1
Working Paper in Islamic Economics and Finance No. 0824
Lending Structure and Bank Insolvency Risk: The Case of Islamic Banks in Malaysia.
Aisyah Binti Abdul Rahman Pusat Perniagaan Pengurusan Perniagaan,
Fakulti Ekonomi dan Perniagaan, UKM
eychah@ukm.my
Mansor H. Ibrahim Ahamad Kameel Mydin Meera International Islamic University of Malaysia
Gombak Kuala Lumpur
This Draft September 2008
Bangi 43600, Selangor, Malaysia Fax: +603-89215789 http://www.ukm.my/ekonis
E-mail: ekonis@ukm.my
ABSTRACT
This study investigates the impact of lending structure on the Islamic banks’ insolvency risk exposure by analyzing four lending structure instruments; 1) the ratio of real estate lending to total asset; 2) lending concentration; 3) short term lending portfolio stability; and 4) medium term lending portfolio stability. Our findings indicate that an increase in real estate lending decreases the Islamic banks’ insolvency risk exposure. Similarly, loan specialization reduces the Islamic banks’ insolvency risk exposure. This can be due to the fact that 48.5% of lending portfolio composition is given to the broad property sectors. Hence, policy makers as well as practitioners should take into account the lower risk impact of real estate lending to the Islamic banks’
insolvency risk exposure in their decision making process.
JEL classification: G21
Keywords: Lending Structure; insolvency risk; Islamic banks
INTRODUCTION
Insolvency risk exposure is a subject that receives much attention in the banking literature.
Indeed, the recurring financial crisis has heightened intention in the subject. While some authors suggest that lending expansion increases bank risk exposure, the others believe that the structure of lending portfolio composition is of crucial importance. Hanson et. al. (2005) suggests that if a firm parameters come from different distributions (different sectors), there will be further scope for risk diversification by changing the portfolio weights, even in the case of sufficiently large portfolio. The implication of their work to the banking sector is that if lending composition from different sectors is fairly diversified, the deterioration of bank capital in an uncertain economy would be lower. Moreover, Blasko and Sinkey Jr (2005) provide empirical evidence for the case of the United States by showing that concentration in mortgage lending challenges the capability of banks to manage interest rate risk, especially during a rising climate. They highlight that without proper regulatory supervisions, banks which specialize in real estate lending could shift their risks onto the government safety net, especially if those banks were established to fulfill certain government objectives. In another study, Madura et al. (1994) find that real estate lending increases thedepository institutions’ risk exposure in the U.S., but such relationship disappears when a similar test is conducted for the case of commercial banks. For the Malaysian context, Nor Hayati and M. Ariff (2003) reveal that loan on broad property sectors, consumer credit, and purchase of securities increases market risk, but not total and unsystematic risk exposure for the case of depository institutions. Against this background, the result for commercial banks in Malaysia is still questionable. Let alone for the case of the Islamic banks. Thus, the objective of this study is to examine the relationship between lending structure and bank insolvency risk exposure for the case of Islamic banks in Malaysia.
As the significance of analyzing the impact of lending structure on risk exposure is obvious, the measurement for lending structure variables is not extensive. For most attention, previous studies focus on real estate lending and loan specialization index. This study adds value to the lending structure measurement by employing four different lending structure models, adopted with modification from Mansor and Ruzita (2003) and Amin Gutierrez de Pinares and Ferrantino (1997; 1999). Even though those studies focus on the structure of export portfolio composition, the objective is similar; that is to examine how the structure of one variable influences the other variable. The four lending structure measurements employed in this study are 1) ratio of real estate lending to total loan, 2) specialization index, 3) Lending composition change, and 4) Variance of traditionality Index
The remainder of this paper is organized as follows: section 2 outlines a literature review on the risk determinants. Section 3 put forwards the research design. Section 4 highlights the findings and finally section 6 concludes.
