May 2018).
• Pavlenko, Tatjana, and Oleksandr Chernyak. 2009. Bayesian Networks for Modeling and Assessment of Credit Concentration Risks. International Statistical Conference Prague. Available online: http://www.czso.cz/conference2009/proceedings/data/methods/pavlenko_paper.pdf (accessed on 21 July 2018).
• Peters, Gareth, Pavel V. Shevchenko, Ruben Cohen, and Diane Maurice. 2017. Statistical Machine Learning Analysis of Cyber Risk Data: Event Case Studies. Available online: https://ssrn.com/abstract=3073704 (accessed on 18 June 2018).
• Proofpoint. 2010. MLX Whitepaper “Machine Learning to Beat Spam Today and Tomorrow”. Available online: https://www.excelmicro.com/datasheets/Proofpoint-White-Paper-MLX-Technology.pdf (accessed on 2 May 2018).
• Pun, Joseph, and Yuri Lawryshyn. 2012. Improving credit card fraud detection using a meta-classification strategy. International Journal of Computer Applications 56: 41–46. [ Google Scholar ] [ CrossRef ]
• Raei, Reza, Mahdi Saeidi Kousha, Saeid Fallahpour, and Mohammad Fadaeinejad. 2016. A Hybrid Model for Estimating the Probability of Default of Corporate Customers. Iranian Journal of Management Studies 9: 651–73. [ Google Scholar ]
• Ray, Sunil. 2015. Understanding Support Vector Machine Algorithm from Examples (along with Code). Available online: https://www.analyticsvidhya.com/blog/2017/09/understaing-support-vector-machine-example-code/ (accessed on 16 August 2018).
• Sala, Jordi Petchamé. 2011. Liquidity Risk Modeling Using Artificial Neural Network. Master’s thesis, Universitat Politècnica de Catalunya, Barcelona, Spain. [ Google Scholar ]
• Saunders, Anthony, Marcia Millon Cornett, and Patricia Anne McGraw. 2006. Financial Institutions Management: A Risk Management Approach . New York: McGraw-Hill. [ Google Scholar ]
• Shalev-Shwartz, Shai, and Shai Ben-David. 2014. Understanding Machine Learning: From Theory to Algorithms . Cambridge: Cambridge University Press. [ Google Scholar ] [ CrossRef ]
• Sharma, Shashank, and Arjun Roy Choudhury. 2016. Fraud Analytics: A Survey on Bank Fraud Prediction Using Unsupervised Learning Based Approach. International Journal of Innovation in Engineering Research and Technology 3: 1–9. [ Google Scholar ]
• Sudjianto, Agus, Sheela Nair, Ming Yuan, Aijun Zhang, Daniel Kern, and Fernando Cela-Díaz. 2010. Statistical Methods for Fighting Financial Crimes. Technometrics 52: 5–19. [ Google Scholar ] [ CrossRef ]
• Tavana, Madjid, Amir Reza Abtahi, Debora Di Caprio, and Maryam Poortarigh. 2018. An Artificial Neural Network and Bayesian Network Model for Liquidity Risk Assessment in Banking. Neurocomputing 275: 2525–54. [ Google Scholar ] [ CrossRef ]
• Vaidya, Avanti H., and Sudhir W. Mohod. 2014. Internet Banking Fraud Detection using HMM and BLAST-SSAHA Hybridization. International Journal of Science and Research (IJSR) 3: 574–9. [ Google Scholar ]
• Van Gestel, Ir Tony, Bart Baesens, Ir Joao Garcia, and Peter Van Dijcke. 2003. A support vector machine approach to credit scoring. In Forum Financier—Revue Bancaire Et Financiaire Bank En Financiewezen . Bruxelles: Larcier, pp. 73–82. Available online: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.93.6492&rep=rep1&type=pdf (accessed on 7 July 2018).
• Van Liebergen, Bart. 2017. Machine Learning: A Revolution in Risk Management and Compliance? Journal of Financial Transformation 45: 60–67. [ Google Scholar ]
• Van-Sang, Ha, and Ha-Nam Nguyen. 2016. Credit Scoring with a Feature Selection Approach Based Deep Learning. In MATEC Web of Conferences . Les Ulis: EDP Sciences, volume 54, p. 05004. Available online: https://www.matec-conferences.org/articles/matecconf/abs/2016/17/matecconf_mimt2016_05004/matecconf_mimt2016_05004.html (accessed on 19 July 2018).
• Villalobos, Miguel Agustín, and Eliud Silva. 2017. A Statistical and Machine Learning Model to Detect Money Laundering: An Application. Available online: http://hddavii.eventos.cimat.mx/sites/hddavii/files/Miguel_Villalobos.pdf (accessed on 21 June 2018).
• Wang, Yongqiao, Shouyang Wang, and Kin Keung Lai. 2005. A New Fuzzy Support Vector Machine to Evaluate Credit Risk. IEEE Transactions on Fuzzy Systems 13: 820–31. [ Google Scholar ] [ CrossRef ]
• Wang, Hong, Qingsong Xu, and Lifeng Zhou. 2015. Large Unbalanced Credit Scoring Using Lasso-Logistic Regression Ensemble. PLoS ONE 10: e0117844. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Wójcicka, Aleksandra. 2017. Neural Networks vs. Discriminant Analysis in the Assessment of Default. Electronic Economy , 339–49. [ Google Scholar ] [ CrossRef ]
• Yang, Zijiang, Wenjie You, and Guoli Ji. 2011. Using partial least squares and support vector machines for bankruptcy prediction. Expert Systems with Applications 38: 8336–42. [ Google Scholar ] [ CrossRef ]
• Yao, Xiao, Jonathan Crook, and Galina Andreeva. 2015. Support vector regression for loss given default modelling. European Journal of Operational Research 240: 528–38. [ Google Scholar ] [ CrossRef ]
• Yao, Xiao, Jonathan Crook, and Galina Andreeva. 2017. Enhancing two-stage modelling methodology for loss given default with support vector machines. European Journal of Operational Research 263: 679–89. [ Google Scholar ] [ CrossRef ]
• Yeh, I. Cheng, and Chehui Lien. 2009. The Comparisons of Data Mining Techniques for the Predictive Accuracy of Probability of Default of Credit Card Clients. Expert Systems with Applications . [ Google Scholar ] [ CrossRef ]
• Yu, Lean, Zebin Yang, and Ling Tang. 2016. A Novel Multistage Deep Belief Network Based Extreme Learning Machine Ensemble Learning Paradigm for Credit Risk Assessment. Flexible Services and Manufacturing Journal 28: 576–92. [ Google Scholar ] [ CrossRef ]
• Zareapoor, Masoumeh, and Pourya Shamsolmoali. 2015. Application of Credit Card Fraud Detection: Based on Bagging Ensemble Classifier. Procedia Computer Science 48: 679–86. [ Google Scholar ] [ CrossRef ]
• Zhang, Wenhao. 2017. Machine Learning Approaches to Predicting Company Bankruptcy. Journal of Financial Risk Management 6: 364–74. [ Google Scholar ] [ CrossRef ]
• Zhang, Heng Guo, Chi Wei Su, Yan Song, Shuqi Qiu, Ran Xiao, and Fei Su. 2017. Calculating Value-at-Risk for High-Dimensional Time Series Using a Nonlinear Random Mapping Model. Economic Modelling 67: 355–67. [ Google Scholar ] [ CrossRef ]
• Zhou, Lifeng, and Hong Wang. 2012. Loan Default Prediction on Large Imbalanced Data Using Random Forests. TELKOMNIKA Indonesian Journal of Electrical Engineering . [ Google Scholar ] [ CrossRef ]

Click here to enlarge figure

Share and Cite

Leo, M.; Sharma, S.; Maddulety, K. Machine Learning in Banking Risk Management: A Literature Review. Risks 2019 , 7 , 29. https://doi.org/10.3390/risks7010029

Leo M, Sharma S, Maddulety K. Machine Learning in Banking Risk Management: A Literature Review. Risks . 2019; 7(1):29. https://doi.org/10.3390/risks7010029

Leo, Martin, Suneel Sharma, and K. Maddulety. 2019. "Machine Learning in Banking Risk Management: A Literature Review" Risks 7, no. 1: 29. https://doi.org/10.3390/risks7010029

Article Metrics

Article access statistics, further information, mdpi initiatives, follow mdpi.

Subscribe to receive issue release notifications and newsletters from MDPI journals

The role of artificial intelligence in developing a banking risk index: an application of Adaptive Neural Network-Based Fuzzy Inference System (ANFIS)

• Open access
• Published: 24 April 2023
• Volume 56 , pages 13873–13895, ( 2023 )

Cite this article

You have full access to this open access article

• Ibrahim Elsiddig Ahmed   ORCID: orcid.org/0000-0002-2656-2023 1 ,
• Riyadh Mehdi   ORCID: orcid.org/0000-0002-0317-9777 2 &
• Elfadil A. Mohamed   ORCID: orcid.org/0000-0001-9281-3815 2  

4316 Accesses

4 Citations

Explore all metrics

Banking risk measurement and management remain one of many challenges for managers and policymakers. This study contributes to the banking literature and practice in two ways by (a) proposing a risk ranking index based on the Mahalanobis Distance (MD) between a multidimensional point representing a bank’s risk measures and the corresponding critical ratios set by the banking authorities and (b) determining the relative importance of a bank’s risk ratios in affecting its financial standing using an Adaptive Neuro-Fuzzy Inference System. In this study, ten financial ratios representing five risk areas were considered, namely: Capital Adequacy, Credit, Liquidity, Earning Quality, and Operational risk. Data from 45 Gulf banks for the period 2016–2020 was used to develop the model. Our findings indicate that a bank is in a sound risk position at the 99%, 95%, and 90% confidence level if its Mahalanobis distance exceeds 4.82, 4.28, and 4.0, respectively. The maximum distance computed for the banks in this study was 9.31; only five out of the forty-five banks were below the 4.82 and one below the 4.28 and 4.0 thresholds at 3.96. Sensitivity analysis of the risks indicated that the Net Interest Margin is the most significant factor in explaining variations in a bank’s risk position, followed by Capital Adequacy Ratio, Common Equity Tier1, and Tier1 Equity in order. The remaining financial ratios: Non-Performing Loans, Equity Leverage, Cost Income Ratio, Loans to Total Assets, and Loans to Deposits have the least influence in the order given; the Provisional Loans Ratio appears to have no influence.

Similar content being viewed by others

Evaluation research on commercial bank counterparty credit risk management based on new intuitionistic fuzzy method

Financial Performance Analysis with the Fuzzy COPRAS and Entropy-COPRAS Approaches

A Study on Risk Factors Impact on the Efficiency of Commercial Banks in China—Based on Fuzzy DEA Model

Explore related subjects.

Artificial Intelligence

Avoid common mistakes on your manuscript.

1 Introduction

Artificial Intelligence (AI) in its various forms of machine learning, natural language processing and robotic process automation—is still in its early stage in terms of financial business applications.

