The current issue and full text archive of this journal is available at www.emeraldinsight.com/1753-8394.

htm

Bank margin determination: a comparison between Islamic and conventional banks in Indonesia
Erwin G. Hutapea
Central Bank of Indonesia, Jakarta, Indonesia, and

Bank margin determination

65

Rahmatina A. Kasri
Islamic Economics and Business Center (PEBS)/Department of Economics, Faculty of Economics, University of Indonesia, Depok, Indonesia
Abstract
Purpose The purpose of this paper is to examine the relationship between Islamic bank margin (BM) and its determinants. It also compares the BM behavior of Islamic and conventional banks in the Indonesian dual banking system. Design/methodology/approach The paper employs a time series approach under the dealership framework of Ho and Saunders. The autoregressive distributed lag model is used to inspect cointegration between BM and its determinants for the period of January 1996 to February 2006 of ve sample banks (two Islamic banks and three conventional banks). Findings The result conrms that there exists a long-running relationship between the Islamic BM and its determinants. In particular, as interest rate volatility increases, Islamic BM responds negatively while that of conventional banks responds positively. The ndings differ from most of the other studies as they found a positive relationship between BM and interest rate volatility. This paper also shows that the margin behavior changes as the basis of bank operations changes from conventional to Islamic principles. Research limitations/implications The paper uses a relatively small sample of three (out of 150) conventional banks as a comparison to two sample Islamic banks. However, as they come from the same peer with the Islamic banks, it is believed that the nding is valid. Islamic banks in Indonesia are not remote from the interest rate volatility in their presence under a dual banking system. It is the displaced commercial risk that threatens Islamic banking protability in a changing market interest rate situation. Practical implications Under a dual banking system, the stability of interest rates and the nancial system is of great importance for the policy maker in developing the Islamic banking industry in Indonesia. As long as the BM is still a major source of income to the Islamic banks, it is necessary for Islamic banks to have prudent risk management to mitigate the negative effect of displaced commercial risk and maintain its protability. Implementation of prot equalization reserves concept is a possible measure for Islamic banks to shield their operation. Originality/value This paper is believed to be the rst study on Islamic BM behavior in Indonesia. It is expected to provide useful information for policy makers and Islamic bank management to develop a sound and protable Islamic banking industry in Indonesia. Keywords Banks, Risk management, Islam, Indonesia Paper type Research paper
International Journal of Islamic and Middle Eastern Finance and Management Vol. 3 No. 1, 2010 pp. 65-82 q Emerald Group Publishing Limited 1753-8394 DOI 10.1108/17538391011033870

1. Introduction The importance of nancial system stability for aggregate economic activity has been increasingly emphasized (Kaminsky et al., 1999; Hoggarth and Saporta, 2001; Borio, 2003).

IMEFM 3,1

66

This is especially true for a developing economy like Indonesia, where banking sector acts as a dominant source of business nancing. Thus, it is important that the intermediary function of banks be carried out at the lowest possible cost in order to achieve greater social welfare. Obviously, lower bank margin (BM) will lead to lower social costs of nancial intermediation (Maudos and de Guevara, 2004, p. 2260). The setting of Indonesian banking system, where Islamic banks operate alongside and compete with their conventional counterparts, gives a direct motivation to study Islamic BM. In fact, in the last ve years the Islamic banks had been growing at 64 percent rate annually, surpassed their counterpart that grew around 30 percent annually (Bank Indonesia, 2007). Thus, we are motivated to examine the determinants of Islamic BM. In particular, we are interested to investigate the effect of interest rate volatility on the Islamic BM. For comparative purpose, we also investigate the determinants of conventional BM. Given that the Islamic banks have to follow the Sharia rules, we expect that the Islamic BM will behave differently with that of their counterpart. In efforts to arrive at conclusive ndings, we use the extended version of the dealership approach developed by Ho and Saunders (1981) study and employ the autoregressive distributed lag (ARDL) bounds testing approach to test the existence of the relationship between BM and its determinants. The data cover monthly series from 1996:01 to 2006:02 of a sample of ve Indonesian banks (two Islamic banks and three conventional banks). As one of the conventional banks converted to an Islamic bank in the reviewed period, we also inspect the existence of a break in the BM behavior. It is expected that this study could give a better understanding on the behavior of the Islamic BM in Indonesia as well as enrich Islamic banking literature. This paper proceeds as follows. In the next section, existing literature and empirical evidence on BM is discussed. Section 3 describes theoretical setting for Islamic BM and empirical approach of the study, followed by Section 4 which discusses empirical results of this study. Section 5 concludes the paper and highlights some relevant policy implications. 2. Literature review 2.1 Denition and models of BM BM is generally dened as the spread between interest revenue on bank assets and interest expense on bank liabilities, which is presented as a proportion of average bank assets or earning assets (Ho and Saunders, 1981; Angbazo, 1997; Saunders and Schumacher, 2000). As a branch of study on bank behavior, prior to 1981, there was no signicant development in the literature on bank margins. In its early time, the literature on BM determination might be traced back to Pyle (1971, as cited in Baltensperger, 1980, pp. 25-9). Afterwards, the literature on bank margins can be classied into two approaches: (1) the dynamic intermediation or dealership approach (Ho and Saunders, 1981; Angbazo, 1997; Saunders and Schumacher, 2000; Valverde and Fernandez, 2005, 2007); and (2) the static micro-model of banking rm (Zarruck, 1989; Wong, 1997). The dealership approach is developed by Ho and Saunders (1981) to study the determinants of bank interest margin. According to them, in playing its role as a dealer

