- Email: [email protected]
Market structure, conduct and performance: Evidence from the Bangladesh banking industry
Available online at www.sciencedirect.com
Journal of Asian Economics 19 (2008) 181–193
Market structure, conduct and performance: Evidence from the Bangladesh banking industry Abdus Samad * Department of Finance and Economics, Utah Valley State College, 800 W. University PKY, Orem, UT 84508, United States Received 26 August 2007; received in revised form 14 December 2007; accepted 14 December 2007
Abstract With regard to the market structure and performance in Bangladesh banking industry, there are two competing hypotheses—the traditional structure–conduct–performance (SCP) hypothesis and the efficiency hypothesis (EH). Using pooled and annual data for the period 1999–2002, this study tests the validity of these two hypotheses. In general, the results of this study support the EH hypothesis as an explanation for market performances in Bangladesh, but for definitive policy purposes, the impact of the banking structure needs to be explored further. Published by Elsevier Inc. JEL classification: G21 Keywords: Bank performance; Market structure; Bangladesh banking
1. Introduction In bank performance literature, two competing hypotheses have been the subject of controversy for more than 40 years. The oldest hypothesis is the traditional structure–conduct–performance (SCP), also known as structureperformance (SP), hypothesis. According to SP hypothesis (Gilbert, 1984; Hannan, 1991), the profitability of a banking firm is dependent upon the market structure and the level of competition. The lower the level of competition in the market, the higher the economic rent for a firm. The basic message of the SP hypothesis is that a higher concentration ratio leads to a higher profitability. The hypothesis that concentration leads to higher profitability has been challenged by an alternative hypothesis, known as the efficiency hypothesis (EH). The EH emphasizes superior efficiency as an explanation for a firm’s profitability. According to EH, ‘‘there is no relationship between concentration and profitability, but rather market share and bank profitability’’ (Smirlock, 1985, p. 69). In other words, the performance of an individual firm depends on the firm’s degree of efficiency. In this hypothesis, the explanation for the relationship between the market structure and performance of a firm is dependent upon the firm’s efficiency. If a firm enjoys a higher degree of efficiency than its competitors (that is, if the firm has a relatively low cost of production structure), the firm can maximize profits and increase its size and market share. This can be done by keeping the
* Tel.: +1 801 863 8368. E-mail address: [email protected]. 1049-0078/$ – see front matter. Published by Elsevier Inc. doi:10.1016/j.asieco.2007.12.007
182
A. Samad / Journal of Asian Economics 19 (2008) 181–193
present market price level. Thus, the increase of profits and market share is the result of efficiency, not of a higher concentration ratio (collusion). The proponents of both hypotheses claim support in their favor. There are three motivating factors that brought about this study. First, given the controversy between the two competing hypotheses, it is important to determine which hypothesis truly approximates and describes the Bangladesh banking industry. Determining the approximate behavioral relationship has important policy implications for Bangladesh because the presumed positive relationship between the market concentration and higher performance, for example, higher profits influences regulatory decisions regarding mergers and de novo entry. Increasing market concentration is also considered a social cost. Second, the relevance of the two hypotheses has been explored and tested in the context of the bank markets in developed countries. This study evaluates the validity of the two hypotheses in the context of the banking market in Bangladesh, a third-world country. Until 1980, the banking system of Bangladesh was quite concentrated and was regulated and protected from foreign competition. The domestic bank market was dominated by the operation of five nationalized commercial banks. This fact was conducive to collusive bank behavior leading to higher profitability. Thus, Bangladesh provides a good test case for evaluating the validity of the two competing hypotheses. Third, since the controversy of SCP and EH has not been resolved, the hypotheses deserve further exploration. This study can contribute to new findings in the banking performance literature. This paper is structured as follows: Section 1 provides a brief description of the institutional structure of the Bangladesh banking market. Section 2 describes SCP and EH in the context of market structure and performance relationship. Section 3 provides data and methodology. Section 4 reports results and conclusions. 2. Institutional structure of the Bangladesh banking market Bangladesh is a small country bordering India. The country seceded from Pakistan and gained independence in 1971. At that time, all commercial banks were nationalized by the government without compensation. The question of compensation did not arise at this time because the owners of the commercial banks were non-Bengalis who left Bangladesh before or during the war of liberation. The banking system of Bangladesh was highly regulated and protected from foreign competition and was concentrated until 1980. There were only five nationalized commercial banks that dominated the domestic banking market in Bangladesh at the time. The first move towards competition in the banking market of Bangladesh was undertaken during the 1980s with the denationalization of the Pubali Bank, one of the five nationalized banks. At this time, the opening of private and foreign banks was allowed and encouraged. At present, there are 41 privately owned commercial banks operating in Bangladesh. Of those banks, 12 are foreign. The structural features of the Bangladeshi banking system are provided in Table 1. An explanation of Table 1 follows: 1. Until 2001, banking activity of commercial banks was dominated by nationalized commercial banks. There were four banks owned by the government of Bangladesh. The number of the nationalized commercial banks was four. Thus, only four commercial banks controlled over 48% of the total assts of the entire banking system in Bangladesh during the year 2000. Privately owned commercial banks controlled a little over 50% of the total assets until 2000. Table 1 Bangladesh banking system total assets and distribution Institutions
Assets in millions TK
Commercial banks
1999
2000
2001
1999
Percentage share of assets 2000
2001
Nationalized commercial banks Domestic private banks Foreign banks All private banks Total banking system
446,350 376,197 88,437 464,634 910,984
506,119 450,481 97,261 547,742 1,053,861
589,366 1,405,466 115,349 1,520,815 2,110,181
48.99 80.96 19.04 51.01 100
48.02 82.24 17.76 51.98 100
27.92 92.41 7.59 72.03 100
Source: Publications from the Ministry of Finance of the government of Bangladesh. Percentage of assets is calculated by the author. The increase of asset growth was the result of many factors. One of the factors was the increase in exports supported by private commercial banks. Bangladesh’s exports increased from $249,860 (Million) in 2000 to $269,150 (million) in 2001. Source: International Financial Statistics, published by IMF. The increase of exports increased bank deposits and assets.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
183
However, there was a significant increase, from 51.98 to 72.03% in the market share by the privately owned commercial banks in 2001. 2. Within the privately owned commercial banks, domestic private commercial banks dominated the banking activity. Domestic private commercial banks controlled a significant percentage of all private bank assets (see Table 1). Privately owned domestic banks controlled 92.41% of all private bank (domestic and foreign) assets in 2001. 3. Although foreign banks have been active in retail banking, their activities were concentrated in the metropolitan area of Dhaka, the capital city of Bangladesh. Banking activity has declined due to severe competition from domestic private banks (see Table 1). The foreign bank market share in the total assets of all private banks declined from 19.04% in 1999 to 7.59% in 2001. 4. The banking market in Bangladesh is quite concentrated and dominated by four nationalized banks. With the growth of private banks, these nationalized banks face strong competition (see Table 1). The asset market share of nationalized commercial banks declined from 48.99% in 1999 to 27.91% in 2001. On the other hand, the asset market share of private commercial banks increased from 51% in 1999 to 72% in 2001. 3. Structure-performance relationship and background of controversy The notion that market structure influences a firm or industry performance originates in the classical model of the theory of firm. In a pure competitive model, there are a large number of firms. The larger the number of firms, the higher the competition and lower the concentration ratio. The lower the concentration ratio, the lower the profit rates for a firm. Each firm’s market share is insignificant and its product homogenous. Because of these factors, firms cannot influence market price. In such a market, concentration of market share is absent. Therefore, productive efficiency is achieved when price equals minimum of ATC and firms earn normal profits. In contrast, firms under a pure monopoly or monopolistic competition, have significant market shares and thus, have control over the price of the product. In such circumstances, price exceeds the marginal cost MC = MR and leads to sub-optimal production and supernormal profits. The notion is that the higher the concentration of market share by a firm, the higher the economic profits. Market concentration is associated with increasingly higher profits known as ‘‘Recardian rent.’’ The notion that there exists a monotonic relationship between market concentration and performance is a matter of empirical evidence. Bain (1951) first tested this hypothesis – market share concentration and performance – and found that increased concentration led to higher profit rates. This result originated in the structure–conduct and performance controversy. The controversy concerning the relationship between the market structure and performance has resulted in two testable hypotheses: the traditional hypothesis known as SCP and the EH also known as efficient structure hypothesis (ESH). 3.1. Structure–conduct–performance hypothesis The SCP hypothesis is based on the following proposition: when a few firms have a large percentage of market shares, this fosters collusion among firms in the industry. This can happen overtly or without public knowledge. Collusive behavior increases as market share is concentrated in the hands of a few firms. The higher the concentration ratio in a market, the higher the profitability performance of the firms. Thus, according to the SCP hypothesis, there is a positive correlation between the degree of market share concentration and the firm’s performance. Due to collusive or monopolistic reasons, ‘‘firms in a concentrated market will earn higher profits than firms operating in a less concentrated one, irrespective of efficiency’’ (Lloyad-Williams, Molyneux, & Thornton, 1994, p. 437). This hypothesis could be supported if the impact of market concentration was found to be significantly positive, irrespective of the efficiency of the firm. There are a number of empirical studies in the banking market that provide support in favor of SCP hypothesis. The prominent early studies that supported the SCP hypothesis were studies by Rose and Fraser (1976), Heggestad and Mingo (1977), Spellman (1981), and Rhoades (1982). Podenda (1986), Lloyad-Williams et al. (1994), and Samad (2005). Podenda studied the California banking market. The results of Podenda’s study are consistent with the idea that ‘‘increased concentration is associated with increasingly high profits’’ (p. 5). Lloyad-Williams et al. (1994) tested the two hypotheses in the context of the Spanish banking industry by using pooled data and found support in favor of the SCP hypothesis. In a major survey done by Gilbert (1984), 44 studies were summarized for the impact of bank
184
A. Samad / Journal of Asian Economics 19 (2008) 181–193
performance to a market concentration. Thirty-two of the 44 studies Gilbert reviewed found that market concentration significantly and positively affected bank performance. Brucker (1970), Edward (1964), Phillips (1967), and Vernon (1971) found evidence in support of the market structure hypothesis (SCH). They found that the coefficient of market structure, CR3, is statistically significant. 3.2. Efficiency structure hypothesis The SCP hypothesis – the positive association between market concentration and performance – has been challenged by the ESH. The basic foundation of the ESH is that market concentration is not the cause of a firm’s superior performance. Instead, ESH finds that the positive direction of concentration and higher performance is the result of a firm’s superior efficiency. It is argued that the higher profits enjoyed by large firms in a concentrated market are the result of economies of scale and the consequences of superior efficiency in larger firms. If a firm enjoys a higher degree of efficiency (in terms of cost and technology) than its competitors, the firm can easily capture a larger market share by lowering its price and earning economic profits. Thus, the driving force behind the process of gaining a large market share, and thus concentration, is the efficiency of the firm. The most efficient firms will gain market share and earn economic profits. In summarizing the basic points of the proponents of EH, Smirlock (1985) wrote: ‘‘concentration is not a random event but rather the result of superior efficiency of leading firms’’ (p. 70). Firms possessing a comparative advantage in production become large and obtain a high market share, and as a consequence, the market becomes more concentrated. Such firms earn Ricardian rent. Therefore, the basic message of ESH is that leading firms’ efficiency leads to increased market share regarding concentration and is positively correlated with higher performance. The main proponents of the EH are Demsetz (1973), McGee (1974), Peltzman (1977), and Brozen (1982). In these studies the authors challenged the SCP hypothesis. According to the studies, market concentration was not a random event; rather it was the result of a firm’s efficiency. Using manufacturing data, the authors tested the two competing hypothesis and found evidence in support of the ESH. Using banking data, Gillini, Smirlock, and Marshall (1984), Smirlock (1985), and Evanoff and Fortier (1988) tested the competing hypotheses, SCP and EH, and found that firm-specific efficiency was a factor for explaining the profitability in the United States banking industry. Fleching (1965), Ware (1972) and Fraser and Alvis (1975) found that market structure is not significant and did not support SCH. Given controversy between the two competing hypotheses, determining the approximate behavioral relationship is important for bank performance literature. The present study is motivated in this direction. The study has important policy implication for Bangladesh because of the presumed positive relationship between market concentration and higher performance. For example, higher profits influence regulatory decisions regarding the issue of mergers and de novo entry. 3.3. Data and methodology This study includes cross-sectional data for all 44 commercial banks operating in Bangladesh. All relevant data for the period 1999–2001 was obtained from the consolidated balance-sheets and income-statements published by the Ministry of Finance, Peoples Republic of Bangladesh1. The methodology used in this paper is based on the approach utilized by Smirlock (1985) and Lloyad-Williams et al. (1994). These studies asserted that the correct approach in testing the competing hypotheses was to ‘‘take both market share and concentration into account at the same time’’ (Weiss, 1974, pp. 225–226). Accordingly, this study uses the following equation:
P ¼ a0 þ a1 CR þ a2 MS þ
1
See Appendix A for details.
