India: Selected Issues and Statistical Appendix

This Selected Issues paper presents an analysis of trends in growth and investment in India in the 1990s, with a focus on the slowdown in growth during the second half of the 1990s. The paper discusses the fiscal situation, outlining the key reasons for the deterioration in fiscal balances, how the fiscal situation compares with other developing countries, and the key lessons from countries that managed successful fiscal consolidation. The paper also contains an assessment of India’s opening to global trade and factors that may be affecting India’s export performance.

Abstract

This Selected Issues paper presents an analysis of trends in growth and investment in India in the 1990s, with a focus on the slowdown in growth during the second half of the 1990s. The paper discusses the fiscal situation, outlining the key reasons for the deterioration in fiscal balances, how the fiscal situation compares with other developing countries, and the key lessons from countries that managed successful fiscal consolidation. The paper also contains an assessment of India’s opening to global trade and factors that may be affecting India’s export performance.

V. The Performance of Indian Banks During Financial Liberalization1

A. Introduction

1. During the past decade, the financial system in India has been undergoing a process of liberalization. Bank deposit and lending rates have been deregulated; reserve requirements have been reduced; and regulations on competition, credit allocation, and prudential supervision have been reformed.

2. This paper examines the impact of these reforms on the performance of commercial banks in India.2 In particular, the behavior of industry concentration, cost of intermediation, and profitability of the banking sector are analyzed by focusing on the following questions:

  • How has the level of the industry concentration evolved over the past decade of financial liberalization?

  • Have the cost of intermediation and profitability of different categories of banks—state, nationalized, old private, new private and foreign—changed significantly over the period?

  • What are the determinants of the cost of intermediation and profitability in the Indian banking system?

  • Does ownership structure matter for the intermediation cost and profitability of Indian banks? Do public banks underperform relative to private and foreign banks?

  • What has been the impact (if any) of entry deregulation on these indicators?

3. These questions are addressed using balance sheet and earnings and expenses data for all Indian commercial banks between 1991/92 and 2000/01. Alternative measures of industry concentration, bank spreads and bank profitability are constructed and used in the empirical analysis. Then the cross-sectional and time-series properties of the data are examined in a panel regression framework, under a variety of model specifications and estimation methods.

4. The main findings can be summarized as follows:

  • Industry concentration has declined during the decade. Nevertheless, the combined market share of the three largest banks remains at about one-third of the total assets of the banking system.

  • The cost of financial intermediation and bank profitability appear to have decreased in recent years. This decline is statistically significant for selected definitions of these variables and in most specifications of the regression analysis.

  • State and nationalized are the two types of public sector banks in India. On average, nationalized banks have significantly lower profitability than private and foreign banks. The same result does not hold for state banks.

  • Operational costs, priority sector lending, non-performing loans, investment in government securities, and the composition of deposits (demand, term, savings) play an important role in explaining the bank-level variation in intermediation costs and profitability.

  • The decrease in industry concentration, following the entry of new foreign and domestic banks, is associated with a significant decline in bank intermediation costs and profitability.

5. The rest of the paper is organized as follows. Section B provides a brief description of financial liberalization in the Indian banking sector. Section C describes some findings of the empirical literature on the effect of financial liberalization and public ownership on bank intermediation costs and profitability. Section D discusses the bank-level data and variable definitions used in the empirical analysis. Section E reports the results from the empirical analysis, and Section 6 concludes.

B. Financial Liberalization in India (1991/92–2000/01)

6. This section highlights the key elements of the gradual liberalization strategy implemented during the past decade. The main reforms included: i) interest rate liberalization; ii) reduction in reserve requirements; iii) entry deregulation; iv) credit policies; and v) prudential supervision.3

7. Most deposit and lending rates of commercial banks have been liberalized.4 Before the onset of financial liberalization, deposit and lending rates in India were heavily regulated. Interest rates were administered for all types of deposits (demand, term, and savings). The lending rate structure, on the other hand, was characterized by six loan size categories, each with a minimum lending rate. The chronology of the main events in this process is shown in Annex I.

8. Reserve requirements of commercial banks were gradually reduced. In particular, the average cash reserve requirement (CRR) has fallen from 15 percent to its current value of 5 percent since the start of the reform period.5 The statutory liquidity requirement (SLR) was decreased from 38.5 percent for domestic liabilities and 30 percent for non-resident liabilities to its current level of 25 percent, which is the minimum ratio of liquid assets to demand and time liabilities allowed under the existing law.