that sets the interest rate on loans and deposits, a bank faces uncertainties and costs since the demand for loans and the supply of deposits are stochastic in the sense that they arrive at different times. Thus, it has to hold a long or short position in the short-term money market to balance this uncertainty that make it exposed to interest rate risk. This inevitably affects the BM. It is further suggested that greater degree of risk aversion, larger size of bank transaction and greater variance of interest rate are associated with larger bank spread. These imply that even if banking market is highly competitive, as long as bank management is risk-averse and faces transaction uncertainties a positive BM will exist as the price of providing deposit-loan immediacy. In contrast, the second approach analyzes banking rms in a static setting where demand for loans and supply of deposits simultaneously clear both markets. It is developed from the criticisms that the rst approach fails to consider some relevant aspects of the banks operation, such as the administrative cost to maintain loan/deposit contracts and the institutional structures of the banking market (Lerner, 1981, p. 601). Zarruck (1989) pioneered the study and found that the risk-averse bank operates with a smaller spread than the risk-neutral bank. This nding was later challenged by Wong (1997) who extended Zarruks work by incorporating credit risk and interest rate risk into the model. In contrast to Zarruks nding, Wong suggested that the optimal bank interest margin is larger in the case of risk-averse banks compared to that of risk-neutral banks. That is, the spread widens as the banks risk aversion rises. Therefore, as this model leads to different results, most empirical studies on BM use the rst approach. 2.2 Empirical research on BM determination Besides developing the theoretical model, Ho and Saunders (1981) examined the validity of their model in 53 sample banks in the USA using quarterly series for the 1976:04-1979:04 period. It is reported that the main determinants of the size of actual BM were transaction uncertainty (pure spread) and implicit interest rate. The pure spread was smaller in the case of large banks compared to that of the small banks, mainly due to the differences in the US banking market structure rather than risk aversion and transaction size of the banks. A number of empirical works on BM have employed and extended the original model of Ho and Saunders (1981). Angbazo (1997) incorporated other types of risk, namely credit risk and liquidity risk, to study 286 US commercial banks in the year 1989-1993. He found that high BM is associated with high credit risk, interest rate risk, and liquidity risk. Capital base, management quality and opportunity cost of reserves are also signicant and positively related to BM. This study is enlarged by Saunders and Schumacher (2000) to cover 614 banks in six selected European countries (UK, Germany, Switzerland, France, Italy, and Spain) and the USA from 1988 to 1995. They found that BM appears to be equally sensitive to both the short-rate and long-rate volatility[1]. These results are conrmed by Maudos and de Guevara (2004) who extended the work of Saunders and Schumacher (2000) by introducing operating cost, degree of market power, and bigger sample size in the period 1993-2000. A study by Valverde and Fernandez (2005) enrich the BM literature by investigating the determinants of BM in banking market of seven European countries (Germany, Spain, France, The Netherlands, Italy, UK, and Sweden) from 1994 to 2001.

Bank margin determination

67

IMEFM 3,1

68

Different from the previous studies, the authors incorporated a broader framework of BM analysis. In particular, they estimated the determinants of the BM using the traditional measure of interest margin (the loan to deposit rate spread), a wider accounting margin (gross income) and a broader BM measure following the New Empirical Industrial Organization perspective (the Lerner index). They also employed two simultaneous equations, i.e. those with level and rst-differenced variables. In the interest margin regression, they reported that the index is not signicant, but both liquidity and interest rate risks are signicant. Operating inefciency and capital to assets ratio are positive and signicant. As a proxy to specialization, they found that the ratio of loans to total assets is negatively and signicantly related to the spread. They suggested that the banks specialized in lending can offer a lower BM due to higher efciency (Valverde and Fernandez, 2005, pp. 8-9). A more recent study from the same authors, Valverde and Fernandez (2007), extended the previous research and further investigated the relationship between specialization and BM. Using the same data set and adding the explanatory variables to include vectors of the pure spreads determinants (risk and market power), other bank-specic factors, bank specialization factors and regulatory/macroeconomic control variables, they found that both market power and risk parameters alter BMs when nancial innovations are introduced. In particular, output diversications permit the banks to increase their revenue and obtain higher market power since revenue from non-traditional business (which includes non-interest income) may compensate for lower interest margins that result from stronger competition in the traditional segments (deposits/loans). These results explain, at least in part, the paradoxical coexistence of decreasing interest margins and increasing market power in European banking sectors which earlier studies have discovered (Valverde and Fernandez, 2007, p. 15). As far as study of banks margin is concerned, the literature on bank margins for Islamic banks is hardly available. Although they are not directly related to the study of BM, some works on Islamic bank performance and protability are found to be helpful in laying down the theoretical foundation for the Islamic BM. Haron and Shanmugam (1995) pioneer the empirical study by investigating Bank Islam Malaysia over the year 1983-1993. Using autoregressive models to examine the relationship between the rates of return and the level of deposit in the bank, the study nds that there is an inverse relationship between the variables. This implies that the Islamic bank customers did not consider returns from the Islamic bank deposit as an incentive to maintain their funds with the bank. However, as this result is contrary to the normal behavior of investors, Haron and Ahmad (2000) expand the study to include all funds deposited in Islamic banks in Malaysia from January 1994 to December 1998. The results indicate that the rates of return have strong positive relationship with the Islamic banks deposits, while the interest rates have strong negative relationship with it. Therefore, it is suggested that the Islamic banks customers in Malaysia are attracted to higher return and guided by the prot motive. A more recent study by Sukmana and Yusof (2005) which employs a wider data set (January 1994 to October 2004) also conrm conclusions of the previous studies. Accordingly, higher BM is required to attract such customers. In the case of Indonesia, using time series analysis for 1994-2002 data, Mangkuto (2004) conrmed that the deposits outstanding of Bank Muamalat Indonesia were negatively responsive to the market interest rate. That is, as interest rates increase,

depositors take their fund from the Islamic bank and move it to the conventional banks, thus reducing the Islamic banks deposit outstanding. He therefore suggested that the Islamic banks depositors were mainly rational depositors who were driven by prot motive in using the banks services. Scope of the study is later widened by Kasri (2008) to cover data of all Islamic banks in Indonesia from March 2000 to August 2007. Using vector autoregressive model model, she revealed that the mudharaba[2] investment deposit in the Islamic banks are cointegrated with return of the Islamic deposit, interest rate of the conventional banks deposit, number of Islamic banks branches, and national income in the long-run. She also reported that rate of return and interest rate move in tandem, indicating that Islamic banks in Indonesia are exposed to benchmark risk and rate of return risk. Further, as the level of Islamic deposit decreased when interest rate increased, she suggested that the banks are also exposed to displaced commercial risk. In Islamic banking literature, the displaced commercial risk occurs due to market pressure that Islamic bank pays a return that exceeds the rate that has been earned on assets nanced by investment account holders when the return on assets is under-performing as compared with competitors rates[3]. This result, hence, conrms the previous nding of Mangkuto (2004). Taken together, these results suggest that the Islamic banks are exposed to various banking risks which will inevitably affect the BM. 3. Empirical approach 3.1 Islamic BM: theoretical setting and empirical specication The nature of Islamic BM is determined by the nature of its component. Islamic banks never have a predetermined-commitment to pay the return or prot to depositors, neither in nominal sum nor in rate, since this violates the Sharia rules regarding riba. Thus, deposit or nancing rates of Islamic banks debt-based products[4] are known ex ante, while those of the equity-based products[5] are known ex post. Taken together, as the deposit rate and nancing rate of the equity-based products of Islamic banks will be known at the end of period, it follows that the Islamic BM is ex post in nature[6]. Conversely, the deposit rate of conventional banks is a predetermined-commitment by the name of interest rate. Since the deposit and lending rates are predetermined as interest rate commitment, the net interest margin of conventional banks is known ex ante. As nancial intermediaries, Islamic and conventional banks normally face the same problems in their operation. Among others, they have to face asymmetrical arrival time of nancing demand and deposit supply, interest rate volatility, default risk on nancings, and liquidity risk. In light of the original model of the dealership approach (Ho and Saunders, 1981) and its extended versions (Angbazo, 1997; Maudos and de Guevara, 2004), the relationship between the Islamic BM and its determinants can be stated as follow: BMt FS t ; X t ; 1t 1