X
aiZi
(1)
A. Samad / Journal of Asian Economics 19 (2008) 181–193
185
where P measures profit rate as firm’s performance, CR denotes concentration ratio representing the measure of market concentration. MS is a measure of market share of the firm and Z is a vector of control variables which are included to take account for firm-specific and market-specific characteristics. Eq. (1) provides the straightforward discrimination between the two hypotheses. The SCP hypothesis can be supported by the coefficient of CR, a1 > 0 and the coefficient of MS, a2 = 0. A coefficient combination of a1 > 0 and a2 = 0 implies that the market share does not affect a firm’s profitability and that profitability is the result of a monopoly behavior measured by concentration. On the other hand, ESH is found to be true if a1 = 0 and a2 > 0. A coefficient combination of a1 = 0 and a2 > 0 implies that firms with a large market share are more efficient than their rivals and thus earn higher profits. In addition market concentration does not affect bank profitability. Thus, a1 > 0 and a2 = 0 supports SCP and a1 = 0 and a2 > 0 supports ESH. The following is a complete OLS regression equation which is estimated in testing the relevance of SCP and ESH for the Bangladesh banking industry: P ¼ a0 þ a1 CR3 þ a2 MS þ a3 CRTA þ a4 LDEP þ a5 ASSET þ a6 OWNER
(2)
where P is a dependent variable, and measures bank performance. Since a bank is a multi-product service industry, prices of certain individual products or services are not a good measure for bank performance2. In a multi-product service industry such as banking, cross subsidization among the products and services is more common than in other single product industries. Similar to LMT (1994), this paper uses profit for measuring bank performance. Profitability is a consolidated figure and takes into account all products and services regarding profits and losses. This overcomes the problem of cross-subsidization. This paper uses two measures of profits: total net profits P A ¼ ROA: ROA ¼ total assets total net profits P C ¼ ROE: ROE ¼ total equity capital There are six independent variables in this model. They are as follows: 1. CR3 = Three bank market concentration ratios. CR3 is used to measure market structure. This study uses two P measures of CR3—one CR3 for assets and the other CR3 for deposits. Where CR3 = assets (deposits) of the top three banks/total assets (deposits) in the market. The sign for a1, coefficient of CR is expected to be positive and significant for the SCP hypothesis. 2. MSi = Market share of ith bank in assets or deposits. In firm-specific market share, MS is used to capture firm efficiency. The sign for a2, coefficient of MS is expected to be positive and significant for ESH. 3. CRTA = Capital and reserve to total assets, where CRTA = total capital and reserve/total assets. 4. LDEPi = Loans to deposit ratio of banki. LDEP = loans of banki/total deposits of banki. Since bank profits, ROA and ROE are not independent of risk, this paper uses two variables – CRTA and LDEP – for measuring bank specific risk. The sign for the CRTA coefficient, a3, is expected to be negative. The higher the capital and reserve as a percentage of total assets, the lower the risk for a bank and, therefore, the lower the expected rate of profits. Similarly, the higher the amount of loans as a percentage of deposits, the higher the risk for a bank. To compensate for the higher risk, the bank is expected to earn a higher rate of return. Therefore, the sign for the LDEP coefficient, a3, is expected to be positive. 5. ASSETSi = ith bank assets. Assets measure the bank size. The asset of each bank is included to take account of the differences in performance brought about by the bank size. The coefficient for ASSETS may show a positive or negative sign depending upon the economies of scale. 6. OWNERi = Ownership of banki. The ownership of ith bank is represented by the binary value = 1 for publicly (government) owned banks 0 for private owned banks. Since the policy of publicly owned banks in Bangladesh is guided by one of the government’s major objectives – removing poverty and supporting the poor – loans are issued to the poor. Such loans to certain classes are expected to be non-performance. Secondly, loans to high-level 2
Although individual product or service price is not a good measure (because taking a single price for a multi-product firm may underestimate or overestimate the performance of a firm), many SCP studies have used it. See Kaufman (1966), Fraser and Rose (1971).
186
A. Samad / Journal of Asian Economics 19 (2008) 181–193
government officials for public sector banks are expected to be in default and become non-performance. Therefore, the sign of the coefficient for OWNER is expected to be negative. 3.4. Heteroscedasticity Since cross-sectional data deals with members of a population at a given point of time, such members may be of different sizes such as small, medium, and large. ‘‘Heteroscedasticity may be a rule rather than the exception is generally expected if small, medium, and large-size firms are sampled together’’ (Guzrarti, 1995, p. 368). As this study involves various categories of banks such as small, medium, and large heteroscedasticity is likely to be common. In order to overcome the heteroscedasticity problem, this study utilizes White’s heteroscedasticity correction method. EVIEW package provides White’s heteroscedasticity corrected outputs. All results of Eq. (2) reported in the next section are based on White’s heteroscedasticity correction. 4. Empirical results Three types of regressions were computed with cross-sectional data for 42 banks in Bangladesh. First, liner regressions were estimated with pooled data for 1999–2002. Regressions 1–12 in Table 2 report the results of the pooled data for 1999–2002. 4.1. Findings from pooled data 1. With the imposition of restrictions CR = 0 and MS 6¼ 0 and vice versa, Eqs. (2), (3), (5), (6), (11), and (12) show that neither the coefficient of CR nor MS is significant whether they are considered for assets and deposit markets in Bangladesh. The results suggest that neither of the two hypotheses – SCP and ESH – can explain the profit performance of Bangladesh banks. However, Eq. (9) provides an exception. The coefficient of MS in a deposit market, in Eq. (9), is statistically significant. The significance of MS in one out of 7 restricted equations lends support in favor of ESH. 2. According to Smirlock (1985), the test of validity for the alternative hypothesis is to estimate (2) without any parameter restriction. When no restriction is imposed, these results are presented in Eqs. (1), (4), (7), and (10). These results show that two out of four, Eqs. (1) and (4), do not support either of the two hypotheses. The other two equations in the deposit market, Eqs. (7) and (10), provide support in favor of ESH because the coefficients of MS in both Eqs. (7) and (10) in Table 2 are statistically significant with the 5% level of significance. On the other hand, the coefficients of CR3 for both these equations are not significant. The result support the validity of the EH and rejects the SCH for Bangladesh. 3. The signs for the MS coefficients are positive and consistent with the exception of Eq. (3). However, the coefficient MS is not significant. Contrary to the model’s expectation, the signs for CR are negative and not statistically significant. The plausible reason for the negative sign is that the three largest banks which accounted for 52% of the market assets were nationalized banks owned by the government of Bangladesh and their profit performance was unsatisfactory. The profits of the Rupali bank, one of the three largest banks, earned TK 308 million losses in 1999. As a result, the ROA is negatively related to CR. 4. In almost all of the regressions 1–12, the signs for the coefficient of CRTA, LDEP, Assets and OWNER are consistent with the expectation of model and are statistically significant. This suggests that bank performances are significantly dependent upon CRTA, LDEP, Assets and OWNER. Since the public sector’s government owned banks were running losses, the coefficient of OWNER is negative and statistically significant. 5. R2 of all the regressions for ROA are in the range between .16 and .83. The F-statistics for these regressions are significant in most cases. 4.2. Result from the non-pooled data Regressions are computed with yearly data. Regressions 1–12 in Table 3 report the result for 1999. Regressions 1– 12 in Table 4 are estimated with the sample for 2000. Regression result for the year 2002 is provided in Table 5.
Eq.
