Non-performing Loans and Moral Hazard in the Indian Banking Sector: A Threshold Panel Regression Approach

Abstract

This article attempts to examine the effect of non-performing assets (NPA) on behaviour of banks in India. The objectives of this article is to test if lending choices of Indian Banks demonstrate moral hazard and to test whether an increase in NPA ratio of banks raises riskier bank lending. We employ a threshold panel data regression model on a data set retrieved from the Reserve bank of India, which covered 45 commercial banks during the period 2009–2015, to test if lending choices of Indian banks demonstrate moral hazard. The results establish that the moral hazard hypothesis does not hold true for the given sample of India Banks, suggesting that an increase in the NPA ratio does not potentially increase riskier lending in sample banks. We find empirical evidence for the notion that ‘too-big-to-fail’ banks possibly have certain incentives to take higher risks and thus have higher NPA ratios. Graphical approach to NPA threshold explanation reveals presence of threshold; however, it could not be statistically established. Future implications of findings are evaluated. The study seminally adds to the empirical literature on use of fixed effects threshold panel data regression model in the context of Indian banks.

Keywords

Non-performing loans moral hazard lending behaviour threshold panel regression

Introduction

The problem of non-performing loans has vexed the Indian banks long before the economic liberalization of 1991 (Ghosh & Saggar, 1998), and in more recent times post the global financial crises of 2008 the problem has burgeoned multiple times (Rajan, 2016a, b). Banking literature determinedly identifies that levitated degree of non-performing loans are antecedent to bank failure, as observed by Demirguc-Kunt (1989) and Barr et al. (1994). Iyer and Peydro (2011) find robust evidence that financial contagion spreads prominently through the channel of interbank linkages; expressly, the impact of failure of one bank may rapidly spread to other financial institutions, causing a chain effect and possibly quivering the strength of the entire financial system locally or even internationally. Frank and Hesse (2009), Cheung et al. (2010) and Chen et al. (2016) observe that the global financial crisis (2008) has demonstrated how delicate the entire worldwide interlinked financial system could be, and that a financial crisis instigated in any country could adversely impact the stability of the global banking system and could be possibly disastrous for the real economy. There are ample empirical studies and statistical data to substantiate the linkages that financial system development, and banking reforms have pointedly upgraded the economic growth in India (Bapat, 2012; Deolalkar, 1998; Ghosh & Saggar, 1998). Regarding developing economies, Villanueva and Mirakhor (1990) point at the importance of non-performing loans and observe that any reform in the banking and financial sector of an economy requires thorough familiarity of non-performing loans and its possible implications for banking sector and ultimately sustainable financial stability. In a study of Indian banks, Berger et al. (2008) observe that government-owned banks often bear abject loan repayment and therefore suffer with high degree of problem loans in their asset portfolios. Kornai (1986) discusses in detail about how institutions/persons behave when the unfavourable downside of their actions is covered by the state. Kornai (1986) discourses that soft budget constraints (how financial constraints of institutions/private firms become thin) are visible if the stringent association of expenditure (costs) and earnings (revenues) are diluted and relaxed, in anticipation (highly probable surety) that excess expenditure over budgeted expenditures will be paid by someone else or some other institution, typically by the government or another government institution. In terms of developing economies, Arun and Turner (2004) observe that, in modern banking systems, the governments usually offer and agree for deposit insurance, in spite of that bank managers have incentives for speculatively risk-taking at government’s expense. The moral hazard problem in the Indian banking sector thus will likely be affected by the soft budget constraints associated with state ownership of banks in India.

Subbarao (2009) deliberates that Indian banks have no direct exposure to the sub-prime mortgage assets or to the financial institutions that failed during the Global Financial crises (2008, 2009) and have very limited exposure towards securitized assets in terms of off-balance sheet items; nonetheless, the contagion of the crisis found entry in India through the financial and real channel. Adding to the above observation, Bajpai (2011) inscribes that though India had almost zero exposure to the institutions, they failed during the global financial crises or had zero exposure to the toxic assets associated with sub-prime mortgages and could not successfully insulate itself from the hostile changes during and immediately after the culmination of the global crisis. It is therefore timely to consider the existence of moral hazard problems and issues in Indian banks and to find out how it may possibly be linked with non-performing loans.

Istrate et al. (2007) discern that India has approached the non-performing loans problem with both interim (relatively short-term) and lasting (long-term) measures. The Indian government resorted to capital injection in its public banks as a short-term measure. As long-term measures, the government has restricted private investment in public banks, has reduced its directed lending directives, has upgraded regulatory oversight, brought multi-pronged deregulation in public banks and has introduced prudential norms in harmony with international Basel norms. Istrate et al. (2007), however, further discern short-term measures, like capital infusion, may only be sufficient in cases where problem loans are predominantly situational. If, however, problem loans are chiefly as a result of systemic causes, then broader reforms are required to prevent the accretion of new non-performing loan stocks.