Bank margin determination

69

where BMt is the reported Islamic BM at time t, St is a vector of the determinants of the pure spread (interest rate volatility and default risk on nancings), Xt is a vector of bank-specic control variables (liquidity risk, capital base, implicit return to depositors, opportunity cost of bank reserves, and management quality), and 1t is residual term.

IMEFM 3,1

70

Following discussion in the previous sections, we argue that there are seven factors critical in determining the BM, namely interest rate volatility, default risk, liquidity risk, capital base, management quality, implicit return to depositors, and opportunity cost of bank reserves. With respect to interest rate volatility, theories and empirical evidences show that conventional banks would ask for a higher margin as compensation when interest rate risk increases. In contrast, we argue that Islamic BM reacts negatively to interest rate volatility. That is, as market interest rate volatility increases, ceteris paribus, Islamic banks need to increase their deposit rate or to decrease their nancing rate. In the deposit market, as market interest rate increases, it is possible that Islamic banks customer withdrawn their fund and transfer it to the conventional counterpart, a risk known as displaced commercial risk[2]. In the nancing (credit market), however, demand for Islamic banks nancing will also increase when market interest rates increase, as their prices are normally lower as compared to the credit interest rate of the conventional banks. Hence, as the market interest rate swings, no matter whether it increases or decreases, Islamic banks are exposed to a certain degree of risk, which possibly comes from the movement of either their depositors or users of funds. Thus, the higher the volatility of market interest rate, the bigger the displaced commercial risks faced by Islamic banks such that the banks need to increase their deposit rate or to decrease their nancing rate or operate at a lower margin. Default risk (of nancing) is the risk of non-repayment on nancing due to the inability of the funds users to fulll their obligations to banks. Both Islamic and conventional banks have to face this risk in their operation. Since nancing is the major source of income for both banks, deterioration of nancing quality will affect the banks protability and subsequently the banks viability. As default risk increases, Islamic banks will put a higher risk premium on nancing, then the nancing price will increase and, ceteris paribus, the BM will increase. Thus, the BM relates positively to the default risk. In a similar way, liquidity risk or the risk of not having sufcient cash or borrowing capacity to meet deposit withdrawals or new nancing demand could force banks to resort to emergency funds at a higher cost. As liquidity risk of a bank increases, it will ask for a higher risk premium, which might be done through increasing nancing rate. As a result, ceteris paribus, the Islamic BM will increase. Hence, the Islamic BM responds positively to the liquidity risk. In the cost of capital perspective, higher portion of equity in the capital structure will lead to higher cost of capital as the cost of equity is higher than the cost of debt. Following the earlier literature, an increase in the capital base may increase the average cost of capital and, to compensate for the cost, the bank will ask for a higher nancing rate. As a result, ceteris paribus, the Islamic BM will increase or, in other words, Islamic BM would respond positively to the capital ratio. Likewise, implicit return to depositor that reects extra payments to depositors through service charge remission or other types of transfers is positively related to the Islamic BM, ceteris paribus. As the implicit return to depositors is a component of the operation costs of the banks, a higher nancing rate will be asked to cover the higher costs. Akin to conventional banks, Islamic banks have to fulll reserve requirement regulation which reduces the banks opportunity to give nancing. Previous literatures suggest that the banks will request a higher nancing rate as the reserves increase in compensation for opportunity forgone. Thus, ceteris paribus, we argue that the Islamic

BM responds positively to the opportunity cost of bank reserves. Finally, management quality reecting the ability of management to minimize costs at a given level of income or to maximize income at a reasonable level of cost would would react negatively to Islamic BM. That is, as efciency of Islamic banks improves, the banks would be able to reduce their costs and thus will ask for a relatively lower margin than that of the inefcient banks. In view of the above, in this paper, we directly estimate the relationship between the Islamic BM and its determinants in the following model: BM Fdefault risk; interest rate volatility; liquidity risk; capital ratio; implicit return; bank reserve; management quality 2

Bank margin determination

71

This empirical model also serves as our hypothesis in studying the behavior of the Islamic BM. The data and empirical variables as proxies to the theoretical variables are discussed in the next section. 3.2 Data and empirical variables In this paper, we use monthly data of banks publication reports covering 1996:01 to 2006:02 (122 observations) from the Bank Indonesia database. We use a sample of ve banks: two Islamic banks, namely Bank Muamalat and Bank Syariah Mandiri (henceforth BM and BSM, respectively), and three conventional banks, namely Bank Bumiputera, Bank Ekonomi Rahardja, and Bank Agroniaga (henceforth BB, BER, and BA, respectively). In order to get a comparable picture of the margin behavior of Islamic and conventional banks, we select the conventional banks of the same asset size as the Islamic banks[7]. It is important to note that in this paper, we dene Islamic BM as the ratio of net-nancing income to average earning assets. Net-nancing income equals income of nancing minus income distributed to depositors. As our study compares the margin behavior of Islamic banks with that of the conventional banks, the Islamic BM is a comparable variable to the net interest margin of conventional banks. Meanwhile, net interest margin is the ratio of net interest income to average earning assets. Denition and measure for each determinant of the conventional bank (CB) and Islamic bank (IB) margin as well as expected relationship between the variables are summarized in Table I. 3.3 Econometric procedures Based on the empirical specications and variables discussed in the previous section, the regression model is specied as follows: BMt a b1 DEFt b2 MKTt b3 LIQt b4 SOLVt b5 IMPLt b6 OCBRt b7 QMt nt 3

where all variables are as previously dened, a is intercept and bi (for i 1, 2, . . . , 7) are slope coefcients. In testing the existence of the relationship between the Islamic BM and its determinants (equation (3)), we apply the recently developed cointegration analysis