Category
Dependent Intercept variable
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Pooled Pooled Pooled Pooled Pooled Pooled Pooled Pooled Pooled Pooled Pooled Pooled
PA PA PA PC PC PC PA PA PA PC PC PC
asset asset asset asset asset asset deposit deposit deposit deposit deposit deposit
0.017 (4.8)*** 0.015 (4.7)*** 0.019 (4.5)*** 0.58 (2.89)** 0.78 (3.13)** 0.41 (3.7)** 0.04 (4.6)*** 0.04 (4.8)*** 0.05 (4.8)*** 0.06 (3.2)** 0.67 (3.2)** 0.6 (3.8)**
CR 0.04 (0.89) 0.03 (0.75)
MS asset
MS deposit
CRTA
0.04 (21.32)*** 0.04 (28.9)*** 0.08 (0.67) 0.05 (22.5)*** 3.1 (98) 17.05 (1.0) 0.28 (2.8)* 2.4 (1.21) 0.31 (2.68)* 7.8 (1.2) 0.25 (3.9)* 0.08 (0.45) 0.25 (2.78)** 0.34 (23.75)*** 0.03 (1.21) 0.03) (24.1)*** 0.22 (4.0)*** 0.04 (24.6)*** 2.3 (0.88) 0.03 (3.6)** 0.42 (2.1)* 0.7 (0.6) 0.21 (2.6)* 0.87 (0.98) 0.3 (2.4)* 0.09 (0.67)
LDEP
Asset
Owner
R2 F
0.005 (3.08)** 0.0003 (4.0)** 0.0004 (3.8)** 0.02 (65) 0.03 (0.87) 0.01 (0.68) 0.006 (3.23)** 0.004 (3.1)** 0.004 (3.4)** 0.3 (2.9)* 0.03 (0.6) 0.04 (0.65)
2.21E-04 (3.78)** 2.1E-05 (0.89) 2.121E-05 (0.89) 0.004 (2.0)* 4.53E-05 (0.67) 8.1E-05 (2.78)* 1.2E-06 (0.9) 2.3E-06 (0.88) 2.4E-05 (0.7.9) 6.4E-07 (0.41) 5.3E-08 (2.2)* 3.2E-07 (0.84)
0.016 (2.26)** 0.02 (3.2)** 0.021 (2.01)* 1.86 (2.4)* 1.75 (2.56)* 1.8 (2.5)* 0.03 (4.7)*** 0.21 (4.2)*** 0.2 (3.3)** 2.1 (2.4)* 2.1 (2.6)* 2.1 (2.4)*
.67 12.2 .65 13.6 .59 15.4 .31 2.9 .16 078 .24 1.9 .65 10.2 .62 12.6 .63 12.7 .19 2.4 .16 1.9 .18 1.5
***Indicates the variable is significant at the .0001 level. **Indicates the variable is significant at the level of .009. *Indicates the variable is significant at the level of .05. The reported R2 is adjusted R2. The results are White heteroskedasticity adjusted. Figures in parentheses are t-statistics.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
Table 2 Regression results for pooled data
187
188
Table 3 Regression results #
Category
Dependent Intercept variable
CR
PA
0.019 (4.09)*** 0.0002 0.02 (0.96) 0.34
2 1999 asset
PA
0.018 (4.57)*** 0.0001 0.02 (0.70) 0.48
3 1999 asset
PA
0.018 (4.18)*** 0.0002
4 1999 asset 5 1999 asset 6 1999 asset
PC PC PC
0.60 (2.54)* 0.01 0.82 (3.27)** 0.002 0.72 (3.27)** 0.002
7 1999 deposit
PA
0.018 (4.05)*** 0.0003 0.006 (0.35) 0.72
M5 deposits
0.08 (0.72) 0.47
0.05 (0.58) 0.56 2.56 (1.00) 0.32 2.19 (1.15) 0.24
19.05 (1.00) 0.32 8.77 (1.41) 0.16
8 1999 deposits PA
0.018 (4.12)*** 0.0002
9 1999 deposit
PA
0.01 (4.12)*** 0.0002
10 1999 deposit
PC
0.07 (3.06)** 0.004
2.64 (0.99) 0.32
11 1999 deposit
PC
0.72 (3.19)** 0.003
0.06 (0.04) 0.96
12 1999 deposit
PC
0.72 (3.17)* 0.003
0.02 (1.58) 0.12
CRTA
LDEP
0.03 (22.31)*** 0.0000 0.03 (23.92)*** 0.0000 0.03 (22.78)*** 0.0000 0.21 (2.19)* 0.03 0.28 (2.45)* 0.02 0.25 (2.46)* 0.01
0.004 (3.04)** 0.004 0.0004 (3.16)** 0.003 0.0004 (3.10)** 0.003 0.01 (0.32) 0.72 0.02 (0.36) 0.72 0.09 (0.41) 0.68
0.03 (22.73)*** 0.0000
0.004 (3.06)** 0.004
Asset
2.44E-07 (3.85)*** 0.0005 0.03 (22.51)*** 0.004 (3.03)** 1.02E-07 (096) 0.37 0.20 (3.73)*** 0.03 (23.26)*** 0.0004 (3.09)** 2.41 E-07 0.0007 0.0000 0.003 (3.67)*** 0.0008 16.82 (1.19) 0.24 0.25 (2.34)* 0.02 9.75E-07 (0.23) 0.81 0.25 (2.41)* 0.02 0.02 (0.40) 0.68 9.74E-06 (1.17) 0.24 10.69 (1.15) 0.26 (2.39)* 0.02 0.02 (0.43) 0.66 2.22 E-07 0.25 (0.04) 0.96 0.22 (2.48)** 0.01
Owner
R2 F
0.019 (2.78)** .61 11.38 0.0000 0.008 0.019 (2.61)* 0.01 .60 13.53 0.0000 0.015 (1.64)
.61 14.57 0.0000
1.82 (2.05)* 0.05 1.58 (2.43)* 0.02 1.889 (2.36)* 0.02 0.02 (4.44)*** 0.0001
.16 .09 .14
1.03 0.41 0.73 0.57 1.30 0.29
.62
9.59 0.0000
0.02 (4.19)*** 0.0002 0.02 (3.33)** 0.002
.61 11.47 0.0000 .62 11.83 0.0000
2.06 (2.42)* 0.02 1.87 (2.40)* 0.02
.18
1.05 0.41
.14
1.00 0.43
2.40 (2.04)* 0.04
.17
1.19 0.33
Figures in parentheses are t-statistics. Figures after parentheses are p-value. ***Indicates the variable is significant at the .0001 level. **Indicates the variable is significant at the level of .009. *Indicates the variable is significant at the level of .05. The reported R2 is adjusted R2. The results are White heteroskedasticity adjusted.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
1 1999 asset
M6 assets
A. Samad / Journal of Asian Economics 19 (2008) 181–193
189
The regression results for the 1999 data are presented in Table 3. 4.3. Findings for the 1999 sample in Table 3 1. With restriction CR = 0, regression equation (9) in Table 3 shows that the coefficient of MS in deposit markets was positive and significant. This clearly supports ESH. Similarly, the results of the regression in Eq. (7), where ROA is regressed on both CR and MS of deposit markets, simultaneously show that the coefficient of CR is negative and insignificant. On the other hand, the coefficient of MS (in regression 7) is positive and significant. Thus, the result of both regressions, regression equations (7) and (9), support evidence in favor of ESH. This suggests that the bank MS is a significant factor for explaining bank performance. 2. The coefficients of CR in all regressions in Table 3 – whether restriction MS = 0 or MS and CR 6¼ 0 – are not significant. This suggests that SCP is not a valid approximation for explaining bank performance in the Bangladesh banking market. 3. With regard to other independent variables such as CRTA, LDEP, ASSET, and OWNER, it is found that they were significant factors for explaining bank performance in the Bangladesh banking market. This is evident from the tstatistic and p-value of the coefficient of CRTA, LDEP, ASSET, and OWNER. The coefficients of CRTA and LDEP in all regressions 1–12 in Table 3 are consistent with the expectation of the model and are significant. The higher the amount of capital and reserve as a percentage of total assets, the lower the risk for the bank, and therefore, the lower the ROA. ROA and CRTA are, thus, negatively related. Table 3 showed the coefficient of LDEP is positive and significant in most equations. The higher the amount of loans as a percentage of deposits, the greater the risk for a bank. In order to compensate for the higher risk, banks earn higher ROA. Thus, ROA and LDEP are negatively related. The coefficient of ASSETS in Eqs. (7) and (9) in Table 3 are negative and significant. The higher the amount of assets for a bank the greater advantage for the diversification of product and loans. The increased diversification implies less risk and less profits. Since the three largest concentrated banks in Bangladesh are owned by the government and run losses, the coefficient of OWNER is consistently negative. 4. R2 of all the regressions for ROA are between .82 and .83 and the F-statistics of these regressions are significant. This suggests that the model explains well and the overall explanatory power of the model is satisfactory. The regression results for the 2000 data are presented in Table 4. 4.4. Findings for 2000 sample in Table 4 1. Similar to findings reported in Table 3, it was found that the coefficient of MS in the deposit market of 2000 (Eq. (9), Table 4) is significant whether the restrictions CR = 0 and CR and MS 6¼ 0 are imposed. The coefficient of MS is positive (0.20) and significant. The significance of the MS coefficient supports ESH for the banking market in Bangladesh, without imposing any restrictions like. When ROA is regressed simultaneously on both MS and CR along with other independent variables, the result (Eq. (7), Table 4) shows that the coefficient of MS is significant while the coefficient of CR is not significant. The result of Eq. (7) reinforces the evidence supporting in favor of the ESH that firm efficiency, not the firm’s concentration, is a significant factor for explaining bank performance. 2. The coefficients of CR in all equations in Table 4 with the imposition of restrictions, CR = 0 and CR and MS 6¼ 0, are not statistically significant. This suggests that the concentration of the bank CR is not a significant factor for explaining bank performance. This explanation rejects the validity of the SCP hypothesis in Bangladesh. 3. The sign for CRTA, LDEP, ASSETS, and OWNER in most equations are consistent with the expectation of the model. The coefficient of CRTA for Eqs. (1), (2), (3), (7), (8), and (9) are positive and significant. The sign is not inconsistent. Banks with a higher amount of capital and reserve, as a percentage of total assets, might expect to earn a higher ROA if they are engaged in risky investment projects. The sign for OWNER is consistently negative and significant in equations where the ROA is considered as a dependent variable. 4. R2 of all the regressions for ROA is in the range between .60 and .62 and the F-statistics of these regressions are significant. This suggests that the model explains the overall explanatory power satisfactorily. The regression results for extending one more year for example 2002 do not change. The regression results for the 2002 data are presented in Table 5.
