Abstract
Banks are exposed to different types of risks in the process of financial intermediation and maturity transformation. The experience of the extant global financial crisis provided ample evidence of interaction among bank risks and perils of ignoring interactions in the changing economic, technological and regulatory environment. In this study, we assess the dynamic interaction among bank risks for the entire banking sector and bank groups based on various bank-specific characteristics in a vector autoregression framework, including variance decomposition and impulse response function analysis. We estimate the market measures of different risks using a multivariate GARCH (1, 1) in mean model. The study uses weekly bank level data from 23 October 2004 to 1 August 2014 for 40 listed Indian banks. The findings suggest that there is a positive interaction between equity risk and all other risks. Credit risk and exchange rate risk have a reciprocal relationship. It has also been observed that equity risk impacts credit risk positively. Interest rate risk seems to be affected by its lagged values and does not appear to be affected by other risks. The study highlights the role of liquidity in reducing bank risk exposures and supports new liquidity standards introduced in Basel III. The results improve the understanding of the interaction among risk exposures, which may enhance the supervisory process in the Basel framework. The risk interactions must be kept in mind for making capital provisioning, and an integrated approach to risk management by banks is more desirable.
Keywords
Introduction
Banks are at the core of the financial system in an emerging economy like India (International Monetary Fund, 2013) where the financial sector is the lifeline of the economy and problems stemming from it can lead to widespread ramifications to other segments in the economy. Over the next few years, several changes are expected in the Indian economy, which warrants a robust, flexible and efficient financial system to achieve the economic goals.
The literature has long recognized the role of the financial sector and its stability in stimulating economic growth (see Angadi, 2003; Levine, 1997, 2005; Sehgal, Ahmad, & Deisting, 2012). Banks performs a variety of functions in the process of financial intermediation and maturity transformation with a liquid liability and an illiquid asset profile, which exposes them to different types of risk. Primarily these are credit and market risks (including equity, interest rate risk and exchange rate risk), apart from other risks emerging from the changing environment such as operational risk, liquidity risk, reputational and regulatory risk.
From this perspective, the understanding of different aspects of risk is an integral part of banking, and of utmost significance to bankers, regulators and policy-makers. The 1998 Basel Accord required banks to hold capital with respect to the risk-weighted assets. The 1996 Market Risk Amendment to the Basel Accord of 1988 incorporated market risk along with credit risk. Basel II in June 2004 widened the horizon for the risks covered under the accord and included operational risk along with market and credit risk. Basel III 1 now additionally recognizes several other risks explicitly, such as liquidity risk, reputational risk and regulatory risk. Due to the gradual, piecemeal inclusion of these risks in the regulatory framework, the interaction among these bank risks have been ignored (Hartmann, 2010). Thus, it has become imperative to revisit risk management practices and plug the gaps in risk management policies. Reserve Bank of India (RBI) 2 defines credit risk as ‘the risk that the obligor (borrower or counterparty) in respect of a particular asset will default in full or in part on the obligation to the bank in relation to the asset’ (2002b, p.1). Market risk is ‘the risk that the value of “on” or “off” balance sheet positions affected by changes in equity and interest rates in an undesirable manner, currency exchange rates, and commodity prices’, as defined by RBI (2002b, p.1).
The recent global financial crisis (GFC) emphasizes the need for an enhanced understanding of the interaction among bank risks and highlights the inefficacy of the existing regulatory framework. Traditionally, for varied reasons, different risks are assumed to have no relation with one another and are measured and managed independently of other risks. Such practices are adopted because of the convenience they offer but are not sufficient (Li, Feng, Wu, & Lee, 2012; Liang et al., 2013). The factors causing different types of risks are similar and often interact with one another, which makes their interrelations more complicated. Ignoring the dynamic interaction among risk drivers when assessing the total risk leads to a biased estimate of the aggregate risk exposure (Alexander & Pézier, 2003; Böcker & Hillebrand, 2008; Drehmann, Sorensen, & Stringa, 2010). The Basel Committee on Banking Supervision (BCBS, 2009) also points out this lack of integration as one of the reasons for the failure of stress testing practices in banks.
The basic premise of this study is to examine the interactions of the risks in the banking system. In assessing capital requirements on the basis of risk, it is important to consider the broader relationship between credit and market risks. At times of crises, these risks can interact and amplify one another. For instance, during the global financial crisis of 2008–2009, credit and market risks surged when liquidity dried up in the financial markets. To deal with the impact of such feedback effects, capital requirements could be structured to take a broader view of risk and the relationships (as well as the potential feedback mechanisms) among different sources of risk in the financial system. This implies that different aspects of risk must first be carefully considered at the level of the individual institution and then analysed at a broader systemic level.