IMEFM 3,1

Expected BM and its determinants BM Denition and measure Ratio of net-nancing income to average earning assets Ratio of allowances for nancing losses to total nancing: [(allowance for nancing losses/ total nancing) 100 percent] Standard deviation of market interest rate: q X  r 2 r2 =n 2 1 Ratio of liquid assets to liabilities: [(liquid assets/liabilities) 100 percent] Ratio of core equity to assets: [(core equity/assets) 100 percent] Ratio of non funding-related expenses minus non nancing-related revenues to average earning assets: [(net nonfunding related expenses/ average earning assets) 100 percent] Ratio of cash and balance of the deposit with the central bank to total assets: [(cash balance at CB/ total assets) 100 percent] Ratio of operating cost to operating income: [operating costs/ operating incomes) 100 percent] relationship CB IB

72
Default risk of nancings (DEF)

Market interest rate volatility (MKT)

Liquidity risk (LIQ) Capital base or solvency ratio (SOLV) Implicit return to the depositors (IMPL)

Opportunity cost of bank reserves (OCBR)

Table I. BM and its determinants

Management quality (QM)

based on the ARDL framework, namely the bounds F-test[8]. The statistic used is the familiar Wald or F-statistic, which is used to test the joint signicance of lagged level variables in a conditional unrestricted equilibrium correction model (UECM) presented in the ARDL format[9]. If the computed Wald or F-statistic is bigger than the upper bound, we reject the null of no cointegration between variables, vice versa. However, if the Wald or F-statistic falls inside these bounds , i.e. bigger than the lower bound but smaller than the upper bound inference about cointegration is inconclusive and knowledge of the order of the integration of the underlying variables is required before conclusive inferences can be drawn (Pesaran et al., 2001, p. 11). This approach has advantages, as it does not involve pre-testing the integration property of the variables under study[10] and can be applied to small samples. Before conducting the F-test for cointegration, we need to form the UECM for Islamic BM equation and select the optimal lag length[11]. Based on the dealership model, the UECM of the Islamic BM equation is as follows:

DBMt a0

n X p1

lp DBMt2p

n X p0

gp DDEFt2p

n X p0

wp DMKTt2p

Bank margin determination

n X p0 n X p0

fp DLIQt2p

n X p0

up DSOLVt2p

n X p0

dp DIMPLt2p
4

cp DOCBRt2p

n X p0

qp DQMt2p p1 BMt21 p2 DEFt21

73

p3 MKTt21 p4 LIQt21 p5 SOLVt21 p6 IMPLt21 p7 OCBRt21 p8 QMt21 1t where all variables are as previously dened, D is the rst difference operator, n is the order of UECM model [ARDL(n, n, . . . , n)], and p1 (for i 1, 2, . . . , 8) is the slope coefcients of lagged levels variables. The null hypothesis of the bounds testing for our model is that there exists no cointegration between the Islamic BM and its determinants. Given that there exists a cointegration between the Islamic BM and its determinants, for the sake of parsimony, we adopt the general-to-specic approach of Hendry (1977) to select a parsimonious specication for the dynamic of the Islamic BM[12], then we are able to establish the long-run relationships of the Islamic BM through normalizing coefcients of regressions of the UECM. 4. Empirical results and discussion 4.1 The bounds testing for cointegration Table II presents the F-statistics for testing the existence of cointegration between BM and its determinants. In general, the results suggest that there is evidence of cointegration between the BM and its determinants for each sample bank. In particular, for BM, we reject the null hypothesis of no cointegration between variables at lags 1-4 and 6 but we fail to reject the null at lags 5, 7, and 8. For BSM, we reject the null hypothesis at all lag lengths. Meanwhile, for BB and BER, we nd evidence of
Lag order 1 2 3 4 5 6 7 8 BM 4.516 * 6.344 * 4.499 * 3.626 * * 2.915 3.323 * * * 1.957 2.211 BSM 12.447 * 8.855 * 7.888 * 4.965 * 9.637 * 6.553 * 3.158 * * * 8.308 * BB 6.994 * 4.495 * 5.521 * 3.695 * * 3.517 * * 2.295 1.804 1.482 BER 4.833 * 5.352 * 5.685 * 5.523 * 4.834 * 1.429 1.763 1.382 BA 3.730 * * 3.307 * * * 4.640 * 2.090 3.705 * * 3.681 * * 3.602 * * 2.238 Table II. F-statistic for testing the existence of a levels bank margin equation

Notes: Statistic is signicant at: *0.01, * *0.05, and * * *0.10 levels, respectively; critical values for the F-test are taken from Narayan (2004) with number of observations T 80 and number of regressors k 7; lower and upper bounds for 0.01, 0.05, and 0.10 levels are 3.021-4.350, 2.336-3.458, and 2.0173.052, respectively

IMEFM 3,1

cointegration at lags 1-5 but we do not nd it at a higher lag. Lastly, for BA, we reject the hypothesis of no cointegration except for lags 4 and 8. 4.2 The long-run equation for bank margin 4.2.1 Long-run equation of Islamic banks. The evidence of cointegration between Islamic BM and its determinants for BM and BSM serves as an empirical ground for existence of the Islamic BM equation and allow us to proceed and estimate the coefcient of the Islamic BM equations. The long-run coefcients of the BM equation for the respective sample bank are reported in Table III. The relationship between the Islamic BM and its determinants for BM, as reected in the long-run equation, is generally as expected. The coefcients of various determinants, namely default risk of nancing, interest rate volatility, liquidity risk, solvency ratio, implicit cost, and opportunity cost of bank reserves carry the expected signs. In particular, the Islamic BM positively responds to the default risk, solvency ratio, implicit costs, and opportunity cost of bank reserves. Conversely, it negatively reacts to interest rate volatility and liquidity risk. However, the coefcient of management quality carries a negative sign, which is not as expected. As described earlier, the management quality is measured by efciency ratio, whereby a higher ratio reects a lower efciency, thus we expected that inefcient management would be reected in the higher margin. This nding suggests that the efciency ratio might not be the best proxy to the management quality; otherwise, it becomes a puzzle in the margin behavior of Islamic banks. Of special interest is the coefcient of interest rate volatility. The nding of negative sign on interest rate volatilitys coefcient for BM reinforces the theoretical foundation for the Islamic BM behavior described earlier. Moreover, with regard to the size of its impact on the Islamic BM, the interest rate volatility stands in second place after the implicit cost (with coefcient of 2 0.0195). This nding implies that as the volatility of interest rates (measured by its standard deviation) increases by 1 percentage point, ceteris paribus, the Islamic BM will decrease by 0.02 percentage point. Further implication of this relationship is that the performance of Islamic banks will deteriorate

