Macro Stress Testing of Indian Bank Groups

Abstract

This study examines how various bank groups operating in India have fared macro stress events and conduct macro stress testing (MST) to trace the impact of certain macroeconomic stress scenarios on the credit quality of five Indian bank groups, that is, the State Bank of India (SBI) and its associates (SBGs), nationalised banks (NBs), old private sector banks (OPBs), new private sector banks (NPBs) and foreign banks (FBs), using panel data from 1997 to 2014.

Credit quality is modelled as a function of both macroeconomic variables (output growth, interest rate, inflation rate and exchange rate) and idiosyncratic variables (profitability and size indicator of bank business activity). The model is estimated by employing a panel cointegration approach, and the impact of adverse scenarios on the estimated credit quality is computed.

Empirical findings show that credit quality is pro-cyclical in nature and rises in the event of a slowdown in the economy. In general, the credit quality of Indian bank groups is found to be inversely and significantly related to the economy’s growth rate, inflation rate, exchange rate and profits of banks and positively and significantly related to the interest rate. Shock analysis also reveals that a downturn in the economy through certain adverse scenarios has a significant adverse impact on the credit quality. The shocks are quickly propagated across banks with substantial heterogeneities present in different bank groups. Thus, macroeconomic policy measures promoting growth with price stability are expected to impact credit quality positively. Further, measures at the bank level can improve credit quality by enhancing their profitability.

JEL Classifications: C32, C58, E170, G21

Keywords

Non-performing Assets (NPAs)Credit Quality Macro Stress Testing Banks

1. Introduction

The investigation of various potential empirical determinants of credit quality has constantly been a vital component of financial stability and bank management and, thus, is an important ingredient of macro stress testing (MST). A declining credit quality (which may be measured as the non-performing asset ratio) continuously drains an institution of its profits, restricts its cash flows by necessitating the creation of higher provisions against non-performing assets (NPAs). This, in turn, adversely impacts equity growth (limiting the capital adequacy requirements), mobilisation of funds, banking system credibility, productivity and expansion. In particular, the rising volume and growth of NPAs relative to the growth in credit have serious implications and thus have been a matter of serious concern for both banks and the economy.

Therefore, it is imperative for analysts, institutions and regulatory authorities to identify factors which may cause a deterioration in credit (or asset) quality and understand the underlying reasons behind them. This brings to the fore the question of whether these are caused by economic/business cycles and/or idiosyncratic reasons, which can be regulated at the institutional level (group and/or aggregate level) or at the policy level or are beyond their control.

In this context, this study conducts an MST exercise at a relatively disaggregated (bank group) level which is motivated by the hypothesis that both macro fundamentals and idiosyncratic features have an effect on credit quality. A scenario analysis is also conducted based on a set of hypothetical stress events for five Indian bank groups, that is, the State Bank of India (SBI) and its associates group (SBGs), nationalised banks (NBs), old private sector banks (OPBs), new private sector banks (NPBs) and foreign banks (FBs). The scenario analysis quantifies the impact of certain stress events on credit risk by taking into account bank group specific variables.

For the empirical estimation of the MST analysis, a panel cointegration technique¹ based on the seminal contribution of Pedroni (1999, 2004) is employed to determine a relationship between the NPA ratio and its potential determinants. The empirical findings show that a downturn in the economy due to certain stress events has a significant adverse impact on the NPA ratio and, thus, provides important insights to policy implications at the institutional level as well as at the level of the authorities. An overhaul of the bank management system in terms of reforms to promote profitability (which may improve efficiency, the operational network and credit portfolio management) is expected to have a positive impact on banks’ balance sheets at the institutional level. Furthermore, any expansionary policy measure to revive the economy by promoting growth with price stability is expected to impact credit quality positively at the disaggregate level of bank groups.

The scenario analysis quantifies the impact of a modest and a severe shock on the estimated credit risk of five bank groups. Shocks are quickly propagated across banks with substantial heterogeneities present in different bank groups. Major policy inferences, however, may be formulated on the basis of how the current situation evolves overtime, depending on whether stress in macroeconomic variables has a significant permanent or temporary negative effect on the NPA ratio for different bank groups.

The rest of the article is organised as follows. The second section gives a background of the Indian banking system. It provides a historical analysis of trends in NPA ratios and lists specific factors contributing to these trends. Third section has a brief discussion on empirical determinants in the literature on MST panel-specific studies. Fourth section specifies the empirical model for MST. Fifth section presents data sources and panel data methodology employed for estimation. Sixth section discusses estimation results and scenario analysis, followed by concluding remarks given in seventh section.

2. The Indian Banking System: A Background and Analysis of NPA Ratios

India has two broad categories of banks: scheduled and non-scheduled banks. Scheduled banks include commercial banks and cooperative banks. Commercial banks, on the basis of ownership, can be subclassified into public sector banks (PSBs), private sector banks (PVBs) and FBs. PSBs are state-run banks with the government shareholding being more than 51 per cent in contrast to PVBs which have a majority of private shareholders. The FBs, however, are those incorporated outside India which operate branches in India. All PSBs have been nationalised since 1969 and include two groups, that is, the SBI and SBGs and the NBs. PSBs dominate the Indian banking system as they own 75 per cent of the assets in the system and control more than 75 per cent of banking business.

The PVBs were split into two groups: OPBs and NPBs in 1990s, following liberalisation in banking policy. OPBs are small-sized banks with a regional focus which were not nationalised during 1969 and 1980.² They maintained their independence, thereafter too. For the majority, their board of directors comprise locally prominent traders and businesspeople. In contrast, NPBs are those banks that have begun operations since 1991, having obtained their banking licenses when economic and financial sector reforms were introduced. The amendment of the Banking Regulation Act in 1993 facilitated the entry of NPBs in Indian banking.

The five bank groups,³ that is, SBGs, NBs, OPBs, NPBs and FBs, account for approximately 99 per cent of total banking sector assets and contribute to more than 80 per cent of total credit outstanding in the economy. The present panel study considers these five bank groups as cross-sectional units and examines how they have weathered macro stress events. The heterogeneity across bank groups which may be partly attributed to different ownerships is an important consideration for this segregation. This is because PSBs, with predominantly government ownership, have the backing of the government. They have a mandate to implement various social development programmes of the government and have high provision requirements due to their staff expenses (including pension liabilities), which may dent their profitability. On the other hand, PVBs are more susceptible to market forces, and FBs are required to maintain profitability in their Indian branches while being regulated by their head offices.

Table 1 and Chart 1 represent a historical analysis of the NPA ratios of the Indian banks. The NPA ratio, in general, is highest for PSBs (with SBGs much higher than the NBs), followed by PVBs (with OPBs much higher than NPBs) and FBs.

More recently, in the post-global financial crisis period, we find NPA ratios are declining for both (old and new) PVBs and FBs due to various reform measures taken over the years. However, this is still not the case for PSBs, despite their significant reforms and improvements in profitability and efficiency (in terms of intermediation costs). This may be a reflection of various social welfare and priority sector commitments⁴ as one of the major objectives of PSBs coupled with their higher operating expenses (due to branch expansion and pension obligations) which may conflict with their profitability.⁵

Table 1

An Analysis of NPA Ratios of Indian Banks

	SCBs	SBG	NBs	OPBs	NPBs	FBs
1997	15.7	15.81	19.05	10.71	2.6	4.3
2007	2.5	2.59	2.69	3.1	1.9	1.8
2014	3.8	5.1	4.2	2	1.8	3.9
Maximum(Year)	15.7(1997)	15.81(1997)	19.05(1997)	13.06(1999)	8.9(2002)	7.6(1999)
Minimum(Year)	2.3(2008)	2.5(2011)	1.7(2009)	1.8(2012)	1.7(2006)	1.8(2007,8)

Source: Authors’ calculations.