According to the Government of India’s Department of Financial Services, between the years 2009 and 2015, the Indian government has infused a substantial volume of ₹697.24 billion as capital in the state-owned banks (Department of Financial Services, 2015). RBI data report that the gross non-performing assets ratio (GNPA ratio—calculated as GNP assets as a ratio of gross advances) for all scheduled commercial banks (not including regional rural banks and co-operative banks) at the end of 2015 was registered at 4.27 (amounting to ₹3,229.16 billion), while the same ratio was registered at 2.31 (amounting to ₹699.53 billion) at the end of 2009, an explicit visible jump in the NPA ratio. The recent data from the RBI registered the GNPA ratio for all Scheduled Commercial Banks at the end of 2018, at a considerably high level of 11.18, amounting to ₹10,361.87 billion. Another major concern for regulators and government is that the distribution of NPA ratio is irregular across bank groups with the bulk of the NPA problem skewed towards the state-owned Indian banks. For instance, at the end of 2015, the GNPA ratio for Indian public sector banks (PSU banks) was registered at 5.26; for another set of state-owned banks, the State Bank of India and its associate banks (SBI Group), the GNPA ratio was registered at 4.28, while it was registered at 2.10 and 3.20 for private sector banks (private banks) and foreign banks operating in India (foreign banks), respectively.

Very few international studies exist, which employ panel threshold regression model, encompassing non-performing loans and bank-specific factors. In the Indian context, such studies are strikingly limited. This article delivers originality in terms of the time period under coverage for this article and is exclusive in the context of threshold regression, involving NPAs and other endogenous banking variables.

In view of the earlier discussion, we apply a threshold panel regression model to Indian commercial banks and test whether troubled banks have incentives to take unnecessary risks, thereby increasing losses and imparting threats of insolvency. Our empirical findings have vital inferences for Indian banks, which are marred with high levels of NPAs and any potential moral hazard problems.

Threshold regression is applied to a data set of 45 Indian commercial banks with coverage years from 2009 to 2015, and we probe if banks’ lending behaviour is sensitive to reaching a particular threshold level of NPAs and do banks with higher NPA ratio tend to adopt a more aggressive and riskier lending strategy? We conjecture that a bank having higher NPA ratio inhabits unnecessary risk-taking in an attempt to cover the damages related with already high NPAs and thus create a vicious circle where, rather than decreasing, the NPAs further rise owing to higher riskier loan growth.

The literature survey conducted for this study could not find any similar studies for the reference time period considered in this article, which covers the duration of the impact of the Global Financial Crises of 2008 and goes on to 2015 when the Reserve Bank of India (RBI) conducted the Asset Quality Review of Indian public sector banks, which resulted in higher NPA reporting in 2016.

This article is divided and presented in the following manner. The next section is a brief review of related studies and empirical literature in this area, followed by objectives and rationale of the study. Subsequent sections explain the methodology and adopted empirical approach and the data used, followed by data analysis and the computation of empirical results. The subsequent section discusses the findings, present conclusion and sums up the article outlining the important contributions of the study to the literature.

Review of Literature

The relation between moral hazard problem, credit growth and NPAs is explained by Jensen and Meckling (1976) who have used a bifurcation in description of the moral hazard problem. According to Jensen and Meckling (1976), a riskier lending decision compared to optimal lending decision may have incentives for the bankers. Moral hazard arises, first, when the bankers have a tendency of receiving private benefits out of vested interests, that is, when managers have vested interests in certain projects, resulting in investments in favoured projects or lapses in proper loan monitoring of those projects. And, second, when bank equity owners may want to invest in riskier loans and subsequently shift the burden towards bank depositors.

A very rampant discussion in academic literature regarding the observance of moral hazard behaviour leads to the conclusion that such a behaviour is implied in nature through a bank’s behaviour. As discussed by Jensen and Meckling (1976), the chief indicator of moral hazard is unwarranted raised riskier loans. Loan growth rate may hold important implications for indication of riskier lending decisions (Foos et al., 2010). They test the impact of loan growth on asset riskiness for 14 major Western countries’ banks covering more than 10,000 different banks for the period 1997–2005. Authors conclude credit growth to be a crucial driver of bank risk. Foos et al. (2010) specifically test three hypotheses and find empirical support for the view that loan growth leads to (a) a rise in loan loss provisions post a 3-year lapse, (b) a comparative decrease in interest income and (c) a decrease in capital ratios. Their further analysis also disclose that loan growth has adverse effect on risk-adjusted interest income.

Academic literature has also discussed the short-run and long-run impact of multiple lags of loan growth on portfolio asset quality. Clair (1992), in his study of Texas banks, observed that loan growth improves credit quality of lending portfolio at the outset, but measured with a lapse of certain time period in effect sinks the credit quality. The observation infers that early and prompt detection of deterioration in credit quality is a challenge for the bankers, particularly when there is a positive initial rise in credit quality. Clair (1992) finds an exception to this observation for banks whose equity position was strong; banks which were rapidly growing and had strong capital ratios did not show evidence of deterioration in asset quality. Clair (1992) thus discusses about the premise that banks with low capital ratios may be more prone to moral hazard issues in the expectation that higher risks may yield higher returns—a premise which was empirically established by Berger and DeYoung (1997), that for thinly capitalized banks, a decrease in bank equity ratios generally predate increases in non-performing loans. Nier and Baumann (2006) attribute implicit government guarantees as a cause for higher problem loans for banks which are considered ‘too big to fail’. Authors conclude that for a given level of equity ratio (capital adequacy), the government-backed banks choose such assets that have a higher probability of default.