LITERATURE REVIEW ON RISK DETERMINANTS
Studies on the determinants of bank risk exposure per se are very limited. So far, only Madura et al (1994) and Nor Hayati and M. Ariff (2003) examine the determinants of risk exposure. On one hand, Madura et al (1994) examine the determinants of the implied risk exposure for the case of deposit-taking institutions and commercial banks in the U.S.1 Their findings for the depository institutions are however not consistent with the findings for the commercial banks. With regards
1The implied risk exposure is estimated based on the call option price.
to the depository institutions, real estate lending and real estate owned are positively related to risk; while non-interest income and capital buffer are inversely related. For the case of the commercial banks, the impact of real estate lending on risk disappears. However, the findings on real estate owned and capital buffer hold. On the other hand, Nor Hayati and M. Ariff (2003) investigate the determinants of risks for the case of Malaysia using a single-factor CAPM. They only focus on the deposit-taking institutions which comprise of commercial banks, merchant banks, and finance companies. Their findings document that various types of risk exposures have different risk factors. For market risk exposure, the determinants are loan default, cost of fund, loan expansion, and loan concentration. With regards to unsystematic risk exposure, the first two variables hold while short term interest rate replaces loan expansion and loan concentration. For the case of total risk exposure, the result is similar to unsystematic risk exposure plus an additional factor, which is loan expansion. Finally, the determinant for equity risk exposure is the regulatory capital. Taken together the findings from both studies, at least two points can be highlighted. First, the risk behavior for the depository institutions and commercial banks differ.
Second, different types of risk exposures have different determinants.
As there is no established theoretical framework for risk exposure, most studies include some banks’ specific characteristics as control variables when investigating a specific factor on risk exposure. Saunders et al (1990) examine the relationship between ownership structure and the U.S. bank risk exposures by taking into account equity capital, fixed asset, and size. In a later study, Anderson and Fraser (2000) apply a slightly different specification for the equity capital.
As oppose to Saunders et al (1990) who employ the ratio of total equity to total asset (TE/TA), they introduce,‘frequency’, which is the ratio of an average daily share volume traded to number of shares outstanding as an alternative proxy for capital buffer.2 While the aforementioned studies analyze for the case of the United States, Konishi and Yasuda (2004) examine the same issue but for the case of Japan. Other than size, they also incorporate three bank specific variables that are related to capital. Besides (TE/TA) and ‘frequency’, they add a dummy variable to cater for the period of capital constraint. In short, studies on ownership structure and risk exposures show that variables related to business activities and capital are included as control variables.
Despite that, studies on loan sales embrace other bank specific variables. M, Kabir (1993) includes variables related to credit, interest rate, and business activities. The credit related variables are loan specialization, loan expansion, and loan default. For the interest rate related variable, he develops the ratio of GAP to total asset.3 Further, size and dividend payout ratio are included to analyze the risk behavior of bank business activities. In another study, Cebenoyan and Strahan (2004) include bank specific variables related to capital, liquidity, and credit. While a standard specification applies for capital and credit, the liquidity factor is proxied by short term investment. In conclusion, studies on loan sales take into consideration the variables related to credit, interest rate, liquidity, capital, and business operations.
Focusing on a different issue, Gallo et al (1996), who examine the impact of mutual fund activities on bank risk, incorporates bank specific variables related to credit, investment, capital, and business operating variables. In another study, Brewer et al (1996) who focus on lending structure only include one control variable that is related to capital. In a recent study, Gonzales (2004) who analyzes the impact of regulatory restriction on risk incorporates bank specific variables related to credit, investment, and business operations.
2This is due to the fact that it denotes the speed of which new info is captured in stock price and correlated to variances in bank balance sheet and off-balance sheet portfolio.
3GAP is RSA-RSL. RSA is rate sensitive asset and RSL is rate sensitive liability.
As the theoretical framework for bank risk exposure is not yet established, this study incorporates all bank specific variables that can be gathered from the annual reports. Variables based on market information such as dividend payout ratio and ‘frequency’ is beyond the scope in this study as most of the Islamic banks in Malaysia have not yet going public.