Notes: (1) SCBs, SBGs, NBs OPBs, NPBs and FBs are aggregate and individual bank groups, that is, scheduled commercial banks, State Bank of India and its associate banks, nationalised banks, old private sector banks, new private sector banks and foreign banks, respectively.

(2) In general, the NPA ratio for NBs > for SBGs > for SCB > for OPBs > for FB > for NPBs.

Chart 1

Source: Authors’ calculation.

In 2014, the NPA ratio were at their highest level at 5.1 per cent for SBGs, followed by NBs at 4.2 per cent, FBs at 3.9 per cent, OPBs at 2 per cent and finally 1.8 per cent for NPBs, while the aggregate NPA ratio is 3.8 per cent.

If we juxtapose the numbers for the return for assets (ROA), then a very transparent relationship emerges between banks’ internal profits and their loan quality portfolio. In 2014, the ROA is least for PSBs (0.45 for NBs and 0.63 for SBGs), followed by OPBs at 1.2, FBs at 1.57 and the highest for NPBs at 1.8.

Various factors, especially after 2012, might have contributed towards the mounting of NPA ratios. These include the slowdown in the economy’s growth rate, governance issues in some banks, an incorrect credit appraisal of borrowers, wilful defaults and a greater allocation of credit to sectors like infrastructure, mining, textiles, steel and aviation which have been adversely affected due to the recent fall in commodity prices (given the direction of global commodity prices).

Empirically, Mohan (2004) and Lokare (2014) have pointed out a much greater rise in non-priority sector advances (than priority sector), lower spreads and lower operating expenses for FBs and NPBs compared to their public-sector counterparts. Inadequate appraisal and lax monitoring of credit proposals might also have contributed to a rising level of NPAs for the individual groups (RBI, 2012).⁶

According to the IMF report (2015),⁷

While the Indian banking system is well capitalized, slow growth and heightened corporate vulnerabilities have led to a deterioration in bank asset quality […] augmenting capital buffers in public sector banks remains essential to ensure the banking sector’s ability to provide adequate credit to support the recovery […]. Over the medium term, the authorities should implement their plan to reduce public shareholding in public sector banks to 52 per cent, and improve corporate governance at public banks […].

The report cautioned Indian banks about the significant increase in NPAs and put the onus on PSBs (with system-wide, gross NPA ratio rising to 4.1 per cent in contrast to PSBs at 4.7 per cent at the end of 2013–14). The major reasons cited relate to bottlenecks in infrastructure projects, primarily due to issues related to the acquisition of land, availability of fuel, delays in securing environmental approvals and poor credit assessments in some banks.

Against this backdrop, the empirical analysis of the present study estimates a long-run relationship which captures the observed trends in NPA ratios and their key determinants precisely as given in the chart, data and concomitants for various bank groups.⁸

The study attempts to analyse the heterogeneities observed in the impacts of stress events across various Indian banks groups, which can be partly attributed to ownership. An understanding of the empirical results of the study, in general, suggests reforms at the bank level to improve performance in terms of profitability. In particular, since NPAs in PSBs are significantly higher than in PVBs, this may be interpreted as an indication for policy and/or institutional reforms for PSBs. These reforms/measures may include their governance, ownership, credit evaluation procedures and lending policies to improve net profits (and reduce higher operating expenses such as branch expansion and pension obligations especially for PSBs vis-à-vis PVBs and FBs). A similar situation, however, does not hold for PVBs regarding NPAs over the period under consideration. Thus, this may present a case for allowing more independence and autonomy to PSBs in their functioning.

3. Empirical Determinants of Credit Risk

This section briefly presents various panel data-based studies⁹ on MST and related issues examining various empirical determinants of credit risk along with their possible implications.

Since the seminal paper of Wilson (1997a, b) on credit risk modelling under adverse macroeconomic conditions, there has been a sizeable volume of applied work on macro stress tests which cover various aspects. A number of comprehensive surveys review various alternative empirical approaches and methodologies, including Allen and Saunders (2003), Sorge (2004), Cihak (2004, 2005, 2007), ECB (2006) for the EMU area, Chan-Lau (2006), Foglia (2009), Drehmann (2008, 2009) and Canova and Ciccarelli (2013) and Demekas (2015).

In panel models, a broad classification of credit risk determinants is based on macroeconomic factors such as aggregate or sector-specific GDP growth rate, output gap, production, employment, consumption, money supply, interest rate, prices, stock market indices, property prices, exports, imports, exchange rate, terms of trade, etc.; bank-specific factors such as size, profitability, efficiency, regulation indicators, etc.; institutional factors such as the structure of the financial system, financial liberalisation, supervision, incentive structures, etc. or a combination of factors of the three categories.

A brief outline and evolution of various empirical studies over the years have been classified into two broad categories based on macroeconomic factors and bank-specific factors and is provided below.

3.1 Macroeconomic Factors

There is consensus on the hypothesis of pro-cyclicity of banks’ behaviour. A study by Quagliariello (2007) explicitly analysed the possibility of pro-cyclicity in banks’ behaviour for a panel of 207 Italian intermediaries and found the business cycle effect on banks’ loan loss provisions and new bad debts to be significant and negative. Other studies examining macroeconomic indicators on growth, price stability and the financial and external sectors include: Borio and Lowe (2002), Hoggarth and Zicchino (2004), Marcucci and Quagliariello (2005), Fofack (2005), Hoggarth, Sorensen and Zicchino (2005), Filosa (2007), End, Hoeberichts and Tabbae (2008), Zeman and Jurca (2008), Asberg and Shahnazarian (2008), Khemraj and Pasha (2009), Louzis, Vouldis and Metaxas (2010), Nkusu (2011), Messai and Jouini (2013) and Roy (2014).

Havrylchyk (2010) did a similar stress testing exercise and considers GDP growth, GFCF, the all-share index, interest rate, inflation, growth in M1, property prices, oil prices, consumption, debt/income, the employment index and REER to examine the resilience of South African banks based on three types of scenarios, and finds the system to be resilient. Beck, Jakubik and Piloiu (2013) examine various macroeconomic determinants of NPAs across 75 countries. They find a significant impact of real GDP growth, share prices, the exchange rate and lending interest rate on NPA ratios and the results are robust to alternative econometric specifications. Likewise, Akinlo and Awolowo (2014) investigate only macro determinants of non-performing loans (NPLs) for the Nigerian banking sector and find GDP, inflation, the exchange rate, interest rate, unemployment rate and stock market index to have a significant impact on NPLs.

3.2 Bank-specific Factors

In order to capture idiosyncratic effects, various panel studies extend Wilson’s (1997a, b) work on credit risk modelling using alternative approaches and methodologies. Some of the key panel data studies include Kwan and Eisenbeis (1995, 1997), Kelly and Mavrotas (2003), Drehmann (2008), End et al. (2008), Fofack (2005), Nkusu (2011), Messai and Jouini (2013), Vazquez, Tabak and Souto (2012), Klein (2013), Warue (2013), Kitamura et al. (2014), Fiordelisi, Ibanez and Molyneux (2010). A fair amount of recent research with an extensive focus on a large number of bank-specific factors include Louzis et al. (2010), which consider the rate of return, rate of equity, solvency ratio, loans-deposit ratio, inefficiency measure (as operating expenses/operating income ratio), credit growth, market power and size (measured by total asset ratios) along with three macroeconomic variables (GDP, unemployment rate and interest rate). They examine their differential impact on three different categories of loans, viz., mortgages, business loans and consumer loans for the nine largest Greek banks. Vazquez et al. (2012) examine a credit risk model using time-series and panel data methodology for Brazilian banks, using the estimated NPLs to calculate tail credit losses, using a credit value-at-risk (VaR) approach. Chelo and Manlagnit (2015) investigate potential correlates of efficiency of banks, their relationships (and the spillover effects) with macroeconomic fundamentals, bank-specific factors and institutional factors. They consider bank-specific factors in terms of intermediation ratios (as the proportion of loans to deposits), deposit-to-liability ratios and the degree of asset market concentration (as the ratio of the assets of the three largest banks to the total assets of commercial banks) for 17 banks in Philippines.