Apart from endogenous factors directly related to bank assets, there are certain other determinants that link the ‘theory of the firm’ and ‘agency concepts’ in management theory with NPA problem in banks. Gorton and Rosen (1995), in their study on US banks, attribute corporate control problems for risky, low return lending decisions in unprofitable banks. Authors suggest that such poor lending decisions lead to an erosion of bank capital, and the tipping point arises when enough capital is eroded, at which point a flurry of unnecessary riskier lending is instigated as is predicted by the moral hazard hypothesis. Gorton and Rosen (1995) conclude that in the context of large US banks, existence of moral hazard can be empirically attributable to corporate control problems.

Yet, another tranche of literature approaches the moral hazard problem in banks with view of financial structure of banks and risk shifting. Duran and Lozano-Vivas (2015) explain that in a certain situation if the capital and assets of a bank are assumed to be specified, then any alteration in the bank’s lending portfolio would deduce that in an attempt to earn higher returns, the shareholders are shifting the risk to debt holders. In due course, higher risk-taking eventually yields to be profitable, then the shareholders enjoy most of the profits, but any probable losses are primarily borne by deposit holders. Duran and Lozano-Vivas (2015) further observe that shareholders have a higher tendency of risk shifting if they have lower equity stakes, building up on the premise that thinly capitalized banks’ equity holders have stronger tendency of unnecessary risk-taking. Eventually, moral hazard conflict between shareholders and depositors (bank creditors) heightens when banks are thinly capitalized and have raised riskier lending.

Certain other studies approach the NPA problems in banks from the perspective of ‘managerial competition theory’. Boyd and De Nicolo (2005) study the effect of bank competition and risk-taking incentives of banks. They suggest that reduced bank competitiveness leads to riskier lending decisions. This happens because as banks’ markets become more concentrated and competition declines, they tend to charge higher interest rates. Higher interest rates eventually result in higher bankruptcy risk for bank borrowers. Moral hazard incentives on the borrowers’ part is visible in such conditions where the borrowers optimally increase their own risk of failure.

More recent studies which employ threshold panel data regression include Piatti and Cincinelli (2019) and Bardhan et al. (2019). Piatti and Cincinelli (2019) employ panel threshold model to a data set of 298 Italian banks for the period 2006–2014 to examine if banks’ credit processing quality results in non-performing loans reaching a certain threshold level and report that if non-performing loans remain below a preset value, then an increase in the quality of credit and loan monitoring reduces the bad loans ratio; however, if the bad loans ratio surpasses a preset threshold, this relationship reverses, and now an increase in monitoring of loans results in an increase in bad loans. Bardhan et al. (2019) employ threshold regression model of Hansen (1999) on an unbalanced panel data set of 82 Indian banks for the period from 1996 to 2011 to examine the role of bank-specific endogenous factors on non-performing loans and report that above a preset value, both—capital adequacy ratio and loan growth rate—wield negative impact on bad loans.

The above-mentioned brief literature survey points that the degree of non-performing loans can be a significant predictor of bank behaviour. When faced with higher problem loans than usually acceptable, the banks may behave differently—leaning towards increased risky assets in their asset portfolios. The magnitude of problem loans therefore is a prominent identifier for recognizing moral hazard in lending decisions for banks. This article, therefore, identifies banks’ risky lending behaviour and identifies moral hazard through analysis of threshold levels of banks’ non-performing loans.

Objectives

The objectives of this article are:

To test if lending choices of Indian banks demonstrate moral hazard and

To test whether an increase in NPA ratio of banks raises riskier bank lending.

Rationale of the Study

The history of the Indian banking sector has been indicative of the fact that banks in India—whether public or private—have enjoyed substantial governmental support in the form of explicit capital infusion as in the case of public sector banks or implicit bail outs such as in the recent cases of IDBI bank (2019) and Yes Bank (2020). Academic literature suggests and empirically discusses that capital infusions and bail outs may prompt a bank to take raised risks, thereby inducing a moral hazard (Haq & Heaney, 2012; Williams, 2014; Zhang et al., 2016).

In the context of Indian banking sector, Chavan and Gambacorta (2016) discuss that banks’ recapitalization funded by public funds are linked with a decline in banks’ asset quality. It may be so that after recapitalization, banks may uncover their already-existing but covered NPAs—which in effect is the underlying effect of the moral hazard problem. Their observation is in concurrence with literature on banks’ recapitalization and bail-out packages (Flannery, 1989; Gennotte & Pyle, 1991; Samantaraya, 2016).