Indian banking industry provides a compelling case to investigate its approach towards risk management. The Reserve Bank of India (RBI) also emphasized that the traditional perception of banks as mere financial intermediaries has transformed and risk management is now their defining attribute (Reserve Bank of India, 2002). RBI recommends banks to consider integrating market risk elements into their credit risk assessment process (Reserve Bank of India, 2003). Implementation of Basel III norms for the Indian banks also necessitates understanding dynamic interactions among bank risks. Emphasis of the RBI on integrated risk management (Reserve Bank of India, 2003) as well as banks’ own concerns for it can be attributed to the economic and financial crisis in some countries, which revealed a strong correlation between unhedged market risk and credit risk. In view of the above situation of the Indian banking industry, this study becomes pertinent. A bank-level study helps to highlight the role of bank-specific characteristics, which explains the heterogeneity among banks and reviews the suitability of a ‘one size fits all’ approach for the Indian banking sector.
So far, research in the Indian context has remained limited to isolated studies of credit and market risks that banks are exposed to. Research on their interaction is scant even internationally. A study of the dynamic interaction of different bank risk exposures is thus necessitated by the GFC and lack of adequate literature on the subject. Rising bank risk owing to the volatile environment have led the regulators, practitioners and researchers to study different bank risks from an integrative perspective, and to accurately measure and effectively manage them. The primary objective of this study is to investigate the interactions between credit and market risk, as well as the impact they have on each other, and to make recommendations for developing a more integrated risk management system for the Indian banking sector. For this purpose, we assume that the market risk of a bank comprises equity risk, interest rate risk and foreign exchange risk (Reserve Bank of India, 2002b). We analyse the interaction between these risks and also examine their linkages with credit risk.
This article is structured as follows: The second section summarizes the related literature; the third section describes the data; The next section covers the methodology adopted; the fifth section presents a discussion about the empirical results along with their implications and the last section provides the summary and concluding observations.
Review of Literature
This section provides an intensive review and evaluation of literature that underpins the research objectives of the present study. Earlier work had two dimensions: first, dealing with the issues related to the measurement of risk exposures and second, focussing on the assessment of the level of interaction among bank risks.
Risk Measurement Issues
One of the major issues faced while studying the interaction among different bank risks in an emerging economy is how to quantify risk exposures of banks. For this purpose, the literature suggests the use of the capital market approach, which measures banks’ risk exposure by studying the sensitivity of equity returns of banks to movements in various risk factors. The underlying assumption is that stock markets correctly capture bank risks. Numerous studies (Atindéhou & Gueyie, 2001; Chamberlain, Howe, & Popper, 1997; Choi, Elyasiani, & Kopecky, 1992; Elyasiani & Mansur, 2004, 2005; Ryan & Worthington, 2004; Shamsuddin, 2009; Sukcharoensin, 2013) have focussed on the capital market approach to risk measurement for the developed markets along with a few attempts in the emerging economies (Hahm, 2004; Hooy, Tan, & Nassir, 2004; Kasman, Vardar, & Tunç, 2011). For the purpose of empirical analysis, the capital market approach has some advantages, especially for an emerging market like India. Availability of accounting data at a lower frequency, especially for banking sector with a short financial history, limits the use of the accounting-based approach. The capital market approach is also considered to be forward-looking (Martin & Mauer, 2003; Muller & Verschoor, 2006), making it more appropriate for policy analysis perspective.
Interaction of Risk Exposures
This section describes how interactions between different types of risk occur and why they are important. The primary objective is to summarize the research on the interaction of these risks and its implications for risk management.
The issue is reflected in the literature, as early as Jarrow and Turnbull (2000, p. 272):
Economic theory tells us that market and credit risk are intrinsically related to each other and, more importantly, they are not separable. If the market value of the firm's assets unexpectedly changes—generating market risk—this affects the probability of default—generating credit risk. Conversely, if the probability of default unexpectedly changes—generating credit risk—this affects the market value of the firm—generating market risk.