In the Indian context, however, applied work on MST has been limited in comparison to other countries. Few panel studies exist at a relatively disaggregate level. Rajaraman, Bhaumik and Bhatia (1999) examine variations in NPAs across Indian banks due to differences in operating efficiency, solvency and regional concentration. While Das (1999) compares various available measures of efficiency for Indian PSBs using a data envelopment analysis model. Their results indicate the existence of a significant negative relationship between the NPA level and efficiency parameters. Furthermore, Rajaraman and Vasishstha (2002) and Ranjan and Dhal (2003) show the existence of a significant bivariate relationship between NPAs of PSBs and inefficiency measures. Recently, Swamy (2012), Roy (2014) and Lokare (2014) investigated various macro-level factors (such as GDP growth, inflation, IIP, savings growth rate and per capita income growth) coupled with micro-level factors (such as market capitalisation growth rate, bank assets, capital adequacy ratio, credit/deposit ratio, lending rates, operating expenses/total assets, priority sector loans/total loans, rural and semi-urban branches/total bank branches and ROA) underlying the asset quality deterioration in Indian banks at the aggregate and group levels. On the basis of ownership, Maji and De (2015) investigate the nexus between the capital and risk of Indian commercial banks (21 PSBs and 20 PVBs) and the impact of other variables like profitability, size and human capital efficiency in a simultaneous two-equation model using three-stage least squares. The study finds an inverse relationship between risk and capital and a negative and significant impact of human capital efficiency on risk for PVBs. For PSBs, however, the impact of human capital efficiency on risk is not statistically significant. The study, therefore, concludes that PVBs are more efficient in utilising human capital for reducing credit risk.

4. The MST Modelling and Scenario Analysis

This study follows a three-step strategy¹⁰ to conduct an MST and scenario analysis.

Step 1: Scenario Designing

To begin with, this exercise identifies a baseline scenario,¹¹ various stress events and constructs adverse scenarios. For stress scenario designs, two types of hypothetical scenarios for change in the individual risk factor (i.e. the economy’s growth rate) are constructed to conduct MST. One is based on the annual output growth rate using an approach similar to Zeman and Jurca (2008) (referred to as Type I). In this context, magnitude of change in real GDP growth rate is based on historical development and historical extreme relative changes in the sample period under consideration.

Alternatively, another hypothetical scenario is considered based on annual output growth rate using an approach (referred to as Type II) similar to FSR of the Indian central bank (RBI) on MST for moderate and severe risk. This approach involves construction of two scenarios based on up to 1 standard deviation (sd) for moderate risk and between 1.25 sd to 2 sd for severe risk (10-year historical data) ¹².

Step 2: The Macroeconomic Credit Risk Model

In order to conduct MST, this study considers credit risk as the largest individual risk in the banking system and extends the credit risk/MST model¹³ as an empirical macro-financial model given as equation (1). The model links credit quality (also referred to as the probability of default or the default rate) as a measure reflecting the credit risk and quality of the aggregate bank portfolio and key macroeconomic variables based on Wilson (1997 a, b) and bank-specific factors.

\begin{matrix} P D_{i} = f (o u t p u t g r o w t h, \inf l a t i o n r a t e, int e r e s t \\ r a t e, e x c h a n g e r a t e, p r o f i t, s i z e i n d i c a t o r) \end{matrix}

(1)

where PD_i (the credit quality indicator) of an individual bank group is specified as a function of key macroeconomic variables representing macroeconomic conditions in an economy and a set of idiosyncratic factors representing the specific group behavioural characteristics. The former includes cyclical variables (some measure of growth of the economy), price stability and financial and external sector¹⁴ indicators, while the latter includes some measure of profitability and a size indicator of a specific bank group. In general, the model indicates that a better state of economy reflected by certain measures of macroeconomic development implies a lower default rate, credit quality (PD) and vice versa. The study uses NPA ratio as an empirical measure of PD.

For cyclical indicators, the study uses the economy’s growth rate,¹⁵ which is expected to have negative impact on NPA ratios. A prolonged economic recession in the economy reflected by a declining GDP is likely to depress investment optimism in the economy and subsequent bank credit creation. This, in turn, may lead to more defaults on loans (and rising NPA ratios) as a concomitant of the slowdown, primarily due to rising unemployment and falling incomes.

For price stability¹⁶ this article has considered inflation. The impact of inflation on NPA ratios may be negative due to the direct as well as indirect effects on borrowing and lending. The direct effect is the price effect which reduces the real interest rate through the Fishers’ effect and thus promotes investment and economic growth in the economy. On the other hand, the indirect effect is the volume effect, due to which the real value of the borrowed amounts of debtors declines and consequently they are expected to gain during periods of inflation.

The interest rate is considered as a financial market indicator¹⁷ and is expected to have a positive association with NPA ratios. It impacts directly through the rise in the cost of borrowed funds and indirectly through the rise in the expected profitability of the corporate sector.

On the external front,¹⁸ the impact of the exchange rate on NPA ratios can be mixed and indirect. A depreciation of the nominal exchange rate (a rise in INR/$) is expected to increase the competitiveness of exports and an increase in export earnings would favourably impact their ability to service their debt, promote export-oriented sectors and economic growth. However, the impact of a depreciation in the INR on imports is expected to work in the opposite direction. It would imply dearer imports, rising oil or fuel prices, and an adverse debt-servicing capacity of borrowers who borrow in foreign currency. It is expected to worsen repayment conditions of importers and borrowers and consequently raise NPA ratios. Therefore, the net impact of a change in the exchange rate on NPA ratios can be positive or negative.

Regarding bank-specific factors,¹⁹ banks’ assets as an indicator of the size and strength of the institution is considered. It is the most commonly used factor empirically,²⁰ is found to be significant with the correct sign according to economic theory and also one for which bank groups data are available over the sample period.

As a measure of profitability,²¹ the return on assets (ROA)²² is chosen for two reasons. First, it has been the most popular measure used empirically,²³ (also found to be significant with the correct signs a priori in almost all studies). Second, a company may have a high ROA even if it has a low profit margin, because it has a high asset turnover. Furthermore, it is quite possible for a company to have an impressive ROE²⁴ without actually being more effective at using shareholder equity for the growth of the company, where the latter may be revealed by a lower ROA. In general, more profitable banks are expected to be more efficient and end up releasing less NPAs.