74

BSM BM (1,0,0,0,0,0,0,5) BB BER BA Variables (6,3,4,2,5,6,5,5) 96:01-99:10 99:12-06:02 (1,5,5,4,0,0,4,5) (2,5,1,4,0,2,0,0) (1,0,0,0,3,0,0,3) DEFt MKTt LIQt SOLVt IMPLt OCBRt QMt Table III. A comparison of long-run equation of BM 0.0040 2 0.0195 2 0.0003 0.0153 0.0371 0.0149 2 0.0143 2 0.0076 0.0021 2 0.0070 0.0236 0.2805 2 0.0511 2 0.0176 0.0034 2 0.0064 2 0.0036 0.0109 0.4499 0.0362 2 0.0080 0.0622 0.0110 20.0057 0.0046 0.4315 0.0034 20.0251 0.0021 0.0048 2 0.0018 0.0048 0.5374 0.0022 2 0.0040 20.0025 0.0385 20.0004 0.0010 0.7266 20.0193 20.0098

Notes: The regression is based on the conditional UECM[9], using an ARDL(n1, n2, . . ., n8) specication, which is appropriate for each sample bank, with dependent variable BMt estimated over 1996:01-2006:02; for BSM, it is estimated in 1996:01-1999:10 and 1999:12-2006:02 due to its conversion to be an Islamic bank in 1999:11

as the volatility of interest rates increases, given the fact that the BM is still the major source of income for the Indonesian Islamic banks. As mentioned earlier, the second Islamic bank of our sample banks, BSM, had operated under a conventional framework prior to its conversion into an Islamic bank. In fact, this setting benets us in observing the existence of switching behavior in the BM of BSM in the period under consideration. Since BSM was converted into an Islamic bank in November 1999, we perform the Chow Break test to check the existence of switching behavior with 1999:11 as the break point. Having the F-statistic equal to 23.95, we reject the null hypothesis of no break in the model at the 0.01 level of signicance. This nding motivates us to break the observation into 1996:01-1999-10 as the conventional period and 1999:12-2006:02 as the Islamic period, and then estimate the BM equation for each period. In general, we nd that the BM equation of BSM in the Islamic period is as expected and thus strengthens the abovementioned empirical evidence of BM. All coefcients of the Islamic BM determinants of BSM carry the expected signs. With respect to interest rate volatility, however, the size of its coefcient is only 2 0.0064, which is far below that of BM of 2 0.0195. Moreover, based on its effect on the Islamic BM, for BSM, interest rate volatility stands in fth position, following the implicit costs, opportunity cost of bank reserve, solvency ratio and management quality. Comparing the BM equation of BSM in its conventional and Islamic periods, we observe some differences in the coefcient sign of some determinants. For instance, the coefcient sign of default risk of nancing changes from negative to positive and the coefcient sign of opportunity cost of bank reserve changes in the same way. Interestingly, the coefcient sign of interest rate volatility changes from previously positive in the conventional period to be negative in the Islamic period. This reveals that, as interest rate volatility increases, conventional banks will ask for a higher margin to compensate for the interest rate risk (reinvestment or renancing risks), while Islamic banks have to increase their deposit rate or decrease their nancing rate (thus the margin decrease) to protect their operation from increasing magnitude of displacement risk. With respect to the banks response to the interest rate movement, this nding implies that the margin behavior changes as the basis of the banks operation switches from conventional practices, where it faces interest rate risk, to the Islamic principles where it faces the displaced commercial risk. Having seen that the coefcient of interest rate volatility is negative in sign for both BM and BSM, we further investigate the empirical justications of the nding. Specically, we examine the behavior of the Islamic BM components, namely price of nancing products (nancing rate) and rate of return to depositors (deposit rate), under different trends of market interest rate. Indeed, with the purpose of properly analyzing the behavior of the components under a different tendency of market interest rate, we break the observation into increasing and decreasing periods of market interest rate[13]. Afterwards, for the respective periods, we calculate the correlation coefcient between market interest rate and nancing rate, measured as a ratio of nancing income to nancing outstanding. We also calculate the correlation coefcient between market interest rate and deposit rate, measured as a ratio of prot distributed-to-depositors to deposits outstanding. The results are reported in Table IV. Some interesting ndings could be highlighted from our investigation. In the increasing interest rate period, we observe that the nancing rates of Islamic banks

Bank margin determination

75

IMEFM 3,1
Correlation BM Financing rate-market rate Deposit rate-market rate BSM Financing rate-market rate Deposit rate-market rate

Increasing regime (96:01-98:08) 2 0.6624 0.6082 0.7006 0.5386

Decreasing regime (98:08-00:02)a 0.5521 0.9118 0.2417 0.3101

Increasing regime (00:02-02:01) 20.4334 0.2423

Decreasing regime (02:01-05:03) 0.5497 0.8062

76

Table IV. Correlation between nancing and deposit rate with market rate

Notes: Financing and deposit rates are measured as ratios of nancing incomes (equivalent to interest incomes) to total nancings outstanding and ratio of prot distributed-to-depositors (equivalent to interest costs) to total deposits outstanding, respectively; afor the BSM, this period is set to 1998:081999:10 in order to avoid the effect of conversion of its operation from conventional bank into Islamic bank on 1999:11

(BM and after-conversion-BSM) move opposite to the market interest rate, as supported by the correlation coefcient of 2 0.66 and 2 0.43. In contrast, the deposit rates change in a similar direction to the market interest rate, as suggested by the correlation coefcient of 0.61 and 0.24. These results imply that, as the market interest rate increases, the Islamic banks are unable to adjust instantaneously their nancing rate, since upwardly adjusting the nancing rate will violate the Sharia rule regarding the alteration of selling price[14] once it was agreed up-front. On the other hand, as the market interest rate increases, to dissuade their depositors from withdrawing their funds (displaced commercial risk), the Islamic banks have to increase their deposit rate. In fact, these two contrary effects will bring down the Islamic BM in periods of increasing interest rates. This evidence restates our nding on the negative sign of the interest rate volatilitys coefcient for BM and BSM. In the period of decreasing market interest rates, we verify that both the nancing and deposit rates move in tandem with the market interest rate. For BM and BSM, the correlation coefcients of nancing rate-market interest rate are 0.55 and 0.55, while the correlation coefcients of deposit rate-market interest rate are 0.91 and 0.81. However, to this point we are unable to explain why the Islamic BM responds negatively to the market interest rate in the decreasing period. Thus, we estimate the effect of market interest rate on the nancing and deposit rates of Islamic banks[15]. The result reveals that as the interest rate decreases by 1 percentage point, ceteris paribus, the deposit rates of BM and BSM will decrease by 0.03 and 0.07 percentage points, respectively, while the nancing rate of the two banks will decrease by 0.008 and 0.04 percentage points, respectively. Thus, we can verify that the deposit and nancing rates respond disproportionately to the market interest rate. As the market interest rate shrinks, the deposit rates decrease by a bigger percentage point compared to that of the nancing rate. This evidence is found in the regressions of both BM and BSM, with the evidence of BM being more apparent. In fact, this is the reason why the Islamic BM does not decrease in the period of decreasing market interest rates. Again, this nding reinforces the negative sign of interest rate volatilitys coefcient in the Islamic BM equation. Another interesting nding in the decreasing rate period would be the fact that the nancing price is able to adjust to the market interest rate in the decreasing period,