9. Entry and ownership restrictions were liberalized. Prior to these reforms, the entry of foreign banks was restricted, and new domestic private banks had not entered the market since the early 1970s. Moreover, private ownership in public sector banks was not allowed. The key changes in the regulations on competition and ownership are described in Annex I.

10. The system of credit delivery has undergone significant changes, including the easing of priority sector lending requirements. During the pre-reform period, the credit decisions of Indian commercial banks were governed by detailed regulations on the provision of cash credit for working capital, credit authorization, holdings of inventory and receivables of various industries, consortium arrangements, etc. The reform efforts in this area were focused on giving banks more discretion in making credit decisions. In addition, the definition of priority sector lending has been expanded gradually, thus making this requirement less restrictive.

11. Prudential supervision norms have been tightening gradually. The specific areas of reform have included the introduction of capital adequacy requirements and the phased improvement of income recognition, asset classification, and provisioning norms (see Chapter VI).

C. Selected Literature Review

12. Financial liberalization has generally been found to have a positive effect on bank performance. Using panel data estimation, Barajas, Steiner, and Salazar (2000) find that financial liberalization and foreign investment in Columbia had a beneficial effect on bank behavior by increasing competition, lowering intermediation costs, and improving loan quality. The positive effect on intermediation spreads is also found in Claessens, et al. (1998), using cross-county data. In Norway, the deregulation of lending rates and volumes improved the efficiency and productivity of banking sector (Berg, Forsund, and Jansen (1992)). According to Zaim (1999), commercial banks in Turkey had a similar experience. In the Indian context, the impact of liberalization on public sector profitability is analyzed in Chaudhuri (2002) and Mohan (2002).

13. Although several studies have found that ownership has a significant effect on bank performance in developing countries, the empirical evidence on the Indian banking sector has been mixed. Barth, et al. (2001) demonstrate that a larger share of state ownership is associated with higher intermediation costs. Using bank-level data from 80 countries, Demirguc-Kunt and Huizinga (1998) report that foreign banks have higher profitability than domestic banks in developing countries. In the case of the Indian banking system, Sarkar, Sarkar, and Bhaumik (1998) find that the differences in performance between public and private banks are not significant. Shirai (2002) concludes that “…even though foreign banks and private sector banks generally perform better than public sector banks in terms of profitability, earnings efficiency and cost efficiency in the initial stage [of reforms], such differences have diminished as public sector banks have improved profitability and cost efficiency.”

14. In this paper, the analysis of the cost of intermediation—measured by several types of bank spreads—is based mainly on Brock and Suarez (2000). In their investigation of the determinants of bank spreads in seven countries in Latin America, Brock and Suarez (2000) point out that “…the study of interest rate and spreads only makes economic sense in a fully liberalized economy….” Therefore, they construct several measures of bank spreads and proceed to analyze the behavior of banks across time and banks.

D. Data

Sample Description

15. The sample comprises all commercial banks in India between 1991/92 and 2000/2001.6 The database was constructed using various issues of Statistical Tables Relating to Indian Banks, Report on Trend and Progress of Banking in India, and Database on Indian Banking, 1987–98. The number of banks in the sample varies across years, owing to the entry and exit of some banks, as well as data availability. The total number of observations in the sample is 882.

16. The structure of the Indian banking sector is characterized by five categories of commercial banks. There are two types of public banks—eight state banks (SBI and seven associates) and 19 nationalized banks. The classification of private banks into “old private” and “new private” is based on the timing of market entry. Following the RBI guidelines of 1993 to promote competition in the banking sector, nine new private banks entered the market in 1994 and 1995.7 A number of foreign banks were allowed entry into the Indian banking system between 1991 and 1998, and consequently, the total number of foreign banks increased from 24 in 1991/92 to 42 in 2000/2001.8 The market shares of the five categories of banks in 2000/01 are shown in Table V.1.

Table V.1.

Market Share by Bank Category in 2000/01

article image

Variable Definitions

Industry concentration and composition

17. The changes in the concentration and composition of the Indian banking industry during the sample period are analyzed using several indicators. To capture the evolution of overall concentration, we construct: i) Herfindahl index of the banking sector in three different ways, using the market share of each bank in the asset, loan, and deposit markets;9 and ii) M-concentration ratios of the one, three and ten largest banks in the asset, loan, and deposit markets. The effect of entry deregulation on the composition of the banking industry is described by tracing the changes over the sample period in: i) the number share of each bank type, i.e., the number of banks of type i (i = state, nationalized, old private, new private and foreign) divided by the total number of banks in a given year; and ii) the market share of each bank type in the asset, loan and deposit markets.

