This paper explores an empirical approach to the analysis of commercial banks' non-performing loans (NPLs) in the Indian context. The empirical analysis evaluates as to how banks’ non-performing loans are influenced by three major sets of economic and financial factors, i.e., terms of credit, bank size induced risk preferences and macroeconomic shocks. The empirical results from panel regression models suggest that terms of credit variables have significant effect on the banks' non-performing loans in the presence of bank size induced risk preferences and macroeconomic shocks. Moreover, alternative measures of bank size could give rise to differential impact on bank's non-performing loans. In regard to terms of credit variables, changes in the cost of credit in terms of expectation of higher interest rate induce rise in NPAs. On the other hand, factors like horizon of maturity of credit, better credit culture, favorable macroeconomic and business conditions lead to lowering of NPAs. Business cycle may have differential implications adducing to differential response of borrowers and lenders.

JEL Classification : G21, E51, G11, C23

Key Words : Bank credit, non-performing loans, terms of credit, panel regression.

Financial stability is considered as sine qua non of sustained and rapid economic progress. Among various indicators of financial stability, banks’ non-performing loan assumes critical importance since it reflects on the asset quality, credit risk and efficiency in the allocation of resources to productive sectors. A common perspective is that the problem of banks’ non-performing loans is ascribed to political, economic, social, technological, legal and environmental

* Rajiv Ranjan is Director and Sarat Chandra Dhal is Assistant Adviser in Department of Economic Analysis and Policy, Reserve Bank of India. The authors are extremely thankful to Shri Manoranjan Mishra, for his valuable suggestions and insightful discussions. The responsibility for the views expressed in the paper rests with the authors only.

(PESTLE) factors across countries (2003, Bhide, et.al., 2002, Das and Ghosh). During the last decade, the PESTLE framework has undergone significant changes, largely, due to structural transformation of emerging economies, including India, amidst reform of financial sector, economic integration induced by rapid increase in the pace of globalisation and advances in information technology. Moreover, Government intervention in the credit market has eased considerably. Advances in technology have facilitated rapid exchange of information across markets, creation of newer financial products, and reduction in transaction costs, thus, contributing to enhanced operational efficiency of banks and financial institutions. The institutional infrastructure has been strengthened in various ways. Countries have adopted international best practices pertaining to prudential regulation and supervision. In the sphere of legal environment, several measures have been undertaken in the areas of debt recovery, securitisation and asset reconstruction, resolution of defaults and non-performing loans, besides changes and amendments to the archaic laws pertaining to banking and financial sector. Overall, these developments have led to structural change in the financial sector, which has created conducive environment for market mechanism, in general, and economic factors, in particular, for playing a critical role in influencing the portfolios of banks and financial institutions.

It is in this context that this study has undertaken an empirical analysis for evaluating the impact of economic and financial factors on banks’ non-performing loans. The distinguishing feature of the study is that it provides a framework for analysis of underlying behaviour of borrowers’ in terms of their loan repayments in response to lending terms of banks and other macroeconomic indicators. The study is organised into four sections. Section I briefly reviews the extant literature focusing on the proximate determinants of non-performing loans (NPLs) in the light of cross-country evidences. Section II presents stylised facts about Non-Performing Assets (NPAs) and terms of credit variables. Section III postulates a theoretical analysis of the problem of NPL in the Indian context, which provides an underlying framework for the empirical analysis in Section IV. Section V concludes with some policy implications.

In the banking literature, the problem of NPLs has been revisited in several theoretical and empirical studies. A synoptic review of the literature brings to the fore insights into the determinants of NPL across countries. A considered view is that banks’ lending policy could have crucial influence on non-performing loans (Reddy, 2004). Reddy (2004) critically examined various issues pertaining to terms of credit of Indian banks. In this context, it was viewed that ‘the element of power has no bearing on the illegal activity. A default is not entirely an irrational decision. Rather a defaulter takes into account probabilistic assessment of various costs and benefits of his decision’. Mohan (2003)¹ conceptualised ‘lazy banking’ while critically reflecting on banks’ investment portfolio and lending policy. In a study of institutional finance structure and implications for industrial growth, Mohan (2004) emphasised on key lending terms of credit, such as maturity and interest-terms of loans to corporate sector. The Indian viewpoint alluding to the concepts of ‘credit culture’ owing to Reddy (2004) and ‘lazy banking’ owing to Mohan (2003a) has an international perspective since several studies in the banking literature agree that banks’ lending policy is a major driver of non-performing loans (McGoven, 1993, Christine 1995, Sergio, 1996, Bloem and Gorters, 2001).