Fridson, Garman and Wu (1997) used quarterly data for the period from 1971 to 1995 and found that there was a moderate, significant positive correlation between default rates and real interest rate, and a strong positive correlation between default rate and lagged 2-year real interest rate. Iscoe, Kreinin and Rosen (1999) suggest a joint market and credit risk model on the basis of the Merton model for the evaluation of risky debt. Barnhill, Papapanagiotou and Schumacher (2000) developed a model for the South African financial environment as of June 1999 in the context of financial system stability assessment. Barnhill and Gleason (2001) compared bank capital requirements estimates with an integrated market and credit risk simulation using the model proposed by Barnhill and Maxwell (1998) to those calculated under the 1988 Basel Accord and the then proposed New Accord (Basel II) for a set of 54 hypothetical banks. Their results emphasize the serious limitation of the then proposed New Accord and point towards the crucial importance of developing conceptual frameworks for undertaking integrated bank risk assessments. Barnhill and Maxwell (2000) measured credit and market risk for the whole portfolio of banks’ fixed income securities with the correlated interest rate, interest rate spread, exchange rate and credit risk by simultaneously simulating both the future financial environment and the credit rating of specific firms, using Barnhill 1998 model. Barnhill et al. (2001) applied the integrated model to Japanese banks. Alexander and Pezier (2003) developed a common risk factor model to characterize the joint distribution of market and credit risk factors for the US market (from 4 January 1999 to 31 December 2002). They found that the aggregate economic capital estimates the benefit of negative risk factor correlations (approximate reduction of 20%). Fiori and Iannotti (2008) used a Factor-Augmented Vector Autoregressive (FAVAR) approach to examine the interactions between credit and market risks, by identifying common risk drivers. Analysing quarterly data for the Italian economy over a period from March 1991 to September 2006, the study concludes that the impact of a shock in the monetary policy, that is, the 50 basis points increase in short-term interest rates, is intensified (increase 6 times) when dynamic interactions are taken into account. Analysing foreign currency loans in Austria, Breuer, Jandačka, Rheinberger and Summer (2010) indicates that simply adding up the separately measured exchange rate and default risk components underestimates the actual level of risk. For example, for a B+ rated obligor, the integrated risk measurement approach leads to an overall risk that is 1.5–7.5 times larger than the risk derived from a compartmentalized approach, in which each risk is measured separately and then added up. This bias becomes more pronounced for portfolios with lower ratings. Another example is provided by Alessandri and Drehmann (2010) who explain that when interest rates rise, more borrowers default and short-term borrowing rates react faster than long-term lending rates. Over time, however, banks may succeed in passing higher interest rates as compensation for increased credit risk on to borrowers and thus recover the interest margins. The authors’ results suggest that significant diversification benefits between interest rate and credit risk in the banking book may be achieved through this mechanism. Put differently, their data and model indicate that simply adding up the separately calculated market and credit risk components can imply a large upward bias in the estimation of overall risk over longer periods.
Recent studies also have underlined the role of interaction in determining capital requirements. Wong and Hui (2009) highlighted the role of interaction among market and credit risk while measuring the adequacy of liquidity in facing stress events using bank-level data for 12 listed banks in Hong Kong for the year 2007. Htay and Syed (2013) provided evidence of a relationship between bank risks by studying the correlation between liquidity risk, operational risk, credit risk and market risk for 10 listed banks in the United Kingdom using data for the period 2002–2011. Htay and Salman (2014) studied the correlations between operational, liquidity, credit and market risk of five listed Malaysian banks during 2002–2011. The results reported a positive correlation between liquidity, credit and market risk.
Previous Indian studies have focussed mainly on one dimension of risk, that is, either on the credit risk or on the market risk, or on the components of market risk. For example, there are studies by Patnaik and Shah (2004), Sy (2005), Makkar and Singh (2013), Prabhavathi (2013) for interest rate risk; Das and Ghosh (2007), Bodla and Verma (2009), Kumar, Arora and Lahille (2011), Arora (2012), Arora (2013) for credit risk and Sharma (2012) for market risk. The literature on the interaction of different types of risks is virtually missing in the Indian context.
Research Gap
Prior research mainly focussed on the measurement and management aspects of specific risks. No comprehensive study seems to have been conducted to examine the interaction amongst the four types of bank risks (i.e., credit, equity, interest rate and exchange rate risk), especially using bank level data. Most of the previous attempts concentrate on subsets of risks but not all of them together. Though the work on estimating risk exposures and measuring their interactions is documented for developed countries, it is virtually absent for any emerging market. Further, the impact of bank-specific factors on these measured interactions among bank risk exposures remains a relatively unexplored topic. In the absence of any significant work for the Indian banking sector, the present study will play a major role in bridging the gap. It aims to estimate the risk exposure of Indian banks to credit risk and components of market risk (equity risk, interest rate risk and exchange rate risk). We propose a comprehensive model to measure the interaction of credit risk and components of market risk. It also contributes further by assessing the bank risk interactions for banks belonging to different categories based on the primary bank characteristics, instead of only focussing on the whole banking industry, and by using bank-level data instead of an aggregate level data. The work accounts for a structural break in the data, owing to the occurrence of GFC during the study period. The study tries to find answers to issues relating to how regulation and supervision should account for these relationships between the risks and make recommendations for developing a more integrated risk management system.
Data and Variable Description
The study uses weekly data for 40 individual banks listed on the Indian stock exchange for the period of about ten years, that is, from 23 October 2004 to 1 August 2014 (see Appendix 1). The year 2004 was selected to capture the post-reform period correctly, that is, after the adoption of prudential norms for the banking sector recommended by Narasimham Committee Report (Zhao, Casu, & Ferrari, 2008). The choice of the date (23 October 2004) was guided by the availability of yield data for the BBB corporate bonds in the Indian context. It is used to develop a credit risk indicator, measured as the difference between BBB and AAA corporate yield spreads. Foreign banks operating in India are not considered in the study because as a group they behave differently on the priority sector of lending. 3 Not only that, but for the risk assessment of these banks, the financial performance of their parent banks is crucial, which lie outside the Indian legal jurisdiction.
Following the capital market approach, the credit risk factor is taken to be the difference in the yield of BBB and AAA corporate bonds; the equity risk variable is calculated as per Basel norms as the weekly change in S&P CNX 500 4 ; the interest rate risk is estimated by the change in yield of the long-term government bonds (10 years) and the exchange rate risk is calculated by the change in the rate of USD and INR. The bank percentage stock return is used as the dependent variable and was estimated using stock price data. The Bloomberg and NSE databases have been used as sources to get estimates of risk variables.