190
Table 4 Regression results #
Dependent Intercept variable
2000 2000 2000 2000 2000 2000 2000
PA PA PA PC PC PC PA
assets asset asset asset assets asset deposit
CR
M6 assets
0.013 (2.22)* 0.010 (0.65) 0.51 0.017 (0.16) 0.87 0.013 (2.52) 0.01 0.016 (0.93) 0.013 (2.27)* 0.02 0.03 (0.40) 0.68 1.03 (2.79)** 0.008 1.98 (0.65) 0.51 17.60 (1.24) 0.22 1.20 (2.76)** 0.009 3.48 (1.30) 0.20 1.04 (2.83)** 0.007 14.65 (1.38) 0.17 0.012 (2.02)* 0.04 0.018 (0.08) 0.49
8 2000 deposit PA
0.013 (2.13)* 0.03
9 2000 deposit PA
0.013 (2.17)* 0.03
M5 deposits
CRTA
0.08 0.08 0.08 0.04 0.05 0.05 0.35 (1.96)* 0.05 0.08
0.004 (0.20) 0.84
0.08 0.30 (2.04)* 0.04
0.08
10 2000 deposit PC
0.93 (2.60) 0.01
5.31 (0.98) 0.33
50.84 (1.45) 0.14
0.05
11 2000 deposit PC
0.99 (2.68)* 0.01
1.98 (0.69) 0.49
35.66 (1.39) 0.17
0.05
12 2000 deposit PC
1.01 (2.20) 0.01
35.66 (1.39) 0.17
0.05
LDEP
Asset
(345.95)*** 0.0000 0.002 (1.46) 0.15 (360.64)*** 0.0000 0.002 (1.50) 0.14 (352)*** 0.0000 0.002 (1.49) 0.14 (2.62)* 0.01 0.08 (1.68) 0.28 (2.61)* 0.01 0.08 (1.08) 0.28 (2.67)* 0.01 0.08 (1.10) 0.27 (331.01)*** 000.00 0.002 (1.47) 0.14 1.63 E-07 (1.50) 0.14 (1.44) 0.15 0.002 (1.44) 0.15 2.03 E-08 (0.17) 0.86 (336.7)*** 0.0000 0.002 (1.48) 0.14 1.62 E-07 (1.50) 0.14 (2.49)* 0.01 0.06 (0.85) 0.40 963 E-06 (1.24) 0.22 (2.445)* 0.01 0.06 (0.07) 0.38 1.68 E-05 (1.20) 0.23 (2.45)* 0.01 0.06 (0.87) 0.38 9.28 E-06 (1.27) 0.21
Owner
R2
F
0.01 (2.80)** 0.007 0.06 (2.96)** 0.005 0.015 (1.91)* 0.06 2.18 (1.87)* 0.07 2.09 (1.64) 0.11 2.51 (1.84)* 0.07 0.02 (2.67)* 0.01
.99 .99 .99 .08 .04 .08 .99
165.34 0.0000 1494.66 0.0000 1494.3 0.0000 056 0.72 0.41 0.79 0.71 6.58 936.08 0.0000
0.01 (2.28)* 0.02
.99
1134.04 0.0000
0.02 (2.65) 0.01
.99 1152 6.0000
3.47 (1.55) 0.12
.12
.71 0.64
2.23 (1.68) 0.16
0.07
.50 (.77)
3.51 (1.51) 0.13
.10
0.69 0.62
Figures in parentheses are t-statistics. Figures below parentheses are p-value. ***Indicates the variable is significant at the .0001 level. **Indicates the variable is significant at the level of .009. *Indicates the variable is significant at the level of .05. The reported R2 is adjusted R2. The results are White heteroskedasticity adjusted.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
1 2 3 4 5 6 7
Category
Eq. Category
Dependent Intercept variable
(1) (2) (3) (4)
PA PC PA PC
Deposit Deposit Deposit Deposit
market market market market
0.01 0.01 0.01 0.01
(2.79) (2.64) (2.29) (2.74)
CR
MS
CRTA
0.005 0.19 (1.60) 0.12 0.028 (18.2) 0.0000 0.007 0.08 (0.41) 0.65 0.03 (15.3) 0.0000 0.003 0.09 (1.09) 0.22 0.03 (2.38) 0.43 0.03 (16.9) 0.0000 0.007 0.11 (1.10) 0.12 0.018 (2.41) 0.006 0.04 (11.3) 0.000
LDEP 0.005 0.008 0.005 0.001
ASSET (2.8) 0.008 (2.96) 0.007 (2.88) 0.007 (2.86) 0.007
0.000002 0.000002 0.000002 0.000005
Owner
R2
(2.19) 0.04 0.001 (2.70) 0.005 .72 (3.46) 0.004 .66 (4.6) 0.0001 .69 (3.96) 0.0008 .68
Figures in parentheses are t-statistics. Figures after parentheses are p-value. ***Indicates the variable is significant at the .0001 level. **Indicates the variable is significant at the level of .009. *Indicates the variable is significant at the level of .05.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
Table 5 Regression results for 2002
191
192
A. Samad / Journal of Asian Economics 19 (2008) 181–193
Eqs. (1) and (2) in Table 5 show the imposition of restriction, CR = 0, MS 6¼ 0 and vice versa. The result shows that the coefficients of both CR and MS are insignificant which suggests that neither SCH nor EH are valid for the banking market in Bangladesh. However, when no restrictions are imposed, Eqs. (3) and (4) in Table 5 show that the coefficients of MS are significant. This result supports the EH for explaining the profit performance in the Bangladesh banking industry. The finding rejects SCH. 5. Conclusions The two hypotheses – SCP and ESH – are examined in the context of the assets and deposit market in Bangladesh. The results were presented from the tests of these hypotheses using both pool and annual data. Although three or four of the largest banks in Bangladesh dominated and controlled 50–62% of the total assets and deposits, their ROA and ROE are not satisfactory. One of the three largest banks in Bangladesh, Rupali bank, was operating with losses. Therefore, the coefficients of CR are, in most equations, negatively related. However, they are not significant. This is supported by the combined result of 36 equations reported in Tables 2–4. Thus, the SCP hypothesis is rejected for explaining bank performance in Bangladesh. The examinations of equations from the sample data revealed that the coefficient of MS in Eqs. (7) and (9) in Tables 2–4 and Eqs. (3) and (4) in Table 5 are positive and statistically significant even after controlling for concentration. This supports the efficiency structure (ESP) hypothesis. The efficiency of the banks measured by MS, not the concentration (CR) was an important factor for bank performance in the Bangladesh banking industry. These findings lend evidence in support of ESH. The result was consistent with the findings of Gilbert (1984), Osborne and Wendel (1983), Smirlock (1985), and Evanoff and Fortier (1988). However, the result should be interpreted cautiously for policy implications. The pooled data which contains robust information does not lend support to either of the hypotheses. Bank specific factors are more consistent in explaining bank performance in Bangladesh. Almost all regression results show that the bank specific variables, such as CRTA, LDEP, ASSETS, and OWNER, are statistically significant factors concerning bank performance in Bangladesh. Since four public sector-owned and regulated banks could not explain bank performance, policy makers should encourage more liberalization, not government control and regulation, of banks in Bangladesh. Liberalization of banks is good for the banking industry of Bangladesh. Appendix A Data is collected from the several bulletins published by the Ministry of Finance, government of the Peoples’ Republic of Bangladesh. These bulletins are published in Bengali language. The name of the bulletin is ‘‘Bank O Arthick Pratisthansamuher karjabali’’ whose close translation is ‘‘Bank and Financial Institutions Activities’’. References Bain, J. S. (1951, August). Relation of profit rate to industry concentration: American manufacturing, 1936–1940. Quarterly Journal of Economics, 293–324. Brucker, E. (1970, December). A microeconomic approach to banking competition. Journal of Finance, 1141. Brozen, Y. (1982). Concentration, merger and public policy. New York: Macmillan. Demsetz, H. (1973). Industry structure, market rivalry, and public policy. Journal of Law and Economics, 16(April), 1–9. Edward, F. R. (1964, August). Concentration in banking and its effect on business loan rates. Review of Economic and Statistics, 294–300. Evanoff, D. D., & Fortier, D. I. (1988). Reevaluation of the structure–conduct performance paradigm in banking. Journal of Financial Services Research, I(June), 313–329. Fraser, D. R., & Rose, P. (1971). More on banking structure and performance; the evidence from Texas. Journal of Financial and Quantitative Analysis, 601–611. Fleching, T. G. (1965, May). The effect of concentration of bank loan rates. Journal of Finance, 298–311. Fraser, D. R., & Alvis, J. B. (1975, Fall). The structure performance relationship in banking: A dichotomous analysis. Review of Business and Economics Research, 37–57. Gilbert, R. A. (1984). Bank market structure and competition. Journal of Money, Credit and Banking, 16(November), 617–645.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
193
Gillini, T., Smirlock, M., & Marshall, W. (1984). Scale and scope economics in the multi-product banking firm. Journal of Monetary Economics, 13(May), 393–405. Guzrarti, D. (1995). Basic econometrics. New York: McGraw-Hill. Hannan, T. J. (1991, Summer). Bank profitability and the threat of entry. Journal of Bank Research, 157–163. Heggestad, A. A., & Mingo, J. J. (1977). The competitive condition of US banking markets and the impact of structural reform. Journal of Finance, 32(June), 649–661. Kaufman, G. (1966, April). Bank market structure and performance: The evidence from Iowa. The Southern Economic Journal, 429–439. Lloyad-Williams, D. M., Molyneux, P., & Thornton, J. (1994). Market structure and performance in Spanish banking. Journal of Banking and Finance, 18, 433–443. McGee, J. (1974). In H. Goldschmith, H. M. Mann, & J. F. Weston (Eds.), Efficiency and economies of size in industrial concentration: The new learning (pp. 55–97). Boston: Little, Brown, and Company. Osborne, D. K. & Wendel, J. (1983). Research in structure, conduct and performance in banking 1964–1979. Research paper 83-003, College of Business Administration, Oklahoma State University. Peltzman, S. (1977). The gains and losses from industrial concentration. Journal of Law and Economics, 20(October), 229–263 (Reserve Bank of San Francisco, (Winter) pp. 5–17). Phillips, A. (1967). Evidence of concentration in banking markets and interest rates. Federal Reserve Bulletin, 53, 916–926. Podenda, Randall J. (1986). Structure and performance: Some evidence from California banking, Economic Review, Federal Reserve Bank of Sanfrancisco (Winter), pp. 5–17. Rose, P. S., & Fraser, D. R. (1976). The relationship between stability and change in market structure: An analysis of bank prices. Journal of Industrial Economics, 24(June), 251–266. Rhoades, S. (1982). Welfare loss, redistribution effects, and restriction of output due to monopoly. Journal of Monetary Economics, 9, 375–387. Samad, A. (2005). Banking structure and performance: Evidence from Utah. Review of Business Research, V(2), 151–156. Smirlock, M. (1985). Evidence on the (non) relationship between concentration and profitability in banking. Journal of Money Credit and Banking, 17(February), 69–83. Spellman, L. J. (1981). Commercial banks and profits of savings and loan markets. Journal of Bank Research, 12(Spring), 32–36. Vernon, J. R. (1971, January). Separation of ownership and control and profits rates, the evidence from banking: Comment. Journal of Financial and Quantitative Analysis, 615–625. Ware, R. F. (1972, March). Banking structure and performance: Some evidence from Ohio. Federal Reserve Bank of Cleveland, 3–14. Weiss, L. W. (1974). The concentration–profits relationship and antitrust. In H. Goldschmith, H. M. Mann, & J. F. Weston (Eds.), Industrial concentration: The new learning (pp. 184–233). Boston: Little, Brown, and Company.