RESEARCH DESIGN
The following linear model is estimated using unbalanced panel data using the generalized least squares (GLS) estimation to test the risk behavior of lending structure. Three models are tested;
namely, none effect, fixed effect, and random effect model. The best model is selected based on Likelihood Ratio and Hausman test.4 To cater for the heteroskedasticity issue, this study incorporates cross-section weight in the GLS estimation. Following Shahida (2006) and Roza Hazli (2007), first order autocorrelation problem is tackled based on the Park’s model by incorporating AR(1) in the regression model. Following Gujarati (2003), multicollinearity problem is handled by omitting one of the paired variables that has correlation coefficient value higher than 0.8. 14 Islamic banks and windows are collected for year 1994-2006.
+ ? + - + + - +/- - -
Insolvency Risk = f (TL, LS, PLL, TE, GAP, INTEXP, INV, LTA, NONII, MGT) Where the four alternative lending structure (LS) variables are:
Ratio of Real estate lending to total loan
Specialization index
Lending Composition Change
Variance of Traditionality Index
and where:
Insolvency Risk = the Zrisk index
TL = Ratio of Total loan to total asset
PLL = Ratio of provision for Loan Loss to total asset TE = Ratio of total equity to total asset
GAP = Ratio of (total market rate assets - total market rate liabilities) to total asset INTEXP = Ratio of (income distributed to depositors and shareholders fund) to total asset INV = Ratio of short term investment securities to total asset
LTA = Logarithm of total asset
NONII = Ratio of non-interest income to total asset MGT = Ratio of earning asset to total asset
TL, LS, PLL, TE, GAP, INTEXP, INV, LTA, NONII, and MGT are proxies for the loan expansion, lending structure, loan default, financial leverage, GAP analysis, cost of fund, liquid asset, size, non-interest income, and management efficiency. To be in line with the literature review, TL, LS, and PLL represent the credit related variables; TE represents capital related variable; GAP and INTEXP represent interest rate related variables; INV represents the liquidity related variable; and finally, LTA, NONII, and MGT represent business operation related variables.
4Please refer to Beaver et al.(1989), Hsio (2002), Green ( )Gujarati (2003), Shahida (2006) and Roza Hazli (2007) for further details on panel data regression technique.
The expected sign of this model is based on the findings and justifications from previous empirical studies. With regards to loan expansion, researchers seem to have a consensus view that loan expansion (TL) is positively related to risk with various reasons. M. Kabir (1993) argues that heavy reliance on loans by commercial banks is considered as having a high degree of financial leverage; thus increases the bank financial risk. Looking from a perspective of probability of default, Madura et al (1994) highlight that giving loans is more risky than holding investment in securities since banks are allowed to invest only in good investment grade securities. This infers that increasing loan as oppose to investment securities leads to a higher risk. On the other hand, Gallo et. al. (1996) justify that loans are relatively illiquid, besides subject to default risk. They believe that both liquidity and default issues are the rationale for a positive relationship between loan expansion and risk. Against this background, this study hypothesizes that TL is positively related to bank insolvency risk exposure.
For the case of lending structure, the expected sign is questionable. Hanson et al (2005) theoretically show that if the firm parameter comes from different sectors, there will be further scope for risk diversification by changing the portfolio weights. However, empirical findings fail to offer consistent results. On one hand, M. Kabir (1993) finds that loan portfolio diversification reduces bank risk exposure. In a similar vein, Madura et al (1994), Cebeyonan and Strahan (2004), and Blasko and Sinkey Jr (2005) show that specialization in real estate lending increases bank risk exposure. On the other hand, Nor Hayati and M. Ariff (2003) reveal that loan specialization in ‘risky sectors’(defined as the summation of BPS, purchased of securities, and consumption credit) reduces the market risk exposure, but not unsystematic and total risk exposure. As the empirical findings are inconclusive, this study seeks to answer the question of lending structure-risk relationship via four different lending structure models that will be discussed shortly afterwards.
With regards to loan default, provision for loan loss (PLL) represents ex-ante loan default while non-performing loan (NPL) represents ex-post loan default. Due to this view, some authors believe that NPL is a better measure of loan default. However, as most Islamic banks start to declare their NPLs in year 2002, we are left with no choice other than adopting PLL as a measure for loan default. As loan default could reduce earnings and dividends, PLL is expected to be positively related to bank insolvency risk exposure.