Since the global financial crisis, risk evaluation and management in banks have gained more prominence, and there has been a constant focus on how risks are being detected, measured, reported and managed.

The use of AI by the banking sector is globally expanding as per the Global Association of Risk Professionals (GARP), and analytics leader SAS survey in 2018, the top area of AI use is the automation of manual processes (52%), followed by credit scoring (45%), data cleansing and enhancement (43%), risk grading (37%), model validation (35%), and model calibration (34%). Some newer-emerging AI applications with 20% or higher response rates were regulatory reporting, loan approvals, collections, and loan pricing.

All indications are that AI technology is here to stay and will become an increasingly important tool in risk monitoring, modeling and analytics. Risk professionals will likely have to broaden their abilities, melding domain expertise with highly quantitative and technical skills. Risk management departments may be re-skilled and reshaped, while quantitative and analytical capabilities are applied more comprehensively in more areas of the organization. McKinsey & Co highlighted that risk functions in banks, by 2025, would need to be fundamentally different from what they are today. The emergence of financial technology (FinTech) has seen a surge in interest and comment with regard to how AI might be developed and incorporated to better serve more traditional financial services and operations (Zhang and Kedmey 2018 ).

Himeur et al. ( 2022 ) explained that AI is facing many practical and operational challenges, ranging from the need for a basic familiarity with these systems to finding the necessary technical talent to managing the quality of big-data inputs and being able to understand and explain how AI models produce their outputs.

The review of the literature has shown that the application of AI (machine learning) in the management of banking risks such as credit risk, market risk, operational risk and liquidity risk has been explored by studies such as; Van Liebergen 2017 ; Deloitte University Press 2017 ; MetricStream 2018 ; Oliver Wyman 2017 . However, all previous studies concentrated on individual components of banking risk rather than the overall ranking of all the risk measures as investigated by this study. A large number of areas remain in bank risk management that could significantly benefit from the study of how AI can be applied to properly address banking risk. Another motivation for this research is the need to consider all types of risk measures because of the increasing credit default and the negative impact of Covid 19 on banking performance and operations. In addition, most of the previous studies concentrated on the measurement of credit risk as the main source of default, but this study will be more comprehensive by including all other possible types and sources of risks such as operational, liquidity, market, and credit risks. This study will add great value to both limited literature and body of knowledge as well as to the banking practice through the ranking of risks variables.

A number of banking risk problems can be solved through the use of machine learning (ML). Measurement of risk and the analysis of key factors, including the study of the interconnections between the factors, can be achieved through the use of ML. For the purposes of developing a comprehensive risk index that considers all the banking risk areas and variables, Artificial Neural Networks (ANN), a genetic algorithm can be applied. ANN can be used in the approximation of the general risk trend and determination of the most influential risk variables. Therefore, the main objective of this study is to apply the most useful AI tool, Adaptive Neural Network-Based Fuzzy Inference System (ANFIS), to accurately compute, analyze, and develop an overall risk index of the Gulf banking sector. Furthermore, the study intends to explore the role of artificial intelligence in developing an index that ranks all the banking risk variables classified into their common categories as per their importance and influence on the banking operations and strategies.

The paper is organized as follows. Section  2 discusses the related literature while Sect.  3 covers the data and method. Section  4 presents the results and discussion. Finally, Sect.  5 provides a summary, conclusions and recommendations for future work.

2 Related literature work

Models for financial distress forecasting of banks are being increasingly used as important tools to identify early warning signals for the whole banking system.

Numerous methods have been proposed for the analysis of banking risk problems, ranging from traditional methods to modern AI and machine learning techniques. To mention a few that include Neural Networks, Fuzzy Logic, Swarm Intelligence, Artificial Neural Networks (ANN) and Genetic Algorithms.

Mishraz et al. ( 2021 ) proposed a study that attempted to predict the financial distress of commercial banks by developing a bankruptcy prediction model for banks in India. In their study, they have developed three models using Logistic, Linear Discriminant Analysis (LDA) and Artificial Neural Network (ANN). The findings of the study indicate that the logistic and LDA models exhibit similar prediction accuracy. The results of the ANN prediction model exhibit better prediction accuracy. These findings clearly show the superiority of the AI techniques compared with the traditional ones.

In the same direction, another study that compared models developed using traditional methods and machine learning one to predict corporate bankruptcy is discussed in (Marso and EL Merouani 2020 ). In their study, they developed two models. The first model is multiple discriminant analysis (MDA), and the second one is an ANN trained by backpropagation (BPNN). The experimental results show that ANN models, on average, are approximately 10% more accurate in relation to MDA in different periods.

Another interesting survey on the applications of non-traditional techniques, such as AI on the forecasting of bankruptcy is given in (Aly et al. 2022 ). In this survey, they have compared the performance and accuracy of both traditional and non-traditional methods. Their result indicated that no single methodology can be the general superior tool for every bankruptcy situation. They have indicated that each methodology exhibits either strong or weak performance based on the situation and the used datasets.

A recent study that investigates corporate failure is discussed in (Jabeur et al. 2021 ). They have proposed a novel approach to classify categorical data using gradient-boosting decision trees, namely, CatBoost. Compared with eight reference machine learning models, their model demonstrates an effective improvement in the power of classification performance.

Researchers have studied the factors that have an impact on corporate failure. For example, González et al. ( 2021 ) proposed a Bayesian one-stage approach to estimate the effect of inefficiency on the time to failure (bankruptcy) of U.S. commercial banks. The result obtained show that their proposal outperforms the two-stage maximum likelihood approach traditionally used in the literature. In addition, empirical evidence suggests that the inefficiency of U.S. commercial banks during the global financial crisis in 2008–2009 played a statistically and economically significant role in determining the time to failure.

The neuro-fuzzy system, one of the AI techniques, has vast applications in business which include stock market prediction (SU and Cheng 2016 ; Vlasenko et al. 2019 ; Mohamed et al. 2021 ).

Rajab and Sharma ( 2018 ) presented a review of the applications of Neuro-Fuzzy Systems (NFS) in business on the basis of the research articles issued in various reputed international journals and conferences during 2005–2015. The result of this survey indicated that finance is among the viable application of Neuro-Fuzzy systems.

An interesting study conducted by Brlečić Valčić ( 2021 ) that used Adaptive Neuro-Fuzzy Inference System (ANFIS) approach, to produce models for investigating the effect of cost structure on sustainable development and business persistence with respect to selected financial indicators. Before building the model using clustering method, Brlečić Valčić ( 2021 ) established a link between the business performance indicators and parameters that increase or decrease a particular cost component.

Other AI methods used for the analysis of corporate financial distress include Interval-valued Fuzzy cognitive maps (IVFCMs) to model additional uncertainty in decision-making tasks with complex causal relationships (Hajek and Prochazka 2018 ). Hajek and Prochazka ( 2018 ) introduced a novel IVFCM with real-coded genetic learning. They have demonstrated that the proposed method is effective for predicting corporate financial distress based on causally connected financial concepts. In comparison with other methods, the proposed one outperforms Fuzzy cognitive maps, fuzzy grey cognitive maps and adaptive neuro-fuzzy systems in terms of root mean squared error.

Neuro-fuzzy also has been used in predicting the future stock price value. For example, Vlasenko et al. ( 2019 ) proposed a novel ensemble neuro-fuzzy model used to address the limitations and improve the previously successfully applied a five-layer multidimensional Gaussian neuro-fuzzy model and its learning. The suggested solution allows skipping the error-prone hyperparameters selection process and shows better accuracy results in real life financial data.

Another example of the use of Neuro-fuzzy in stock price prediction is given in (Su and Cheng 2016 ). In their study they have proposed a novel ANFIS (Adaptive Neuro Fuzzy Inference System) time series model based on integrated nonlinear feature selection (INFS) method for stock prediction. In their proposed model an integrated nonlinear feature selection method to select the important technical indicators objectively has been suggested. Next the method used ANFIS to build time series model and test forecast performance, then utilized adaptive expectation model to strengthen the forecasting performance. The performance evaluation of the proposed model using the TAIEX and HSI stock market transaction data outperformed the listing models in accuracy, profit evaluation and statistical test.

The surveyed literature indicates that there are many built models that aim to predict corporate bankruptcy. These models ranged from traditional one to the recent AI and machine learning. The survey results clearly indicate that the performance of the AI and Machine learning in bankruptcy prediction are better than the traditional one. Kar et al ( 2022 ) state that the Lack of an AI adoption strategy and lack of AI talent are the most significant barriers to AI applications.

Over the last decade, there has been a plethora of works in the literature in regard to corporate and bank bankruptcy prediction using AI and machine learning methods. However, researches exploring the application of ANFIS for banks bankruptcy prediction and the identification of banking risk index are scarce. The only work that uses neuro-fuzzy system for bankruptcy prediction we found is discussed in (Hajek and Prochazka 2018 ). Some studies applied AI techniques to recognize and predict human actions and activities (Gupta et al. 2022 ; Neu et al. 2022 ) focused on deep learning.

2.1 Risk areas and variables

2.1.1 capital adequacy risk.

Adequate capital base acts as a financial safety against a variety of risks a bank is exposed to in its daily operations. Capital adequacy reflects whether the bank has enough capital to absorb unanticipated losses and declines in asset values that could otherwise cause a bank to fail, and provide protection to depositors and debt holders in the event of liquidation. The balance sheet of the bank cannot be expanded beyond the level determined by the capital adequacy ratio (CAR). The important parameters which provide an insight into the capital adequacy of the bank that are used in the current study are (i) Tier1 ratio, (ii) Total capital ratio.

Tier 1 Capital Ratio compares a bank's core equity capital to total risk-weighted assets. A firm's core equity capital is known as its Tier 1 capital and is the measure of a bank's financial strength based on the sum of its equity capital and disclosed reserves, and non-redeemable, non-cumulative preferred stock. A firm's risk-weighted assets include all assets that the firm holds that are systematically weighted for credit risk. Basel committees recommend the weights which are further modified by the central banks, if required, for different asset classes, such as cash and coins, which have zero risk, versus a letter or credit, which carries more risk. A Bank must have a Tier 1 capital ratio of 8% (as per Basel guidelines) or greater. More than 8% is indicator of good risk position.

Total capital ratio (CAR): Total Capital ratio or Capital adequacy ratio is the ratio which determines the bank's capacity to meet the time liabilities and other risks such as credit and operation risk (Yang et al. 2020 ). Banks are required to maintain this ratio which is supposed to provide "cushion" for potential losses, and thereby protects the bank's depositors and other lenders. In this ratio two types of capital are measured: tier one capital, which can absorb losses without a bank being required to cease trading, and tier two capital, which can absorb losses in the event of a winding-up and so provides a lesser degree of protection to depositors. Basel stated a minimum acceptable ratio of 12%.

Common Equity Tier 1 Capital (CET1): Common Equity Tier 1 Capital is the highest quality of capital available reflecting the permanent and unrestricted commitment of funds that are freely available to absorb losses. It essentially includes ordinary share capital, retained earnings and reserves less prescribed deductions. The Basel standard is 10% and above.