In the seminal study on ‘credit policy, systems, and culture’, Reddy (2004) raised various critical issues pertaining to credit delivery mechanism of the Indian banking sector. The study focused on the terms of credit such as interest rate charged to various productive activities and borrowers, the approach to risk management, and portfolio management in general. There are three pillars on which India’s credit system was based in the past; fixing of prices of credit or interest rate as well as quantum of credit linked with purpose; insisting on collateral; and prescribing the end-use of credit. Interest rate prescription and fixing quantum has, however, been significantly reduced in the recent period. The study also highlighted the issues in security-based or collateralised lending, which need careful examination in the context of growing services sector. Given the fungibility of resources, multiple sources of flow of resources, as well as application of funds, the relevance and feasibility of end-use restrictions on credit need a critical review. The link between formal and informal sectors shows that significant divergence in lending terms between the two sectors still persists, despite the fact that the interest rate in informal markets is far higher than that of the formal sectors- the banking sector. The convergence between formal and informal sectors could be achieved by pushing the supply of credit in the formal sector following a supply leading approach to reduce the price or interest rate. Furthermore, in the context of NPAs on account of priority sector lending, it was pointed out that the statistics may or may not confirm this. There may be only a marginal difference in the NPAs of banks’ lending to priority sector and the banks lending to private corporate sector. Against this background, the study suggested that given the deficiencies in these areas, it is imperative that banks need to be guided by fairness based on economic and financial decisions rather than system of conventions, if reform has to serve the meaningful purpose. Experience shows that policies of liberalisation, deregulation and enabling environment of comfortable liquidity at a reasonable price do not automatically translate themselves into enhanced credit flow.

Although public sector banks have recorded improvements in profitability, efficiency (in terms of intermediation costs) and asset quality in the 1990s, they continue to have higher interest rate spreads but at the same time earn lower rates of return, reflecting higher operating costs (Mohan, 2004). Consequently, asset quality is weaker so that loan loss provisions continue to be higher. This suggests that, whereas, there is greater scope for enhancing the asset quality of banks, in general, public sector banks, in particular, need to reduce the operating costs further. The tenure of funds provided by banks either as loans or investments depends critically on the overall asset-liability position. An inherent difficulty in this regard is that since deposit liabilities of banks often tend to be of relatively shorter maturity, long-term lending could induce the problem of asset-liability mismatches.

The maturity profile of commercial bank deposits shows that less than one fifth is of a tenor of more than three years. On the asset side, nearly 40 per cent has already been invested in assets of over three year maturity. Banks also have some capacity to invest in longer term assets, but this capacity will remain highly limited until the fiscal deficit remains as high as it is and the Government demand for investment in long dated bonds remains high. Some enhancement of their capacity to invest in infrastructure, industry and agriculture in longer gestation projects can be achieved by allowing a limited recourse to longer term bond issues.

In an another study, Mohan (2003) observed that lending rates of banks have not come down as much as deposit rates and interest rates on Government bonds. While banks have reduced their prime lending rates (PLRs) to some extent and are also extending sub-PLR loans, effective lending rates continue to remain high. This development has adverse systemic implications, especially in a country like India where interest cost as a proportion of sales of corporates are much higher as compared to many emerging economies.