but not in the increasing period. As argued earlier, any alteration of selling price in the xed-price mode of nancing will breach the contract as it violates the Sharia rule. That is, in light of the Islamic principles, it is prohibited to make any increase to the selling price, i.e. adding the mark-up as interest rate increases or for whatever reasons, as this is unjust and will increase the burden of the customers. However, giving a discount or rebate under the name of the banks kindness without altering the selling price is permissible as long as the rebate is not contracted and should be at the banks discretion[16]. In fact, to protect their operation from the displaced commercial risk of nancing in the decreasing period, in which the customers re-nance their nancing in the Islamic banks with that of conventional banks since now the market interest rate is lower, it is a common practice for the Islamic banks to give a discount or rebate to their customers. It is also worth to note that when the nancing rates of both of the banks decrease only by 0.008 and 0.04 percent, their deposit rates decrease by a bigger magnitude, i.e. 0.03 and 0.07 percent, respectively, while the deposit contract is based on the prot-loss-sharing (PLS) contract. Recall that there are two factors working here in reducing the deposit rates to a bigger extent as compared to the nancing rates. The rst factor is the decrease in the nancing rate itself as discussed previously, while the other factor is the movement of depositors from conventional banks to Islamic banks since now the return of the rst is lower than that of the second, ceteris paribus. This deposits movement will bring down further the rate of return to depositors as now the denominator (amount of deposits) increases while the numerator (prot distributed to depositors) is relatively slow to adjust[17]. The deposits movement will continue until the difference of deposit rates between the Islamic and conventional banks is not signicant enough to invite arbitrage opportunity action. Finally, should we look at the correlation coefcient of BSM in its conventional period, some interesting ndings are noticeable. We verify that, in the conventional period, the nancing rate behaves in a different manner relative to that in its Islamic period. Indeed, the dissimilarity is found in the sign of correlation coefcient of the nancing rate and market interest rate. As discussed earlier, after it became an Islamic bank, this coefcient is found to be negative in the period of increasing interest rates. However, the coefcient is always positive in the conventional period irrespective of whether it is in the period of increasing or decreasing interest rates. That is, as conventional banks have no constraints in adjusting their credit rate to follow the market rate, the correlation coefcient is always positive in the different tendency of market interest rate. 4.2.2 Long-run equation of conventional banks: a comparison. Some comparisons of the long-run BM behavior of Islamic and conventional banks could be pointed out from Table III results. In general, both banks share similar behavior in the relationship between the BM and majority of its determinants, namely default risk of nancing, liquidity risk, solvency ratio, implicit cost, opportunity cost of bank reserves and management quality. In particular, although the coefcients of those determinants vary in size, they show similar signs for all sample banks, irrespective of whether they are conventional or Islamic banks[18]. This evidence reinforces the validity of Ho and Saunders (1981) dealership model and its extended versions. With respect to the interest rate volatility, while its coefcient shows a negative sign for the Islamic banks, it is positive for the conventional banks. These ndings reveal that the conventional banks have no constraint in adjusting their credit (nancing) and deposit rates to

Bank margin determination

77

IMEFM 3,1

follow the market interest rate movement, regardless of whether it is in the period of increasing or decreasing interest rates. However, as discussed earlier, the Islamic banks are unable to adjust their nancing price in the period of increasing interest rates, since this will violate the very central principle in their operations, namely the Sharia rule. 5. Conclusion and policy recommendation We have examined the behavior of the BM in Indonesia using monthly data over the period of 1996:01-2006:02. Applying a relatively new cointegration technique, namely the bounds testing approach within the ARDL framework, we specically identify a long-run relationship between the Islamic BM and its determinants, namely default risk of nancing, interest rate volatility, liquidity risk, solvency ratio, implicit cost, opportunity cost of bank reserves, and management quality. In particular, this paper conrms the Ho and Saunders dealership approach and its extended versions in BM determination. In the theoretical setting, we argued that Islamic banks in Indonesia are not remote from the interest rate volatility in their presence under a dual banking system. It is the displaced commercial risk, rather than reinvestment and renancing risk, that threatens Islamic banking operations in a changing market interest rate situation. For Islamic banks case, our nding differs from Angbazo (1997) and Valverde and Fernandez (2005, 2007). Indeed we nd a negative relationship between Islamic BM and interest rate volatility. This is in line with the nding of Kasri (2008) as well as our empirical evidence that the Islamic BM behaves differently relative to that of conventional banks, i.e. with respect to their response to the market interest rate movement. Moreover, we documented that BM behavior switched as the underlying operational principles of the bank changed from conventional to Islamic principles. It is an opportunity for future research to compare our nding with BM behavior in other dual banking system such as Malaysia, Bahrain, and Pakistan. As we make use of time series approach, it will also be interesting to employ different methodology like two-step approach and a pool-data analysis to investigate the issue. Our nding has particular relevance to the formulation of banking policy in developing Islamic banks in Indonesia. As far as the Islamic banks are concerned, knowing that the Islamic BM responds negatively to the volatility of the market interest rate implies that the stability of market interest rates becomes of great concern in developing the Islamic banking industry. As long as the BM is still a major source of income to the Islamic banks, the instability of the market interest rates will have a negative impact on the protability and thus sustainability of the Islamic banks, regardless its share to the Indonesian banking system currently. In light of this understanding, Bank Indonesia as the monetary authority as well as the banking authority (at least at the present time) should take into consideration the impact of market interest rate volatility on the Islamic banks, in designing its monetary policy. It is also necessary for Islamic banks to have prudent risk management to mitigate the negative effect of displaced commercial risk and maintain its protability. As such, the prot equalization reserve (PER) concept introduced by Islamic Financial Services Board (IFSB) is a possible solution that could be adopted by Islamic banks in Indonesia. The PER is the amount set aside by the Islamic nancial institutions out of their gross income in order to maintain a certain level of rate of return for their