2.1.2 Credit risk

Credit can be defined as the risk of potential loss to the bank if a borrower fails to meet its obligations (interest, principal amounts) (Thomas et al. 2022 ; Elsiddig and Sara 2015 ). Credit risk is the single largest risk banks face (Apostolik et al. 2009 ). The Basel Accord allows banks to apply the internal ratings-based approach to assess credit risk. They can internally develop their own credit risk models for calculating expected loss. As lending is the primary business for the banks. Asset quality is an important parameter to test the financial credibility of the banks and their risk exposure. Asset quality can be tested using different measures such as (i) Non-Performing loans to gross loans, (ii) loan loss provisions to net interest revenue.

Non-performing Loans to gross loans (NPL/Loans): This ratio is an important sub-parameter to measure asset quality of the bank as it determines the loan quality. Higher the ratio more problematic the loans are and vice versa. A decreasing trend for this ratio over the years is desirable, as it indicates that the Banks are following more cautious approach to risk management and there is a fall in the problem loans.

Loan loss provisions to net interest revenue (Loan LP): loan loss provisions are set aside by banks as an allowance against bad loans. This ratio clearly indicates the quality of the existing loans thereby the financial strength of the banks. The lower ratio is more preferable as it stands for less loss as compared to the interest revenue.

2.1.3 Liquidity

Liquidity is the ability of the bank to meet financial obligations as they become due, without incurring unacceptable losses (Chen 2022 ). Liquidity corresponds the bank’s ability to make payments to its customers punctually; their inability to make these payments will detrimentally affect the bank’s solvency. Liquidity management for a bank extends to both its loan customers and depositors. Since fractional reserve banking means that banks keep only a fraction of their deposits available for immediate withdrawal, improperly managing the bank’s liquidity risk could lead to serious consequences like liquidity crisis which may lead to bank run.

Net Loans / Total Assets (Loan/TA): The ratio of net loans to total assets indicates what percentage of the assets of the bank are tied up in loans. The higher the ratio the less liquid the bank is. A low ratio of loans to deposits indicates excess liquidity, and potentially low profits, compared to other banks. A high loan-to-deposit ratio presents the risk that some loans may have to be sold at a loss to meet depositors' claims.

Net Loans / Total Deposits (Loan/Dep): This ratio measures the degree of illiquidity of the bank as it indicates the percentage of the total deposits which are locked into non-liquid assets. A high figure denotes lower liquidity and high risk.

2.1.4 Earning quality risk

Earning quality ratios are used to measure the ability of the bank to earn profit compared to expenses. It shows the bank's overall efficiency and performance as it examines the bank’s investment decisions as compared to their debt situations. The Sub-parameters chosen to measure earning quality in this study are (i) net interest margin, (ii) financial risk.

Net Interest Margin (NIM): The core functions of banks are accepting deposits and lending. For that reason, net interest margin acts as a prime ratio in measuring the bank’s performance because it is the difference between what they receive from what they pay. Bank management is expected to keep a stable net interest margin as it illustrates the extent to which bank is exposed to interest rate fluctuation and also reflective of bank’s management’s ability to effectively manage interest rate risk. A positive and high ratio is considered to be desirable as it implies the bank had made optimal lending decisions and is successful in getting the timely interest on loans back from the customers.

Financial Risk (LEV): This ratio reflects the risk associated with the sources of finance and capital structure. Reliance on debt finance results in high leverage ratio and stands for high risk, whereas more equity finance as compared to total assets is a sign of low risk and is a base for generating more earnings from the internal sources of finance. This ratio has a positive relationship with the degree of financial risk and earnings quality.

2.1.5 Operational risk

Operational risk is defined by BCBS as the risk of loss resulting from “inadequate or failed internal processes, people and systems or from external events” and is a “fundamental element of risk management” at banks. This definition includes legal risk, but excludes strategic and reputational risk (Roumani et al. 2020 ). It is considered inherent in all banking products, activities, processes and systems (Basel Committee on Banking Supervision 2011 ). The best financial measure of it is through the comparison between cost and generated income.

Cost to income ratio: This ratio is an operating measure as it indicates the bank’s ability to manage its cost against income. Minimizing cost against income denotes efficiency hence lower the ratio higher is the efficiency. It measures how costs are changing compared to income. The cost income ratio, defined by operating expenses divided by operating income, can be used for benchmarking by the bank when reviewing its operational efficiency. Due to the inverse relationship between the cost income ratio and the bank's profitability, highly efficient banks will have low ratio and they generate higher profits.

3 Data and methodology

The data on different types of risks is collected from the banks operating in the Gulf Council Countries (GCC). The data set covers the last five years (2016 to 2020). The main source of this secondary data is the financial statements and reports of the GCC banks available on the DataStream and the banks’ websites.

Bostrom ( 2014 ), stated that a full AI solution would be automated in terms of data identification, data testing, and making decisions based on the data testing. In practice, AI might involve additional techniques in addition to ML, such as including hard-coded and logic rules. ML on the other hand normally involves manual data identification and testing by the data scientist, and human decisions as to how to apply the outputted information. Given the lack of technological and organizational readiness for pure AI, and the reality that most claimed AI is in fact ML, the study will apply ML to risk measurement, analysis, and prediction. The study will test the accuracy of the measures of different types of banking risk and develop an overall measure of the risk, then will use ML to develop a model for predicting banking risk.

Some of the recent studies have revealed that emerging artificial intelligent techniques, such as Decision Tree (DT), Support Vector Machine (SVM), Genetic Algorithm (GA) and Artificial Neural Networks (ANN) are advantageous to statistical models and optimization technique for credit risk evaluation. In contrast with statistical methods, AI methods do not assume certain data distributions. These methods automatically extract knowledge from training samples. According to previous studies, AI methods are superior to statistical methods in dealing with corporate credit risk evaluation problems, especially for nonlinear pattern classification.

3.1 Dataset

Ten financial ratios of 45 Gulf banks were collected for a 5-year period resulting in a dataset, Gulf Banks, consisting of 223 observations. As shown in Table 1 , the ten financial ratios are: Tier 1 Equity Ratio (TEIR1), Capital Adequacy Ratio (CAR), Common Equity Tier 1 Ratio (CET1), Provisional Loans Ratio (PLoans), Non-Performing Loans Ratio (NPLoans), Loans to Total Assets Ratio (LTAR), Loan Deposit Ratio (DepLoans), Equity Leverage Ratio (EQLEVR), Net Interest Margin Ratio (NIMR), Cost Income Ratio (CIR).

The risks variables have been assigned as measures for the following 5 categories of risk.

X1 = CAPR is capital risk.

X2 = Credit risk.

X3 = LIQR is liquidity risk.

X4 = EQ Risk is earning quality risk.

X5 = OPR is operational risk  =  Cost/income.

3.2 Proposed risk ranking index metric

Based on the Basel Accords and GCC Central Banks’ requirements for the minimum and maximum ratios, we have established a 10-dimensional point that reflects the minimum value for each financial ratio; the maximum ratios were replaced by their inverse. We referred to this point as the “critical risk ratios” (CRR) as it represents the minimum ratios that a bank must adhere to ensure survival. The more the distance between the point representing the financial ratios of a bank from CRR, the better the bank’s financial position and the less the risk of its insolvency. The recommended values of the ten financial ratios used for the critical point are: TIER1, 8.5 min; CAR, 10.5 min, CET1, 7.0 min; PLoans, 2.5 max; NPLoans, 5.6 max; LTAR, 70.0 max; DEPLoans, 85.0 max, EQ_LEVR, 3.5 min, EQ_NIM, 0 min; CIR, 50.0 max.

In this work, we have chosen to use the Mahalanobis distance (MD) as a measure of the Risk Ranking Index (RRI) as an indicator of a bank’s financial position. The MD is a measure of the distance between a multi-dimensional point (P) and a distribution of such points (D) (Mahalanobis 1936 ). It is a multi-dimensional generalization of the distance between P and D measured in terms of the number of standard deviations. This distance is zero for P at the mean of D and grows as P moves away from the mean along each principal component axis. If each of these axes is re-scaled to have unit variance, then MD corresponds to standard Euclidean distance in the transformed space. The MD is thus unitless, scale-invariant, and considers the correlations of the data set (Hadi and Simonoff 1993 ; Hill et al. 2006 ). In addition, the Mahalanobis distance is commonly used to identify outliers (Hadi and Simonoff 1993 ). The RRI indicator for a particular bank is computed as the MD between the multidimension point of the financial ratios of that particular bank and the CRR point. A bank is in a good financial position if it appears as an outlier with respect to the CRR point. A confidence level, such as, 95%, is used to construct the outlier’s boundary.

The Mahalanobus distance is given by Eq. ( 1 ) [ Hadi and Simonoff 1993 ]:

where X p1 , X p2 is a pair of multidimensional points, and C is the sample covariance matrix

Using the R programming language, we have computed the distance of each multidimensional point representing the ten financial ratios from the CRR point. Figure  1 shows the distribution of the distance values with the following statistics: mean = 6.81, standard deviation = 1.046, minimum distance = 3.96 maximum distance = 9.31. The distance computed represent the relative strength of the financial position of the bank with respect to the critical point, CRR. Thus, the bank with maximum distance has the strongest financial position and the reverse is true for the bank having the minimum distance. The cut-off distance (outliers boundary) from the CRR point, at the 95% confidence level, was found to be 4.28. There is only one instance where a distance of 3.96 was found to be less than the cut-off distance, yet close to the outliers’ boundary indicating that the bank still has low risk of going insolvent. In fact, at the 85% confidence level, the cut-off point is 3.81, and all banks are sufficiently distant from the critical point.

Distribution of the distance from the critical point

A bank is in good standing with respect to risk if its multidimensional point representing its financial ratios is an outlier with respect to its distance from the critical point. The cutoff distance that classifies a point as an outlier can be computed from the Chi-squared distribution based on the confidence level required and the degrees of freedom measured by the number of input variables. Figure  2 shows the bank risk index as measured by the Mahalanobis distance for the 233 financial ratios analyzed in this work and the cutoff distances of 4.82, 4.28, and 4.0 at the 99%, 95%, and 90% confidence levels, respectively. Five cases were identified to be below the cutoff distance of 4.82 at the 99% confidence level, and one case was below the 95% and 90% levels with a distance of 3.96. No case was found below the cutoff distance at the 89.5% confidence level. Thus, the best financial ratios are those having a Mahalanobis distance of at least 4.82 from the critical financial ratios.