The purpose of this article therefore is to test empirically the conjecture that explicit or implicit guarantee to banks may induce moral hazard, which can be captured quantitatively in their lending behaviour. Thus, we test thresholds in NPA movement to determine the behaviour of banks’ through their lending decisions.

Methodology: Data source, Sample frame and Empirical model

Theoretical Framework and Empirical Model

The inherent reference of a threshold can be perceived in Jensen and Meckling (1976) Theory of the Firm, as they mention of managers’ deviations from interests of other stakeholders for personal incentives. Berger and DeYoung (1997) conclude that for the banking industry as a whole, decreases in measured cost efficiency generally predate increases in non-performing loans. They establish that banks’ bad management practices cause (a) excess expenditures and (b) substandard underwriting and poor monitoring practices, causing increased problem loans.

The moral hazard premise mentions that if non-performing loans are worsening, then the managers would try to cover the losses by resorting to increased lending and raising their overall risk levels in the process. To identify if such a behaviour exists (eventually identifying if moral hazard exists), one way is to scrutinize if there exists a certain level of threshold value of NPA ratio, such that beyond that level of NPAs, banks opt to take elevated levels of risks in lending decisions, eventually causing a worsened NPA ratio.

This study therefore deploys a threshold regression model to identify moral hazard problems in sample banks. The threshold panel regression model aims to divide individual observations into regimes (classes), which are conditioned on the value of a predefined variable. The model used in this article is based on Hansen (1999), which is a proven effective instrument for investigating likely asymmetric effects. This article employs the methodology as has been deployed by Zhang et al. (2016) in the context of moral hazard and non-performing loans for a sample of Chinese commercial banks.

Given a balanced panel data (y: dependent variable; x: independent variable; γ—threshold value; i for cross-sectional index; and t for the time series part), the structural equation can be written as:

Y_{i, t} = C_{i} + β_{1} x_{i, t} I (q_{i, t} \leq γ) + β_{2} x_{i, t} I (q_{i, t} > γ) + ε_{i, t}

(1)

where I(.) represents the indicator function such that;

I(.) = 1; if statement in bracket is true;

I(…) = 0; if statement in bracket is false; and

q_i,t = predefined threshold variable.

The threshold value is selected endogenously. A partial threshold effect is also allowed.

Based on basic structural Equation (1) the advanced estimation equation can be depicted as:

{NPA}_{i, t} = \sum_{j - 0}^{M} β_{1, j} {LGR}_{i, t - j} ({NPA}_{i, t - 1} \leq γ) + \sum_{j - 0}^{M} β_{2, j} {LGR}_{i, t - j} ({NPA}_{i, t - 1} > γ) + θ^{'} X_{i, t} + ε_{i, t}

(2)

where NPA = non-performing assets; LGR = loan growth rate; M represents lags in LGR data, M = 0 represents no lag in LGR data; and γ = threshold value.

Table 1.

Descriptive Statistics

Variables	N	Mean	Std. Deviation	Minimum	Maximum
NPA	315	3.526169	14.59301	0.2003318	260
CAR	315	14.50032	14.38451	7.52	260
LGR	315	126.059	1491.936	−97.0242	26000
DGR	315	25.43461	94.9574	−90.9544	1543.156
ER	315	7.696064	14.53416	3.543706	260
Size	315	6.757695	14.32488	4.23222	260

Source: The authors.

Notes: NPA = Non-performing assets (gross NPA ratio), CAR = capital adequacy ratio, LGR = loan growth rate, DGR = deposit growth rate, Size = log of total assets, ER = equity ratio.

The level of threshold variable is preset to be the last period’s NPA ratio. X represents a vector containing explanatory variables.

If banks functioning above the threshold value γ in Equation (2), the decision process will be given by β₂ and not by β₁.

The first explanatory variable is the credit growth rate. Based on the study by Foos et al. (2010), which conclude that above-normal credit growth might result in substantial consequent loan losses with a lag of 2–4 years, the present study hypothesizes a significant association of banks’ credit growth rate levels and the level of NPA ratio in Indian banks.

The second explanatory variable is bank size. Multiple studies have argued that bank size substantially impacts NPA ratio (Dhananjaya, 2019; Rajan & Dhal, 2003; Salas & Saurina, 2002). Based on Louzis et al. (2012), a positive association between bank size and NPA ratio is expected.

Equity ratio (computed as capital to total assets) is employed as one of the possible NPA determinants, with a negative expected impact. This expectation is based on Swami et al. (2019) and Salas and Saurina’s (2002) argument that a higher level of capital adequacy ratio (CAR) (or equity ratio) will generally indicate that a bank will have lower NPAs and is compared to its peer as a safer enterprise.

Bank deposits are considered another explanatory variable based on arguments by Soedarmono et al. (2012) that there is a positive association between deposit growth rate and loan loss provisions.