Accordingly, an empirical macro-financial model for the ith bank group at time t is specified as:

N = f (y, i, w, e, R, a)

N_{i t} = β_{0 i} + β_{1} y_{t} + β_{2} i_{t} + β_{3} π_{t} + β_{4} e_{t} + β_{5} R_{i t} + β_{6} a_{i t} + ε_{i t}

(2)

where N_it = NPA ratio of the ith bank group at time t, y_t = economy’s growth rate in time t, π_t = inflation rate in time t, i_t = nominal interest rate in time t, e_t = nominal exchange rate in time t, R_it = return on assets (ROA) of the ith bank group at time t, a_it = assets of the ith bank group at time t and β_0i is an intercept term. It captures the impact of unobservable bank-specific effects and allows for heterogeneity in the means of the series across bank groups. It is time-invariant and accounts for any group-specific characteristics that are not included in the regression (Table 2).

Step 3: Scenario Analysis

The third step examines the quantitative impact of the baseline scenario and two alternative scenarios (Type I and Type II for modest and severe shock) constructed in step 1, on the estimated NPA ratio of bank groups based on the estimated MST equations. So, after estimation of the macro-financial model (2) specified in step 2, the NPA ratio (N) is estimated for five bank-groups based on the scenario analysis given in step 1 and the impact of stress in the macroeconomy on the credit quality numerically²⁵ is traced.

Table 2

Expected Impact of Variables on NPA ratio

MODEL: N = f (y, i, π, e, R, a)
Variables	Indicator	Empirical Variable	Expected Sign
Dependent	N_it	NPA ratio of the ith bank group at time t
Explanatory variables
Macroeconomic	Cyclical/growth	Growth rate	–
	Price	Inflation rate	–
	Financial	Interest rate	+
	External	Exchange rate	+/–
Idiosyncratic (bank specific)	Size	Assets	–
	Profitability	Return on assets	–

Source: Authors’ calculations.

5. Data and Methodology

5.1 Data Sources

The model estimation is based on annual data²⁶ covering a sample period from 1997 to 2014. Data for all the variables have been taken from official sources: The Reserve Bank of India data warehouse, as well as statistical tables relating to banks and the Ministry of Statistics and Programme Implementation.

Table A1 (in the appendix) provides the data definitions and sources of the variables used in the study.

5.2 Econometric Methodology—Panel Cointegration Technique

The analysis is based on the panel data technique²⁷ which measures the same set of cross-sectional units (bank groups in this study) at different points in time. This reduces the problem of multicollinearity among the regressors, controls for individual heterogeneity when examining temporal effects on behaviour and improves the efficiency of econometric estimates.

For empirical estimation, a panel cointegration methodology is used and three familiar steps are followed: examining the stochastic properties of the variables involved by means of panel unit root tests; testing for panel cointegration in order to assess for the presence of a long-run relationship; and applying the fully modified OLS (FMOLS) estimates to the cointegration vector in the empirical model. The subsequent sections succinctly outline a brief description of this methodology.

5.2.1 Panel Unit Root Tests

To begin with, the non-stationarity properties of the variables are tested by using panel unit root tests,²⁸ as they are expected to be more powerful than individual unit root tests for individual time-series and cross-sectional data. The study uses two tests proposed by Maddala and Wu (1999) and Im, Pesaran and Shin (2003). These tests are generally preferred over various other available panel unit root tests because they allow for heterogeneity of the autoregressive (AR) roots under the alternative hypothesis and combine individual unit root tests to derive a panel-specific result. Underlying differences in the construction of these two tests are attributed to the choice of statistics. In the construction of the IPS tests, an average of individual ADF t-statistics is taken, while the ADF-Fisher test considers an average of individual p-values across cross-section units. Further, the Fisher test does not require a balanced panel and can be based on different lag lengths in the individual ADF-regressions.

5.2.2 Panel Cointegration Tests

The next step is to carry out tests for panel cointegration to find the existence of a statistically acceptable cointegration relationship between the NPA ratio and its potential determinants as specified in our empirical model (2). The end objective is to estimate the long-run relationships between the said variables and employ them to conduct MST for Indian bank groups.

Since panel cointegration among non-stationary variables avoids spurious regression and inconsistency problems at the time of estimation, the study employs the Pedroni (1999, 2001, 2004) technique which extends the Engle and Granger (1987) framework (two-step, residual-based cointegration tests) involving panel data. Pedroni²⁹ (2004) examines the most general case of the residual-based tests based on the null hypothesis of no cointegration for panels with heterogeneous dynamics and heterogeneous slope coefficients. Pedroni proposes seven statistics for testing the same null of no cointegration against two alternative³⁰ hypotheses—with a common AR and the other allowing for heterogeneity of the AR roots. Four of the tests assume a common AR root (within-dimension or panel statistics test) and other three allow for heterogeneity of the AR roots (between-dimension or group statistics test).

5.2.3 Pedroni Panel Estimation Technique (FMOLS)

The objective of panel cointegration estimation is to examine the existence and nature of long-run relations among variables on a single cross-sectional unit, as predicted by economic theory in respect of their signs, estimates and significance of the relevant coefficients.

McCoskey and Kao (1999) propose a test of the OLS estimator with a bias-corrected t-statistic in finite sample size, but that is not an improvement over the OLS estimator in general. An alternative estimator, developed by Kao, Chiang and Chen (1999), the FMOLS and Dynamic OLS (DOLS) estimator³¹ is shown to provide more accurate results in panel cointegration models. Further, Pedroni (2000) proposes a Fully Modified Group Mean Estimator (GFMOLS) derived from the simple averages of the individual FMOLS for each cross-section. This approach allows for possible endogeneity in the variables, corrects for serial correlation and provides consistent estimates of β’s coefficients in relatively small samples. Hence, the FMOLS estimator is used in this study for estimation of panel cointegration.

6. Estimation Results

6.1 Panel Unit Root Tests

Table 3 reports results on two panel unit root tests and reveals that all the variables are I (1), except for the assets. For the assets variable³², the first differences and change in assets variable (da) is taken, which is I (1). Thus, there is evidence to show that the variables in the regression model exhibit non-stationarity. Therefore, the application of simple OLS to the stacked model in (1) may lead to biased and inconsistent estimates. Hence, it is necessary to employ appropriate panel cointegration techniques in order to determine whether a long-run relationship exists between non-stationary variables in level form.

Table 3

Panel Unit Root Tests (Assuming Individual Intercept)

Variable	IPS		ADF-FISCHER		Conclusion
Variable	Levels	First Difference	Levels	First Difference	Conclusion
a	7.596(1.000)	1.971(0.975)	0.012(1.000)	2.912(0.983)	I(2)
da	1.971(0.976)	–4.113*(0.000)	2.912(0.983)	35.282*(0.000)	I(1)
e	2.160(0.984)	–3.514*(0.000)	1.534(0.999)	29.991*(0.001)	I(1)
i	–0.620(0.268)	–2.688*(0.004)	9.806(0.458)	23.539*(0.008)	I(1)
N	0.031(0.513)	–1.627***(0.052)	7.388(0.688)	22.467**(0.012)	I(1)
R	–0.098(0.460)	–2.725*(0.003)	7.836(0.644)	24.526*(0.006)	I(1)
π	–0.829(0.204)	–6.177*(0.000)	10.954(0.361)	52.006*(0.000)	I(1)
y	–0.508(0.306)	–3.574*(0.000)	9.218(0.512)	30.474*(0.000)	I(1)

Source: Authors’ calculations.

Notes: (1) p-values are indicated in parentheses.

(2) Superscripts *, ** and *** indicate that the null hypothesis of the unit root is rejected at the 1, 5 and 10 per cent levels, respectively.

(3) For assets (A), the calculated IPS and ADF-Fischer statistics for the second difference are 4.113 (0.000) and 35.282 (0.000), respectively.