Bank risk index as measured by the Mahalanobis distance and cutoff distances from the critical point at 99%, 95%, and 90% confidence levels

3.3 Variables influencing the risk ranking index

In order to conduct a sensitivity analysis to determine the relative importance of the ten financial ratios in Table 1 , we need to fit a model that captures the functional relationship between the RRI and the ten financial ratios. In this work, we have opted to use ANFIS to build the model. ANFIS is a hybrid analytical method that combines the merits of the neural network and the theory of fuzzy logic systems in its prediction mechanism (Jang 1993 ; Negnevitsky 2017 ). While neural networks control the representation of information and the physical architecture of the model, fuzzy logic systems imitate human reasoning and increase the model’s ability to manage uncertainty within the system (Jang 1993 ; Negnevitsky 2017 ). ANFIS basically learns the features of a given data and alters the system parameters to suit the required error criterion of the system in order to generate an output. Thus, in order to avoid making any assumptions with regard to the complexity, uncertainty, and the linearity or otherwise of the cause effect relationship between the Risk Ranking Index (RRI) and its determinants an attempt is made in this paper to develop ANFIS-based model for developing a model describing the relationship between the input variables (financial ratios) and the RRI computed as the Mahalanobus distance rather than opting for conventional methods such Multiple Linear regress (MLR) which basically estimates the parameters of a predetermined functional form of the relationship between the dependent and independent variables (Pal and Bharati 2019 ).

Table 2 shows the critical, best, and worst financial ratios identified in this work for banks working in the Gulf region. Some of the ratios in the worst case violate the specified ratios to keep the bank in a sound financial position. The Non-Performing Loans are just above the maximum under the worst instance with 5.8 compared to the critical ratio of 5.60. In addition, Loans to Assets and Loans to Deposits have obviously violated their maximum critical ratios with the worst scores of 85.5 and 93.2, respectively.

Fuzzy system models fall into two categories, which differ fundamentally in their abilities to represent different types of information. The first category includes linguistic models, which have been referred to as Mamdani fuzzy models. They are based on collections of if–then rules with vague predicates and use fuzzy reasoning (Mohamed et al. 2021 ). In these models, fuzzy quantities are associated with linguistic labels, and a fuzzy model is essentially a qualitative representation of the underlying system (Tanaka and Sugeno 1998 ). The second category of fuzzy models is based on the Takagi–Sugeno (TS) method of reasoning (Terano et al. 1994 ; Yager and Filev 1994 ; Sproule et al. 2002 ; Babaei and Bamdad  2020 ). These models are formed by logical rules that have a fuzzy antecedent part and a functional consequent. Fuzzy models based on the TS method of reasoning integrate the capability of linguistic models for qualitative and quantitative information representation (Rutkowski 2004 ). The main difference between Mamdani and Sugeno fuzzy systems is that the Sugeno output membership functions are either linear or constant. A Sugeno Fuzzy Inference System (FIS) is superior to a Mamdani FIS with regard to computational efficiency, accuracy, robustness in the presence of noisy input data (Subhedar and Birajdar 2012 ; Hamam and Georganas 2008 ; and MathWorks 2020 ). In addition, Sugeno FIS works well with optimization and adaptive techniques, guarantees continuity of the output surface, and well suited to mathematical analysis. As a result, this work uses ANFIS methodology based on a Sugeno FIS model. A description of the model structure and learning mechanism is provided in the subsections below.

3.4 Adaptive neural network-based fuzzy inference system

A shortcoming of fuzzy systems is the lack of ability to learn and adapt to changes in their environment. The opposite is true with neural networks, they can learn from data, but their reasoning is embedded in the connection weights of their neurons (Mitra and Hayashi 2000 ). The integration of fuzzy inference systems and neural networks can provide a platform to build intelligent systems by replacing the weakness of one system by the strength of the other (Negnevitsky 2017 ). Jang ( 1993 ) proposed a neural network that is functionally equal to a Sugeno fuzzy inference model called the Adaptive Neural Network-Based Fuzzy Inference System (ANFIS) that integrates the mechanisms of neural networks and fuzzy inference systems to utilize the capabilities of both.

3.4.1 ANFIS architecture

Jang’s ANFIS ( 1993 ) is normally represented by a six-layer feedforward neural network. Figure  3 below shows the ANFIS structure that corresponds to a first-order Sugeno fuzzy model with two inputs and two membership functions per input from Negnevitsky ( 2017 ). The network has fixed and adaptive types of neurons. Fixed neurons are represented as a circle and the adaptive ones depicted as a square. The following exposition is adapted from Negnevitsky ( 2017 ).

An adaptive Sugeno neuro-fuzzy inference system Architecture

For a first-order Sugeno fuzzy model, a two-rule base is expressed as follows:

Gaussian, triangular, sigmoidal, S-shaped, among other membership functions, can be used with an ANFIS. Sugeno neuro-fuzzy algorithm performs best when used with a Gaussian type membership function. With Gaussian fuzzy sets, the algorithm is capable of utilizing all information contained in the training set to calculate each rule conclusion unlike using triangular partitions (Jain and Martin 1998 ). Sambariya and Prasad ( 2017 ) have found that a Gaussian membership function performs best when the number of membership functions is 3 or 5.

Let the membership functions of fuzzy sets A i , B i for i  = 1, 2 be two Gaussian membership functions μ Ai and μ Bj respectively. In evaluating the rules, a product T-norm (logical and ) is chosen. Evaluating the rule premises using product T-norm results in:

Evaluating the implication and the rule consequents give:

Leaving the arguments out:

The above equation can be rewritten as:

3.4.2 ANFIS learning

ANFIS uses a combination of least-squares estimator and the gradient descent algorithm to learn its parameters (Jang 1993 ). Initially, activation functions are assigned to each membership function neuron. The centres of the membership functions are set so that the range of input is divided equally, and the widths and slopes are set to allow sufficient overlapping of the respective functions (Negnevitsky 2017 ). For each epoch, training is conducted in two steps: forward pass and a backward pass. In the forward pass, the ANFIS learning mechanism uses training patterns to estimate the parameters of the consequents of the rules by a least-squares algorithm. Once the rule's consequent parameters are established, the network can compute the error. In the backward pass, the errors are propagated back, and the parameters of the membership functions are adjusted using the back-propagation learning algorithm (Negnevitsky 2017 ). In the ANFIS training algorithm suggested by Jang ( 1993 ), both antecedent parameters and consequent parameters are optimized through the learning process. In the forward pass, the consequent parameters are adjusted while the antecedent parameters remain fixed. In the backward pass, the antecedent parameters are modified while the consequent parameters are kept fixed. Membership functions can be described by a human expert and kept fixed throughout the training process when the input–output data set is relatively small (Negnevitsky 2017 ).

3.5 Programming environment

The following Matlab functions have been used to implement the Sugeno type neuro-fuzzy system (MathWorks 2020 ):

fismat  =  genfis(trnDataInput, trnDataOutput, optGenfis) , genfis generates a fuzzy inference system using fuzzy c-means (fcm) clustering to extracting a set of rules that model the data behavior with the function fcm() . The function fcm() determines the number of rules and membership functions for the antecedents and consequents. The arguments for genfis are as follows (Sambariya and Prasad 2017 ; Talpur 2022 ):

trnDataInput : a matrix where each row contains the input values, financial ratios, of a data point. The matrix trnDataInput has one column per input variable.

trnDataOutput : a matrix where each row contains the output values of a data point. In this work, each input vector has one output, Risk Ranking Index , and thus trnDataOutput is a column vector.

optGenfis  =  genfisOptions(clusteringType),  creates a default option set for generating a fuzzy inference system structure using  genfis() . In this work, we have used subtractive clustering to find the cluster centres.

The model has been built using k-fold cross validation data (k = 5) in building the fuzzy inference system. To enforce the use of validation data the anfisOptions function was used to set the ValidationData option: optAnfis = anfisOptions('InitialFIS', fismat, 'ValidationData', valData);

fismat1  =  anfis(trainingData, optAnfis), this function fine-tune a Sugeno-type fuzzy inference system, fismat, using training data and optAnfis generated by anfisOptions(). The options allow the user to specify: an initial FIS object to tune; validation data to prevent overfitting to training data; and training algorithm options such as EpochNumber, ErrorGoal, InitialStepSize, StepSizeDecreaseRate, StepSizeIncreaseRate, OptimizationMethod among others (MathWorks 2020 ). Default values are used for options not overridden.

predictedOutput  =  evalfis(inputData, fismat1) uses the fuzzy inference system fismat1 and test data as input data to predict the bank’s Risk Ranking Index .

4 Results and discussion

The following subsections discuss the effectiveness of the model as a tool to predict and explain variations in a bank’s Risk Indicator .

4.1 Model performance as a predictive tool

The model was trained using k-fold cross-validation with k = 10 to reduce overfitting. The root means square error (RMSE) of the model’s predictions obtained was 0.27. Figure  4 shows the actual RRI and model-predicted RRI. The percentage of predicted values that are within one RMSE of the true respective RRI value is 78% and 94% within two RMSEs. Figure  5 depicts the percentage of the RRI predicted scores as a function of the deviation from the actual RRI. We can see that all predicted RRIs fall within 0.75 of their respective true value, indicating that the model has implicitly captured the functional relationship between the financial ratios and the RRI. In the next step, the model is used to conduct sensitivity analysis to rank the individual risk ratios in terms of their importance in explaining variations in a bank’s risk position.

Actual and predicted risk ranking indicator (RRI)

Percentage of predictions scores as a function of the deviation from the actual risk ranking index value

4.2 Model explanatory performance

Sensitivity analysis was conducted to determine the causal importance of each input variable (financial ratio). Many methods have been proposed for neural network‐based sensitivity analysis (Cao et al. 2016 ). The partial derivative algorithm (Dimopoulos et al. 1995 ) and the input perturbation algorithm (Zeng and Yeung 2003 ) have been shown to have superior performance compared to other techniques (Gedeon. 1997 ; Wang et al. 2000 ). However, two major weaknesses can be found in the partial derivatives method. First, it cannot implement neural networks with non-differentiable activation functions and second, it is inadequate for calculating the magnitude effect of the input variable in output sensitivity assessment (Cheng and Yeung 1999 ).

In this work, we have chosen the perturbation method for the reasons cited above. This method perturbs a given input variable by adding noise while keeping all other inputs unchanged. The change ratio of the output variable regarding the perturbation in the input variable is calculated. The process is repeated for a number of different noise levels. The input variable with the most significant change ratio is the one that has the strongest explanatory effect on the output of the system being analyzed (Lamy 1996 ). The crucial issues, however, are: (i) selecting a reasonable index for measuring the change in the output and (ii) the range of input perturbation levels. Bai et al. ( 2011 ) has investigated several approaches to neural network sensitivity and showed that the formula given by Reddy et al. ( 2006 ) described in Eq. ( 9 ) measures correctly both the direction and magnitude of the sensitivity of a neural network output with respect to a perturbation in a particular input variable value:

S j is a sensitivity index of the output with respect to input j,

N is the number of input training vectors, \(\widehat{y}, and y\) measures the network output with and without perturbation using the training data, and \({\widehat{u}}_{i}, and {u}_{i}\) are input variable i with and without noise, respectively.