Journal of Asian Economics 19 (2008) 181–193
Market structure, conduct and performance: Evidence from the Bangladesh banking industry Abdus Samad * Department of Finance and Economics, Utah Valley State College, 800 W. University PKY, Orem, UT 84508, United States Received 26 August 2007; received in revised form 14 December 2007; accepted 14 December 2007
Abstract With regard to the market structure and performance in Bangladesh banking industry, there are two competing hypotheses—the traditional structure–conduct–performance (SCP) hypothesis and the efficiency hypothesis (EH). Using pooled and annual data for the period 1999–2002, this study tests the validity of these two hypotheses. In general, the results of this study support the EH hypothesis as an explanation for market performances in Bangladesh, but for definitive policy purposes, the impact of the banking structure needs to be explored further. Published by Elsevier Inc. JEL classification: G21 Keywords: Bank performance; Market structure; Bangladesh banking
1. Introduction In bank performance literature, two competing hypotheses have been the subject of controversy for more than 40 years. The oldest hypothesis is the traditional structure–conduct–performance (SCP), also known as structureperformance (SP), hypothesis. According to SP hypothesis (Gilbert, 1984; Hannan, 1991), the profitability of a banking firm is dependent upon the market structure and the level of competition. The lower the level of competition in the market, the higher the economic rent for a firm. The basic message of the SP hypothesis is that a higher concentration ratio leads to a higher profitability. The hypothesis that concentration leads to higher profitability has been challenged by an alternative hypothesis, known as the efficiency hypothesis (EH). The EH emphasizes superior efficiency as an explanation for a firm’s profitability. According to EH, ‘‘there is no relationship between concentration and profitability, but rather market share and bank profitability’’ (Smirlock, 1985, p. 69). In other words, the performance of an individual firm depends on the firm’s degree of efficiency. In this hypothesis, the explanation for the relationship between the market structure and performance of a firm is dependent upon the firm’s efficiency. If a firm enjoys a higher degree of efficiency than its competitors (that is, if the firm has a relatively low cost of production structure), the firm can maximize profits and increase its size and market share. This can be done by keeping the
* Tel.: +1 801 863 8368. E-mail address: [email protected]. 1049-0078/$ – see front matter. Published by Elsevier Inc. doi:10.1016/j.asieco.2007.12.007
182
A. Samad / Journal of Asian Economics 19 (2008) 181–193
present market price level. Thus, the increase of profits and market share is the result of efficiency, not of a higher concentration ratio (collusion). The proponents of both hypotheses claim support in their favor. There are three motivating factors that brought about this study. First, given the controversy between the two competing hypotheses, it is important to determine which hypothesis truly approximates and describes the Bangladesh banking industry. Determining the approximate behavioral relationship has important policy implications for Bangladesh because the presumed positive relationship between the market concentration and higher performance, for example, higher profits influences regulatory decisions regarding mergers and de novo entry. Increasing market concentration is also considered a social cost. Second, the relevance of the two hypotheses has been explored and tested in the context of the bank markets in developed countries. This study evaluates the validity of the two hypotheses in the context of the banking market in Bangladesh, a third-world country. Until 1980, the banking system of Bangladesh was quite concentrated and was regulated and protected from foreign competition. The domestic bank market was dominated by the operation of five nationalized commercial banks. This fact was conducive to collusive bank behavior leading to higher profitability. Thus, Bangladesh provides a good test case for evaluating the validity of the two competing hypotheses. Third, since the controversy of SCP and EH has not been resolved, the hypotheses deserve further exploration. This study can contribute to new findings in the banking performance literature. This paper is structured as follows: Section 1 provides a brief description of the institutional structure of the Bangladesh banking market. Section 2 describes SCP and EH in the context of market structure and performance relationship. Section 3 provides data and methodology. Section 4 reports results and conclusions. 2. Institutional structure of the Bangladesh banking market Bangladesh is a small country bordering India. The country seceded from Pakistan and gained independence in 1971. At that time, all commercial banks were nationalized by the government without compensation. The question of compensation did not arise at this time because the owners of the commercial banks were non-Bengalis who left Bangladesh before or during the war of liberation. The banking system of Bangladesh was highly regulated and protected from foreign competition and was concentrated until 1980. There were only five nationalized commercial banks that dominated the domestic banking market in Bangladesh at the time. The first move towards competition in the banking market of Bangladesh was undertaken during the 1980s with the denationalization of the Pubali Bank, one of the five nationalized banks. At this time, the opening of private and foreign banks was allowed and encouraged. At present, there are 41 privately owned commercial banks operating in Bangladesh. Of those banks, 12 are foreign. The structural features of the Bangladeshi banking system are provided in Table 1. An explanation of Table 1 follows: 1. Until 2001, banking activity of commercial banks was dominated by nationalized commercial banks. There were four banks owned by the government of Bangladesh. The number of the nationalized commercial banks was four. Thus, only four commercial banks controlled over 48% of the total assts of the entire banking system in Bangladesh during the year 2000. Privately owned commercial banks controlled a little over 50% of the total assets until 2000. Table 1 Bangladesh banking system total assets and distribution Institutions
Assets in millions TK
Commercial banks
1999
2000
2001
1999
Percentage share of assets 2000
2001
Nationalized commercial banks Domestic private banks Foreign banks All private banks Total banking system
446,350 376,197 88,437 464,634 910,984
506,119 450,481 97,261 547,742 1,053,861
589,366 1,405,466 115,349 1,520,815 2,110,181
48.99 80.96 19.04 51.01 100
48.02 82.24 17.76 51.98 100
27.92 92.41 7.59 72.03 100
Source: Publications from the Ministry of Finance of the government of Bangladesh. Percentage of assets is calculated by the author. The increase of asset growth was the result of many factors. One of the factors was the increase in exports supported by private commercial banks. Bangladesh’s exports increased from $249,860 (Million) in 2000 to $269,150 (million) in 2001. Source: International Financial Statistics, published by IMF. The increase of exports increased bank deposits and assets.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
183
However, there was a significant increase, from 51.98 to 72.03% in the market share by the privately owned commercial banks in 2001. 2. Within the privately owned commercial banks, domestic private commercial banks dominated the banking activity. Domestic private commercial banks controlled a significant percentage of all private bank assets (see Table 1). Privately owned domestic banks controlled 92.41% of all private bank (domestic and foreign) assets in 2001. 3. Although foreign banks have been active in retail banking, their activities were concentrated in the metropolitan area of Dhaka, the capital city of Bangladesh. Banking activity has declined due to severe competition from domestic private banks (see Table 1). The foreign bank market share in the total assets of all private banks declined from 19.04% in 1999 to 7.59% in 2001. 4. The banking market in Bangladesh is quite concentrated and dominated by four nationalized banks. With the growth of private banks, these nationalized banks face strong competition (see Table 1). The asset market share of nationalized commercial banks declined from 48.99% in 1999 to 27.91% in 2001. On the other hand, the asset market share of private commercial banks increased from 51% in 1999 to 72% in 2001. 3. Structure-performance relationship and background of controversy The notion that market structure influences a firm or industry performance originates in the classical model of the theory of firm. In a pure competitive model, there are a large number of firms. The larger the number of firms, the higher the competition and lower the concentration ratio. The lower the concentration ratio, the lower the profit rates for a firm. Each firm’s market share is insignificant and its product homogenous. Because of these factors, firms cannot influence market price. In such a market, concentration of market share is absent. Therefore, productive efficiency is achieved when price equals minimum of ATC and firms earn normal profits. In contrast, firms under a pure monopoly or monopolistic competition, have significant market shares and thus, have control over the price of the product. In such circumstances, price exceeds the marginal cost MC = MR and leads to sub-optimal production and supernormal profits. The notion is that the higher the concentration of market share by a firm, the higher the economic profits. Market concentration is associated with increasingly higher profits known as ‘‘Recardian rent.’’ The notion that there exists a monotonic relationship between market concentration and performance is a matter of empirical evidence. Bain (1951) first tested this hypothesis – market share concentration and performance – and found that increased concentration led to higher profit rates. This result originated in the structure–conduct and performance controversy. The controversy concerning the relationship between the market structure and performance has resulted in two testable hypotheses: the traditional hypothesis known as SCP and the EH also known as efficient structure hypothesis (ESH). 3.1. Structure–conduct–performance hypothesis The SCP hypothesis is based on the following proposition: when a few firms have a large percentage of market shares, this fosters collusion among firms in the industry. This can happen overtly or without public knowledge. Collusive behavior increases as market share is concentrated in the hands of a few firms. The higher the concentration ratio in a market, the higher the profitability performance of the firms. Thus, according to the SCP hypothesis, there is a positive correlation between the degree of market share concentration and the firm’s performance. Due to collusive or monopolistic reasons, ‘‘firms in a concentrated market will earn higher profits than firms operating in a less concentrated one, irrespective of efficiency’’ (Lloyad-Williams, Molyneux, & Thornton, 1994, p. 437). This hypothesis could be supported if the impact of market concentration was found to be significantly positive, irrespective of the efficiency of the firm. There are a number of empirical studies in the banking market that provide support in favor of SCP hypothesis. The prominent early studies that supported the SCP hypothesis were studies by Rose and Fraser (1976), Heggestad and Mingo (1977), Spellman (1981), and Rhoades (1982). Podenda (1986), Lloyad-Williams et al. (1994), and Samad (2005). Podenda studied the California banking market. The results of Podenda’s study are consistent with the idea that ‘‘increased concentration is associated with increasingly high profits’’ (p. 5). Lloyad-Williams et al. (1994) tested the two hypotheses in the context of the Spanish banking industry by using pooled data and found support in favor of the SCP hypothesis. In a major survey done by Gilbert (1984), 44 studies were summarized for the impact of bank
184
A. Samad / Journal of Asian Economics 19 (2008) 181–193
performance to a market concentration. Thirty-two of the 44 studies Gilbert reviewed found that market concentration significantly and positively affected bank performance. Brucker (1970), Edward (1964), Phillips (1967), and Vernon (1971) found evidence in support of the market structure hypothesis (SCH). They found that the coefficient of market structure, CR3, is statistically significant. 3.2. Efficiency structure hypothesis The SCP hypothesis – the positive association between market concentration and performance – has been challenged by the ESH. The basic foundation of the ESH is that market concentration is not the cause of a firm’s superior performance. Instead, ESH finds that the positive direction of concentration and higher performance is the result of a firm’s superior efficiency. It is argued that the higher profits enjoyed by large firms in a concentrated market are the result of economies of scale and the consequences of superior efficiency in larger firms. If a firm enjoys a higher degree of efficiency (in terms of cost and technology) than its competitors, the firm can easily capture a larger market share by lowering its price and earning economic profits. Thus, the driving force behind the process of gaining a large market share, and thus concentration, is the efficiency of the firm. The most efficient firms will gain market share and earn economic profits. In summarizing the basic points of the proponents of EH, Smirlock (1985) wrote: ‘‘concentration is not a random event but rather the result of superior efficiency of leading firms’’ (p. 70). Firms possessing a comparative advantage in production become large and obtain a high market share, and as a consequence, the market becomes more concentrated. Such firms earn Ricardian rent. Therefore, the basic message of ESH is that leading firms’ efficiency leads to increased market share regarding concentration and is positively correlated with higher performance. The main proponents of the EH are Demsetz (1973), McGee (1974), Peltzman (1977), and Brozen (1982). In these studies the authors challenged the SCP hypothesis. According to the studies, market concentration was not a random event; rather it was the result of a firm’s efficiency. Using manufacturing data, the authors tested the two competing hypothesis and found evidence in support of the ESH. Using banking data, Gillini, Smirlock, and Marshall (1984), Smirlock (1985), and Evanoff and Fortier (1988) tested the competing hypotheses, SCP and EH, and found that firm-specific efficiency was a factor for explaining the profitability in the United States banking industry. Fleching (1965), Ware (1972) and Fraser and Alvis (1975) found that market structure is not significant and did not support SCH. Given controversy between the two competing hypotheses, determining the approximate behavioral relationship is important for bank performance literature. The present study is motivated in this direction. The study has important policy implication for Bangladesh because of the presumed positive relationship between market concentration and higher performance. For example, higher profits influence regulatory decisions regarding the issue of mergers and de novo entry. 3.3. Data and methodology This study includes cross-sectional data for all 44 commercial banks operating in Bangladesh. All relevant data for the period 1999–2001 was obtained from the consolidated balance-sheets and income-statements published by the Ministry of Finance, Peoples Republic of Bangladesh1. The methodology used in this paper is based on the approach utilized by Smirlock (1985) and Lloyad-Williams et al. (1994). These studies asserted that the correct approach in testing the competing hypotheses was to ‘‘take both market share and concentration into account at the same time’’ (Weiss, 1974, pp. 225–226). Accordingly, this study uses the following equation:
P ¼ a0 þ a1 CR þ a2 MS þ
1
See Appendix A for details.