Previous literature (Flannery & James, 1984; Fraser, Madura, & Weigand, 2002; Yong, Faff, & Chalmers, 2009) suggests segregation of banks according to their balance sheet characteristics to identify potentially weak banks. Hirtle and Lopez (1999) discusses the use of CAMEL ratings as a quick and simple measure, containing information indicating the financial soundness of the bank. Bongini, Laeven and Majnoni (2002) find that CAMEL ratings 5 are strongly related to distress for a sample of East Asian banks. The key bank characteristics employed in the present study are as follows—capitalization, asset quality, management efficiency, earning level and liquidity, all of which form a part of CAMEL approach.
Following Oura, González-Hermosillo, Chan-Lau, Gudmundsson and Valckx (2013), capitalization ratio (equity to total assets) is considered as an indicator to assess and identify bank vulnerabilities. Higher equity base reduces the perceived risk exposure of the bank. Asset quality is gauged by the level of nonperforming assets (NPA; Agusman, Monroe, Gasbarro, & Zumwalt, 2008; Iannotta, Nocera, & Sironi, 2007; Klomp & Haan, 2012). Poor asset quality reflected in higher NPA would mean poor asset quality, higher provisioning and write-offs, reduced profitability and earnings and increased risk.
Marinković and Radović (2014) suggests the level of net interest margins (NIM) as an indicator of management efficiency. The higher the NIM (excess interest income over interest expense scaled by total asset), higher is the efficiency and thus lower is the risk exposure. Earnings level of the bank is represented by the level of profits earned by the bank, indicated by return on assets (ROA; see Brewer & Lee, 1986; Goodhart, 2010). Ex ante, the relationship between profitability and risk is not clear. A more profitable bank would possess more resources to mitigate the risk (Demsetz, Saidenberg, & Strahan, 1996; Keeley, 1990; Martynova, Ratnovski, & Vlahu, 2015; Repullo, 2004). On the other hand, it is also possible that higher profitability is the result of higher risk undertaken (Blum, 1999; Hellmann, Murdock, & Stiglitz, 2000; Matutes & Vives, 2000).
Renewed focus on the liquidity of banks and its impact on the bank risk after the crisis motivated us to include credit deposit ratio as a measure of liquidity in the study (van den End, 2016). Too high credit–deposit ratio would mean less liquidity available with the bank. Too low would mean the bank is paying a very high premium to maintain the level of liquidity (Imbierowic & Rauch, 2014).
Additional factors like bank size, ownership, diversification and price-to-book ratio, apart from the CAMEL ratings are also included in the analysis. Bank size is measured using natural logarithm of total assets (Arora, 2012; Das & Ghosh, 2007; Hakenes & Schnabel, 2011; Ranjan & Dhal, 2003) and the sample is divided into small and large banks using a median breakpoint. It is generally believed that large banks are less risky as they possess more systemic risk and hence enjoy greater government protection, and are supposed to be ‘too big to fail’ (Adrian & Brunnermeier, 2011). Further, they have more opportunities for diversification, both in their activities and geographical reach (DeYoung & Roland, 2001). However, complex large entities, which are difficult to monitor, can actually be exposed to a higher risk, which means that bank size is an important factor affecting risk. The ownership structure of the bank is also shown to impact the level of banks risk exposure (Arora, 2013; Das & Ghosh, 2007; Iannotta et al., 2007). The literature indicates that public banks are set up with the motive of increasing social wellbeing and inclusive economic development (Stiglitz, 1993). As such, they assist projects which are socially beneficial but do not have access to other financial resources, which makes public sector banks riskier.
The impact diversification has on the level of risk has been extensively studied earlier. DeLong (2001) concluded for a sample of the US banks that diversification fails to create shareholders’ value. DeYoung and Roland (2001) associate fee-based activities with increased revenue volatility and thus risk. In accordance with Baele, De Jonghe and Vander Vennet (2007), the ratio of non-interest income to total operating income is used to study the diversification effects on risk. Masera & Mazzoni (2014) suggest price-to-book ratio as more effective in forecasting banks’ future distress than the accounting ratios are. To analyze this argument in the Indian context, it is included as one of the classifying variables.
The bank characteristics, along with their measures, their reference in prior work and the data source are provided in Exhibit 1.
Methodology
This article adopts a two-step method. In the first step, risk exposures are estimated, and in the second step, the dynamic interaction between different risks is analyzed for the entire banking sector as well as for individual banks based on their characteristics.
To generate a measure of risk exposure of banks to credit risk (CR), equity risk (MR), interest rate risk (IR) and exchange rate risk (ER), multivariate approach is used. In the model, bank stock returns are made a function of different risk factors and its conditional volatility. GARCH (1, 1) in mean model along with the lagged interest volatility is used to estimate the conditional variance of bank stock returns.