To obtain an objective assessment of the sensitivity to perturbations in the input variables, the optimum range of input perturbation ratio should be determined (Bai et al. 2011 ). If the perturbation is overlarge, the sensitivity spectra may appear clipped. Generally, the farther a perturbation moves from the base case value, the less reliable the results become. However, if the perturbation is undersized, the sensitivity spectra may have no noise and sometimes no signal (Bai et al. 2011 ). Their study has found that a reasonable range of the input perturbation ratio is [− 20%, 20%] as there is no significant difference in the sensitivity of measurement within this range.

4.3 Discussion of the findings

After the Neuro-Fuzzy Model training had been completed, the sensitivity spectra values at increasing levels of input perturbation levels ranging from 0 to 20 % were calculated in steps of 0.01 according to formula ( 9 ) using training data. Table 3 and Fig.  6 show the sensitivity index of the model’s output for each input variable; these sensitivity index measurements indicate that by far Net Interest Margin Ratio (NIMR) is the most significant factor in explaining variations in bank risk position. NIM is an indicator of earnings’ quality and computed as the difference between the money that a bank is earning in interest on loans and the amount it is paying in interest on deposits. The justification for this high sensitivity is derived from variations in this ratio among the different banks in the different countries. NIMR is affected by factors of supply and demand for loans and the banking regulations that can increase or decrease the demand for deposit accounts and the demand for loans. This variable captures other sources of risk as it is an indicator of bank’s profitability and growth as well. This finding is consistent with previous studies such as; Saksonova, ( 2014 ) who finds that net interest margin is the most appropriate criterion for evaluating the effectiveness and stability of banks’ operations. Observe a negative relationship between NIM and the yield curve slope. Therefore, top ranking of earning quality risk is logical because it is very sensitive to the changes in loans and deposits.

The second sensitivity index is the Capital Adequacy Ratio (CAR). It is noticed that the GCC banks are adequately capitalized and they meet the international criteria and many of the banks score higher than the target of all capital adequacy ratios. As evidenced by the model the CAR (0.3506) is more sensitive than the other two capital (CET1, 0.1189 and TEIR1, 0.1062). This result is consistent with Basel and Oudat ( 2020 ) who found a positive significant relationship between capital adequacy and banks’ performance in Bahrain. This finding is consistent with Basel and Oudat ( 2020 ), in which the capital risk is the most significant type of risk. The importance of capital adequacy risk is supported by Yang et al. ( 2020 ) who stated that banks with adequate capital are able to mitigate capital risk. The ranking of the CAR as a second important variable by this index is supported by its impact on different risk variables and other performance indicators. This was proved by El Ansari et al. ( 2019 ) when their study showed a positive association between CAR level and many of the performance indicators.

Non-Performing Loans Ratio (NPLoans) is one of the measures of credit risk. The model shows that NPLoans with a sensitivity score of 0.0814 is somehow sensitive to the banking risk. Rajab and Sharma ( 2018 ) found the non-performing loan ratio to be one of the influential types of risk.

The liquidity risk is measured by two ratios that consider the relationship of the banking loans to asset (LTAR 0.023) and to deposits (DepLoans, 0.0166). Both variables of liquidity score very low impact on the overall banking risk. This result is supported by Karamoy et al. ( 2020 ), who concluded that there is no significant impact of liquidity risk on banking performance. The GCC banks have low liquidity risks as they can be classified as conservative banks. This supports the findings of this model as it ranks loan to assets and loans to deposits at the end just before provisional loans ratio and explains their less sensitivity to the risk index.

The Provisional Loans Ratio (PLoans) appears to have no role to play as an explanatory variable of changes in a bank’s risk position. The loan loss provision is an allowance for uncollectible loans. Generally, the GCC banks haven’t much uncollectible loans. Therefore, this ratio is expected to have no real impact on banking risk index. It is important to note that these results are pertinent to the banks composing the dataset; however, the model can be used for any group of banks and financial ratios to analyze the relative strength of their financial position. Table 2 shows the relative ranking of the ten risk ratios based on their sensitivity index for an increase of 10% in the ten ratios.

Equity Leverage Ratio (ELEVR): Majority of the GCC banks are in a good capital position to meet the enhanced capital requirements of Basel III as most of them have a large portion of capital structure financed by equity. Therefore, the results of the model seem to be realistic as the ratio is not highly sensitive to risk and ranked in the sixth place with 0.0581.

The Cost Income Ratio (CIR): The model proves that the GCC banks are efficient in managing their cost in relation to the income they generate. A low sensitivity to risk of CIR (0.0278) is an indicator of low risk associated with this variable of operational risk group. The model results indicate that the performance of the banking sector from both operating expense reduction and revenue generation lead to a lower risk on the operational side. This ratio is classified as a lower weight ratio suggest that banks are well-capitalized and well-regulated with little variations among banks. BIS ( 2013 ) stated that 15% of banks funding are short-term in nature, although their cost-to-income are among the lowest ratios.

Both credit risk ratios are indexed at the end because 45% of GCC banks are Islamic banks. The existence of Islamic banks mitigates to a large extent the variables of credit risks. This ranking is consistent with Srairi ( 2013 ) who finds that Islamic banks display a lower exposure to credit risk as compared to the conventional ones.

Sensitivity values of the financial ratios

5 Conclusions and recommendations

Although several studies have examined risk in emerging and advanced economies, studies on GCC banks are still limited. Over and above the study of all risk variables remains an unaddressed area of research, specifically in this part of the world. This paper has developed a banking risk index that incorporates the five main categories of banking risk with ten sub-variables measuring the risk components of each risk category. The study utilized data extracted from almost all the GCC banks (45 banks) over a 5-year period to develop the proposed risk index based on the Mahalanobis distance. ANFIS methodology based on a Segeno FIS model was used to build a prediction model to capture the relationship between the RRI and the various risk variables used. Using the ANFIS model, sensitivity analysis was conducted to determine the causal importance of each risk input variable. The main findings start with the development of a risk ranking index that measures the distance between the bank position, represented in terms of the 10 risk measures, from the critical values of these financial ratios stipulated by financial authorities and norms. Our findings indicate that a bank is in a sound financial position at the 99% and 90% confidence level if the value of the RRI is greater than 4.89 and 4.0, respectively. Sensitivity analysis of the functional model shows that by far Net Interest Margin Ratio (NIMR) is the most significant factor in explaining variations in bank risk position, followed by Credit Adequacy Ratio (CAR); while the Provisional Loans Ratio variable appears to have no role to play as an explanatory variable of changes in a bank’s risk position.

As a recommendation, the study shows that there are opportunities for future research to study the relationship between the different kinds of risks and banking performance. Further studies may be needed to segregate the risk of Islamic banks from conventional ones and develop separate indexes. The same study can be replicated at the country level to investigate the impact of certain local regulations on banking risk.

Data availability

The data that support the findings of this study are available from the corresponding author upon any request.

Abbreviations

Adaptive Neuro-Fuzzy Inference System

Artificial Neural Network

Back Propagation Neural Network

Capital Adequacy Ratio

Common Equity Tier 1 Capital

Cost Income Ratio

Critical Risk Ratios

Loan Deposit Ratio

Decision Tree

Equity Leverage Ratio

Fuzzy Inference System

Genetic Algorithm

Global Association of Risk Professionals

Gulf Council Countries

Hang Seng Index

Integrated Nonlinear Feature Selection

Interval-valued Fuzzy Cognitive Maps

Linear Discriminant Analysis

Loans to Total Assets Ratio,

Mahalanobis Distance

Multiple Discriminant Analysis

Machine Learning

Net Interest Margin

Net Interest Margin Ratio

None- Performing Loans Ratio

Provisional Loans Ratio

Root Mean Square Error

Risk Ranking Index

Support Vector Machine

Tier 1 Equity Ratio

Aly S, Alfonse M, Salem ABM (2022) Bankruptcy prediction using artificial intelligence techniques: a survey. In: Magdi DA, Helmy YK, Mamdouh M, Joshi A (eds) Digital transformation technology. Lecture notes in networks and systems, vol 224. Springer, Singapore

Google Scholar  

Apostolik R et al (2009) Foundations of banking risk: an overview of banking, banking risks, and risk-based banking regulation, New York: Wiley

Babaei G, Bamdad S (2020) A neural-network-based decision-making model in the peer-to-peer lending market. Intell Syst Account Finance Manag 7:142–150

Article   Google Scholar  

Bai R, Jia H, Cao P (2011) Factor sensitivity analysis with neural network simulation based on perturbation system. J Comput 6(7):1402–1407

Bank for International Settlements BIS (2013) Annual report. Basel: BIS

Basel Committee on Banking Supervision (2011) Core Principles for effective banking supervision. https://www.bis.org/publ/bcbs213.pdf

Basel A, Oudat MS (2020) Financial risk and the financial performance in listed commercial and investment banks in Bahrain bourse. Int J Innov Creat Change 13(12):160–180

Brlečić Valčić S (2021) Utilisation of ANFIS in analysing impact of cost structure on Croatian maritime company’s development. Sci J Marit Res 35:179–185

Bostrom N (2014) Superintelligence: Paths, dangers, strategies. Oxford University Press

Cao M, Alkayem NF, Pan L, Novák D (2016) Advanced methods in neural networks-based sensitivity analysis with their applications in civil engineering. In: Rosa JLG (ed) Artificial neural networks-models and applications. IntechOpen, London, pp 335–353. https://doi.org/10.5772/64026

Chapter   Google Scholar  

Chen TY (2022) Decision support modeling for multiple criteria assessments using a likelihood-based consensus ranking method under Pythagorean fuzzy uncertainty. Artif Intell Rev 55:4879–4939. https://doi.org/10.1007/s10462-021-10122-z

Cheng AY, Yeung DS (1999) Sensitivity analysis of neocognitron. IEEE Trans Syst Man Cybern Part c: Appl Rev 29(2):238–249

Deloitte University Press (2017) Global risk management survey, 10th ed. Deloitte University Press. https://www2.deloitte.com/tr/en/pages/risk/articles/global-risk-management-survey-10th-ed.html . Accessed 4 Oct 2018

Dimopoulos Y, Bourret P, Lek S (1995) Use of some sensitivity criteria for choosing networks with good generalization ability. Neural Process Lett 2(6):1–4. https://doi.org/10.1007/BF02309007

El-Ansary O, El-Masry A, Yousery Z (2019) Determinants of Capital Adequacy Ratio (CAR) in MENA region: Islamic vs. Conventional Banks. Int J Account Financ Rep 9(2):287–313. https://doi.org/10.5296/ijafr.v9i2.14696

Elsiddig I, Sara S (2015) Assessing interest rate risk using income gap. EPRA International Journal of Economic and Business Review, 3(2):2–12

Gedeon TD (1997) Data mining of inputs: analyzing magnitude and functional measures. Int J Neural Syst 8(2):209–218. https://doi.org/10.1142/S0129065797000227

González JS, Restrepo-Tobón D, Hassan RA (2021) Inefficiency and bank failure: a joint Bayesian estimation method of stochastic frontier and hazards models. Econ Model 95:344–360. https://doi.org/10.1016/j.econmod.2020.03.002