Data Source and Sample Frame (Descriptive Variables)

Data are obtained from the RBI’s database on Indian economy. Hansen’s (1999) threshold model requires a balanced set of panel data, and therefore some banks (observations) were dropped from the sample, making the sample comprising 45 commercial banks for the period from 2009 to 2015. Data set comprise 6 banks of the State Bank of India and its associate banks, 20 public sector banks and 19 private sector banks. The set of foreign banks operating in India, regional rural banks and cooperative banks is excluded from the sample to maintain uniformity and homogeneity in data.

Analysis

Empirical Results

The threshold variable in the threshold regression model is preset to the last period’s NPA ratio (1-year lagged NPA) in an attempt to identify banks with high NPA ratios that may conceivably behave contrarily from those with low NPA ratios. Theoretical discussions have led to the supposition that bank managers have incentives to take mindless risks in lending business, owing to losses in bank operations when the bank has relatively large NPA ratios. Such incentives, being qualitative in nature, may be indirectly discerned based on bank behaviour, and any deviation from the expected behaviour is tantamount to characterization of moral hazard.

This study employs four models (threshold panel models), viz. models 1, 2 and models 3, 4, based on Equation (2) mentioned earlier. We employ baseline models for the sake of basic conception where M is set to 0 (contemporaneous values of loan growth rate—LGR). Model 2 is similar to model 1 but replaces ER with CAR to check stability of model 1. Dependent variables in all models are represented in current NPA ratios. Models 3 and 4 are similar to models 1 and 2; model 3 replaces threshold variable from 1-year lagged NPA to 1-year lagged CAR, model 3 uses CAR as one of the exploratory variable, while model 4 replaces CAR with ER to test the stability of model 3.

Threshold Estimations

Threshold estimations are preformed based on Hansen (1999) and tests the null hypothesis (H₀: linear regression model or no threshold model) that no threshold effect is present in Models 1–4. The alternative hypothesis thus becomes a test of single-threshold regression model. If a single-threshold effects model is found significant, then tests for second or third threshold effects may be performed. Table 2 describes the variables used in the threshold models in this study.

Table 2.

Estimation Parameters

Estimation Model	Dependent Variable	Independent Variable	Regime Dependent Variable (based on Hansen, 1999)	Threshold Variable
Model 1	NPA ratio	CAR, DGR, Size	Loan growth rate	LNPA (1 year lagged NPA ratios)
Model 2	NPA ratio	ER, DGR, Size	Loan growth rate	LNPA (1 year lagged NPA ratios)
Model 3	NPA ratio	CAR, DGR, Size	Loan growth rate	LCAR(1 year lagged CAR ratios)
Model 4	NPA ratio	ER, DGR, Size	Loan growth rate	LCAR (1 year lagged CAR ratios)

Source: The author.

Table 3.

Estimation of Single-threshold Effects Model

Model	Threshold (Gamma)	Conf. Interval 95%	Residual Sum of Squares	p-Value
1 (NPA)	1.4619	{1.4425; 1.4709}	522.2118	0.6967
2 (NPA)	1.4619	{1.4425; 1.4709}	519.0293	0.7167
3 (CAR)	13.12	{12.7650;13.1300}	518.4724	0.1367
4 (CAR)	13.12	{13.1100;13.1300}	523.0035	0.2167

Source: The authors.

Threshold panel regression requires that empirical analysis must begin with spotting threshold effects and setting up of threshold for respective models. Table 3 accounts the calculated threshold values for the Models 1–4 and corresponding calculated p-values. ‘Single’ corresponds to H₀ (linear model) and Ha (single threshold model) (Wang, 2015). The reported p-values are much above the significant p-values, and hence the null hypothesis cannot be rejected in models 1–4. These outcomes endorse the non-existence of the threshold effect in contrast to the linear model.

Regression Results

Table 4.

Model 1: NPA Ratio as Threshold

NPA	Coef.	Std. Err.	T	p > t	[95% Conf. Interval]
CAR	−0.0500836	0.031855	−1.57	0.117	−0.1128069 to 0.0126398
DGR	0.0005448	0.0009631	0.95	0.342	−0.0009788 to 0.002814
Size	1.34715	0.2148767	6.27	0.000*	0.9240525 to 1.770247
LGR	−0.0158404	0.0146811	−1.08	0.282	−0.0447478 to 0.013067
Constant	−4.58976	1.368229	−9.03	0.001*	−7.28384 to −1.895685

Source: The authors.

Note: *Significant at α = 0.05.

Table 5.

Model 2: NPA Ratio as Threshold

NPA	Coef.	Std. Err.	T	p > t	[95% Conf. Interval]
ER	0.0870013	0.057498	1.51	0.131	−0.0262137 to 0.2002163
DGR	−0.0015666	0.0013027	−1.20	0.230	−0.0041317 to 0.0009984
Size	1.30086	0.2142264	6.07	0.000*	0.8790427 to 1.722677
LGR	0.0000455	0.0009328	0.05	0.961	−0.0017912 to 0.0018822
Constant	−5.527426	1.267193	−4.36	0.000*	−8.02256 to −3.032292

Source: The authors.