6.2 Panel Cointegration Tests

In the next step to carry out tests for panel cointegration between the NPA ratio and its potential determinants specified in our empirical model (2), the optimal lag length is chosen based on the Schwarz Information Criterion (SIC). Table 4 reports the results on panel cointegration test on seven statistics proposed by Pedroni (2004) for testing the null hypothesis of no cointegration against two alternative hypotheses—with a common AR root (within-dimension or panel statistics) and taking into account the heterogeneity of the AR roots (between-dimension or group statistics). The results indicate that the majority of the tests (i.e., four statistics) reject the null hypothesis of no cointegration at conventional levels of significance³³ and, thus, provide evidence for a long-run relationship between NPA ratios for various bank groups and the explanatory variables.

Table 4

Panel Cointegration Tests H₀: No Cointegration

Model: N = f (y, i, π, e, R)
Dependent variable: NPA ratio (N)
Pedroni Cointegration Tests	Statistic	p–value
Alternative Hypothesis: Common AR Coefficients (within dimension)
Panel v-Statistics	–1.181	0.881
Panel rho-Statistics	0.511	0.695
Panel PP-Statistics	–2.315*	0.010
Panel ADF-Statistics	–2.302*	0.010
Alternative Hypothesis: Individual AR Coefficients (between dimension)
Group rho-Statistics	1.576	0.942
Group PP-Statistics	–1.977**	0.024
Group ADF-Statistics	–1.986**	0.023

Source: Authors’ calculations.

Notes: (1) Superscripts * and ** indicate rejection of null hypothesis of no-cointegration at the 1 and 5 per cent levels of confidence, respectively.

(2) The total number of observations, NT = 90.

(3) Deterministic trend specification includes individual intercept.

(4) The lag length selection is based on automatic SIC with a max. lag of 2.

(5) For semi-parametric corrections, Newey–West automatic bandwidth selection and Barlett kernel is used.

6.3 Estimated MST Models for Individual Bank Groups

Table 5 presents results on the grouped FMOLS, pooled FMOLS and pooled (weighted) FMOLS,³⁴ which indicate that the signs of all the variables are consistent with economic theory. All the coefficients are significant at the 1 per cent level, according to grouped FMOLS. For pooled FMOLS, the GDP growth rate and exchange rate are significant at 1 per cent and the inflation rate is significant at 5 per cent. For the pooled (weighted) FMOLS, all but the exchange rate is significant at 1 per cent and the exchange rate variable is significant at 5 per cent.

Table 5

Estimation Results

Dependent Variable	NPA Ratio
Dependent Variable	Grouped FMOLS	Pooled FMOLS	Pooled (Weighted) FMOLS
y	–0.908*(0.000)	–0.952*(0.000)	–1.046*(0.000)
π	–0.363*(0.000)	–0.409**(0.041)	–0.466*(0.009)
i	0.443*(0.000)	0.706*(0.002)	0.591*(0.000)
R	–5.127*(0.000)	–1.826(0.138)	–2.652*(0.000)
e	–0.256*(0.000)	–0.246*(0.007)	–0.191**(0.045)

Source: Authors’ calculations.

Note: Superscripts * and ** indicate statistical significance of the coefficients at 1 and 5 per cent, respectively.

The individual FMOLS results for the estimated NPA ratios for the five bank groups also render signs for the coefficients for all the variables in sync with economic theory, with the majority of the variables being statistically significant at the conventional level of significance. Table 6 reports details on the estimated equations for individual bank groups, while the estimated MST models for the five Indian bank groups are reported in Table 7.

Six important findings stand out on a closer examination and a comparison of the coefficients of the determinants of the credit quality for the five bank groups. First, the impact of macroeconomic growth is more severe and significant for PSBs than for PVBs and FBs. Hence, they are expected to be more vulnerable to any significant downturn in the economy.

Second, the direct impact of monetary policy may be gauged by the coefficient of the nominal interest rate in the equations. This is an important result, with implications for a rise in the policy rate in a tightened monetary policy and its transmission to the bank default rate via pass-through effects to other rates in the economy, including lending rates. The coefficients and t-statistics indicate a favourable positive and significant impact of an expansionary monetary policy on NBs compared to other groups (in which case the impact is positive but insignificant).

Third, inflation has a negative and significant impact on the NPA ratios for all the groups (except for NPBs where it is, of course, negative but insignificant). Interestingly, the most affected are the OPBs while the least affected are the NPBs.

Table 6

Estimated MST Specifications for Five Bank Groups

Dependent Variable	NPA Ratio
Dependent Variable	N1	N2	N3	N4	N5
Const	48.408*(0.000)	33.485*(0.003)	28.882*(0.001)	12.691**(0.020)	10.517*(0.002)
y	–1.332*(0.000)	–1.267*(0.000)	–1.205*(0.000)	–0.586*(0.007)	–0.145(0.235)
π	–0.508**(0.032)	–0.500**(0.039)	–0.612**(0.017)	–0.04(0.809)	–0.156***(0.10)
i	0.442(0.354)	1.058*(0.003)	0.527***(0.053)	0.109 (0.524)	0.077(0.518)
R	–7.682**(0.040)	–7.558**(0.031)	–3.691**(0.045)	–3.512*(0.001)	–3.190*(0.000)
e	–0.586*(0.000)	–0.383*(0.003)	–0.270**(0.031)	–0.036 (0.622)	–0.003(0.928)

Source: Authors’ calculations.

Notes: (1) N1, N2, N3, N4 and N5 denote NPA ratios for individual bank groups, that is, SBGs, NBs, OPBs, NPBs and FBs, respectively.

(2) p-values are indicated in the parentheses.

(3) Superscripts *, ** and *** indicate statistical significance of the coefficients at 1, 5 and 10 per cent, respectively.

Fourth, the impact of both the exchange rate and return on assets are found to be negative and significant with the greatest magnitude for the SBGs and smallest for FBs (except OPBs for the returns to assets and NPBs and FBs for the exchange rate variable, where it is negative in line with economic theory but insignificant).

Fifth, the group-specific, constant term embodies specific characteristics and individual heterogeneity across the five bank groups. It captures the positive and significant impact of various unobservable individual bank group-specific features in all of the five estimated specifications.

Finally, a notable feature of the empirical results is that bank-specific factors contribute to NPA ratios at the highest magnitude (the value of the coefficient β5) in comparison to other macroeconomic factors, across all individual groups. The profitability indicator is significant and negative in all the equations. An increase in profits is expected to reduce the NPA ratio. This finding is in line with majority of empirical studies such as Fofack (2005), Flamini, McDonald and Schumacher (2009), Khemraj and Pasha (2009), Dinos and Ashta (2010) and Warue (2013). Warue’s (2013) recommendation³⁵ holds for this panel study as well, ‘[…] commercial banks portfolio management strategies focus more on the bank-specific factors which the management has more control over and seek practical and achievable solutions to redress NPLs problems’.

In general, a very interesting pattern of the absolute values of various coefficients in the above equations has been observed. Except for the interest rate and inflation rate variables, the magnitude of the impact falls in the following order: the impact for SBGs > NBs > OPBs > NPBs > FBs. This pattern may be understood as discerning the direct and indirect impact of contractionary monetary policy shocks,³⁶ (through changes in the interest rate and inflation rate) which may severely distress the credit quality for individual bank groups if their capitalisation (say) is substantially low.³⁷

Any contractionary monetary policy, say an increase in the benchmark repo rate, is expected to impact economic growth significantly via high interest rates. Though its impact on inflation may or may not be as pronounced, it is expected to slow down the credit growth rate and increase the cost of funds, both of which may adversely affect the profitability of banks. Conversely, an expansionary monetary policy may improve the credit quality of banks by having a favourable impact on the economy’s growth and the profitability of the banks.