78

depositors (IFSB, 2005). The concept is in line with the nding and suggestion of Valverde and Fernandez (2007) in which (Islamic) banks could enlarge their range of products and services in attempts to increase the revenue from non-traditional business (which includes non-nancing income) which may compensate for lower BM due to stronger competition in the traditional segment.
Notes 1. The short-run rate is dened as the annual standard deviation of weekly interest rate on three-month securities while the long-rate volatility is dened as that of the one-year securities. An 1 percent increase in the volatility of interest rates increased BM by about 0.2 percent. 2. Mudharaba is a type of time deposit based on prot-loss sharing contract offered by Islamic banks in Indonesia. 3. Further, according to IFSB, displaced commercial risk is dened as an exposure when the Islamic banks are under market pressure to pay a return that exceeds the rate that has been earned on assets nanced or when the return on assets is under-performing as compared with competitors rates. As such, the Islamic banks may decide to waive their rights to part or their entire mudarib share of prots in order to satisfy and retain their fund providers and dissuade them from withdrawing their funds (IFSB, 2005, Guiding Principles of Risk Management for Islamic Financial Institution, p. 23). 4. Products of Islamic banks can be classied into debt- and equity-based products. The debt-based (xed-return) products are commonly based on the sale (trading) and lease of tangible assets with delayed payment, among others murabahah (mark-up sale), ijarah (leasing), istisna (sale to manufacture transaction) and salam (sale with future delivery). 5. The variable-return products employ PLS contracts such as mudarabah (trust nancing) and musharakah ( joint nancing). 6. Although in the extreme case, where all nancing products are xed-return (debt-based) and thus the nancing rate is known ex ante, notice that the Islamic BM remains ex post since the deposit rate is unknown ex ante. 7. Please see the Appendix for the detailed comparison of the banks. 8. In empirical studies of time series, estimation of relationship in levels such as equation (3) is justied as long as variables appearing in the equation are stationary. Should some variables in the system or all of them be non-stationary, then they should be cointegrated. Otherwise, we will face the problem of spurious regression. Notice that in the presence of non-stationary variables, Granger and Newbold (1974) suggest that there might be a spurious regression with high R 2 and t-statistics that appear to be signicant, but the results has no economic meaning (in Enders, 1995, p. 216). Thus, it is important in the rst place to verify the cointegration properties of the variables under study. In general, this analysis is based on the use of two main approaches in the cointegration analysis have been widely used, namely the two-step residual-based test as developed by Engle and Granger (1987) and the maximum likelihood-based test as developed by Johansen (1988) and Johansen and Juselius (1990). However, those approaches require a pre-testing integration order of the studied variables. Otherwise, pre-testing bias problem could happen. In light of those problems, Pesaran and Shin (1997) and Pesaran et al. (2001) developed a new technique based on F-statistic in the autoregressive distributed lag specication (henceforth ARDL) framework. 9. There are two sets of critical values for the F-test, which assumes all the regressors are, on the one hand, purely I(1) and these are referred to as the upper bound critical values, and, on the other hand, purely I(0) and these are referred to as the lower bound critical values.

Bank margin determination

79

IMEFM 3,1

10. In particular, the cointegration test is directly applicable irrespective of whether the underlying regressors are purely I(0), purely I(1) or mixture of I(0) and I(1) (Pesaran et al., 2001, p. 1). 11. We make use of the information criteria, namely, Akaikes information criterion (AIC) or Schwarzs information criterion (SC). However, as emphasized by Pesaran et al. (2001), it is important to note that the assumption of serially uncorrelated errors is essential for the validity of the bounds tests. Therefore, this concern must be incorporated in determining the appropriate lag length of the UECM. In this paper, we calculate the AIC, SC, and F-statistic by imposing the same lag length for all rst differenced variables in equation (4). 12. In particular, we start from the base model as suggested by AIC or SC, then sequentially reduce the insignicant lag-length for each of the lagged rst-differenced variables until the last-lag is signicant. 13. It seems easier for us to understand the response of the deposit rate and nancing rate under the increasing or decreasing interest rate era rather than under increasing or decreasing interest volatility. In fact, the nding of coefcient correlation between the interest rate and its volatility (standard deviation) equal to 0.66 serves as a basis for us to proceed. 14. This argument is reliable, as the dominant contract in nancing product is murabahah (xed-price nancing). Even if the contributions of mudarabah and musharakah (PLS nancing) in the assets portfolio of Islamic banks are signicant, this argument is still workable on the basis that the business sector generally performs poorer in an increasing interest rate period. Hence, the nancing income of the Islamic bank will remain constant or might decrease as the prot of the business sector deteriorates. 15. Note that before we proceed, we have tested whether the market interest rate causes in the Granger sense the deposit rate and nancing rate or not to provide basis for the investigation. From the Granger test, we nd that the market interest rate Granger-causes the deposit rate and nancing rate of BM at 0.01 levels, respectively. For BSM, we nd the same result at 0.01 and 0.10 levels, respectively. 16. At least for the practice of Islamic banks in Indonesia as it is allowed by the Fatwa of National Sharia Council No. 46, February 22, 2005. 17. It has been known in the banking practices that there is a lag between the times the bank receives deposits and the times the bank lend out the money. This happens to conventional banks and Islamic banks as well. 18. An exception is given to BA since this bank shows a peculiar sign in the coefcients of default risk of nancing and opportunity cost of bank reserves. However, the other ve regressors have similar sign with those of remaining conventional banks. References Angbazo, L. (1997), Commercial bank net interest margins, default risk, interest-rate risk, and off-balance sheet banking, Journal of Banking & Finance, Vol. 21, pp. 55-87. Baltensperger, E. (1980), Alternative approaches to the theory of the banking rm, Journal of Monetary Economics, Vol. 6, pp. 1-37. Bank Indonesia (2007), Laporan Perkembangan Perbankan Syariah 2007 ( Islamic Banking Development Report 2007 ), Bank Indonesia, Jakarta. Borio, C. (2003), Towards a macroprudential framework for nancial supervision and regulation, BIS Working Paper 03/128, Bank for International Settlements, Basel, February. Enders, W. (1995), Applied Econometric Time Series, Wiley, Singapore. Engle, R.F. and Granger, C.W.J. (1987), Cointegration and error-correction: representation, estimation and testing, Econometrica, Vol. 55, pp. 251-76.