X
aiZi
(1)
A. Samad / Journal of Asian Economics 19 (2008) 181–193
185
where P measures profit rate as firm’s performance, CR denotes concentration ratio representing the measure of market concentration. MS is a measure of market share of the firm and Z is a vector of control variables which are included to take account for firm-specific and market-specific characteristics. Eq. (1) provides the straightforward discrimination between the two hypotheses. The SCP hypothesis can be supported by the coefficient of CR, a1 > 0 and the coefficient of MS, a2 = 0. A coefficient combination of a1 > 0 and a2 = 0 implies that the market share does not affect a firm’s profitability and that profitability is the result of a monopoly behavior measured by concentration. On the other hand, ESH is found to be true if a1 = 0 and a2 > 0. A coefficient combination of a1 = 0 and a2 > 0 implies that firms with a large market share are more efficient than their rivals and thus earn higher profits. In addition market concentration does not affect bank profitability. Thus, a1 > 0 and a2 = 0 supports SCP and a1 = 0 and a2 > 0 supports ESH. The following is a complete OLS regression equation which is estimated in testing the relevance of SCP and ESH for the Bangladesh banking industry: P ¼ a0 þ a1 CR3 þ a2 MS þ a3 CRTA þ a4 LDEP þ a5 ASSET þ a6 OWNER
(2)
where P is a dependent variable, and measures bank performance. Since a bank is a multi-product service industry, prices of certain individual products or services are not a good measure for bank performance2. In a multi-product service industry such as banking, cross subsidization among the products and services is more common than in other single product industries. Similar to LMT (1994), this paper uses profit for measuring bank performance. Profitability is a consolidated figure and takes into account all products and services regarding profits and losses. This overcomes the problem of cross-subsidization. This paper uses two measures of profits: total net profits P A ¼ ROA: ROA ¼ total assets total net profits P C ¼ ROE: ROE ¼ total equity capital There are six independent variables in this model. They are as follows: 1. CR3 = Three bank market concentration ratios. CR3 is used to measure market structure. This study uses two P measures of CR3—one CR3 for assets and the other CR3 for deposits. Where CR3 = assets (deposits) of the top three banks/total assets (deposits) in the market. The sign for a1, coefficient of CR is expected to be positive and significant for the SCP hypothesis. 2. MSi = Market share of ith bank in assets or deposits. In firm-specific market share, MS is used to capture firm efficiency. The sign for a2, coefficient of MS is expected to be positive and significant for ESH. 3. CRTA = Capital and reserve to total assets, where CRTA = total capital and reserve/total assets. 4. LDEPi = Loans to deposit ratio of banki. LDEP = loans of banki/total deposits of banki. Since bank profits, ROA and ROE are not independent of risk, this paper uses two variables – CRTA and LDEP – for measuring bank specific risk. The sign for the CRTA coefficient, a3, is expected to be negative. The higher the capital and reserve as a percentage of total assets, the lower the risk for a bank and, therefore, the lower the expected rate of profits. Similarly, the higher the amount of loans as a percentage of deposits, the higher the risk for a bank. To compensate for the higher risk, the bank is expected to earn a higher rate of return. Therefore, the sign for the LDEP coefficient, a3, is expected to be positive. 5. ASSETSi = ith bank assets. Assets measure the bank size. The asset of each bank is included to take account of the differences in performance brought about by the bank size. The coefficient for ASSETS may show a positive or negative sign depending upon the economies of scale. 6. OWNERi = Ownership of banki. The ownership of ith bank is represented by the binary value = 1 for publicly (government) owned banks 0 for private owned banks. Since the policy of publicly owned banks in Bangladesh is guided by one of the government’s major objectives – removing poverty and supporting the poor – loans are issued to the poor. Such loans to certain classes are expected to be non-performance. Secondly, loans to high-level 2
Although individual product or service price is not a good measure (because taking a single price for a multi-product firm may underestimate or overestimate the performance of a firm), many SCP studies have used it. See Kaufman (1966), Fraser and Rose (1971).
186
A. Samad / Journal of Asian Economics 19 (2008) 181–193
government officials for public sector banks are expected to be in default and become non-performance. Therefore, the sign of the coefficient for OWNER is expected to be negative. 3.4. Heteroscedasticity Since cross-sectional data deals with members of a population at a given point of time, such members may be of different sizes such as small, medium, and large. ‘‘Heteroscedasticity may be a rule rather than the exception is generally expected if small, medium, and large-size firms are sampled together’’ (Guzrarti, 1995, p. 368). As this study involves various categories of banks such as small, medium, and large heteroscedasticity is likely to be common. In order to overcome the heteroscedasticity problem, this study utilizes White’s heteroscedasticity correction method. EVIEW package provides White’s heteroscedasticity corrected outputs. All results of Eq. (2) reported in the next section are based on White’s heteroscedasticity correction. 4. Empirical results Three types of regressions were computed with cross-sectional data for 42 banks in Bangladesh. First, liner regressions were estimated with pooled data for 1999–2002. Regressions 1–12 in Table 2 report the results of the pooled data for 1999–2002. 4.1. Findings from pooled data 1. With the imposition of restrictions CR = 0 and MS 6¼ 0 and vice versa, Eqs. (2), (3), (5), (6), (11), and (12) show that neither the coefficient of CR nor MS is significant whether they are considered for assets and deposit markets in Bangladesh. The results suggest that neither of the two hypotheses – SCP and ESH – can explain the profit performance of Bangladesh banks. However, Eq. (9) provides an exception. The coefficient of MS in a deposit market, in Eq. (9), is statistically significant. The significance of MS in one out of 7 restricted equations lends support in favor of ESH. 2. According to Smirlock (1985), the test of validity for the alternative hypothesis is to estimate (2) without any parameter restriction. When no restriction is imposed, these results are presented in Eqs. (1), (4), (7), and (10). These results show that two out of four, Eqs. (1) and (4), do not support either of the two hypotheses. The other two equations in the deposit market, Eqs. (7) and (10), provide support in favor of ESH because the coefficients of MS in both Eqs. (7) and (10) in Table 2 are statistically significant with the 5% level of significance. On the other hand, the coefficients of CR3 for both these equations are not significant. The result support the validity of the EH and rejects the SCH for Bangladesh. 3. The signs for the MS coefficients are positive and consistent with the exception of Eq. (3). However, the coefficient MS is not significant. Contrary to the model’s expectation, the signs for CR are negative and not statistically significant. The plausible reason for the negative sign is that the three largest banks which accounted for 52% of the market assets were nationalized banks owned by the government of Bangladesh and their profit performance was unsatisfactory. The profits of the Rupali bank, one of the three largest banks, earned TK 308 million losses in 1999. As a result, the ROA is negatively related to CR. 4. In almost all of the regressions 1–12, the signs for the coefficient of CRTA, LDEP, Assets and OWNER are consistent with the expectation of model and are statistically significant. This suggests that bank performances are significantly dependent upon CRTA, LDEP, Assets and OWNER. Since the public sector’s government owned banks were running losses, the coefficient of OWNER is negative and statistically significant. 5. R2 of all the regressions for ROA are in the range between .16 and .83. The F-statistics for these regressions are significant in most cases. 4.2. Result from the non-pooled data Regressions are computed with yearly data. Regressions 1–12 in Table 3 report the result for 1999. Regressions 1– 12 in Table 4 are estimated with the sample for 2000. Regression result for the year 2002 is provided in Table 5.
Eq.