The mean equation is as follows:
where R j,t is the weekly stock return of bank j at time t; R e,t is the weekly rate of return on the broad-based market index (S&P CNX 500) at time t; R r,t is the weekly rate of change in the 10-year long-term government bond yield at time t; R f,t is the weekly rate of change of the Indian rupee against the US dollar at time t; R c,t is the weekly change in the difference between BBB- and AAA-rated corporate bond yields at time t (Bessler & Kurmann, 2014); h j,t is the conditional variance of bank stock returns; and ε j,t is a serially uncorrelated normally distributed random error term. Coefficients β e , β r , β f and β c represents equity, interest rate, exchange rate and credit risk, respectively, and β 0 is a constant term.
The variance equation is:
where lagged squared error term, prior period conditional variance and last period’s conditional interest rate volatility determines the conditional variance of the bank stock returns.
The above equations have been estimated on a 52-week estimation window, rolling 13 weeks each time. This process results in quarterly estimates of beta coefficients, which are used as a measure of respective risk exposures. These beta measures of risk have been averaged for all banks to get a measure of each risk for the entire banking industry. Similar averaging of betas is done for banks across different categories, which are formed by various characteristics. The results of estimations with 52 weeks rolling window for the total banking industry and for banks across categories are presented in Table 1.
Mean Bank Risk Exposures at the Aggregate Level as well as Different Characteristic Based Bank Groups
As a first step, sample series are tested for stationarity using Augmented Dickey–Fuller (ADF) test, with the null hypothesis of the existence of unit root. As a robustness check, Phillips and Perron (PP) test of stationarity has also been performed. Vector autoregression (VAR) estimates the dynamic interrelation between variables while being indifferent about the choice of the dependent variable. Thus to gain further insight into how different risks interact with each other, we model all the four risks in a VAR framework and analyze the short-run and long-run dynamics of the system by computing variance decomposition (VDCs) and impulse response functions (IRFs). We introduce a dummy variable in the VAR equations, which takes the value equal to 0 before 23 October 2008 and equal to 1 after 23 October 2008 to account for the effect of the global financial crisis, as suggested by previous studies (Filardo et al., 2010; International Monetary Fund, 2010; World Bank, 2010). An important preliminary step in VAR model building and analysis is the selection of the VAR lag order. In this study, we have chosen the lag order based on the commonly used lag order selection criteria—Akaike Information Criteria (AIC). The above-mentioned methodology has been adopted for the entire banking industry and then has been replicated for different banks groups.
The descriptive statistics for key bank characteristics for whole banking sector in the pre- and post-crisis periods are given in Table 2.
Descriptive Statistics of Key Bank Specific Characteristics for the Entire Banking Industry for Both Pre and Post Crisis Periods
As can be observed from the Table 2, mean values for only EA, CD and TA show an increase in the post-crisis period, showing an enhanced regulatory oversight post the crisis on the capitalization and liquidity levels of the banks. NPA declines in the post-crisis period, pointing to, probably, bank’s efforts to improve their asset quality. A fall in the mean NIM, ROA, PB, NON of the bank post the crisis period can be observed too.
Empirical Results
We begin the discussion of empirical results with the stationarity test. The ADF and PP test results reveal that the first difference of all risk exposure series is stationary. We conclude that all series are integrated of order 1, I(1), that is, they become stationary when transformed to first-differences. To study the dynamic interaction of risk exposures, we choose to implement the most parsimonious multivariate VAR model with four lags as selected by minimization of AIC criterion for the entire banking industry. Lags for different bank groups are also identified by repeating a similar exercise. The four lags are chosen for all bank groups except less liquid banks, where three lags are selected for VAR model.
Vector Auto Regression estimates are reported in Table 3.
At the level of whole banking industry, CR is affected by its lagged values and also by lagged values of MR and ER, with the impact of IR being insignificant. Further, at bank group level, CR’s own lagged values are significant for all bank groups except in the case of large banks. Lagged values of CR are the only relevant variables that affect current CR for well-capitalized and more efficient banks. The increase in MR results in an immediate increase and later a decrease in CR for all bank groups except those which are well-capitalized, more efficient and high PB banks. That an increase in IR increases CR is observed for less capitalized, less profitable, less liquid, low PB and low NPA banks irrespective of their ownership structure, while no significant impact is observed for the rest of the groups. ER is a major factor affecting CR exposure of all bank groups except for highly diversified, more efficient, well-capitalized banks irrespective of their liquidity levels. The results show that the complete banking industry and all bank groups (except well-capitalized, well diversified, more efficient, less liquid and high PB banks) have higher credit risk post the crisis.