The problem of NPAs is related to several internal and external factors confronting the borrowers (Muniappan, 2002). The internal factors are diversion of funds for expansion/diversification/modernisation, taking up new projects, helping/promoting associate concerns, time/cost overruns during the project implementation stage, business (product, marketing, etc.) failure, inefficient management, strained labour relations, inappropriate technology/technical problems, product obsolescence, etc., while external factors are recession, non-payment in other countries, inputs/power shortage, price escalation, accidents and natural calamities. In the Indian context, Rajaraman and Vasishtha (2002) in an empirical study provided an evidence of significant bivariate relationship between an operating inefficiency indicator and the problem loans of public sector banks. In a similar manner, largely from lenders’ perspective, Das and Ghosh (2003) empirically examined non-performing loans of India’s public sector banks in terms of various indicators such as asset size, credit growth and macroeconomic condition, and operating efficiency indicators.

Sergio (1996) in a study of non-performing loans in Italy found evidence that, an increase in the riskiness of loan assets is rooted in a bank’s lending policy adducing to relatively unselective and inadequate assessment of sectoral prospects. Interestingly, this study refuted that business cycle could be a primary reason for banks’ NPLs. The study emphasised that increase in bad debts as a consequence of recession alone is not empirically demonstrated. It was viewed that the bank-firm relationship will thus, prove effective not so much because it overcomes informational asymmetry but because it recoups certain canons of appraisal.

In a study of loan losess of US banks, McGoven (1993) argued that ‘character’ has historically been a paramount factor of credit and a major determinant in the decision to lend money. Banks have suffered loan losses through relaxed lending standards, unguaranteed credits, the influence of the 1980s culture, and the borrowers’ perceptions. It was suggested that bankers should make a fairly accurate personality-morale profile assessment of prospective and current borrowers and guarantors. Besides considering personal interaction, the banker should (i) try to draw some conclusions about staff morale and loyalty, (ii) study the person’s personal credit report, (iii) do trade-credit reference checking, (iv) check references from present and former bankers, and (v) determine how the borrower handles stress. In addition, banks can minimise risks by securing the borrower’s guarantee, using Government guaranteed loan programs, and requiring conservative loan-to-value ratios.

Bloem and Gorter (2001) suggested that a more or less predictable level of non-performing loans, though it may vary slightly from year to year, is caused by an inevitable number of ‘wrong economic decisions’ by individuals and plain bad luck (inclement weather, unexpected price changes for certain products, etc.). Under such circumstances, the holders of loans can make an allowance for a normal share of non-performance in the form of bad loan provisions, or they may spread the risk by taking out insurance. Enterprises may well be able to pass a large portion of these costs to customers in the form of higher prices. For instance, the interest margin applied by financial institutions will include a premium for the risk of non-performance on granted loans.

Bercoff, Giovanniz and Grimardx (2002) using accelerated failure time (AFT) model in their study of Argentina’s banking sector’s weakness measured by the ratio of non-performing loans to total loans found that both bank specific indicators such as asset growth, the ratio of net worth to net assets, the ratio of operating cost to assets, exposure to peso loans, and institutional characteristics relating to private bank and foreign bank and macroeconomic variables including credit growth, foreign interest rate, reserve adequacy (imports/reserves) and monetary expansion (M2/reserves), besides the tequila effect were reasons behind the banking fragility. Their empirical results suggested that bank size measured by log of assets had a positive effect but asset growth had a negative effect on NPLs. The variables such as operating cost, exposure to peso loans, credit growth, and foreign interest rate had negative effect on NPLs. The macroeconomic variables such as money multiplier, and reserve adequacy, institutional characteristics and tequila effect had positive influence on NPLs.

Fuentes and Maquieira (1998) undertook an indepth analysis of loan losses due to the composition of lending by type of contract, volume of lending, cost of credit and default rates in the Chilean credit market. Their empirical analysis examined different variables which may affect loan repayment: (a) limitations on the access to credit; (b) macroeconomic stability; (c) collection technology; (d) bankruptcy code; (e) information sharing; (f) the judicial system; (g) prescreening techniques; and (h) major changes in financial market regulation. They concluded that a satisfactory performance of the Chilean credit market, in terms of loan repayments hinges on a good information sharing system, an advanced collection technology, macroeconomic performance and major changes in the financial market regulation. In another study of Chile, Fuentes and Maquieira (2003) analysed the effect of legal reforms and institutional changes on credit market development and the low level of unpaid debt in the Chilean banking sector. Using time series data on yearly basis (1960-1997), they concluded that both information sharing and deep financial market liberalisation were positively related to the credit market development. They also reported less dependence of unpaid loans with respect to the business cycle compared to interest rate of the Chilean economy.