For the case of financial leverage, total equity (TE) is perceived to provide buffer against loss.
Hence, this study believes an inverse relationship exists between TE and risk. For the GAP ratio, it is well acknowledged that a positive GAP indicates that a particular bank is an asset sensitive bank while a negative GAP indicates that it is a liability sensitive bank. A positive GAP bank (or an asset sensitive bank) is exposed to risk that interest rate will fall whereas a negative GAP bank (or a liability bank) is exposed to risk that interest rate will increase. Thus, the greater the absolute value of GAP, the more the bank is exposed to changes in interest rate. Besides, the mismatch of RSA (rate sensitive asset) and RSL (rate sensitive liability) is subject to bank insolvency since bank share price is influenced by movements in interest rates, inferring a higher GAP contributes to a higher bank risk. Thus, this study hypothesizes that GAP is positively related to bank insolvency risk exposure.
Despite the GAP ratio analysis, Madura et al (1994) argue that bank risk depends on the proportion of funds obtained in the deposit account (proxied by INTEXP), which does not been captured in the GAP analysis. They underline that the higher the deposit, the higher the interest expense, the higher the volatility of net interest income, thus the riskier is the bank. Therefore, this study hypothesizes a positive relationship between INTEXP and insolvency risk exposure.
For the case of liquid asset, risk is linked to it from the perspective of deposit withdrawal. As it is well noted that having cash idle is an‘opportunity cost’, banks usually hold short term investment securities (INV) to backup for extraordinary deposit withdrawal. Furthermore, banks are restricted to hold only good investment grade securities, which have lower liquidity risk exposure. Thus, it is anticipated that INV is negatively related to bank insolvency risk.
With regards to size (LTA), majority authors argue that the greater the size, the greater will be the potential to diversify business risk from various perspectives. For instance, Sounder et al (1990) mention that the larger the bank, the more information is likely to be gathered, thus reducing information risk. They also believe that regulators are unwilling to let big banks fail, hence big banks is synonymous with low risk. In a similar vein, M. Kabir (1993) justifies that banks with larger assets are more able to diversify; but instead of looking at information risk, he focuses on operating risk that is associated with product or market lines. He believes that larger banks are more able to utilize personnel skill, particularly when engaging in off-balance sheet activities.
From a different point of view, Anderson and Fraser (2000) believe that bigger banks are more flexible to adjust unexpected liquidity and capital shortfall. Thus if loan composition is the same but differ only in term of asset size, bigger banks should have lower risk as compared to smaller banks, conjecturing an inverse relationship between size and risk. However, if the loan portfolio composition is different, the big banks overall risk might be higher than the smaller ones.
According to them, this is due to the fact that big banks have a tendency to hold riskier loan or to embark in off-balance sheet activities, thus leading to a higher overall risk. Similarly, Gonzales (2004) points out that with the existence of the economy of scale, increase market power, and the
‘too big to fail’ policy for big banks, big banks tend to enter into risky activities, which suggests a positive relationship between the two. Given this scenario, it is expected that LTA could be either positive or negatively related to bank insolvency risk exposure.
For the case of non-interest income (NONII), one way to reduce bank business risk is by diversifying from its intermediation role. The degree of banks’ involvement in non-traditional activities can be measured by non-interest income as it incorporates income from fee-based transaction, investment in financial assets, and income other than financing facilities. Previous research points out that the higher the non-interest income, the more diversified the bank is, thus the lesser the business risk. Following the previous findings, this study expects an inverse relationship between NONII and bank insolvency risk exposure.
With regards to management efficiency (MGT), recent research has conceptually hypothesized that the quality of management is related to bank risk taking behavior. This is due to the fact that management efficiency influences business activities when decision made by the management affects the bank risk and return trade-off.5 Angbazo (1997) and Nor Hayati and M. Ariff (2003) employ the ratio of earning asset to total asset as a proxy for management efficiency. Hence, an inverse relationship between MGT and bank insolvency risk exposure infers an efficient management team, and vice versa.