Note: *Significant at α = 0.05.

Table 6.

Model 3: CAR Ratio as Threshold

NPA	Coef.	Std. Err.	T	p > t	[95% Conf. Interval]
CAR	−0.0471673	0.0295	−1.60	0.111	−0.1052534 to 0.0109189
DGR	0.0006175	0.0008846	0.70	0.486	−0.0011243 to 0.0023592
Size	1.069822	0.0399748	26.76	0.000	0.991111 to 1.148534
LGR	0.012012	0.0163028	0.74	0.462	−0.0200885 to 0.0441126
Constant	−2.868467	0.3068907	−9.35	0.000*	−3.472742 to −2.264191

Source: The authors.

Note: *Significant at α = 0.05.

Table 7.

Model 4 : CAR Ratio as Threshold

NPA	Coef.	Std. Err.	T	p > t	[95% Conf. Interval]
ER	0.0644706	0.0555536	1.16	0.247	−0.0449158 to 0.173857
DGR	0.0004618	0.0008876	0.52	0.603	−0.0012859 to 0.0022096
Size	0.9628361	0.061973	15.54	0.000*	0.8408096 to 1.084863
LGR	0.0152499	0.0163327	0.93	0.351	−0.0169095 to 0.0474094
Constant	−3.325235	0.198195	−16.78	0.000*	−3.715486 to −2.934985

Source: The authors.

Note: *Significant at α = 0.05.

The second part of the empirical analysis reports regression results. The threshold estimation tests could not conclude the presence of a threshold effect. Tables 4 –7 report the regression results for the four models. Tables 4-5 (NPA ratio as the threshold) exhibit regression results for models 1 and 2. Regression effects report that the only important exploratory variable is the bank size (positive and significant for dependent variable NPA ratio). The larger the bank, the greater its NPA ratio would be. The loan growth rate (regime-dependent variable) was not found to be statistically significant in models 1 and 2.

Tables 6 and 7 (CAR as the threshold) exhibit regression results for models 3 and 4. Regression effects report similar results as models 1 and 2 (Tables 4 and 5). Bank size turns out to be the only important exploratory variable (positive and significant for dependent variable NPA ratio) in models 4 and 5. The larger the bank, the greater its NPA ratio would be. Similar to regression results in models 1 and 2, the Loan Growth Rate was not found to be statistically significant in models 3 and 4.

Figures 1 and 2 (LR statistic—likelihood ratio statistic on vertical axis and dependent variable NPA parameters on horizontal axis) are the graphical representations of first and second NPA thresholds (confidence interval), respectively, for model 1. Existence of NPA threshold is visible in both Figures 1 and 2; however, statistical significance of the existence of a threshold could not be established. Similarly, Figures 3 and 4 graphically represent the first and second NPA thresholds (confidence interval), respectively, for model 2. Similar to Figures 1 and 2, the existence of NPA threshold is visible in both the Figures 3 and 4; however, it could not be established empirically.

Figures 5 and 6 are the graphical representation of first and second NPA thresholds (confidence interval), respectively, for model 3. Figures clearly depict non-existence of any NPA threshold for the computed model. Similarly, Figures 7 and 8 graphically represent the first and second NPA thresholds (confidence interval), respectively, for model 4. Figures depict non-existence of any NPA threshold for the computed model.

Discussion and Conclusions

The empirical results presented in this study cover data from 2009 to 2015, casing the period post the global financial crises era. The study conforms with the findings of Louzis et al. (2012) with reference to bank size and NPA ratios; the bigger a bank, higher are the NPA ratio of the bank. Thus, we find support for the notion that ‘too-big-to-fail’ banks may have incentives to take higher risks as has also been reported by Nier and Baumann (2006). There is a significant presence of state-owned banks in the Indian banking system. The state-owned banks are affected directly by the government policies. Despite the above specifics, the present study did not conform to the moral hazard hypothesis in Indian commercial banks for the said period. The linearity in the panel data concludes the non-existence of NPA threshold in the lending behaviour of Indian banks, thus concluding that a rising NPA ratio does not raise riskier lending in Indian commercial banks. The results of this study are also in contrast to the results reported by Nguyen and Nghiem (2015) in the context of moral hazard in Indian banks, which reported the presence of moral hazard in Indian banks (a decrease in capital ratio is generally followed by an increase in the insolvency risk).

The following practical and policy implications arise from the findings of this study. (a) Despite having significant presence of state-owned banks in Indian banking space, which indicates possible fertile grounds for manifestation of the moral hazard problems, Indian banks do not engage in riskier lending decisions arising out of soft budget constraints, clearing indicating the resilience and robustness of Indian banking coupled with the vigorous prudential regulatory norms and (b) if bigger banks tend to have higher NPA ratios, the Indian banking space thus requires small and mid-size banks, and the findings are a clear indication that policy-driven public sector bank mergers may create ‘too-big-to-fail’ entities with possible higher NPAs.