Altman, Resti and Sironi (2001) analysed corporate bond recovery rate adducing to bond default rate, macroeconomic variables such as GDP and growth rate, amount of bonds outstanding, amount of default, return on default bonds, and stock return. It was suggested that default rate, amount of bonds, default bonds, and economic recession had negative effect, while the GDP growth rate, and stock return had positive effect on corporate recovery rate.

Lis, et.al.,(2000) used a simultaneous equation model in which they explained bank loan losses in Spain using a host of indicators, which included GDP growth rate, debt-equity ratios of firms, regulation regime, loan growth, bank branch growth rates, bank size (assets over total size), collateral loans, net interest margin, capital-asset ratio (CAR) and market power of default companies. They found that GDP growth (contemporaneous, as well as one period lag term), bank size, and CAR, had negative effect while loan growth, collateral, net-interest margin, debt-equity, market power, regulation regime and lagged dependent variable had positive effect on problem loans. The effect of branch growth could vary with different lags.

Kent and D’Arcy (2000) while examining the relationship between cyclical lending behaviour of banks in Australia argued that, the potential for banks to experience substantial losses on their loan portfolios increases towards the peak of the expansionary phase of the cycle. However, towards the top of the cycle, banks appear to be relatively healthy - that is, non-performing loans are low and profits are high, reflecting the fact that even the riskiest of borrowers tend to benefit from buoyant economic conditions. While the risk inherent in banks’ lending portfolios peaks at the top of the cycle, this risk

tends to be realized during the contractionary phase of the business cycle. At this time, banks’ non-performing loans increase, profits decline and substantial losses to capital may become apparent. Eventually, the economy reaches a trough and turns towards a new expansionary phase, as a result the risk of future losses reaches a low point, even though banks may still appear relatively unhealthy at this stage in the cycle.

Jimenez and Saurina (2003) used logit model for analysing the determinants of the probability of default (PD) of bank loans in terms of variables such as collateral, type of lender and bank-borrower relationship while controlling for the other explanatory variables such as size of loan, size of borrower, maturity structure of loans and currency composition of loans. Their empirical results suggested that collateralised loans had a higher PD, loans granted by savings banks were riskier and a close bank-borrower relationship had a positive effect on the willingness to take more risk. At the same time, size of bank loan had a negative effect on default while maturity term of loans, i.e., short-term loans of less than 1-year maturity had a significant positive effect on default.

The brief review of the literature is used to formulate theoretical analysis of non-performing loans undertaken in Section III. The following section highlights the underlying cross-section differences across banks in India using some stylized facts about banks non-performing assets, credit portfolio, and terms of credit, particularly, cost conditions.

Section II Stylised Facts About NPAs in India

Globally, the level of non-performing loans is estimated at about US $1.3 trillion during 2003, of which the Asian region accounts for about US $ 1 trillion, or about 77 per cent of global NPLs (Table 1). Within Asia, Japan and China have NPLs at about US $ 330 billion and US $ 307 billion thus, together accounting for 49 per cent of global NPLs. Other hot spots of NPLs in the Asian

region are Taiwan (US $ 19 billion), Thailand (US$ 18.8 billion), Indonesia (US $ 16.9 billion) and Philippines (US $ 9 billion). India’s NPLs of the financial sector as a whole is reported at about US$ 30 billion, which works out to a little over 2 per cent of global NPLs² . On the basis of information provided in the Report on Trend and Progress of Banking in India, 2002-03, the level of gross NPAs of public sector banks is estimated at Rs. 68,714 crore, which is equivalent to about US $ 15 billion, i.e., about 1 per cent of global NPAs.