Gupta N, Gupta SK, Pathak RK et al (2022) Human activity recognition in artificial intelligence framework: a narrative review. Artif Intell Rev. https://doi.org/10.1007/s10462-021-10116-x

Hadi A, Simonoff J (1993) Procedures for the identification of multiple outliers in linear models. J Am Stat Assoc 88:1264–1272

Article   MathSciNet   Google Scholar  

Hajek P, Prochaska O (2018) Interval-valued fuzzy cognitive maps with genetic learning for predicting corporate financial distress. Filomat 32(5):1657–1662

Article   MathSciNet   MATH   Google Scholar  

Hamam A, Georganas ND (2008) A comparison of Mamdani and Sugeno fuzzy inference systems for evaluating the quality of experience of haptic-audio-visual applications. IEEE Int Workshop Haptic Audio-Visual Environ Games 2008:87–92. https://doi.org/10.1109/HAVE.2008.4685304

Hill T et al (2006) Statistics: methods and applications: a comprehensive reference for science, industry, and data mining. Statsoft Inc, Tulsa

Himeur Y, Elnour M, Fadli F et al (2022) AI-big data analytics for building automation and management systems: a survey, actual challenges and future perspectives. Artif Intell Rev. https://doi.org/10.1007/s10462-022-10286-2

Jabeur SB, Gharib C, Mefteh-Wali S, Arfi WB (2021) CatBoost model and artificial intelligence techniques for corporate failure prediction. Technol Forecast Soc Change 166:120658. https://doi.org/10.1016/j.techfore.2021.120658

Jain LC, Martin NM (1998) Fusion of neural networks, fuzzy systems and genetic algorithms: industrial applications. CRC Press, Boca Raton

Jang JSR (1993) ANFIS: adaptive-network-based fuzzy inference system. IEEE Trans Syst Man Cybern 23(3):665–685

Kar S, Kar AK, Gupta MP (2022) Modeling drivers and barriers of artificial intelligence adoption: insights from a strategic management perspective. Intell Syst Account Finance Manag 28(4):217–238. https://doi.org/10.1002/isaf.1503

Karamoy H, Tulung JE (2020) The impact of banking risk on regional development banks in Indonesia. Banks Bank Syst. https://doi.org/10.21511/bbs.15(2).2020.12

Lamy D (1996) Modeling and sensitivity analysis of neural network. Math Comput Simul 40:535–548

Mahalanobis PC (1936) On the generalized distance in statistics. Proc Natl Inst Sci India 2(1):49–55

MathSciNet   MATH   Google Scholar  

Marso S, EL Merouani M (2020) Bankruptcy Prediction using Hybrid Neural Networks with Artificial Bee Colony. Eng Lett 28(4):1134–1140

MathWorks, https://www.mathworks.com/help/fuzzy/genfis.html . Accessed Mar 2020

MathWorks, “Fuzzy Logic Toolbox, User’s Guide,” R2020a. [Online]. https://www.mathworks.com/help/fuzzy/

The MathWorks Inc. (2020). MATLAB (R2020a), Natick, Massachusetts: The MathWorks Inc. https://www.mathworks.com

The MathWorks Inc. (2022). MATLAB version: 9.13.0 (R2022b), Natick, Massachusetts: The MathWorks Inc. https://www.mathworks.com

MetricStream (2018) The chief risk officer’s role in 2018 and beyond managing the challenges and opportunities of a digital era new roles of the CRO: https://www.metricstream.com/insights/chief-risk-officer-role-2018.htm

Mishraz N, Ashok S, Tandon D (2021) Predicting financial distress in the Indian banking sector: a comparative study between the logistic regression, LDA and ANN models. J Glob Bus Rev. https://doi.org/10.1177/09721509211026785

Mitra S, Hayashi Y (2000) Neuro-fuzzy rule generation: survey in soft computing framework. IEEE Trans Neural Netw 11(3):748–768. https://doi.org/10.1109/72.846746

Mohamed EA, Ibrahim Elsiddig Ahmed IE, Riyadh Mehdi R, Hanan Hussain H (2021) Impact of corporate performance on stock price predictions in the UAE markets: Neuro-fuzzy model. Intell Sys Acc Fin Mgmt 28(1):52–71

Negnevitsky M (2017) Artificial intelligence: a guide to intelligent systems. Addison Wesley, Harlow, UK

Neu DA, Lahann J, Fettke P (2022) A systematic literature review on state-of-the-art deep learning methods for process prediction. Artif Intell Rev 55:801–827. https://doi.org/10.1007/s10462-021-09960-8

Oliver Wyman (2017) Next Generation Risk Management. https://www.oliverwyman.com/content/dam/oliverwyman/v2/publications/2017/aug/NextGenerationRiskManagementTargetingATechnologyDividend.pdf

Pal M, Bharati P (2019) Introduction to correlation and linear regression analysis. In: Pal M, Bharati P (eds) Applications of regression techniques. Springer, Singapore, pp 1–18

Chapter   MATH   Google Scholar  

Rajab S, Sharma V (2018) A review on the applications of neuro-fuzzy systems in business. Artif Intell Rev 49:481–510. https://doi.org/10.1007/s10462-016-9536-0

Reddy NS, Lee CS, Kim JH, Semiatin SL (2006) Determination of the beta-approach curve and beta-transus temperature for titanium alloys using sensitivity analysis of a trained neural network. Mater Sci Eng A 434(1–2):218–226. https://doi.org/10.1016/j.msea.2006.06.104

Roumani YF, Nwankpa JK, Tanniru M (2020) Predicting firm failure in the software industry. Artif Intell Rev 53:4161–4182. https://doi.org/10.1007/s10462-019-09789-2

Rutkowski L (2004) Fuzzy inference systems. In: Gaweda AE (ed) Flexible neuro-fuzzy systems: structures, learning and performance evaluation. The international series in engineering and computer science, vol 771. Springer, Boston, pp 27–50

Saksonova S (2014) The role of net interest margin in improving banks asset structure and assessing the stability and efficiency of their operations. Procedia–soc Behav Sci 150:132–141

Sambariya CDK, Prasad R (2017) Selection of membership functions based on fuzzy rules to design an efficient power system stabilizer. Int J Fuzzy Syst 19(3):813–828

Sproule BA, Naranjo CA, Turksen IB (2002) Fuzzy pharmacology: theory and applications. Trends Pharmacol Sci 23:412–417

Srairi S (2013) Ownership structure and risk-taking behaviour in conventional and Islamic banks: evidence from MENA countries. Borsa Istanbul Rev 13:115–127

Su C, Cheng C (2016) A hybrid fuzzy time series model based on ANFIS and integrated nonlinear feature selection method for forecasting stock. Neurocomputing 205:264–273

Subhedar M, Birajdar G (2012) Comparison of Mamdani and Sugeno inference systems for dynamic spectrum allocation in cognitive radio networks. Wireless Pers Commun 71:805–819

Talpur N, Abdulkadir SJ, Alhussian H et al (2022) Deep Neuro-Fuzzy system application trends, challenges, and future perspectives: a systematic survey. Artif Intell Rev. https://doi.org/10.1007/s10462-022-10188-3

Tanaka K, Sugeno M (1998) Introduction to fuzzy modelling. In: Nguyen HT, Sugeno M (eds) Fuzzy systems. The springer handbook series on fuzzy sets, vol 2. Springer, Boston, pp 63–89. https://doi.org/10.1007/978-1-4615-5505-6_3

Terano T, Asai K, Sugeno M (1994) Applied fuzzy systems. AP Professional, Boston

MATH   Google Scholar  

Thomas SS, Bansal M, Ahmed IE (2022) Investigating the disclosure compliance of Basel III in emerging markets: a comparative study between UAE and Indian banks. J Financial Regul Compliance. https://doi.org/10.1108/JFRC-02-2022-0018

Van Liebergen B (2017) Machine learning: a revolution in risk management and compliance? J Financ Transform 45:60–67

Vlasenko A, Vlasenko N, Vynokurova O, Bodyanskiy Y, Peleshko D (2019) A Novel ensemble neuro-fuzzy model for financial time series forecasting. Data 4:126. https://doi.org/10.3390/data4030126

Wang W, Jones P, Partridge D (2000) Assessing the impact of input features in a feedforward neural network. Neural Comput Appl 9(2):101–112. https://doi.org/10.1007/PL00009895

Yager RR, Filev DP (1994) Essentials of Fuzzy Modeling and Control. Wiley

Yang HF, Liu CL, Chou RY (2020) Bank diversification and systemic risk. Q Rev Econ Finance 77:311–326

Zeng X, Yeung DS (2003) A quantified sensitivity measure for multilayer perceptron to input perturbation. Neural Comput 15(1):183–212

Article   MATH   Google Scholar  

Zhang XPS, Kedmey D (2018) A budding romance: finance and AI. IEEE Multi Media 25(4):79–83

Download references

Acknowledgement

This work was financially supported by Ajman University under IRG: 2021-IGR-CBA-8.

Author information

Authors and affiliations.

College of Business Administration, Member of AI and DT Research Centers, Ajman University, Ajman, UAE

Ibrahim Elsiddig Ahmed

Artificial Intelligence Research Center (AIRC), College of Engineering & Information Technology, Ajman University, Ajman, UAE

Riyadh Mehdi & Elfadil A. Mohamed

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Ibrahim Elsiddig Ahmed .

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Ahmed, I.E., Mehdi, R. & Mohamed, E.A. The role of artificial intelligence in developing a banking risk index: an application of Adaptive Neural Network-Based Fuzzy Inference System (ANFIS). Artif Intell Rev 56 , 13873–13895 (2023). https://doi.org/10.1007/s10462-023-10473-9

Download citation

Accepted : 22 March 2023

Published : 24 April 2023

Issue Date : November 2023

DOI : https://doi.org/10.1007/s10462-023-10473-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Banking risk index
• Artificial intelligence
• Neuro-fuzzy systems
• Find a journal
• Publish with us
• Track your research

• SUGGESTED TOPICS
• The Magazine
• Newsletters
• Managing Yourself
• Managing Teams
• Work-life Balance
• The Big Idea
• Data & Visuals
• Case Selections
• HBR Learning
• Topic Feeds
• Account Settings
• Email Preferences

How Banks Can Finally Get Risk Management Right

• James C. Lam

SVB was a cautionary tale.

Banks have three lines of defense for managing risk — and then regulators are the fourth line of defense. In the case of Silicon Valley Bank, all four failed. If banks want to manage risk better, one good place to start is making sure a Chief Risk Officer is in place and a board-level risk committee is in place. And the people on that committee should have real experience in managing enterprise risk.

Here we go again. Banks ought to have the best risk management. But whatever safeguards were in place didn’t prevent Silicon Valley Bank from failing, destroying over $40 billion in shareholder value, and forcing unprecedented government intervention to protect depositors.  

• JL James C. Lam  is President, James Lam & Associates, a risk management consulting firm. Previously, he served as Partner of Oliver Wyman and chief risk officer at Fidelity Investments and GE Capital Market Services. He is the author of  Implementing Enterprise Risk Management: From Methods to Applications .