Results of Dynamic Interaction Between Banks’ Risks, Based on VAR Approach
For the entire banking sector, MR is affected positively by its lagged value along with the lagged value of all the other risks as well. On analysing bank groups, it is observed that except for highly profitable, high PB and high NPA banks, all bank groups exhibit a significant positive impact of MR’s own lagged values. Effect of CR on MR is significant in all cases except for highly diversified, less liquid private sector banks. An increase in CR leads to an increase in MR in the subsequent quarter for more efficient and high PB banks, and an immediate decrease in MR for well-capitalized, highly liquid and high NPA banks, irrespective of the profitability levels. IR affects MR of all bank groups except in the case of less diversified banks regardless of their ownership structure and liquidity level. ER has a significant impact on MR exposure of small sized less efficient less profitable low PB private banks irrespective of their asset quality. GFC has led to an increase in MR for the whole banking industry as well as less diversified, less capitalized, low NPA and high PB banks irrespective of operating efficiency and profitability levels.
For the entire banking sector, IR is impacted only by its lagged value (lag 1) positively. No visible interaction is observed between IR and lagged values of other risks. In the case of bank groups, results are similar for high NPA banks irrespective of their profitability levels. Interestingly, no risk variable is found to have a significant impact on the IR for well-capitalized private sector banks. A significant immediate negative impact of CR exists for highly liquid small size banks, while an immediate positive impact on less liquid banks is observed. MR has a significant positive impact at lag 1, and negative consequences at later lags on IR for large size less liquid public sector banks irrespective of the level of diversification, operating efficiency, PB ratio. Further, a significant positive impact (at lags 1 and 2) and negative consequences (at lags 3 and 4) of ER exists for less efficient, less liquid, high PB and low NPA banks. The coefficient of crisis dummy is positively significant only for less efficient, highly diversified, high PB banks suggesting an increase in the banks’ IR after the GFC.
CR, with a positive coefficient, is the only other variable apart from ER’s lagged values to have a significant impact on total banking industry ER exposure. Further, looking at the bank groups, it can be seen that ER is significantly influenced by CR in the case of more efficient high NPA and low PB public sector banks, irrespective of the profitability levels. A significant effect of MR on ER is observed for less efficient, less profitable, low PB, low NPA private sector banks. The impact of IR on ER is significant in a positive direction at later lags and negative direction at initial lags for all bank groups except more efficient, more profitable, less liquid, high NPA banks irrespective of PB ratio. ER is impacted in a significantly positive manner by its lagged values in the immediate lags and negatively at later lags for all bank groups. Less diversified, less profitable and low PB banks experience an increase in their ER exposure post the GFC.
Results of the variance decomposition analysis are reported in Table 4. For the entire banking industry, much of the variance in CR is explained by its shock in the short run. Over the short term, results for all bank groups are similar to the total banking industry. For the long term, MR explains the maximum variance followed by CR’s own variance and then ER and IR. In the long term, maximum of the variance in CR is explained by its own variance for small sized well-capitalized high PB and high NPA banks irrespective of their operating efficiency, diversification, and profitability. On the other hand, for highly liquid, less capitalized, large size, low PB and low NPA banks irrespective of ownership, MR explains the maximum CR’s variance in the long run followed by its lagged value, ER, and IR. Further, for less liquid banks, the ER explains the CR variance followed by its lagged value, MR and IR.
For the entire banking industry, MR’s innovation explains most of its variance in both short and long terms. Among the banks segregated into different groups, over the short term, MR of all bank groups is primarily affected by its variance followed by CR and there is no impact of IR and ER. Different results emerge in the long run where variance in MR is mainly explained by ER for profitable, less liquid, high NPA banks; by IR for well-capitalized banks; by CR for well-diversified banks irrespective of operating efficiency and by its variance for all other bank groups.
When observed for the whole banking industry, the variance of IR is mainly explained by its variance both in the short and long terms. Apart from its innovation, ER in the long term and MR in the short term explain some portion of IR variance. One can observe while analysing the bank groups that in the short term, IR variance is mostly defined by its innovations for all banks groups, followed by MR for less liquid, large size, private banks and followed by CR for highly liquid, high PB, and high NPA banks irrespective of operating efficiency and profitability levels. ER appears to have a significant impact on IR only for small banks. For all other bank groups, the role of other risks, apart from IR’s innovation, is limited in the short run. Over the long term, MR becomes a dominant factor explaining IR variance for less liquid large banks, CR for highly liquid banks irrespective of the level of operating efficiency and profitability and ER for high NPA small size banks. For all the other bank groups, in the long run, IR explains most of its variance, followed by CR for well-capitalized, low NPA banks irrespective of the level of diversification and followed by ER for small equity to asset ratio, public sector banks irrespective of PB ratio. In the case of private banks, MR supports IR’s innovation to explain IR variance.
Variance Decomposition Results
For the whole banking industry, ER is the dominant factor explaining its variance in both the short and long terms. IR has no visible impact in the short term, but in the long term, it displays a significant impact on ER, apart from ER’s own innovation, followed by MR and CR. In the short term, ER variance is mostly explained by its own shocks for all banks groups, followed by MR for well-diversified, more efficient, high PB banks irrespective of liquidity and otherwise by CR. IR appears to have no significant impact on ER for any bank group in the short run. On the other hand, in the long run, it can be seen that IR explains most of the exchange rate variance for less diversified, low PB banks irrespective of capitalization levels. Furthermore, CR explains most of the variance in ER for less profitable banks, and for all the other bank groups ER remains the dominant variable explaining its variance.