Category
Dependent Intercept variable
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Pooled Pooled Pooled Pooled Pooled Pooled Pooled Pooled Pooled Pooled Pooled Pooled
PA PA PA PC PC PC PA PA PA PC PC PC
asset asset asset asset asset asset deposit deposit deposit deposit deposit deposit
0.017 (4.8)*** 0.015 (4.7)*** 0.019 (4.5)*** 0.58 (2.89)** 0.78 (3.13)** 0.41 (3.7)** 0.04 (4.6)*** 0.04 (4.8)*** 0.05 (4.8)*** 0.06 (3.2)** 0.67 (3.2)** 0.6 (3.8)**
CR 0.04 (0.89) 0.03 (0.75)
MS asset
MS deposit
CRTA
0.04 (21.32)*** 0.04 (28.9)*** 0.08 (0.67) 0.05 (22.5)*** 3.1 (98) 17.05 (1.0) 0.28 (2.8)* 2.4 (1.21) 0.31 (2.68)* 7.8 (1.2) 0.25 (3.9)* 0.08 (0.45) 0.25 (2.78)** 0.34 (23.75)*** 0.03 (1.21) 0.03) (24.1)*** 0.22 (4.0)*** 0.04 (24.6)*** 2.3 (0.88) 0.03 (3.6)** 0.42 (2.1)* 0.7 (0.6) 0.21 (2.6)* 0.87 (0.98) 0.3 (2.4)* 0.09 (0.67)
LDEP
Asset
Owner
R2 F
0.005 (3.08)** 0.0003 (4.0)** 0.0004 (3.8)** 0.02 (65) 0.03 (0.87) 0.01 (0.68) 0.006 (3.23)** 0.004 (3.1)** 0.004 (3.4)** 0.3 (2.9)* 0.03 (0.6) 0.04 (0.65)
2.21E-04 (3.78)** 2.1E-05 (0.89) 2.121E-05 (0.89) 0.004 (2.0)* 4.53E-05 (0.67) 8.1E-05 (2.78)* 1.2E-06 (0.9) 2.3E-06 (0.88) 2.4E-05 (0.7.9) 6.4E-07 (0.41) 5.3E-08 (2.2)* 3.2E-07 (0.84)
0.016 (2.26)** 0.02 (3.2)** 0.021 (2.01)* 1.86 (2.4)* 1.75 (2.56)* 1.8 (2.5)* 0.03 (4.7)*** 0.21 (4.2)*** 0.2 (3.3)** 2.1 (2.4)* 2.1 (2.6)* 2.1 (2.4)*
.67 12.2 .65 13.6 .59 15.4 .31 2.9 .16 078 .24 1.9 .65 10.2 .62 12.6 .63 12.7 .19 2.4 .16 1.9 .18 1.5
***Indicates the variable is significant at the .0001 level. **Indicates the variable is significant at the level of .009. *Indicates the variable is significant at the level of .05. The reported R2 is adjusted R2. The results are White heteroskedasticity adjusted. Figures in parentheses are t-statistics.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
Table 2 Regression results for pooled data
187
188
Table 3 Regression results #
Category
Dependent Intercept variable
CR
PA
0.019 (4.09)*** 0.0002 0.02 (0.96) 0.34
2 1999 asset
PA
0.018 (4.57)*** 0.0001 0.02 (0.70) 0.48
3 1999 asset
PA
0.018 (4.18)*** 0.0002
4 1999 asset 5 1999 asset 6 1999 asset
PC PC PC
0.60 (2.54)* 0.01 0.82 (3.27)** 0.002 0.72 (3.27)** 0.002
7 1999 deposit
PA
0.018 (4.05)*** 0.0003 0.006 (0.35) 0.72
M5 deposits
0.08 (0.72) 0.47
0.05 (0.58) 0.56 2.56 (1.00) 0.32 2.19 (1.15) 0.24
19.05 (1.00) 0.32 8.77 (1.41) 0.16
8 1999 deposits PA
0.018 (4.12)*** 0.0002
9 1999 deposit
PA
0.01 (4.12)*** 0.0002
10 1999 deposit
PC
0.07 (3.06)** 0.004
2.64 (0.99) 0.32
11 1999 deposit
PC
0.72 (3.19)** 0.003
0.06 (0.04) 0.96
12 1999 deposit
PC
0.72 (3.17)* 0.003
0.02 (1.58) 0.12
CRTA
LDEP
0.03 (22.31)*** 0.0000 0.03 (23.92)*** 0.0000 0.03 (22.78)*** 0.0000 0.21 (2.19)* 0.03 0.28 (2.45)* 0.02 0.25 (2.46)* 0.01
0.004 (3.04)** 0.004 0.0004 (3.16)** 0.003 0.0004 (3.10)** 0.003 0.01 (0.32) 0.72 0.02 (0.36) 0.72 0.09 (0.41) 0.68
0.03 (22.73)*** 0.0000
0.004 (3.06)** 0.004
Asset
2.44E-07 (3.85)*** 0.0005 0.03 (22.51)*** 0.004 (3.03)** 1.02E-07 (096) 0.37 0.20 (3.73)*** 0.03 (23.26)*** 0.0004 (3.09)** 2.41 E-07 0.0007 0.0000 0.003 (3.67)*** 0.0008 16.82 (1.19) 0.24 0.25 (2.34)* 0.02 9.75E-07 (0.23) 0.81 0.25 (2.41)* 0.02 0.02 (0.40) 0.68 9.74E-06 (1.17) 0.24 10.69 (1.15) 0.26 (2.39)* 0.02 0.02 (0.43) 0.66 2.22 E-07 0.25 (0.04) 0.96 0.22 (2.48)** 0.01
Owner
R2 F
0.019 (2.78)** .61 11.38 0.0000 0.008 0.019 (2.61)* 0.01 .60 13.53 0.0000 0.015 (1.64)
.61 14.57 0.0000
1.82 (2.05)* 0.05 1.58 (2.43)* 0.02 1.889 (2.36)* 0.02 0.02 (4.44)*** 0.0001
.16 .09 .14
1.03 0.41 0.73 0.57 1.30 0.29
.62
9.59 0.0000
0.02 (4.19)*** 0.0002 0.02 (3.33)** 0.002
.61 11.47 0.0000 .62 11.83 0.0000
2.06 (2.42)* 0.02 1.87 (2.40)* 0.02
.18
1.05 0.41
.14
1.00 0.43
2.40 (2.04)* 0.04
.17
1.19 0.33
Figures in parentheses are t-statistics. Figures after parentheses are p-value. ***Indicates the variable is significant at the .0001 level. **Indicates the variable is significant at the level of .009. *Indicates the variable is significant at the level of .05. The reported R2 is adjusted R2. The results are White heteroskedasticity adjusted.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
1 1999 asset
M6 assets
A. Samad / Journal of Asian Economics 19 (2008) 181–193
189
The regression results for the 1999 data are presented in Table 3. 4.3. Findings for the 1999 sample in Table 3 1. With restriction CR = 0, regression equation (9) in Table 3 shows that the coefficient of MS in deposit markets was positive and significant. This clearly supports ESH. Similarly, the results of the regression in Eq. (7), where ROA is regressed on both CR and MS of deposit markets, simultaneously show that the coefficient of CR is negative and insignificant. On the other hand, the coefficient of MS (in regression 7) is positive and significant. Thus, the result of both regressions, regression equations (7) and (9), support evidence in favor of ESH. This suggests that the bank MS is a significant factor for explaining bank performance. 2. The coefficients of CR in all regressions in Table 3 – whether restriction MS = 0 or MS and CR 6¼ 0 – are not significant. This suggests that SCP is not a valid approximation for explaining bank performance in the Bangladesh banking market. 3. With regard to other independent variables such as CRTA, LDEP, ASSET, and OWNER, it is found that they were significant factors for explaining bank performance in the Bangladesh banking market. This is evident from the tstatistic and p-value of the coefficient of CRTA, LDEP, ASSET, and OWNER. The coefficients of CRTA and LDEP in all regressions 1–12 in Table 3 are consistent with the expectation of the model and are significant. The higher the amount of capital and reserve as a percentage of total assets, the lower the risk for the bank, and therefore, the lower the ROA. ROA and CRTA are, thus, negatively related. Table 3 showed the coefficient of LDEP is positive and significant in most equations. The higher the amount of loans as a percentage of deposits, the greater the risk for a bank. In order to compensate for the higher risk, banks earn higher ROA. Thus, ROA and LDEP are negatively related. The coefficient of ASSETS in Eqs. (7) and (9) in Table 3 are negative and significant. The higher the amount of assets for a bank the greater advantage for the diversification of product and loans. The increased diversification implies less risk and less profits. Since the three largest concentrated banks in Bangladesh are owned by the government and run losses, the coefficient of OWNER is consistently negative. 4. R2 of all the regressions for ROA are between .82 and .83 and the F-statistics of these regressions are significant. This suggests that the model explains well and the overall explanatory power of the model is satisfactory. The regression results for the 2000 data are presented in Table 4. 4.4. Findings for 2000 sample in Table 4 1. Similar to findings reported in Table 3, it was found that the coefficient of MS in the deposit market of 2000 (Eq. (9), Table 4) is significant whether the restrictions CR = 0 and CR and MS 6¼ 0 are imposed. The coefficient of MS is positive (0.20) and significant. The significance of the MS coefficient supports ESH for the banking market in Bangladesh, without imposing any restrictions like. When ROA is regressed simultaneously on both MS and CR along with other independent variables, the result (Eq. (7), Table 4) shows that the coefficient of MS is significant while the coefficient of CR is not significant. The result of Eq. (7) reinforces the evidence supporting in favor of the ESH that firm efficiency, not the firm’s concentration, is a significant factor for explaining bank performance. 2. The coefficients of CR in all equations in Table 4 with the imposition of restrictions, CR = 0 and CR and MS 6¼ 0, are not statistically significant. This suggests that the concentration of the bank CR is not a significant factor for explaining bank performance. This explanation rejects the validity of the SCP hypothesis in Bangladesh. 3. The sign for CRTA, LDEP, ASSETS, and OWNER in most equations are consistent with the expectation of the model. The coefficient of CRTA for Eqs. (1), (2), (3), (7), (8), and (9) are positive and significant. The sign is not inconsistent. Banks with a higher amount of capital and reserve, as a percentage of total assets, might expect to earn a higher ROA if they are engaged in risky investment projects. The sign for OWNER is consistently negative and significant in equations where the ROA is considered as a dependent variable. 4. R2 of all the regressions for ROA is in the range between .60 and .62 and the F-statistics of these regressions are significant. This suggests that the model explains the overall explanatory power satisfactorily. The regression results for extending one more year for example 2002 do not change. The regression results for the 2002 data are presented in Table 5.