Overall, empirical findings may be interpreted as indicating the existence of heterogeneity across the five bank groups, as the constant term is positive and significant (capturing unobservable group-specific effects), and the ROA is negative and significant (capturing observable group-specific features) in all of five estimated equations. Furthermore, heterogeneity across the groups due to ownership basis may reflect in terms of directed credit programmes such as excessive priority sector lending targets,³⁸ a rise in provisioning requirements for NPAs and restructured loans coupled with a mandate to expand in relatively less profitable areas, lower profit margins and higher operating expenses (due to pension obligation and branch expansion) for PSBs vis-à-vis PVBs and FBs.

6.4 Scenario Analysis

Tables 8 and 9 illustrate the quantitative impact of a baseline scenario and two alternative adverse scenarios, Type I and Type II (for modest and severe shocks), on the estimated NPA ratios of bank groups based on the estimated MST equations (1)–(5) given in Table 7.

The results show that a downturn in the economy, as expected, deteriorates the credit quality in the economy over a one-year horizon compared to the baseline scenario. This is reflected by a rising NPA ratio across bank groups. More specifically, the impact of a stressed GDP growth rate on credit quality is found to be most adverse for NBs in the Type I and Type II scenarios, for both modest and severe shocks (and the least for NPBs). One may also note from five individual estimated specifications given in Tables 6 and 7 that the coefficient of the nominal interest rate is largest and significant for NBs; the coefficient of the ROA for NBs is also larger and more significant than for the private and foreign counterparts. This points to the fact that, among the bank groups, NBs are much more vulnerable to shocks than other bank groups. However, this also throws light on the positive role that can be played by the authorities (by reviving the economy’s growth rate) and bankers (promoting profitability by improving ROA) to de-stress (or relieve) banks in terms of reducing NPAs.

Table 7

Estimated MST Equations for Five Bank Groups

1.	SBGs	$N_{1 t} = - 1 .332 y_{1 t} + 0 .442 i_{1 t} - 0 .508 π_{1 t} - 0 .586 e_{1 t} - 7 .682 R_{1 t} + 48 .408$
2.	NBs	$N_{} = - 1 .267 y_{2 t} + 1 .058 i_{2 t} - 0 .500 π_{2 t} - 0 .383 e_{2 t} - 7 .558 R_{2 t} +33 .485$
3.	OPBs	$N_{3 t} = - 1 .205 y_{3 t} +0 .527 i_{3 t} - 0 .612 π_{3 t} - 0 .270 e_{3 t} - 3 .691 R_{3 t} +28 .882$
4.	NPBs	$N_{4 t} = - 0 .586 y_{4 t} + 0 .109 i_{4 t} - 0 .04 π_{4 t} - 0 .036 e_{4 t} - 3 .512 R_{4 t} +12 .692$
5.	FBs	$N_{5 t} = - 0 .145 y_{5 t} + 0 .077 i_{5 t} - 0 .156 π_{5 t} - 0 .003 e_{5 t} - 3 .190 R_{5 t} + 10 .517$

Source: Authors’ calculations.

Note: SBGs, NBs, OPBs, NPBs and FBs denote Indian bank groups, that is, the State Bank of India and its associate banks group, nationalised banks, old-private sector banks, new private sector banks and foreign banks, respectively.

Table 8.

Scenario Assumptions and Analysis of Macro Credit Stress Testing (Type I)

Scenario	One-Year Horizon 2014–15
Scenario	y	N1	N2	N3	N4	N5
Baseline	5.7	1.8	7.4	3.6	1.27	3.8
Moderate stress	3.72	4.45	10.01	6.04	2.43	4.17
Severe stress	2.73	5.77	11.26	7.23	3.01	4.31

Source: Authors’ calculations.

Notes: (1) Stress scenario construction is based on up to 1 sd for moderate risk and 1.25–2 sd for severe risk.

(2 ) N1, N2, N3, N4 and N5 denote the estimated NPA ratio for individual bank group, that is, SBG, NB, OPBs, NPBs and FBs, respectively.

Table 9

Scenario Assumptions and Analysis of Macro Credit Stress Testing (Type II)

Scenario	One-Year Horizon 2014–15
Scenario	y	N1	N2	N3	N4	N5
Baseline	5.7	1.8	7.4	3.6	1.27	3.8
Moderate stress	4.74	3.09	8.71	4.81	1.84	4.02
Severe stress	3.78	4.37	9.93	5.96	2.4	4.16

Source: Authors’ calculations.

Notes: (1) Stress scenario construction is based on historical extreme relative changes but is even more severe³⁹.

(2) N1, N2, N3, N4 and N5 denote the estimated NPA ratios for individual bank group, that is, SBG, NB, OPBs, NPBs and FBs, respectively.

The results have serious implications for the stability and soundness of the system. This is because a worsening of credit quality may cause banks to incur losses (on account of non-receipt of interest and repayments on their loan obligations) and maintain higher statutory provisions against rising NPAs. Both of these impact and limit the capital adequacy ratio of banks, and so, may jeopardise the system if their capital adequacy ratio falls below the international BASEL threshold norms.

Since banks act as a conduit of monetary policy, their risk portfolios reflect crucial information about their efficiency, soundness and management. They also manifest how well banks are performing their role as financial intermediaries in mobilising financial resources and allocating funds in the economy. The results have profound implications for bank management as well as policy authorities in an evolving and dynamic macroeconomic environment. The shocks are quickly propagated across banks with substantial heterogeneities present in different bank groups.

Thus, bank efficiency reforms at the institutional level to improve their performance in terms of profitability and macroeconomic policy measures—promoting growth with price stability, are expected to impact positively at the disaggregate level of bank groups.

7. Conclusion

This study applies panel data cointegration methodology to stress-test five Indian banking groups. The Pedroni (2000) FMOLS technique has been used to estimate the MST equations. The empirical model identifies and analyses the transmission of macroeconomic shocks across bank groups.

The empirical findings show that NPAs are pro-cyclical in nature—they rise in the event of any overall slowdown in the economy. In particular, the credit quality of the Indian bank groups is found to be inversely and significantly related to the economy’s growth rate, inflation rate, exchange rate and profits of the group, and positively and significantly related to the interest rate.

More specifically, the results indicate positive and significant impact of various unobservable individual bank group-specific features captured through the constant term. This finding is also a reflection of specific characteristics and individual heterogeneity across five bank groups. In addition, the impact of the idiosyncratic factor is also found to be significant and quite pronounced, in comparison to the impact of other macroeconomic factors, on the NPA ratio across all the individual groups. Both findings pave the way for a policy prescription based more on individual bank portfolio management strategies. Banks should focus on bank-specific factors over which management has more control and seek practical solutions given the broad guidelines and regulatory framework of the authorities to solve NPA problems.

Furthermore, a scenario analysis quantifies the impacts of baseline, moderate and severe shocks to the economy’s growth rate on the estimated NPA ratios for various bank groups. The shock analysis shows that any slowdown causing a downturn in the economy through certain adverse scenarios has a significant adverse impact on the NPA ratio in a one-year horizon. The impact on NBs is found to be most severe compared to the other groups.

Thus, measures at the institutional level may be based on banks’ focus on lending to more profitable segments, sound credit appraisals, internal ratings-based granting of credit and more standardised banking norms with an emphasis on a reduction in gross NPAs to improve their performance in terms of profitability. This coupled with the macroeconomy-wide twin objectives of growth with price stability is expected to have a positive impact on credit quality at the disaggregate level of bank groups.