Results for the impulse responses function analysis support the results of VAR and variance decomposition and are available on request. These are not shown here due to space constraints.
We now attempt to explain our empirical findings related to the interaction between various bank risks. One can understand the relationship between different risks through various functions banks undertake with regard to maturity transformation and financial intermediation; for instance, transformation functions in the context of size (small size deposits collected and huge amount of loans given), maturity (banks’ liabilities are for a short period and they normally lend medium- and long-term loans) and risk (banks minimize risks). MR and CR exhibit a positive relationship with each other (Htay & Syed, 2013; Jarrow & Turnbull, 2000): when the equity values fall, MR rises and simultaneously it reduces the value of collateral, thereby increasing CR. To appreciate the feedback effect of CR on MR, one needs to understand that with an increase in CR of the bank, the NPAs also increase, which leads to an enhancement in the provisions made for NPAs. This affects the equity value of the bank, translating into a higher MR for the bank. The impact of ER on CR can be understood in light of the arguments given by Breuer et al. (2010). When rupee depreciates, a bank with a net short open position would experience a rise in the ER, but simultaneous improvement in the ability of non-financial corporate borrowers to repay reduces the CR of the bank. In contrast, an increase in CR results in an increase in ER, primarily because as the number of defaults increases, especially in foreign currency loans, the bank will be exposed to a higher ER due to unmatched open exchange position. Theory suggests that analysing economic perspective of IR is more comprehensive, where an increase in the interest rates result in the change in the value of assets and liabilities (banks usually lend for long terms and finance the loan with short-term deposits, which implies a compression in their net interest income and margins), leading to a higher interest rate risk. With rising interest rates, the economy experiences a slowdown, which is reflected in the equity valuations of the market, and MR increases. When ER increases for the bank, the MR faced by the bank also increase. This phenomenon can be understood by looking at the underlying factor affecting both risk exposures. When rupee depreciates, ER for a bank rises (assuming net short open position); simultaneously, the stock market looks unattractive for FII, leading to a fall in equity valuations and increasing the MR faced by the bank.
Summary and Concluding Observations
The lessons from the GFC, rethinking of banking regulation at the international level and bank failures having systemic effects have highlighted the issues that are affecting the current regulations of the banking sector. India, with a bank dominant financial system, has been affected relatively less by the crisis but there is an urgent need to realign the banking structure and to adopt risk management practices/strategies in a fast changing environment. Interactions among several bank risks have been highlighted in the literature and also cited as a primary reason for the crisis.
The main aim of the study is to first measure the bank risk exposure by the capital market approach and then assessing the level of interaction between these risks by using VAR analysis along with variance decomposition and IRF analysis. The analysis has been done using bank-level data for 2004–2014, covering the GFC period, both for the Indian banking industry as an aggregate and for bank groups segregated according to bank-specific characteristics such as size, ownership, diversification and valuation along with CAMEL rating factors.
The results of the study suggest that for all risk exposures, an increase in their lagged values results in higher risk exposures in future, which suggests that to reduce the risk exposure, what is needed is a continuous and consistent effort over a period of time towards the reduction of risk. Looking at the bank-specific characteristics, capitalization (Das & Ghosh, 2006; Fiordelisi et al., 2011; Klomp & Haan, 2012; Nguyen & Nghiem, 2015; Oura et al., 2013) and diversification (Pennathur et al., 2012) levels appears to be a crucial factor which determines the level of interaction and the direction of interaction among bank risk exposures for banks. Well-capitalized and well-diversified banks are observed to either have no interaction among risk exposures or to reduce the impact of adverse interaction among risks. Also there has been no significant rise in the risk exposures post the GFC for high EA banks. Interestingly, a bank with high non-interest income experience an increase in MR and IR post GFC and less diversified banks on the other hand witness a rise in their CR and ER. For banks with high NPAs, CR is the dominant risk, having an impact on other risks as well. When high NPA banks reduce their CR exposure, the impact on the total risk is muted, owing to the rise in their MR and ER exposure. Therefore, integrated credit risk management becomes imperative. This is in line with the fact that NPA are considered a measure of credit risk by several studies like Das and Ghosh (2007), Agusman et al. (2008), Agoraki, Delis and Pasiouras (2011) and Klomp and Haan (2012). As banks increase their operating efficiency, limited interaction is observed among bank risk exposures (Boutin-Dufresne et al., 2013; Fiordelisi et al., 2011; Nguyen & Nghiem, 2015). Also, as profitability reduces, banks witness increased adverse interaction among risks (Martynova, 2015; Repullo, 2004). Banks with high liquidity benefit from risk diversification as integrated risk measurement will lead to a reduction in the overall risk (Imbierowic & Rauch, 2014).