190
Table 4 Regression results #
Dependent Intercept variable
2000 2000 2000 2000 2000 2000 2000
PA PA PA PC PC PC PA
assets asset asset asset assets asset deposit
CR
M6 assets
0.013 (2.22)* 0.010 (0.65) 0.51 0.017 (0.16) 0.87 0.013 (2.52) 0.01 0.016 (0.93) 0.013 (2.27)* 0.02 0.03 (0.40) 0.68 1.03 (2.79)** 0.008 1.98 (0.65) 0.51 17.60 (1.24) 0.22 1.20 (2.76)** 0.009 3.48 (1.30) 0.20 1.04 (2.83)** 0.007 14.65 (1.38) 0.17 0.012 (2.02)* 0.04 0.018 (0.08) 0.49
8 2000 deposit PA
0.013 (2.13)* 0.03
9 2000 deposit PA
0.013 (2.17)* 0.03
M5 deposits
CRTA
0.08 0.08 0.08 0.04 0.05 0.05 0.35 (1.96)* 0.05 0.08
0.004 (0.20) 0.84
0.08 0.30 (2.04)* 0.04
0.08
10 2000 deposit PC
0.93 (2.60) 0.01
5.31 (0.98) 0.33
50.84 (1.45) 0.14
0.05
11 2000 deposit PC
0.99 (2.68)* 0.01
1.98 (0.69) 0.49
35.66 (1.39) 0.17
0.05
12 2000 deposit PC
1.01 (2.20) 0.01
35.66 (1.39) 0.17
0.05
LDEP
Asset
(345.95)*** 0.0000 0.002 (1.46) 0.15 (360.64)*** 0.0000 0.002 (1.50) 0.14 (352)*** 0.0000 0.002 (1.49) 0.14 (2.62)* 0.01 0.08 (1.68) 0.28 (2.61)* 0.01 0.08 (1.08) 0.28 (2.67)* 0.01 0.08 (1.10) 0.27 (331.01)*** 000.00 0.002 (1.47) 0.14 1.63 E-07 (1.50) 0.14 (1.44) 0.15 0.002 (1.44) 0.15 2.03 E-08 (0.17) 0.86 (336.7)*** 0.0000 0.002 (1.48) 0.14 1.62 E-07 (1.50) 0.14 (2.49)* 0.01 0.06 (0.85) 0.40 963 E-06 (1.24) 0.22 (2.445)* 0.01 0.06 (0.07) 0.38 1.68 E-05 (1.20) 0.23 (2.45)* 0.01 0.06 (0.87) 0.38 9.28 E-06 (1.27) 0.21
Owner
R2
F
0.01 (2.80)** 0.007 0.06 (2.96)** 0.005 0.015 (1.91)* 0.06 2.18 (1.87)* 0.07 2.09 (1.64) 0.11 2.51 (1.84)* 0.07 0.02 (2.67)* 0.01
.99 .99 .99 .08 .04 .08 .99
165.34 0.0000 1494.66 0.0000 1494.3 0.0000 056 0.72 0.41 0.79 0.71 6.58 936.08 0.0000
0.01 (2.28)* 0.02
.99
1134.04 0.0000
0.02 (2.65) 0.01
.99 1152 6.0000
3.47 (1.55) 0.12
.12
.71 0.64
2.23 (1.68) 0.16
0.07
.50 (.77)
3.51 (1.51) 0.13
.10
0.69 0.62
Figures in parentheses are t-statistics. Figures below parentheses are p-value. ***Indicates the variable is significant at the .0001 level. **Indicates the variable is significant at the level of .009. *Indicates the variable is significant at the level of .05. The reported R2 is adjusted R2. The results are White heteroskedasticity adjusted.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
1 2 3 4 5 6 7
Category
Eq. Category
Dependent Intercept variable
(1) (2) (3) (4)
PA PC PA PC
Deposit Deposit Deposit Deposit
market market market market
0.01 0.01 0.01 0.01
(2.79) (2.64) (2.29) (2.74)
CR
MS
CRTA
0.005 0.19 (1.60) 0.12 0.028 (18.2) 0.0000 0.007 0.08 (0.41) 0.65 0.03 (15.3) 0.0000 0.003 0.09 (1.09) 0.22 0.03 (2.38) 0.43 0.03 (16.9) 0.0000 0.007 0.11 (1.10) 0.12 0.018 (2.41) 0.006 0.04 (11.3) 0.000
LDEP 0.005 0.008 0.005 0.001
ASSET (2.8) 0.008 (2.96) 0.007 (2.88) 0.007 (2.86) 0.007
0.000002 0.000002 0.000002 0.000005
Owner
R2
(2.19) 0.04 0.001 (2.70) 0.005 .72 (3.46) 0.004 .66 (4.6) 0.0001 .69 (3.96) 0.0008 .68
Figures in parentheses are t-statistics. Figures after parentheses are p-value. ***Indicates the variable is significant at the .0001 level. **Indicates the variable is significant at the level of .009. *Indicates the variable is significant at the level of .05.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
Table 5 Regression results for 2002
191
192
A. Samad / Journal of Asian Economics 19 (2008) 181–193
Eqs. (1) and (2) in Table 5 show the imposition of restriction, CR = 0, MS 6¼ 0 and vice versa. The result shows that the coefficients of both CR and MS are insignificant which suggests that neither SCH nor EH are valid for the banking market in Bangladesh. However, when no restrictions are imposed, Eqs. (3) and (4) in Table 5 show that the coefficients of MS are significant. This result supports the EH for explaining the profit performance in the Bangladesh banking industry. The finding rejects SCH. 5. Conclusions The two hypotheses – SCP and ESH – are examined in the context of the assets and deposit market in Bangladesh. The results were presented from the tests of these hypotheses using both pool and annual data. Although three or four of the largest banks in Bangladesh dominated and controlled 50–62% of the total assets and deposits, their ROA and ROE are not satisfactory. One of the three largest banks in Bangladesh, Rupali bank, was operating with losses. Therefore, the coefficients of CR are, in most equations, negatively related. However, they are not significant. This is supported by the combined result of 36 equations reported in Tables 2–4. Thus, the SCP hypothesis is rejected for explaining bank performance in Bangladesh. The examinations of equations from the sample data revealed that the coefficient of MS in Eqs. (7) and (9) in Tables 2–4 and Eqs. (3) and (4) in Table 5 are positive and statistically significant even after controlling for concentration. This supports the efficiency structure (ESP) hypothesis. The efficiency of the banks measured by MS, not the concentration (CR) was an important factor for bank performance in the Bangladesh banking industry. These findings lend evidence in support of ESH. The result was consistent with the findings of Gilbert (1984), Osborne and Wendel (1983), Smirlock (1985), and Evanoff and Fortier (1988). However, the result should be interpreted cautiously for policy implications. The pooled data which contains robust information does not lend support to either of the hypotheses. Bank specific factors are more consistent in explaining bank performance in Bangladesh. Almost all regression results show that the bank specific variables, such as CRTA, LDEP, ASSETS, and OWNER, are statistically significant factors concerning bank performance in Bangladesh. Since four public sector-owned and regulated banks could not explain bank performance, policy makers should encourage more liberalization, not government control and regulation, of banks in Bangladesh. Liberalization of banks is good for the banking industry of Bangladesh. Appendix A Data is collected from the several bulletins published by the Ministry of Finance, government of the Peoples’ Republic of Bangladesh. These bulletins are published in Bengali language. The name of the bulletin is ‘‘Bank O Arthick Pratisthansamuher karjabali’’ whose close translation is ‘‘Bank and Financial Institutions Activities’’. References Bain, J. S. (1951, August). Relation of profit rate to industry concentration: American manufacturing, 1936–1940. Quarterly Journal of Economics, 293–324. Brucker, E. (1970, December). A microeconomic approach to banking competition. Journal of Finance, 1141. Brozen, Y. (1982). Concentration, merger and public policy. New York: Macmillan. Demsetz, H. (1973). Industry structure, market rivalry, and public policy. Journal of Law and Economics, 16(April), 1–9. Edward, F. R. (1964, August). Concentration in banking and its effect on business loan rates. Review of Economic and Statistics, 294–300. Evanoff, D. D., & Fortier, D. I. (1988). Reevaluation of the structure–conduct performance paradigm in banking. Journal of Financial Services Research, I(June), 313–329. Fraser, D. R., & Rose, P. (1971). More on banking structure and performance; the evidence from Texas. Journal of Financial and Quantitative Analysis, 601–611. Fleching, T. G. (1965, May). The effect of concentration of bank loan rates. Journal of Finance, 298–311. Fraser, D. R., & Alvis, J. B. (1975, Fall). The structure performance relationship in banking: A dichotomous analysis. Review of Business and Economics Research, 37–57. Gilbert, R. A. (1984). Bank market structure and competition. Journal of Money, Credit and Banking, 16(November), 617–645.
A. Samad / Journal of Asian Economics 19 (2008) 181–193
193
Gillini, T., Smirlock, M., & Marshall, W. (1984). Scale and scope economics in the multi-product banking firm. Journal of Monetary Economics, 13(May), 393–405. Guzrarti, D. (1995). Basic econometrics. New York: McGraw-Hill. Hannan, T. J. (1991, Summer). Bank profitability and the threat of entry. Journal of Bank Research, 157–163. Heggestad, A. A., & Mingo, J. J. (1977). The competitive condition of US banking markets and the impact of structural reform. Journal of Finance, 32(June), 649–661. Kaufman, G. (1966, April). Bank market structure and performance: The evidence from Iowa. The Southern Economic Journal, 429–439. Lloyad-Williams, D. M., Molyneux, P., & Thornton, J. (1994). Market structure and performance in Spanish banking. Journal of Banking and Finance, 18, 433–443. McGee, J. (1974). In H. Goldschmith, H. M. Mann, & J. F. Weston (Eds.), Efficiency and economies of size in industrial concentration: The new learning (pp. 55–97). Boston: Little, Brown, and Company. Osborne, D. K. & Wendel, J. (1983). Research in structure, conduct and performance in banking 1964–1979. Research paper 83-003, College of Business Administration, Oklahoma State University. Peltzman, S. (1977). The gains and losses from industrial concentration. Journal of Law and Economics, 20(October), 229–263 (Reserve Bank of San Francisco, (Winter) pp. 5–17). Phillips, A. (1967). Evidence of concentration in banking markets and interest rates. Federal Reserve Bulletin, 53, 916–926. Podenda, Randall J. (1986). Structure and performance: Some evidence from California banking, Economic Review, Federal Reserve Bank of Sanfrancisco (Winter), pp. 5–17. Rose, P. S., & Fraser, D. R. (1976). The relationship between stability and change in market structure: An analysis of bank prices. Journal of Industrial Economics, 24(June), 251–266. Rhoades, S. (1982). Welfare loss, redistribution effects, and restriction of output due to monopoly. Journal of Monetary Economics, 9, 375–387. Samad, A. (2005). Banking structure and performance: Evidence from Utah. Review of Business Research, V(2), 151–156. Smirlock, M. (1985). Evidence on the (non) relationship between concentration and profitability in banking. Journal of Money Credit and Banking, 17(February), 69–83. Spellman, L. J. (1981). Commercial banks and profits of savings and loan markets. Journal of Bank Research, 12(Spring), 32–36. Vernon, J. R. (1971, January). Separation of ownership and control and profits rates, the evidence from banking: Comment. Journal of Financial and Quantitative Analysis, 615–625. Ware, R. F. (1972, March). Banking structure and performance: Some evidence from Ohio. Federal Reserve Bank of Cleveland, 3–14. Weiss, L. W. (1974). The concentration–profits relationship and antitrust. In H. Goldschmith, H. M. Mann, & J. F. Weston (Eds.), Industrial concentration: The new learning (pp. 184–233). Boston: Little, Brown, and Company.