The Insolvency Risk Specification
Zrisk Index is a comprehensive measure of bank insolvency risk exposure. In contrast to the other risk measures that look at the volatility of ROA or ROE, the novelty of Zrisk index is that it captures the impact of capital in the sense that a bank will become insolvent if its capital exhausted.
5Refer to Angbazo (1997) and Nor Hayati and M.Ariff (2003)
Zrisk index was developed by Hannan & Hanweck (1988), who measure the perceived insolvency risk of banks. This index has been widely employed by various banking researchers such as Liang & Savage (1990), Eisenbeis & Kwast (1991), Sinkey & Nash (1993), Nash &
Sinkey (1997), Blasko & Sinkey Jr. (2005), and Ahmad et. al (2005). As the Zrisk index is constructed based on financial ratios, it is best suited for this study since it can capture all Islamic banks and windows in representing the risk taking behavior for the case of Islamic banks.6
Theoretically, Hannan and Hanweck (1988) assumed that insolvency occurs when current losses exhaust capital, thus, the probability of insolvency (Zrisk), may be obtained by noting that it is equivalent to the likelihood that:
Whereby, ROA is the return on assets and CAP is the ratio of equity capital to total assets.
Taking into consideration of the assumption that insolvency occurs when current losses exhaust capital, they show that:
represents the number of standard deviations between the expected value of the return to assets, E(ROA), and that negative values of CAP is equivalent to ROA = - CAP, which would result in insolvency. Consequently, the empirical form of Zrisk index is expressed as follows:
Where the same definition applies for E(ROA) and CAP, while σROAis the standard deviation of ROA. CAP is often used as an indicator for risk in banks because high levels of capital provide protection against large decline in income. Hence, better capitalized firms will, other things equal, incur less risk of insolvency because of loan losses, lower revenues, higher cost of funds, etc. Thus, a lower Zrisk index implies a riskier bank while a higher Zrisk implies a safer bank.
The Lending Structure Specification
In analyzing the impact of lending structure on risks, four measurements are developed; 1) the ratio of real estate sectors to total loan, 2) change in lending composition (LCC, hereafter), 3) variance of traditionality index (VART, hereafter), and 4) concentration or diversification index (SPEC, hereafter). The detail explanations of each are as follows:
a) Ratio of Real Estate Lending To Total Loan
Several studies have made an attempt to investigate the impact of real estate lending on bank risk, but unfortunately there is no standard definition of real estate lending. Madura et al (1994) and Blasko and Sinkey Jr (2005) analyze the real estate sector (RE) per se. In Malaysia, real estate sector falls under broad property sector (BPS, here after). Roza Hazli (2007) employs the ratio of lending on BPS to total loan as a proxy of real estate lending. Further, Nor Hayati and M. Ariff
6 Take note that most Islamic banks are small and not publicly traded in the stock exchange except Bank Islam Malaysia Berhad (BIMB, hereafter), which was firstly traded in the Kuala Lumpur Stock exchange (KLSE) starting 17thJanuary 1992.
(2003) incorporate the BPS into their ‘risky sector’ lending. They define lending on ‘risky sectors’ as a summation of loan distributed to BPS, purchased of securities and consumption credit. Against this background, this study employs three real estate lending measures; 1) Real estate sectors (RE); 2) Broad property sectors (BPS); and 3) ‘risky sector’ (RISKY). All measures are deflated by total asset.
b) Lending Composition Change (LCC)
LCC is used to measure the stability of lending composition in the short run. Twelve lending sectors are collected to construct lending indices representing characteristics of lending composition change. They are agriculture, hunting, forestry and fishing; mining and quarrying;
manufacturing; electricity, gas and water; broad property sectors (construction, real estate, purchase of residential landed property, and purchase of non-residential of landed property);
wholesale, retail trade, restaurants and hotels; transport, storage and communication; finance, insurance and business services; purchase of securities; purchase of transport vehicles;
consumption credit; and others. The classification of broad property sector (construction, real estate for residential and non-residential landed property) is based on the definition given by Bank Negara Malaysia. The LCC is computed as follows:
where sitis the share of broad property sectors i in total lending in year t.7 For example, if lending shares of all 12 sectors remain exactly the same, LCC will have a value of 1. On the other hand, LCC equals 0 if a bank lending sector, none of which were loaned in the previous year.
c) Specialized Index (SPEC)
SPEC is employed to measure the specialization in the lending portfolio composition. It is constructed as below:
where, sit is the lending share of industry i in total lending at year t. A score approaching 1 suggest a high degree of concentration while a score approaching 0 indicates a high degree of diversification.
d) Variance of Traditionality Index (VART)
VART is used to measure the stability of lending portfolio composition in an intermediate term.