Partner Center

The effect of credit risk management and bank-specific factors on the financial performance of the South Asian commercial banks

Asian Journal of Accounting Research

ISSN : 2459-9700

Article publication date: 14 October 2021

Issue publication date: 27 May 2022

Among all of the world's continents, Asia is the most important continent and contributes 60% of world growth but facing the serving issue of high nonperforming loans (NPLs). Therefore, the current study aims to capture the effect of credit risk management and bank-specific factors on South Asian commercial banks' financial performance (FP). The credit risk measures used in this study were NPLs and capital adequacy ratio (CAR), while cost-efficiency ratio (CER), average lending rate (ALR) and liquidity ratio (LR) were used as bank-specific factors. On the other hand, return on equity (ROE) and return on the asset (ROA) were taken as a measure of FP.

Design/methodology/approach

Secondary data were collected from 19 commercial banks (10 commercial banks from Pakistan and 9 commercial banks from India) in the country for a period of 10 years from 2009 to 2018. The generalized method of moment (GMM) is used for the coefficient estimation to overcome the effects of some endogenous variables.

The results indicated that NPLs, CER and LR have significantly negatively related to FP (ROA and ROE), while CAR and ALR have significantly positively related to the FP of the Asian commercial banks.

Practical implications

The current study result recommends that policymakers of Asian countries should create a strong financial environment by implementing that monetary policy that stimulates interest rates in this way that automatically helps to lower down the high ratio of NPLs (tied monitoring system). Liquidity position should be well maintained so that even in a high competition environment, the commercial is able to survive in that environment.

Originality/value

The present paper contributes to the prevailing literature that this is a comparison study between developed and developing countries of Asia that is a unique comparison because the study targets only one region and then on the basis of income, the results of this study are compared. Moreover, the contribution of the study is to include some accounting-based measures and market-based measures of the FP of commercial banks at a time.

• South Asian countries

Credit risk

Bank-specific factors.

• Generalized method of moment

Siddique, A. , Khan, M.A. and Khan, Z. (2022), "The effect of credit risk management and bank-specific factors on the financial performance of the South Asian commercial banks", Asian Journal of Accounting Research , Vol. 7 No. 2, pp. 182-194. https://doi.org/10.1108/AJAR-08-2020-0071

Emerald Publishing Limited

Copyright © 2021, Asima Siddique, Muhammad Asif Khan and Zeeshan Khan

Published in Asian Journal of Accounting Research . Published by Emerald Publishing Limited. This article is published under the Creative Commons Attribution (CC BY 4.0) licence. Anyone may reproduce, distribute, translate and create derivative works of this article (for both commercial and non-commercial purposes), subject to full attribution to the original publication and authors. The full terms of this licence may be seen at http://creativecommons.org/licences/by/4.0/legalcode

Introduction

Around the globe, depository institutions perform a crucial job in bringing financial stability and economic growth by mobilizing monetary resources across multiple regions ( Accornero et al. , 2018 ). The commercial plays an intermediary role by collecting the excessive amount from savers and issuing loans to the borrowers. In return, banks can earn a high interest rate ( Khan et al. , 2020 ; Ghosh, 2015 ). Banks tried to increase their financial performance (FP) by issuing loans while playing their intermediary role; banks have a high chance of facing credit risk. Accornero et al. (2018) found that the country's banking industry mostly collapses due to high credit risk. Sometimes, it leads to the failures of the whole financial system. Credit risk is expected to be arises when a borrower cannot meet their obligation about future cash flows. Commercial banks' FP is affected by two factors: one is external and the other is internal. Bank-specific factors are internal and able to control factors of the commercial banks. Ofori-Abebrese et al. (2016) pointed out that adverse selection and moral hazards were created due to mismanagement of internal factors. The abovementioned financial problems are turmoil period in the banking/financial sector.

Among the entire continent of the world, Asia is the most crucial continent and contributing 60% of world growth but facing the serving issue of high nonperforming loans (NPLs). It is well known that a high ratio of NPLs weakens the economy or country's financial position. The growth level in South Asia was the highest in 2015, and the ratio is 9.3%, which is the highest among all continents. According to the Asian Development Bank (2019), the NPLs in the south are approximately $518bn, which is relatively high compared to previous years. The soaring of NPLs in South Asian countries enforces a massive burden on commercial banks' financial position (mainly banks' lending process effected). The massive increase in NPL is observed after the global financial crisis (2007–2008). According to Masood and Ashraf (2012) , the credit risk high ratio of NPLs is the main reason for most of the financial crisis because NPLs alarmingly high during the Asian currency crisis in 1997 and subprime crises in 2007, and some loans are declared bad debts. The alarmingly high ratio of NPL resulted in an increasing depression in the financial market, unemployment and a slowdown of the intermediary process of banks (see Figure 1 ).

The World Bank statistics of different regions show that NPLs exist in almost all regions. Still, the ratio of NPLs is relatively high in the South Asian area compared to other regions. Therefore, the study is conducted in South Asia. Two proxies of credit risk are used in this study: NPLs and capital adequacy ratio (CAR). Moreover, the study also incorporates bank-specific factors to increase FP.

Various studies ( Louzis et al. , 2012 ; Ofori-Abebrese et al. , 2016 ; Hassan et al. , 2019 ) are conducted to address the issue, but literature shows that the results of these studies are inconclusive and also ignore the most important region of South Asia. Therefore, the study objective is to investigate that credit risk and banks specific factors affect FP of commercial banks in Asia or not? We have selected two from South Asia, Pakistan and India, as sample countries. In 2019, the NPLs were 13% and 10% in Pakistan and India, respectively. This ratio is relatively high as compared to the other countries of the world. Due to these reasons, we have mainly selected India and Pakistan from South Asian countries ( Siddique et al. , 2020 ). The present study uses secondary panel data set of 19 commercial banks from 2009 to 2018.

Two serious threats may exist: The first is autocorrelation and the second is endogeneity. If the data do not meet these CLRM assumptions, then the regression results are not best linear unbiased prediction (BLUE) ( Sekaran, 2006 ; Kusietal, 2017 ). And in this situation, apply pooled regression is applied, and then the results were biased because the coefficient results cannot give accurate meaning. After all, pool regression ignores year and cross section-wise variation. Therefore, in this study, an instrumental regression can be used that handle all these issues. Generalized method of moments (GMM) is used to analyze the data to overcome endogeneity. Our study is unique by addressing the autocorrelation and endogeneity issue at a time. Our study results show that credit risk measure NPLs decrease the FP due to having negative relation, while CAR has a positive relation with South Asian banks’ FP. The remainder of the research study is organized as follows: Section 2 consists of a detailed literature review; Section 3 consists of data and methodology. Sections 4 contains information about finding and suggestions. Finally, Section 5 discusses the conclusion.

Literature review

The Literature Review has mainly divided into two crucial sections; First part consists of the literature review related to credit risk and FP. The other part is related to the literature review of bank-specific variables and FP. In the hypothesis development, we have used commercial banks' profitability that represents the FP of commercial banks.

Credit risk and financial performance

While operating in the banking industry, three categories of risks that the bank has to face include environmental, financial and operational risks. Banks generate their incomes by issuing a massive amount of credit to borrowers. Still, this activity involves a significant amount of credit risk. When borrowers of the banking sector default cannot meet their debt obligation on time, it is called credit risk ( Accornero et al. , 2018 ). When there is a large amount of loan defaulter, then it adversely affects the profitability of the banking sector. Berger and DeYoung (1997) pointed out that the absence of effective credit risk management would lead to the incidence of banking turmoil and even the financial crisis. Siddique et al. (2020) explain that NPLs are related to information asymmetric theory, principal agency theory and credit default theory. When asymmetric information unequal distribution of information of high NPLs is spread, there is a chance that banks or financial declared bankrupt. According to Pickson and Opare (2016), the principal agency must separate corporate ownership from managerial interest. Because each management has its interest, they want more prestige, pay increment and want the stock options for management. Effective management of credit risk or nonperformance exposure in the banking sectors increases profitability. It enhances the development of banking sectors by adequate allotment of working capital in the economy ( Ghosh, 2015 ).

There is a growing literature ( Louzis et al. , 2012 ) on credit risk and its empirical relationship with the monetary benefits of the banking sector. Ekinci and Poyra (2019) investigate the relationship between credit risk and profitability of deposit banks in Turkey. The data sample used 26 commercial banks from 2005 to 2017. All data of this study are secondary and collected from annual reports of commercial Turkey banks. The proxies of profitability were taken as return on equity (ROE) and return on the asset (ROA), while NPLs of commercial banks were used as a proxy to measure credit risk. The research paper reveals that credit risk and ROA are negatively correlated as well as the relation between credit risk and ROE is also significantly negative relation. Therefore, the study suggests that the Turkey government tightly monitors and controls the alarmingly soaring ratio of NPLs. Upper management introduced some new measures to trim the credit risk.

There is a negative and significant relationship between NPLs and commercial banks' FP.

There is a positive and significant relationship between capital adequacy ratio and commercial banks' FP.

Bank-specific variables and financial performance

Bank-specific variables or internal factors are the product of business activity. Diversifiable risk is associated with these factors ( Louzis et al. , 2012 ) and can be reduced by efficient management. This risk is controllable compared to an external factor, which cannot be diversified because this risk is market risk ( Ghosh, 2015 ; Rachman et al. , 2018 ). If a firm can manage its internal factor effectively, then the firm can be high profitability, while, on the other hand, these factors are mismanaged. It would adversely affect the firm's balance sheet and income statement ( Ofori-Abebrese et al. , 2016 ). Different authors ( Akhtar et al. , 2011 ; Louzis et al. , 2012 ; Chimkono et al. , 2016 ; Hamza, 2017 ) discuss different bank-specific variables and firm performance in their studies. The bank-specific variables used in this study are cost-efficiency ratio (CER), average lending rate (ALR) and liquidity ratio (LR). Aspal et al. (2019) used two types of factors (macro and bank-specific factors) and inspected their connection with the FP of the commercial bank in India. Gross domestic product (GDP) and inflation are used as proxies of macroeconomic factors.

In contrast, a bank-specific variables’ proxy includes capital adequacy ratio, asset quality, management efficiency, liquidity and earnings quality. Data of 20 private banks have been used from 2008 to 2014. The panel data pointed out that one macroeconomic factor is significant (GDP), and another factor (Inflation) is insignificant. All bank's specific factors (earning quality, asset quality, management efficiency and liquidity) significantly affect the FP except the CAR (insignificant). Hasanov et al. (2018) conducted their study to explore the nature of the interrelation between bank-specific (BS) and macroeconomic determinants with the banking performance of Azerbaijan (oil-dependent economy). The study used the GMM to analyze the panel data set. The results show that bank loans, size, capital and some macro factors (inflation, oil prices) were positive and significantly interconnection with the FP of banks; on the other hand, liquidity risk, deposits and exchange rates are significantly affected negatively bonded with the FP.

There is a negative and significant relationship between the CER and commercial banks' FP.