The scope for further future studies arises with the availability of data for successive years, when more consistent and robust threshold values could be obtained and would imply existence or non-existence of moral hazard based on threshold panel regression, setting the agenda for useful future research plan. This study becomes the basis for detection of moral hazard problems in Indian banks. The threshold panel data models and empirical findings may work as a robust benchmark for future studies.

Economic Implications

The results of the study indicate that bigger banks in India may lead to the creation of ‘too-big-to-fail’ entities. The case of certain banks such as the Yes Bank, in particular, is a recent example where banks, even though in the private sector, are bailed out by policy decisions by the regulators and other supervisory authorities, so that a contagion may not spread. The importance of this study is also highlighted by the present economic situation in the Indian banking space, posed particularly by the restrictions imposed and subsequent bail out of Yes Bank, where initial literature suggests clear existence of moral hazard and free-rider problem to exaggerated devastating levels. The recent mergers of public sector banks in India raise these concerns in academic literature. The creation of these merged entities opens further research areas in the context of synergies and fall-out of these mergers. Furthermore, it is suggested through this study that a revision of the current prudential and risk framework related to lending be thoroughly reviewed by all stakeholders in light of the newly created merged public sector banks in India.

Footnotes

Acknowledgement

The authors are grateful to the anonymous referees of the journal for their extremely useful suggestions to improve the quality of the article. Usual disclaimers apply.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

ORCID iD

Robin Thomas

References

Arun

T. G.

, & Turner

(2002). Financial liberalisation in India. Journal of International Banking Regulations , 4(2), 183–188.

Arun

T. G.

, & Turner

J. D.

(2004). Corporate governance of banks in developing economies: Concepts and issues. Corporate Governance: An International Review , 12(3), 371–377.

Bajpai

(2011). Global financial crisis, its impact on India and the policy response (Working Papers Series, Working Paper No. 5), Columbia Global Centers, South Asia, Columbia University.

Bapat

(2012). Efficiency for Indian public sector and private sector banks in India: Assessment of impact of global financial crisis. International Journal of Business Performance Management , 13(3–4), 330–340.

Bardhan

, Sharma

, & Mukherjee

(2019). Threshold effect of bank-specific determinants of non-performing assets: An application in Indian banking. Journal of Emerging Market Finance , 18(1 suppl), S1–S34.

Barr

R. S.

, Seiford

L. M.

, & Siems

T. F.

(1994). Forecasting bank failure: A non-parametric frontier estimation approach. Recherches Economiques deLouvain/Louvain Economic Review , 60(4), 417–429.

Berger

A. N.

, & DeYoung

(1997). Problem loans and cost efficiency in commercial banks. Journal of Banking & Finance , 21,(6) 849–870.

Berger

A. N.

, Klapper

L. F.

, Peria

M. S. M.

, & Zaidi

(2008). Bank ownership type and banking relationships. Journal of Financial Intermediation , 17(1), 37–62.

Boyd

J. H.

, & De Nicolo

(2005). The theory of bank risk taking and competition revisited. The Journal of Finance , 60(3), 1329–1343.

10.

Chavan

, & Gambacorta

(2016). Bank lending and loan quality: The case of India. (BIS Working Paper No. 595). Monetary and Economic Department.

11.

Chen

, Filardo

, He

, & Zhu

(2016). Financial crisis, US unconventional monetary policy and international spillovers. Journal of International Money and Finance , 67, 62–81.

12.

Cheung

, Fung

, & Tsai

S. C.

(2010). Global capital market interdependence and spillover effect of credit risk: Evidence from the 2007–2009 global financial crisis. Applied Financial Economics , 20(1–2), 85–103.

13.

Clair

R. T.

(1992). Loan growth and loan quality: Some preliminary evidence from Texas banks. Economic Review, Federal Reserve Bank of Dallas, Third Quarter , 1992, 9–22.

14.

Demirguc-Kunt

(1989). Deposit-institution failures: A review of empirical literature. Economic Review , 25(4), 2–18.

15.

Deolalkar

G. H.

(1998). The Indian Banking Sector. Rising to the challenge in Asia: A study of financial markets. Study commissioned by Asian Development Bank. Publication Stock No. 061099 ISBN 971-561-240-7.

16.

Department of Financial Services. (2015, September). Capital infusion in public sector banks by the government . Ministry of Finance, Government of India. https://data.gov.in/catalog/capital-infusion-public-sector-banks-government

17.

Dhananjaya

(2019). Corporate distress and non-performing assets in India. Global Business Review , doi:10.1177/0972150918812553

18.

Duran

M. A.

, & Lozano-Vivas

(2015). Moral hazard and the financial structure of banks. Journal of International Financial Markets, Institutions and Money , 34, 28–40.

19.

Flannery

M. J.

(1989). Capital regulation and insured banks choice of individual loan default risks. Journal of Monetary Economics , 24(2), 235–258.

20.

Foos

, Norden

, & Weber

(2010). Loan growth and riskiness of banks. Journal of Banking & Finance , 34(12), 2929–2940.