Cost of intermediation and profitability

18. The cost of financial intermediation is measured by four different bank spreads (see Annex II). The most commonly used definition in the literature is the net interest margin (Spread1), i.e., the difference between interest earned and interest expended, normalized by total assets. Since the net interest margin (Spreadl) may not accurately represent the marginal costs and benefits of borrowing and lending, three other bank spread measures are used in the empirical analysis. For example, the definition of Spread4 focuses on the loan and deposit business of the banks only, using the difference between the interest earned on loans (normalized by total loans) and interest expended on deposits (normalized by total deposits). The definitions of Spread2 and Spread3 differ from the net interest margin (Spreadl) in their adjustment for balance sheet composition by normalizing interest earned by total loans and interest expended by total deposits (instead of total assets), thus abstracting from non-loan assets (investments in government securities, balances with RBI and liabilities, etc.) and non-deposit liabilities (borrowings, reserves, etc.) The difference between Spread2 and Spread3 is the inclusion of income from commissions, exchange, and brokerage in Spread2.10

19. The empirical analysis uses two standard measures of bank profitability, before and after provisioning. Bank profitability before provisions and contingencies (Profit2) is equal to the difference between earnings (interest earned plus other income) and expenses (interest expended plus operating expenses), normalized by total assets. Note that Profitl is equal to Profit2 minus (the normalized by total assets) provisions and contingencies.

20. The definitions of profitability and bank spreads are closely related. For example, the profitability before provisioning (Profit2) is the sum of the net interest margin (Spread1) and the surplus of other income over operating expenses. Therefore, if the net interest margin declines, then bank profitability will decrease unless operating costs fall sufficiently (or other income increases) to compensate for this fall.

E. Results

Changes in Industry Concentration and Composition

21. The concentration of the Indian banking sector has declined during the past decade. This fact is illustrated in Figure V.1a, which shows the evolution of three types of Herfindahl indices (asset, loan and deposit). The largest change in concentration—comparing the values at the end-points of the sample period—is observed in the case of the loan-based Herfindahl index (Herf_Advances), which dropped from 0.098 in 1991/92 to 0.07 in 2000/01. The asset-based Herfindahl index (Herf_Assets), on the other hand, declined from 0.1 in 1991/92 to 0.08 in 2000/01. The fall in the deposit-based Herfindahl index (Herf_Deposit) is about three times smaller (0.01). It is interesting to note that the asset-based and deposit-based Herfindahl indices assumed their lowest values in 1997/98, unlike Herf Advances, which continued its decline.

Figure V.1.
Figure V.1.

Herfindahl and M-Concentration Indices

Citation: IMF Staff Country Reports 2002, 193; 10.5089/9781451818567.002.A005

22. The behavior of the Herfindahl indices is dominated by the change in the largest-bank share of total assets, loans and deposits in the banking sector (see Figure V.1b). In particular, the asset market share of SBI fell from 28 percent in 1991/92 to 24 percent in 2000/01, whereas its loan market share declined from 27 percent in 1991/92 to 22 percent in 2000/01. In contrast, the deposit market share of SBI recovered at the end of the sample period to its starting value of 23 percent in 1991/92. The increase in the deposit market share of SBI since 1997/98 materialized at the expense of the nationalized banks.11

23. Additional information about the changes in the market structure is provided by the remaining two M-concentration ratios. Figure V.1c and Figure V.1d show that despite the decline in their relative share, the combined assets of the three largest banks still comprise about one-third of the total assets of the banking system in 2000/01. The asset, loan, and deposit market shares of the ten largest banks, on the other hand, declined continuously during the sample period to just under 60 percent in 2000/01. All of these banks are public.

24. The composition of the banking sector changed with the emergence of new private and foreign banks (Figure V.2). Although the number shares of old private, nationalized, and state banks decreased at the expense of foreign and new private banks (Figure V.2a), their market shares did not adjust as much (Figure V.2b–2d). Measured in terms of the total assets of the banking system, the market share of nationalized banks decreased by 6 percentage points, while new private banks gained a market share of 6 percent.12 The largest decline in the market share of state and nationalized banks occurred in the market for advances—8 and 4 percentage points, respectively. This market share loss was to the benefit of new private banks (6 percentage points), old private banks (3 percentage points) and foreign banks (3 percentage points). New private banks expanded the most in the deposit market as well, as their market share increased at the expense of nationalized banks.

Figure V.2.
Figure V.2.