80

Granger, C.W.J. and Newbold, P. (1974), Spurious regressions in econometrics, Journal of Econometrics, Vol. 2, pp. 111-20. Haron, S. and Ahmad, N. (2000), The effects of conventional interest rates and rate of prot on funds deposited with Islamic banking system in Malaysia, International Journal of Islamic Financial Services, Vol. 1 No. 4, pp. 1-7. Haron, S. and Shanmugam, B. (1995), The effect of rates of prot on Islamic banks deposit: a note, Journal of Islamic Banking and Finance, Vol. 12 No. 2, pp. 18-28. Hendry, D.F. (1977), On the time series approach to econometric model building, in Sims, C.A. (Ed.), New Methods in Business Cycle Research, Minneapolis, MN, Federal Reserve Bank of Minneapolis, pp. 183-202, reprinted in Hendry, D.F. (1993), Econometrics: Alchemy or Science?, Blackwell Publishers, Oxford, 2000. Ho, T. and Saunders, A. (1981), The determinants of bank interest margins: theory and empirical evidence, Journal of Financial and Quantitative Analyses, Vol. 16, pp. 581-600. Hoggarth, G. and Saporta, V. (2001), Cost of banking system instability: some empirical evidence, Bank of England Financial Stability Review, No. 10. IFSB (2005), Guiding Principles of Risk Management for Institutions (Other than Insurance Institutions) Offering Only Islamic Financial Services, Islamic Financial Services Board, Kuala Lumpur. Johansen, S. (1988), Statistical analysis of cointegration vectors, Journal of Economic Dynamis and Control, Vol. 12, pp. 231-54. Johansen, S. and Juselius, K. (1990), Maximum likelihood estimation and inference on cointegration with application to the demand for money, Oxford Bulletin of Economics and Statistics, Vol. 52, pp. 169-209. Kaminsky, G.L., Lizondo, S. and Reinhart, C.M. (1997), Leading indicators of currency crises, IMF Working Paper 97/79, International Monetary Funds, Washington, DC. Kasri, R.A. (2008), The dynamics of Islamic banking growth determinants: the Indonesia experience, paper presented at the International Accounting Conference (INTAC) IV, International Islamic University of Malaysia (IIUM), Kuala Lumpur, June 24-26. Lerner, E.M. (1981), Discussion: the determinants of bank interest margins: theory and empirical evidence, Journal of Financial and Quantitative Analyses, Vol. 16, pp. 601-2. Mangkuto, I.J. (2004), Pengaruh tingkat suku bunga deposito konvensional dan tingkat pendapatan deposito mudharaba terhadap pertumbuhan deposito di Bank Muamalat Indonesia (The impact of interest rate and return of mudharabah-investment account on the growth of mudharabah-investment account in Bank Muamalat Indonesia), unpublished Master dissertation, Universitas Indonesia, Indonesia. Maudos, J. and de Guevara, J.F. (2004), Factors explaining the interest margin in the banking sectors of the European Union, Journal of Banking & Finance, Vol. 28, pp. 2259-81. Narayan, P.K. (2004), Reformulating critical values for the bounds F-statistics approach to cointegration: an application to the tourism demand model for Fiji, discussion papers, Department of Economics, Monash University, Melbourne. Pesaran, M.H. and Shin, Y. (1997), An autoregressive distributed lag modeling approach to cointegration analysis, DAE working paper, Department of Applied Economics, University of Cambridge, Cambridge. Pesaran, M.H., Shin, Y. and Smith, R.J. (2001), Bound testing approaches to the analysis of level relationships, Journal of Applied Econometrics, Vol. 16, pp. 289-326. Saunders, A. and Schumacher, L. (2000), The determinants of bank interest rate margins: an international study, Journal of International Money and Finance, Vol. 19, pp. 813-22.

Bank margin determination

81

IMEFM 3,1

82

Sukmana, R. and Yusof, R.M. (2005), Are funds deposited in Islamic banks guided by prot motive? An empirical analysis on Malaysia, paper presented at the 4th Global Conference on Business and Economics, Oxford University, Oxford, 26-28 June. Valverde, S.C. and Fernandez, F.R. (2005), The determinants of bank margins revisited: a note on the effect of diversication, Working Paper No. 05/11, Departamento de Teora e Historia Economica, Universidad de Granada, Granada. Valverde, S.C. and Fernandez, F.R. (2007), The determinants of bank margins in European banking, Journal of Banking & Finance, Vol. 31 No. 7, pp. 2043-63. Wong, K.P. (1997), On the determinants of bank interest margins under credit and interest rate risks, Journal of Banking & Finance, Vol. 21, pp. 251-71. Zarruck, E.R. (1989), Bank margin with uncertain deposit level and risk aversion, Journal of Banking & Finance, Vol. 13, pp. 797-810. Further reading Bahmani-Oskooee, M. and Chi, W.N.R. (2002), Long-run demand for money in Hong Kong: an application of the ARDL model, International Journal of Business and Economics, Vol. 1 No. 2, pp. 147-55. Gujarati, D.N. (2005), Basic Econometrics, 4th ed., McGraw-Hill, East Windsor, NJ. Appendix
BSMa 3,011 2,186 2,678 2,354 460 0.697 6.425 2.799 32.865 74.819 0.551 5.014 87.08 75

Name of bank Description Assets Financings Earning assets Deposits Core equity Variables c BM DEF MKT SOLV LIQ IMPL OCBR QM Number of observation
b

BM 2,028 1,589 1,738 1,523 226 0.502 5.577 2.799 14.264 19.815 0.675 7.590 100.41 122

BB 1,816 1,186 1,510 1,394 183 0.474 2.411 2.799 12.849 40.713 0.375 4.653 93.78 122

BER 5,229 1,673 3,688 4,074 251 0.502 7.694 2.799 4.656 57.325 0.322 4.536 89.39 122

BA 918 706 848 666 120 0.471 4.696 2.799 18.409 28.278 0.264 3.965 91.95 122

Table AI. Description of sample banks and variables mean

Notes: aAll gures for BSM are after it converted to be an Islamic bank (1999:11 to 2006:02); ball gures in this section are in billions of IDR; call variables are in percentage

Corresponding author Rahmatina A. Kasri can be contacted at: rahmatina@ui.edu To purchase reprints of this article please e-mail: reprints@emeraldinsight.com Or visit our web site for further details: www.emeraldinsight.com/reprints