21.

Frank

, & Hesse

(2009). Financial spillovers to emerging markets during the global financial crisis (No. 9-104). International Monetary Fund.

22.

Gennotte

, & Pyle

(1991). Capital controls and bank risk. Journal of Banking & Finance , 15(4–5), 805–824.

23.

Ghosh

, & Saggar

(1998). Narrow banking: Theory, evidence and prospects in India. Economic and Political Weekly , 34(9), 1091–1103.

24.

Gopalakrishnan

, Jacob

, & Pandey

(2018). Lender moral hazard in state-owned banks: Evidence from an emerging economy (No. WP 2018-07-01). Indian Institute of Management Ahmedabad, Research and Publication Department.

25.

Gorton

, & Rosen

(1995). Corporate control, portfolio choice, and the decline of banking. The Journal of Finance , 50(5), 1377–1420.

26.

Hansen

B. E.

(1999). Threshold effects in non-dynamic panels: Estimation, testing, and inference. Journal of Econometrics , 93(2), 345–368.

27.

Haq

, & Heaney

(2012). Factors determining European bank risk. Journal of International Financial Markets, Institutions and Money , 22(4), 696–718.

28.

Istrate

, Das Gupta

, & Weissburg

(2007). Toward developing a structured approach to the diagnosis and resolution of nonperforming loans: The case of China and India. Review of Policy Research , 24(4), 345–365.

29.

Iyer

, & Peydro

J. L.

(2011). Interbank contagion at work: Evidence from a natural experiment. The Review of Financial Studies , 24(4), 1337–1377.

30.

Jensen

M. C.

, & Meckling

W. H.

(1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics , 3(4), 305–360.

31.

Kornai

(1986). The soft budget constraint. Kyklos , 39(1), 3–30.

32.

Louzis

D. P.

, Vouldis

A. T.

, & Metaxas

V. L.

(2012). Macroeconomic and bank-specific determinants of non-performing loans in Greece: A comparative study of mortgage, business and consumer loan portfolios. Journal of Banking & Finance , 36(4), 1012–1027.

33.

Nguyen

T. P. T.

, & Nghiem

S. H.

(2015). The interrelationships among default risk, capital ratio and efficiency: Evidence from Indian banks. Managerial Finance , 41(5), 507–525.

34.

Nier

, & Baumann

(2006). Market discipline, disclosure and moral hazard in banking. Journal of Financial Intermediation , 15(3), 332–361.

35.

Piatti

, & Cincinelli

(2019). Does the threshold matter? The impact of the monitoring activity on non-performing loans: Evidence from the Italian banking system. Managerial Finance , 45(2), 190–221.

36.

Rajan

R. G.

(2016a). Issues in banking today. Speech at CII’s first banking summit by Dr. Raghuram G. Rajan, Governor, Reserve Bank of India, 11 February, Mumbai.

37.

Rajan

R. G.

(2016b). Resolving stress in the banking system. Speech by Dr. Raghuram G. Rajan, Governor, Reserve Bank of India, at ASSOCHAM – Interactive Meet with Industry & Trade, 22 June, Bengaluru.

38.

Rajan

, & Dhal

S. C.

(2003). Non-performing loans and terms of credit of public sector banks in India: An empirical assessment. Reserve Bank of India Occasional Papers , 24(3), 81–121.

39.

Salas

, & Saurina

(2002). Credit risk in two institutional regimes: Spanish commercial and savings banks. Journal of Financial Services Research , 22(3), 203–224.

40.

Samantaraya

(2016). Procyclical credit growth and bank NPAs. Economic and Political Weekly , 51(12), 112–119.

41.

Soedarmono

, Tarazi

, Agusman

, Monroe

G. S.

, & Gasbarro

(2012). Loan loss provisions and lending behaviour of banks: Asian evidence during 1992–2009. doi: 10.2139/ssrn.2049217

42.

Subbarao

(2009). Impact of the global financial crisis on India: Collateral damage and response. In Speech delivered at the Symposium on The Global Economic Crisis and Challenges for the Asian Economy in a Changing World. Organized by the Institute for International Monetary Affairs, Tokyo (Vol. 18).

43.

Swami

O. S.

, Nethaji

, & Sharma

J. P.

(2019). Determining risk factors that diminish asset quality of Indian commercial banks. Global Business Review . doi:10.1177/0972150919861470

44.

Villanueva

, & Mirakhor

(1990). Strategies for financial reforms: Interest rate policies, stabilization, and bank supervision in developing countries. Staff Papers , 37(3), 509–536.

45.

Wang

(2015). Fixed-effect panel threshold model using Stata. The Stata Journal , 15(1), 121–134.

46.

Williams

(2014). Bank risk and national governance in Asia. Journal of Banking & Finance , 49, 10–26.

47.

Zhang

, Cai

, Dickinson

D. G.

, & Kutan

A. M.

(2016). Non-performing loans, moral hazard and regulation of the Chinese commercial banking system. Journal of Banking & Finance , 63, 48–60.