It is calculated using five-year intervals for each sector. For example, the traditionality index for the year 1995 is computed using lending data from 1993 to 1997; for 1996, using data from 1994- 1998, and so on. The TI formula is as follows:
whereby, the cumulative lending experience (Cit) for each industry is calculated as:
7broad property sector is defined as summation of construction, real estate for residential and non- residential landed property sectors.
where t0and t1are initial and terminal periods of the data and eitis lending of industry i in year t.
Since VART is a variance of TI across sectors, a high variance indicates an episode of divergent pattern of lending during the 5 year period. Meanwhile a low variance suggests a stability of lending composition.
ANALYSIS OF RESULTS
Table 1 reports the descriptive statistics of the variables employed in this study. The mean values of RE, BPS, and RISKY show that at least more than 40% of loan are distributed to real estate lending. Table 2 presents the correlation matrix of the independent variables. BPS, RE, RISKY, LCC, SPEC, and VART are the lending structure variables, which are tested alternately. As all independent variables posit the correlation coefficient values of less than 0.8, it can be inferred that multicollinearity issue is not a serious problem in this study. Table 3 reports the results of the best panel regression model. Based on the Likelihood ratio and Hausman test, fixed effect model is the best model. Thus, the discussion of the finding is based on the fixed effect model.
The results for none effect and random effect model are displayed in Appendix 1.
Recall that Zrisk index is a safety index and lower value means “lower safety” or higher insolvency risk exposure. Hence, the intuition on the insolvency risk exposure is opposite to the sign of the relationships in table 3. For instance, a positive relationship with the Zrisk index implies an inverse relationship to the insolvency risk exposure. Referring to the ratio of real estate lending (model 1(a),(b), and (c)), the results show that an increase in real estate lending decreases the Islamic bank’s insolvency risk exposure (increases the Zrisk index). Surprisingly, this infers that real estate sectors, which are considered as‘risky sectors’ bysome authors are not applicable for the case of the Islamic bank insolvency risk exposure. With regards to the magnitude of BPS, RE, and RISKY, our finding shows that the broad property sectors depicts the highest coefficient, followed by the real estate sector, and finally the risky sectors. Intuitively, a 1% increase in BPS, RE, and RISKY would lead to a reduction in the insolvency risk exposure by 3.17%, 1.90%, and 1.60%, respectively. As the difference between BPS and RE is lending to construction sector, this infers that construction plays a crucial role in reducing the Islamic banks insolvency risk exposure. Also, the specialization index (SPEC) indicates that loan specialization reduces the Islamic banks insolvency risk exposure. This finding is consistent with the results of BPS, RE, and RISKY. As real estate lending reduces the Islamic banks’ insolvency risk, the more banks concentrate into that sector, the lower will be the insolvency risk exposure. With respect to short and medium term lending portfolio stability (LCC and VART), both are not significant. Against these unexpected findings, two potential arguments can be the reasons: First, it could be that the Islamic banks have properly approved real estate loans to customers that are less likely to default. Second, it could be that the fixed rate of the BBA contracts is affordable and appropriate especially during the 1997 financial crisis when most borrowers from the conventional banks suffer as a result of a high based lending rate (BLR).