Footnotes

Appendix

Table A1.

Data Definitions and Sources

Variable	Definition	Source
a	Assets (in millions) of a bank group	www.rbi.org
A	Aggregate assets of the banking system
da	Change in assets (year-on-year) = a (t) – a (t–1)	–
d	Dummy variable for March, 2004, for change in definition of NPAs	–
e	Nominal exchange rate, INR/US$ (at end-year)	www.rbi.org
g_a	Growth in assets = a (t) – a (t-1)/a (t–1)
i	Interest rate on central government dated securities (weighted average) (% per annum)	www.rbi.org
N	NPA ratio (%) is the ratio of gross NPAs to gross advances	www.rbi.org
NPA	Non-performing asset on which either the principal or the interest is overdue for at least two quarters or 180 days. From March 2004, banks adopted a ‘90-day overdue’ norm for calculating NPAs	www.rbi.org
ROA	Return on assets, ratio of net profits to assets of the individual bank	www.rbi.org
R	ROAs for a bank group is obtained as weighted average of ROAs of individual banks in the group, weights being the proportion of total assets of the bank as percentage to total assets of all banks in the corresponding bank group^a	www.rbi.org
RS	Relative size of the bank group in the total assets of the banking system = a/A
W	Wholesale price index, annual average for all commodities (base 2004–05 = 100)	www.rbi.org
p	WPI inflation rate = WPI(t)-WPI(t-1)/WPI(t-1)×100	–
Y	GDP at factor cost (at constant prices) (base 2004–05 = 100) (INR billion)	www.rbi.org
y	Real GDP growth rate = Y(t)-Y(t-1)/Y(t-1)×100 (% per annum)	www.rbi.org

Source: Authors’ own.

Note: ^aExplanatory notes for statistical tables relating to banks in India.

Acknowledgements

The authors are grateful to the anonymous referee for constructive suggestions.

Notes

References

Akinlo

& Awolowo

(2014). Determinants of non-performing loans in Nigeria. Accounting & Taxation, 6(2), 21–28.

Allen

& Saunders

(2003). A survey of cyclical effects in credit risk measurement models (BIS Working Paper No. 126). Basel, Switzerland: BIS.

Altunbas

Carbo

Gardener

E. P. M.

& Molyneux

(2007). Examining the relationships between capital, risk and efficiency in European banking. European Financial Management, 13(1), 49–70.

Asberg

& Shahnazarian

(2008, January). Macroeconomic impact on expected default frequency (Sveriges Riksbank Working Paper No. 219). Stockholm: Financial Stability Department, Sveriges Riksbank.

Baltagi

(2013). Econometric analysis of panel data (5th ed.). Chichester: John Wiley & Sons Ltd.

Beck

Jakubik

& Piloiu

(2013). Non-performing loans: What matters in addition to the economic cycle? (European Central Bank Working Paper No. 1515). Frankfurt: European Central Bank.

Borio

& Lowe

(2002). Asset prices, financial and monetary stability: Exploring the nexus (BIS Working Paper NO. 114). Basel, Switzerland: Bank for International Settlements.

Canova

& Ciccarelli

(2013). Panel vector autoregressive models: A survey (ECB Working Paper No. 1507). Frankfurt: European Central Bank.

Chan-Lau

J. A.

(2006). Market-based estimation of default probabilities and its application to financial market surveillance (IMF Working Paper No. WP/06/104). New Hampshire, US: IMF.

10.

Chelo

& Manlagnit

(2015). Basel regulations and banks’ efficiency: The case of the Philippines. Journal of Asian Economics, 39(2015), 72–85.

11.

Christiano

Motto

& Rostagno

(2010). Financial factors in economic fluctuations (European Central Bank, Working Paper No.1192). Frankfurt: European Central Bank.

12.

Cihak

(2004). Stress testing: A review of key concepts (Czech National Bank, Research Policy Note No. 2/2004). Retrieved from http://www.cnb.cz/en/pdf/IRPN_2_2004.pdf.

13.

Cihak

(2005). Stress testing of banking systems. Czech Journal of Economics and Finance, 55, 9–10.

14.

Cihak

(2007, March). Introduction to applied stress testing (IMF Working Paper No. WP/07/59). New Hampshire, US: IMF.

15.

Das

(1999). Profitability of the public sector banks: A decomposition model. Reserve Bank of India Occasional Papers, 20(1), 55–81.

16.

Demekas

D. G.

(2015, June). Designing effective macroprudential stress tests: Progress so far and the way forward (IMF Working Paper No. 15/146). New Hampshire, US: IMF.

17.

Dinos

& Ashta

(2010). Financial crisis: Lessons from microfinance in Russia, Ecuador, Bolivia. Swiss Press.

18.

Drehmann

(2008). Stress tests: Objectives, challenges and modelling choices. Economic Review (Sveriges Riksbank), 2, 60–92.

19.

Drehmann

(2009). Macroeconomic stress testing banks: A survey of methodologies. In Quagliariello

(Ed.), Stress testing the banking system: Methodologies and applications. Cambridge, UK: Cambridge University Press.

20.

Dua

& Kapur

(2015). Macro stress testing and resilience assessment of Indian banking (Mimeo). New Delhi: Delhi School of Economics.

21.

Dua

& Raje

(2014). Determinants of yields on government securities in India. Margin: The Journal of Applied Economic Research, 8(4), 375–400.

22.

End

J. W. V. D.

Hoeberichts

& Tabbae

(2008). Modeling scenario analysis and macro stress-testing (DNB Working Paper No. 119). Netherlands: De Nederlandsche Bank (Netherlands Central Bank).

23.

Engle

R. F.

& Granger

C. W. J.

(1987). Cointegration and error correction: Representation, estimation and testing. Econometrica, 55, 251–276.

24.

European Central Bank (ECB). (2006, June). A country-level macro stress-testing practices. Financial Stability Review (pp. 147–154). Frankfurt: European Central Bank.

25.

Filosa

(2007). Stress testing of the Italian banking system: A VAR approach. (Heterogeneity and Monetary Policy No. 0703, pp. 1–46).

26.

Fiordelisi

Ibanez

D. M.

& Molyneux

(2010). Efficiency and risk in European banking (European Central Bank Working Paper No. 1211/June 2010). Frankfurt: European Central Bank.

27.

Flamini

McDonald

& Schumacher

(2009). The determinants of commercial bank profitability in Sub-Saharan Africa (International Monetary Fund WP/09/15, pp. 1–15). New Hampshire, US: IMF.

28.

Fofack

(2005). Nonperforming loans in Sub-Saharan Africa: Causal analysis and macroeconomic implications (World Bank Policy Research Working Paper No. 3769). Washington DC, US: World Bank.

29.

Foglia

(2009). Stress testing credit risk: A survey of authorities approaches. International Journal of Central Banking, 5(3), 9–45.

30.

Graeve

Kick

& Koetter

(2008). Monetary policy and bank distress: An integrated micro-macro approach (Deutsche Bundesbank Discussion Paper Series 2: Banking and Financial Studies No. 03/2008).

31.

Havrylchyk

(2010). A macroeconomic credit risk model for stress testing the South African banking sector (South African Reserve Bank Working Paper, WP/10/03). Retrieved from http://mpra.ub.uni-muenchen.de/21639/

32.

Hoggarth

& Zicchino

(2004). Stress testing the UK banking system using a VAR approach (Mimeo). London: Bank of England.

33.

Hoggarth

Sorensen

& Zicchino

(2005). Macro stress testing UK banks using a VAR approach (Bank of England Working Paper No. 282). London: Bank of England.