There is a positive and significant relationship between the ALR and commercial banks' FP.

Francis et al. (2015) define liquidity in their study and, according to the liquidity of an asset, determined by how quickly this asset can be converted or transferred into cash. Liquidity is used to fulfill the short-term liabilities rather than the long term ( Siddique et al. , 2020 ; Raphael, 2013 ). Adebayo et al. (2011) mentioned in their study that when banks are unable to pay the required amount to their customers, it is considered bank failure. Sometimes liquidity risk affects the whole financial system of a country. Different studies are conducted on the issue of liquidity and performance, but different studies show different results. FP and liquidity, on the other hand, a chunk of studies ( Francis et al. , 2015 ; Hamza, 2017 ) revealed significant negative tie-up between liquidity and FP, while some other studies pointed out that there is no significant relationship between liquidity and FP. Therefore, the studies show a contradictory result, so the current study takes the bank-specific measures (LR, ALR study and CER) and checks its interconnection with commercial banks' FP.

There is a positive and significant relationship between the LR and commercial banks' FP.

Data and methodology

Our current study has one problem variable, financial performance (FP), while regressors variables are credit risk and bank-specific variables. Our model is consistent with Chimkono et al. (2016) , where ROA and ROE will be used as a measure of FP, while credit risk will be measured by NPL ratio, CAR and three specific variables: CER, LR and ALR.

Various studies ( Hamza, 2017 ; Belas, 2018 ) emphasize some macro and micro variables that need to be controlled when measuring FP because these factors are the influential factors. Three control variables: size of the bank, age of the banks and Inflation are used in this study and shown as yes in the tables. We have chosen these three control and most relevant variables because these variables represent both micro and economic situations. Data have been collected from two South Asian countries Pakistan and India. The nature of data is panel data and the number of banks from Pakistan (10 commercial banks) and India (9 commercial banks) is 19. The data have been collected from bank financial statements throughout 2009 to 2018, so the data of this study are a panel in nature. The final number of observations is 190 (19*10 = 190) for the analysis of this study (see Table 1 ).

Operational definition

The probability of lenders being the default, high credit risk higher FP of banks ( Louzis et al. , 2012 ).

Bank-specific factors are those which are under the control of the management of commercial banks ( Chimkono et al. , 2016 ).

Nonperforming loans

A loan becomes nonperforming when the duration of the loan has been passed, and after that duration, banks 90 days are passed unable to receive the principal amount of loan and interest payment ( Hamza, 2017 ).

Methodology

The current study investigates the interrelationship between credit risk, bank-specific factors and FP. Panel data set is used in our study, and two serious threats usually faced when using panel data set: (1) autocorrelation and (2) endogeneity. For this purpose, a GMM can be used. GMM model has many advantages on simple ordinary least square regression. And when in any study GMM model applies, it allows by adding the fixed effect model; this model can be able to tackle the problem of heterogeneity, and it also removes the problem of endogeneity by introducing some instrumental variables.

Model specification

The regression model is as follows:.

γ 0  = intercept; γ 1 - γ 8  = estimated coefficient of independent variables and control variables.

ε it represents error terms for those variables that are omitted or added intentionally/unintentionally.

According to Lassoued (2018) , panel data regression has two significant problems: autocorrelation and endogeneity, and this problem is existed due to the fixed effect. Therefore, our study checked the basic two assumptions of ordinary least squares.

Testing for autocorrelation

The fifth assumption of CLRM is that data should be free from autocorrelation. Sekaran (2006) pointed out the relationship between two different error terms should be zero; it means that there is no autocorrelation between error terms. There are different tests for testing autocorrelation, but the Wooldridge test is used in the present paper to test the autocorrelation.

Table 2 shows that the p -value of the Wooldridge test result is zero, so it means that all p -values are less than 0.05. It means that reject the null hypothesis. And the null hypothesis is that our data have no autocorrelation, but the results show that our data have autocorrelation problems.

Testing for endogeneity

The seventh assumption of CLRM is that data have no issue of endogeneity. Sekaran (2006) found that the relationship between the error term and explanatory or independent variable should be zero. If this relationship is not zero, then the problem of endogeneity exists. Brooks (2014) pointed out that Hausman test results probabilities can be used to test the endogeneity, and the null hypothesis of this test is that errors are uncorrelated. He also pointed out that if the probabilities are more than 0.10, then accept the null hypothesis. It means that there is no problem of endogeneity, and if the values are less than 0.1, then our data have the problem of endogeneity. Appendix 1 shows that some values of the Hausman test are less than 0.10, so it means that data have the problem of endogeneity. Our panel data results prove that our data have the problem of autocorrelation and endogeneity. Some CLRM model assumptions are not met, so ordinary least square regression results are not BLUE. And GMM model can be applied to any study because this model can be able to tackle the problem of autocorrelation, and it also removes the problem of endogeneity by introducing some instrumental variables.

Findings and discussion

The present research paper provides empirical evidence on the interconnection between credit risk and bank-specific/internal factors on FP commercial banks. To analyze the data set, first, the study applies the descriptive analysis to identify the big picture of the data, then the correlation section and at the end, regression results are discussed. Table 3 presents the descriptive statistics of the all variables used in the study: credit risk indicator which are the ratio of NPL, CAR; indicators of bank-specific factors (CER, ALR, LR); some control variables SIZE, AGE, INF and the measure of FP: ROA, ROE. The mean value of ROA and ROE is 0.986 and 7.964 with a standard deviation of 1.905 and 39.175, respectively, which shows that ROE has much higher variation than ROA. The standard deviation of NPL is 9.659, which is double that of CAR, whose standard deviation is 4.183 among all bank-specific factors (see Table 4 ).

Factor (CER, ALR, LR) LR has high dispersion (14.177) because there is a remarkable difference between minimum 25.027 and maximum value (107.179) of LR. ROA has 0.986 with a range between 10.408 and −6.234 with a standard deviation of 1.905, and it shows that there is a low level of dispersion in developed countries. The dispersion of ROE 39.175 is highest among all other variables, which means that some outliers exist in the ROE variable.

Correlation analysis is used to check the linear relationship between the two explanatory variables ( Brooks, 2014 ). If the sample size of any approaches to 100, greater than 100 and the correlation coefficient is 0.20, then the correlation is significant at 5% ( Lassoued, 2018 ). Most of the variables in the current study are significant at 5%.NPLs, and CER loans are negatively correlated with almost all independent variables, which supports the literature point that NPLs and CER are negatively associated with FP and bank-specific factors. The negative correlation of NPLs with ROE is loan −0.378, and this correlation is high as compared to other countries. At the same time, all bank-specific factors, CER, ALR and LR are mostly positively correlated with most of the other, almost all dependent and independent variables, while AGE and INF are mostly negatively correlated with the other variables of the study.

Regression results and discussion

Tables 5 and 6 have shown the regression results of pooled regression and GMM models. Tables include all independent, control variable coefficients, t -statistics, standard error and probability values. Additionally, tables have the values of R 2 , adjusted R 2 and Durbin Watson statistics. The adjusted R 2 under pooled regression are 0.250 and 0.231 in both models (ROA and ROE). While adjusted R 2 under the GMM are 0.358 and 0.249 in both models ROA and ROE.

It means the GMM more and better explains our model than pooled regression. Moreover, we also apply a Hausman test on both models. The p -value of both models is less than 0.05, so our data have the problem of endogeneity null hypothesis. To eliminate the endogeneity issue, the GMM coefficient was measured.

NPL has a significant and negative measure of FP: ROA and ROE. In contrast, CAR has significant and positive with all proxies of FP: ROE and ROA, which supports H1 and H2 of the paper. Our finding is consistent with Masood and Ashraf (2012) who conducted their study on credit risk and FP and found a significant negative relationship between NPL and FP, so NPLs hinder banks' profitability. Therefore, NPLs affect the whole financial system of a country especially in developing countries. The findings of CAR matched with Accornero et al. ’s (2018) study and pointed out that CAR has a significantly positive link with FP. CER has a significant negative relationship with ROA and ROE, which is consistent with the study of Francis et al. (2015) who pointed out a significant negative relationship between CER and ROE. Therefore, banks need to adapt strategies to control these costs and tried to increase their profitability. ALR had a significant and positive relationship with both measures of FP. ALR is significant at 1% with ROA and 10% significant with ROE. The result is supported by the study of Chimkono et al. (2016) who found a positive relationship between the ALR and FP of commercial banks.

LR has a significantly negative relationship with ROA and ROE. This finding is consistent with Siddique et al. (2020) who pointed out a significant negative relationship between LR and ROE; the more liquidity is maintained, the lesser the profitability level. In short, most of the independent variables are significant at 5% and 1%, and control variables are also significant in both models size of the bank and inflation except AGE. This result is matched with Ghenimi et al. ’s (2017) findings that prove that total assets or investment increment are directly proportional to the FP. Both variables of credit risk NPL and CAR are significant with the FP of commercial banks in both models. Banks try to reduce bank-specific factors risk, and by doing so, ultimately the amount of bad debt decreased, and another benefit is that it also reduces the amount of loan loss provision.

The current study empirically investigates the causal interrelation between credit risk, bank-specific factors and FP of commercial banks in two South Asian countries (Pakistan and India). The study's finding suggests that managers in South Asian countries should be focused on increasing capital adequacy to enhance the monetary gain (FP) while for the contraction of NPLs by implementing modern techniques and strategies for credit risk (NPLs) management. One indicator of the bank-specific variable (ALR) has a significant and positive interrelation with the FP of commercial banks. In contrast, CER and LR have a significant and positive relationship with the FP of commercial banks of South Asia. Control variables of the study (size of the bank and inflation) are also significant in both models except AGE. There are several policy implications that commercial banks of South Asian countries should be followed. NPLs are soaring due to the following reasons: less supervision and monitoring of customers, the problem of the market and lack of customer knowledge related to loans. Bank management should be efficient in judging that their customers have viable means of repayment or not. Moreover, banks can offer expert opinion to the professional loan take on feasible techniques of efficiently endow the borrowing to secure the required return on total firms investment is acquired. Liquidity position should be well maintained so that even in a high competition environment, the commercial can survive in that environment.

The scope of the study is only limited to commercial banks, but this model can also be applied to Islamic banks. And future researchers can also apply this model to a comparison-based study of commercial and Islamic banks. Data of this study have been collected only from 19 banks; future research can also increase the number of banks and increase the number of years to conduct their study. And if the number of banks and the number of the year increased, the results are a more reliable and accurate representation of the population. The data of this study have been taken only from two countries of South Asia, but this study can be extended by adding more countries in Asia. When we add the number of countries, the results are a better and accurate representation of developing and developed countries of Asia. This model can also be applied to some other continents because the macro environment and bank-specific factors are pretty different from continent to continent Appendix A1 .

NPLs-continent wise

Summary of explanatory variables and dependent variables

Results for autocorrelation for South Asia countries

Descriptive statistics

Correlation figures

ROA model (pooled regression and fixed effect GMM result)