In terms of the ratio of NPLs to total assets of banking sector, there is evidence that the performance of the Asian region is far lower than Europe and US (Table 2). The trend in the NPL ratios of select countries, particularly the emerging countries of East Asia, during the last five-six years provides an interesting insight about the impact of structural reform on the accumulation of non-performing loans by banks. In emerging countries like Thailand and Indonesia, structural reform after the Asian crisis had an immediate impact of removing the accumulated excesses from the financial sector and consequently induced a sharp reduction in NPAs

ratio. In Thailand, the NPLs ratio declined by about 25 percentage points from about 43 per cent in 1998 to about 18 per cent in 2003. In a similar manner, the NPAs ratio in Indonesia declined by about 30 percentage points from about 49 per cent in 1998 to 19 per cent in 2003.

In absolute terms, the volume of gross NPAs of public sector banks had increased continuously, except for the year 1995 (Table 3). During the entire period, Gross non-performing assets (GNPA) have increased at a trend rate of 4 per cent. The annual growth rate of gross NPAs of public sector banks shows a decelerating trend since 2000 (Chart1). In terms of various NPA ratios, such as GNPAs

to gross advances, GNPAs to total assets, and net NPAs to total assets and net NPAs to net advances, public sector banks have achieved remarkable improvement (Table 4).

The ratio of gross NPAs to advances has declined at a trend rate of about 9.6 per cent from the level of about 23-25 per cent during 1993-94 to about 9 per cent by end-March 2003. On the other hand, the net NPAs to assets ratio has declined by 50 per cent from the level of 4 per cent in 1995 to about 2 per cent in 2003. The

GNPAs ratio declined by 5 percentage points in the year 1994-95, and another 1.5 per centage point in the subsequent year 1995-96. Thus, in two-year span, the ratio of GNPAs was reduced by 6.5 percentage points, reflecting the immediate effect of structural reforms. Since 1996, the reduction of NPAs has been almost gradual, reflecting on reform induced market forces determining the level of NPAs.

Across the banks, the frequency distribution of gross and net NPAs to advances ratio of public sector banks provides some useful insights. During 1995-96, in the category of gross-NPAs to gross advances ratio upto 6 per cent, there was only one bank, but by end-March 2003, there were 5 such banks (Table 5). For the category of gross-NPA to gross advance ratio upto 8 per cent, the number of banks have increased from 1 during 1995-96 to 10 by end-March 2003. As at end-March 2003, about one-third of public sector banks had the ratio of GNPAs to gross advances upto 7-8 per cent, about 60 per cent banks had NPAs upto 9-10 per cent, three-fourth had NPAs upto 12 per cent and about 90 per cent banks had NPAs upto 15 per cent. There was, however, not a single bank having NPAs more than 20 per cent.

The distribution of banks’ net NPAs to net advances ratio provides a more consolidated picture (Table 6). By end-March 2003, about two-third of banks had NPAs upto 5 per cent, compared to only 11 per cent of banks during 1995-96. About 93 per cent of banks had NPAs within the range of 8-10 per cent and no bank had NPAs above 12 per cent.

A comparative position of GNPAs to advances ratio and the terms of credit variables including maturity (share of term loans in total advances), interest cost of deposits, operating expenses to asset ratio, total expenses to asset ratio, and credit-deposit ratio of public sector banks in India during the last ten years presents an interesting picture (Table 7). There is evidence that, for the public sector bank group, the

GNPA ratio has declined from 23 per cent in 1993 to 9 per cent in 2003, amidst an increase in the loan-maturity from 31 per cent to about 39 per cent, declining cost condition as reflected in declining interest cost of deposits, operating expenses and total expenses, and reduction in credit-deposit ratio during the same period.