Number and Market Shares by Bank Category

Citation: IMF Staff Country Reports 2002, 193; 10.5089/9781451818567.002.A005

Changes in Bank Spreads and Profitability13

25. Several patterns emerge from examining the behavior of bank spreads and profitability:

  • The net interest margin (Spreadl) has declined in recent years. The negative and significant coefficients of Year99, Year00, and YearOl in Table V.2ac lend support to this observation.

  • The change in Spread4 follows a similar time pattern to the net interest margin (Spread1), although its level is consistently higher than the latter. As in the case of Spreadl, the coefficients of the year dummies Year99, YearOO, and YearOl in Table V.2a–c are negative and statistically significant. However, the mean value of Spread4 in the sample is almost twice as high as the mean value of the net interest margin, Spreadl. This difference indicates that spreads on activities related to lending and deposits are much higher than the simple net interest margin.

  • Spread3, however, has not shown a significant decline. Moreover, the level of Spread3 is much higher than those of the net interest margin (Spread1) and Spread4 (see Figure V.3). As in the case of Spread4, this spread focuses on the loans and the deposits on the balance sheet side, while using the total interest earned and interest expended of the banks. The rationale behind the definition of Spread3 is to attempt to capture the bank spread that measures the marginal cost of intermediation. Spread3 has not decreased significantly in recent years, as indicated by the lack of statistical significance of the relevant year dummies in Table V.2a–c. Compared to the net interest margin and Spread4, the difference in the evolution of Spread3 can be explained mainly by the increasing share of bank investments in government securities as a proportion of total assets, occurring at the expense of total loans.

  • The results for Spreadl are almost identical to those for Spread3. The difference in levels between the two bank spreads is due to the incorporation of income from commissions, fees, and brokerage in Spread2. Otherwise, the evolution of Spread2 over time is fairly similar to the time pattern exhibited by Spread3.

  • Bank profitability indicators after provisioning (Profitl) and before provisioning (Profit2) have generally decreased in recent years. This conclusion is stronger for bank profitability after provisioning, as indicated by the negative and statistically significant coefficients of the year dummies Year99, Year00, and Year01 in Table V.2a–c.

  • Finally, the bank spreads and profitability indicators of the five bank categories have converged in recent years. For example, the median values of Spreadl for all banks are fairly similar in fiscal year 2000/01.14

Table V.2a.

Panel Regressions with Time Dummies Only 1, 2

article image

The significance levels of 1 percent, 5 percent and 10 percent are denoted by (*), (**) and (***), respectively.

The reference categories in the regression are Year97and 01d_private.

Table V.2b.

Panel Regressions with Time and Category Dummies1, 2

article image

The significance levels of 1 percent 5 percent and 10 percent are denoted by (*), (**) and (***), respectively.

The reference categories in the regression are Year97and 01d_private.

Table V.2c.

Panel Regressions with Time and Category Dummies and Their Interactions1/2/

article image

The significance levels of 1 percent, 5 percent and 10 percent are denoted by (*), (**) and (***), respectively.

The reference categories in the regression are Year97and 01d_private.

Figure V.3.
Figure V.3.

Median Bank Spread and Profitability Measures

Citation: IMF Staff Country Reports 2002, 193; 10.5089/9781451818567.002.A005

Figure V.4.
Figure V.4.

Mean Bank Spread and Profitability Measures

Citation: IMF Staff Country Reports 2002, 193; 10.5089/9781451818567.002.A005

Determinants of Bank Spreads and Profitability15

26. Operating costs, priority sector lending, non-performing loans, investment in government securities, and the composition of deposits are among the determinants of bank spreads and profitability in the Indian banking sector. The cost ratio, defined as the ratio of operating costs to total assets, is a key explanatory variable for bank spreads in India. Banks with higher administrative costs have significantly higher spreads and lower profitability. High levels of priority sector lending16 are generally associated with significantly higher bank spreads. Banks with higher levels of non-performing loans have significantly lower profitability. In some specifications, a larger share of investment in government securities (as a proportion of total assets) is linked to higher spreads (see Table V.3a). Finally, banks with a higher share of current deposits (as a proportion of total deposits) have significantly lower bank spreads and higher profitability.17

Table V.3a.

Determinants of Bank Intermediation Cost and Profitability: Specification 1 (Herf)1, 2

article image

The significance levels of 1 percent, 5 percent and 10 percent are denoted by (*), (**) and (***), respectively.

The reference categories in Ihe regression are Year 97 and Old_privale.