34.

Pesaran

& Shin

(2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115, 53–74.

35.

Jain

Parida

T. K.

& Ghosh

S. K.

(2015, October 31). Rethinking priority sector lending for banks in India (IIBF Macro Research Paper for the year 2014–15). Mumbai: IIBF.

36.

Kao

Chiang

& Chen

(1999). International R&D spillovers: An application of estimation and inference in panel cointegration. Oxford Bulletin of Economics and Statistics, 61, 691–709.

37.

Kelly

& Mavrotas

(2003). Savings and financial sector development: Panel cointegration evidence from Africa (WIDER Discussion Paper No. 2003/12). Helsinki, Finland: World Institute for Development Economics Research, United Nations University.

38.

Khemraj

& Pasha

(2009). The determinants of non-performing loans: An econometric case study of Guyana (pp. 3–10). Georgetown, Guyana: University of Guyana Press.

39.

Kitamura

Kojima

Nakamura

Takahashi

& Takei

(2014). Macro stress testing at the Bank of Japan (Reports and Research Papers). Tokyo, Japan: Financial system and Bank Examination Department, Bank of Japan.

40.

Klein

(2013). Non-performing loans in CESEE: Determinants and impact on macroeconomic performance (IMF Working Paper No. WP/13/72). New Hampshire, US: IMF.

41.

Kwan

& Eisenbeis

(1995). Bank risk, capitalization and operating efficiency (The Wharton School Working Paper No. 96–35). Pennsylvania, US: The Wharton Financial Institutions Center, University of Pennsylvania.

42.

Kwan

& Eisenbeis

(1997). Bank risk, capitalization and operating efficiency. Journal of Financial Services Research, 12(2–3), 117–131.

43.

Lokare

S. M.

(2014, February). Re-emerging stress in the asset quality of Indian banks: Macro-financial linkages (Department of Economic and Policy Research Working Paper). Mumbai: Department of Economic and Policy Research, Reserve Bank of India.

44.

Louzis

D. P.

Vouldis

A. T.

& Metaxas

V. L.

(2010, September). Macroeconomic and bank-specific determinants of non-performing loans in Greece: A comparative study of mortgage, business and consumer loan portfolios (Bank of Greece, Euro System Working Paper No. 118).

45.

Maddala

G. S.

& Wu

(1999). A comparative study of unit root tests with panel data and a new simple test. Oxford Bulletin of Economics and Statistics, 61(Special Issue), 631–652.

46.

Maji

S. G.

& De

U. K.

(2015). Regulatory capital and risk of Indian banks: A simultaneous equation approach. Journal of Financial Economic Policy, 7(2), 140–156. DOI: http://dx.doi.org/10.1108/JFEP-06-2014-0038

47.

Marcucci

& Quagliariello

(2005). Is bank portfolio riskiness pro-cyclical? Evidence from Italy using a vector autoregression (Discussion Papers in Economics, No. 2005/09). New York: University of York.

48.

McCoskey

& Kao

(1999). Comparing panel data cointegration tests with an application to the ‘twin deficits problem’ (Working Paper). New York: Center for Policy Research, Syracuse University.

49.

Messai

A. M.

& Jouini

(2013). Micro and macro determinants of non-performing loans. International Journal of Economics and Financial Issues, 3(4), 852–860.

50.

Mohan

(2004, March). Finance for industrial growth (Reserve Bank of India Bulletin, Speech article).

51.

Nkusu

(2011). Nonperforming loans and macrofinancial vulnerabilities in advanced economies (IMF Working Paper No. 11/161). New Hampshire, US: IMF.

52.

Pedroni

(1995). Panel cointegration: Asymptotic and finite sample properties of pooled time series tests, with an application to the PPP hypothesis. (Working paper, pp. 95–103). Indiana University.

53.

Pedroni

(1999). Critical values for cointegration tests in heterogeneous panels with multiple regressors. Oxford Bulletin of Economics and Statistics, 61, 653–670.

54.

Pedroni

(1997). Panel cointegration; asymptotic and finite sample properties of pooled time series tests, with an application to the PPP hypothesis: New results (Working paper). Indiana University.

55.

Pedroni

(2000). Fully modified OLS for heterogeneous cointegrated panels. Advances in Econometrics, 15, 93–130.

56.

Pedroni

(2001). Panel cointegration; asymptotic and finite sample properties of pooled time series tests, with an application to the PPP hypothesis. (Revised working paper). Indiana University.

57.

Pedroni

(2004). Panel cointegration: Asymptotic and finite sample properties of pooled time series tests with an application to the PPP Hypothesis. Econometric Theory, 20, 597–625. DOI: 10+10170S0266466604203073

58.

Pesaran

M. H.

(2015). Chapter 26, p. 634 and Chapter 31, p. 817. In Pesaran

M.H.

(Ed.), Time series and panel data econometrics. Oxford, UK: Oxford University Press.

59.

Quagliariello

(2007). Banks’ riskiness over the business cycle: A panel analysis on Italian intermediaries. Applied Financial Economics, 17(2), 119–38.

60.

Quagliariello

(2008). Stress-testing the Banking System: Methodologies and Applications . Cambridge, UK: Cambridge University Press.

61.

Rajaraman

& Vashistha

(2002, February 2). Non-performing loans of PSU banks: Some panel results. Economic & Political Weekly, 1–7.

62.

Rajaraman

Bhaumik

& Bhatia

(1999). NPA variations across Indian commercial banks. Economic & Political Weekly, 34(3–4), 161–168.

63.

Ranjan

& Dhal

S. C.

(2003). Non-performing loans and terms of credit of public sector banks in India: An empirical assessment. RBI Occasional Papers, 24(3), 1–41.

64.

Reserve Bank of India. (2012). Report on trends and progress, 2011–12 (p. 37). Mumbai: RBI.

65.

Reserve Bank of India Report. (2012, July 20). Working group to review the existing prudential guidelines on restructuring of advances by banks/financial institutions (Chairman: Mahapatra

66.

Roy

S. G.

(2014). Determinants of non-performing assets in India—Panel regression. Eurasian Journal of Economics and Finance, 2(3), 69–78.

67.

Sorge

(2004). Stress-testing financial systems: An overview of current methodologies (Bank for International Settlements Working Paper No. 165). Basel, Switzerland: BIS.

68.

Stolz

Moretti

& Swinburne

(2008). Stress testing at the IMF (IMF Working Papers, pp. 1–23). Retrieved from https://ssrn.com/abstract=1266539

69.

Swamy

(2012). Impact of macroeconomic and endogenous factors on non-performing bank assets. The International Journal of Banking and Finance, 9(1), 27–47.

70.

Vazquez

Tabak

B. M.

& Souto

(2012). A macro stress test model of credit risk for the Brazilian banking sector. Journal of Financial Stability, 8(2), 69–83.

71.

Warue

B. N.

(2013). The effects of bank-specific and macroeconomic factors on nonperforming loans in commercial banks in Kenya: A comparative panel data analysis. Advances in Management & Applied Economics, 3(2), 135–164.

72.

Williamson

(1987). Financial intermediation, business failures, and real business cycles. Journal of Political Economy, 95(6), 1196–1216.

73.

Wilson

(1997a). Portfolio credit risk (1). Risk, 10(9), 111–117.

74.

Wilson

(1997b). Portfolio credit risk (2). Risk, 10(10), 56–61.

75.

Zeman

& Jurca

(2008). Macro stress testing of the Slovak banking sector (National Bank of Slovakia Working Paper No. 1/2008). Bratislava: National Bank of Slovakia.