The data on maturity terms of bank loans, defined over time dimension, is available for public sector banks group as a whole for the last three years (Table 8). The share of short-term loans upto one year, accounts for about 40-42 per cent of total loans followed by loans with one to three years of maturity, which account for 33-36 per cent of bank loans and three to five years of maturity accounting for the remaining 23-25 per cent. The share of bank loans above one year maturity accounts for about 60 per cent. Although, the data for last three-year in respect of loan maturity shows no clear trend across the maturity spectrum, there is evidence that banks’ prefer largely term loans for more than 1 year maturity.

Data on maturity of loan defined as the share of term-loans in total advances are available for reasonably long period of time across

banks. In this regard, information for the public sector as a whole shows that there is, on an average, an increasing trend in the share of term loans in total advances since 1996, reflecting banks’ approach to a better portfolio management (Table 9). However, the cross-section statistics such as minimum and maximum values and standard deviation of loan-maturity across banks provide evidence of differential portfolio strategies of banks over time. The differential range of maturity, which was more pronounced during early 1990s, has narrowed during the late 1990s due to increasing competition. During the early 1990s, at the bottom end, some banks could provide only 10-15 per cent of term loans, while at the upper end, some banks

could provide as high as 60-65 per cent of term loans in the credit portfolio. Nevertheless, the upswings and downswings in the cross section volatility statistics of loan maturity over time reflect bank specific dynamic portfolio management strategies.

The movement of banks’ prime lending rate (PLR) over the years reflects the general cost condition for borrowers. The prime lending rate of State Bank of India (SBI) has declined by about 750 basis points from the peak of 19 per cent in 1991-92 to 11.50 per cent during 2002-03 (Table 10). For other commercial banks too, there has been decline in PLR. From the level of 1995-96, SBI’s PLR has declined by about 300-450 basis points and similarly for other banks, the PLR has declined somewhat more by about 475 basis points. Another striking feature of distribution of loans across various ranges of interest rates is that the decline in the estimated weighted average lending rate is less than the PLR. For instance, during the period 1995-96 to 2001-02, the PLR of banks has declined by 450-500 basis points, while weighted lending rate has declined by about 300 basis points.

The prime-lending rate, however, is not a fully representative cost of lending for the borrowers as banks lend to different borrowers at different lending rates. Moreover, banks have freedom to lend below PLR to some borrowers. Data on distribution of credit across various ranges of interest rate shows that loans provided at PLR could be about 20-30 per cent of total loans (Table 11).

As the PLR may not be a representative cost of lending, a measure of underlying cost including actual interest costs of deposits and operating expenses, which are available from balance sheets of various banks provide useful insights into the cost structure of public sector banks. In terms of interest cost of deposits, there is evidence that, on an average, effective cost of deposits for public sector banks as a whole ³ , has remained almost steady during 1991-92 to 2000-01 (Table 12). Since 2001-02, however, the interest cost of deposits has declined by about 80 basis points. What is most striking in this regard

is that interest cost varies across banks since the range of cross-section minimum and maximum values of interest cost differs over time. The range of interest rate has shown a marked decline after 2001-02 compared to the period of the early 1990s.

On the other hand, the average operating cost ratio for banks has declined marginally in the recent years as compared to the period of the 1990s (Table 13). Across banks, the range of minimum and

maximum operating cost remains steady indicating the continuity of cross section variability. Reflecting the movement in interest cost of deposits and operating costs, average total cost ratio shows a steady movement for most part of the 1990s and showing at times some firmness in the trend (Table 14). The variability across banks continues to be high.

The average credit deposit ratio shows a declining trend during the 1900’s and an increasing trend thereafter. However, there is significant difference across banks as reflected in the cross-section minimum and maximum values of credit-deposit ratio. Although the variability in credit-deposit ratio has somewhat declined, it has remained steady since 1999-2000 (Table 15).

The ratio of credit to GDP ratio indicates as to how banks have been supporting economic activity of borrowers in general. Since 1990-91, the ratio of bank credit to real GDP shows a continuous improvement except for the year 1994 and 1997, when there was a

marginal decline. The ratio of bank credit to nominal GDP, however, shows a mixed trend. It declined during first half of the 1990s and improved towards the late 1990s, especially since 1999-2000 (Table 16).