27. A surprising empirical finding is that the number of branches does not have a significant effect on the profitability of Indian banks, but is positively and significantly related to the net interest margin (Spreadl). This robustness of this result is demonstrated by the insignificant coefficient of the variable Branch in all four specifications of the model. This coefficient does not become significant even after dropping the market share and operating costs variable from the model specifications in order to avoid possible multicollinearity. However, the number of branches is found to have a positive and significant impact on the net interest margin (Spreadl). In other words, banks with a more extensive branch network tend to have higher net interest margins (see Table V.3b). One possible explanation for these results could be that banks with large branch networks maintain their profits by charging higher net interest margins in some geographic areas, where there are few other bank branches, i.e., less competition.

Table V.3b.

Determinants of Bank Intermediation Cost and Profitability: Specification 2 (Fnum and Dnum)1, 2

article image

The significance levels of 1 percent, 5 percent and 10 percent are denoted by (*), (**) and (***), respectively.

The reference categories in the regression are Year97and 01d_privatc.

Ownership

28. New private banks have significantly lower bank spreads and higher profitability than old private banks. For example, the coefficient of the dummy variable New_private is negative and significant in Table V.3b. In addition, new private banks have higher profitability than old private banks, both before and after provisioning. However, these results are not robust.

29. Foreign banks have generally higher bank spreads and higher profitability before provisioning than old private banks. This result is supported by the significant coefficients of Foreign in Table V.2b and Table V.3a–d. Although many foreign banks have high operating expenses,18 these banks have preserved their higher profitability by maintaining relatively high spreads. This phenomenon could be partly explained if one presumes that foreign banks offer their customers different type of services than domestic banks. Consistent with this explanation is the fact that foreign banks continue to have higher levels of income from other sources (commissions, brokerage fees, exchange transactions, etc.) than domestic banks, although the median ratio of “other income” of these banks has declined in recent years.

30. Nationalized banks have lower profitability than private and foreign banks. As demonstrated in last two columns of Tables V.2b and V.2c and Table V.3a–d, the coefficient of the dummy variable Nationalized is consistently negative and significant across model specifications, particularly in the equations for profitability after provisioning (Profit1). The result can be largely attributed to their high level of operating costs, mainly because of large wage bill expenses. The median employment costs of nationalized banks have persistently exceeded those of other bank categories. A different (non-regression) examination of the data also reveals that nationalized banks perform worse than other types of banks. A ranking of banks by profitability before provisioning (Profit2) in 2000/01 indicates that most nationalized banks are at the lower end of the profitability distribution. More specifically, 75 percent of all nationalized banks have profitability before provisioning in the lowest two quantiles of the distribution (see Figure V.5).

Figure V.5.
Figure V.5.

Proportion of Barks (by Category) Across Quantiles of the Profitability Ranking Distribution in 2000/01

Citation: IMF Staff Country Reports 2002, 193; 10.5089/9781451818567.002.A005

31. State banks do not exhibit lower profitability than old private banks. The coefficient of the variable State is insignificant in most of the profitability regressions. Complementary evidence on the relative profitability of state banks is presented in Figure V.5, which indicates that most of the state banks are in the top or middle quantiles of the profitability distribution (before provisioning) in 2000/01.

Entry Deregulation

32. The entry of new foreign and domestic banks and the concurrent decrease in industry concentration are associated with a significant decline in bank spreads and profitability. The explanatory variables used to capture this effect are the Herfindahl index (Herf), the market shares of domestic and foreign banks (Dmarket and Fmarket), and the number shares of domestic and foreign banks (Dnum and Fnum). The positive and significant coefficient of the variable Herf in some model specifications (see Table V.3a) suggests that lower industry concentration is linked to lower spreads and profitability. The negative and significant coefficient of the variable Fnum in Table V.3b indicates that the entry of these banks could explain part of the decrease in bank spreads. In contrast, the effect of Dnum on bank spreads is insignificant. However, the increase in the market share of the new private banks (Dmarket) is related to a significant fall in the overall level of bank spreads and profitability (see Table V.3c). An important caveat of these findings is that the contribution of foreign and domestic entry may be overstated, given that the effect of other reforms is not explicitly controlled for in the regression.19

Table V.3c.

Determinants of Bank Intermediation Cost and Profitability: Specifications (Fmarket and Dmarket)1,2

article image

The significance levels of 1 percent, 5 percent and 10 percent are denoted by (*), (**) and (***), respectively.

The reference categories in the regression are Year97and Old private.

Table V.3d.

Determinants of Bank Intermediation Cost and Profitability: Specification 4 (Npl)1,2

article image

The significance levels of 1 percent, 5 percent and 10 percent tire denoted by (*), (**)and (***), respectively.

The reference categories in the regression ure Year97and 01d-private.