In the light of these findings, one can observe that for banks, liquidity appears to be a crucial factor affecting the risk exposure and higher liquidity with the bank ensures more cushioning and better risk-absorbing capacity. The results of the study also corroborate BASEL III initiative of introducing measures of liquidity risk (liquidity coverage ratio (LCR) for short-term liquidity and net stable funding ratio (NSFR) for long-term liquidity) in the regulatory framework. The global banking regulation has well recognized the crucial role of liquidity in ensuring the strength and stability of the banking system post the crisis.
There is mixed evidence for the impact of bank size on risk. Private banks appear to be better risk managers than public sector banks (Agusman et al., 2008; Bhaumik & Piesse, 2008; Iannotta et al., 2007; Klomp & Haan, 2012). In the case of public sector banks, increase in CR results in rising MR and ER exposures. Also, increase in the MR leads to an increase in their IR. Therefore, ignoring the interaction among risk exposures will lead to an underestimation of total risk and of the required capital. One can have important insights from their operations. Lastly, banks with high PB ratio appear to have no interaction among risk exposures (Masera & Mazzoni, 2014b).
With the banking environment changing rapidly, it becomes imperative for the banks to rationalize costs and enhance earnings, which in turn would improve their profitability, efficiency and capitalization. Banks can formulate a plan of action incorporating the following points: mobilizing low-cost deposits, both current and savings accounts; implementing initiatives to reduce costs, covering technological and human resources along with operational processes; diversifying and generating income from non-traditional sources like fee- and fund-based services, such as investment banking, insurance, housing finance, securities brokerage and other non-traditional financial services, compensating them for the shrinking interest spreads; innovating newer products and services so as to provide better customer services and leverage higher market share. Consequent to this, it is also important to improve the capital base to safeguard banks against adverse situations. Increasing profits is also a vital aspect affecting the financial health and the capital formation ability. Banks should carefully analyse their income and expenditure to be able to choose the path that they may adopt for restructuring their balance sheets.
Our results show that interactions among bank risks matter and may lead to overstatement/understatement of risk if such interactions are ignored. Adopting an integrated approach to risk measurement using the Markowitz’s portfolio theory framework, total bank risk will be a function not only of individual risk but also of their interactions.
The findings of the study are pertinent for bankers, policy-makers, regulators and academia. First, it highlights the measurement issue where the current approach measures risk in isolation instead of as integrated, which leads to the wrong estimation of risk. Second, it warrants the need for an enhanced understanding of the dynamic interaction among bank risk exposures, which is necessary for developing an integrated risk management framework. The outcome of the study suggests that the future banking reforms should take into account the interactions amongst different types of risk or the probability of failing to deliver the desired results is high. Moreover, it may also be useful to look at the strategies to be adopted by the bank in an integrated manner to mitigate multiple risks simultaneously. Policy-makers should focus on creating an enabling framework for banks and encourage them to take integrated risk management approach. The results also highlight that all the stakeholders should take cognisance of the fact that minimum capital requirement should not be set for different risks separately, and it should instead account for the dynamic interactions between risks. Acknowledging this would ensure better capital provisioning by banks. Efficient and effective risk management in banks would enable them to benefit from enhanced public trust.
Further research should focus on new risks such as operational risk identified in Basel II, as well as liquidity risk, systemic risk, regulatory risk, reputational risk, etc., highlighted in Basel III. Moreover, a similar analysis can be performed across countries to facilitate cross-country comparison. To create a more responsive, efficient and integrated risk management approach, future studies should address the concerns mentioned above.
Footnotes
Exhibit 1
| Bank characteristics | Measure | Source | Studies |
| Capitalization | Equity to Assets ratio (EA) | Prowess | Fiordelisi, Marques-Ibanez and Molyneux (2011), Oura et al. (2013); Nguyen and Nghiem (2015) |
| Asset Quality | Ratio of Net NPA to Net Advances (NPA) | RBI | Iannotta et al. (2007), Klomp and Haan (2012), Agusman et al. (2008) |
| Management Efficiency | Net Interest Margins (NIM) | RBI | Fiordelisi et al. (2011); Boutin-Dufresne, Peña, Williams, and Zawisza (2013); Marinkovi´c and Radovi´c (2014); Nguyen and Nghiem (2015) |
| Earnings/ Profitability | Return on Assets (ROA) | RBI | Brewer and Lee (1986), Goodhart (2010) |
| Liquidity | Credit to Deposit ratio (CD ratio) | RBI | Avramova and Le Leslé (2012); van den End (2016) |
| Ownership | As per RBI Categories | RBI | Iannotta et al. (2007), Bhaumik and Piesse (2008); Das and Ghosh (2007), Arora (2013) |
| Size | Natural log of total assets | RBI | Ranjan and Dhal (2003), Das and Ghosh (2007), Hakenes and Schnabel (2011), and Arora (2012) |
| Diversification | Ratio of Non-Interest Income to Total Assets | RBI | DeYoung and Roland (2001), Baele et al. (2007), Chiorazzo, Milani and Salvini (2008); Pennathur, Subrahmanyam and Vishwasrao (2012) |
| Valuation | Price to Book ratio (PB ratio) | Prowess | Masera and Mazzoni (2014) |
Appendix 1
List of Public Sector and Private Listed Banks in India