While loan expansion (TL) is not significant, the finding for provision of loan default (PLL) merits some discussion. PLL is negatively related to insolvency risk exposure for model 1(b), model 2, and model 3. (positively related to the Zrisk index). Recall back that the inclusion of PLL as oppose to NPL as a proxy for loan default is to broaden the sample size.8 Because of the
8 Using NPL as a proxy for loan default will greatly reduce the number of observation as most of the Islamic windows start reporting NPL in 2002.
unexpected finding, we repeat the test using NPL and lag of NPL.9 Despite a limited number of observation, the finding for NPL is consistent with PLL. Interestingly, our findings for lag NPL show positive association, but not statistically significant.10 Against this background, we can infer that the positive association between lag NPL and Islamic bank insolvency risk seems to suggest that the expected positive relationship materialized in a lagged time period.
For the capital related variable (TE), our results robustly show a negative relationship between capital buffer and bank insolvency risk (positive relationship – if looking at the index). This finding conforms to the banking theory that hypothesizes when capital increases, the cushion against loss increases, thus the risk to become insolvent decreases.
With regards to interest rate variables, both GAP and INTEXP variables are not significant. For the former, it implies that the mismatch of rate sensitive asset and rate sensitive liability is not a factor to Islamic banks’ insolvency risk exposure. For the INTEXP variable, it suggests that bank insolvency risk does not depend on the proportion of fund obtained from the depositors and shareholders’ fund. In summary, the Islamic banks’ insolvency risk exposure is not sensitive to interest rate related variables.
For the liquidity variable, INV, the positive relationship infers that when Islamic banks increase their holding in short term investment, their insolvency risk exposure increases. The short term investment can be categorized into three types; namely, securities held for trading (dealing securities), securities held to maturity, and securities available for sale. Islamic banks have several types of dealing securities like Islamic accepted bills, Shahadah ad-Dayn, Bankers acceptance, and Unit trust that have different types of underlying contracts, which may increase the insolvency risk of Islamic banks. Besides, the less developed Islamic money market instruments may strengthen the risk impact due to the lack of risk management tools available for the Islamic banks to hedge against.
For the business operation related variables, size (LTA) is not a factor to Islamic banks’
insolvency risk. For NONII, an inverse relationship with the Islamic bank’s insolvency risk implies that as the Islamic banks move away from its traditional function, its insolvency risk exposure reduces. This infers that an increase in fee-based transaction and brokering activities reduces the Islamic bank insolvency risk exposure
9Findings for NPL and Lag NPL are displayed in table 3(e) and (f), respectively in Appendix 1.
10The limited number of observations may contribute to the insignificant results as most Islamic windows report its NPL starting 2002.
12 Variables
Mnemoni
c Mean Std. Dev
Risk Indicator (dependent variable)
Bank Insolvency Exposure ZRISK 12.2694 11.47985
Lending Structure
Ratio of Real Estate Sector to total loan RE 0.401439 0.306099 Ratio of Broad Property Sector to total loan BPS 0.485398 0.310472 Ratio of Risky Sector to total loan RISKY 0.568131 0.291074
Change of Lending Composition LCC 0.806747 0.205981
Degree of Specialization of Lending SPEC 0.466079 0.275984
Variance of traditionality Index VART 0.032296 0.02797
Credit related Variables
Ratio of Total Loans to Total Asset TL 0.467578 0.298844 Ratio of Provision of Loan Loss to Total
Asset PLL 0.005924 0.0092
Capital Related Variables
Ratio of Total Equity to Total Asset TE 0.098742 0.116212 Interest Rate Related Variables
Ratio of Gap to Total asset GAP -0.14752 0.345904
Ratio of Interest Expense to Total asset INTEXP 0.023001 0.014151 Liquidity Related Variable
Ratio of Short-term Investment to Total
Asset INV 0.209334 0.214718
Business Operation Related Variables
Log of Total Asset LTA 5.958006 0.895551
Ratio of Non-interest Income to Total Asset NONII 0.0162 0.015399 Ratio of Earning Asset to Total Asset EA 0.844961 0.177193
13 Notes: Correlation Matrix is based on common sample
BPS 1
RE 0.91978 1
RISKY 0.925042 0.852939 1
LCC 0.597008 0.542751 0.65488 1
SPEC 0.789725 0.747199 0.816475 0.501032 1
VART -0.31323 -0.26616 -0.37316 -0.19448 -0.31602 1
TL -0.46804 -0.43565 -0.5264 -0.28049 -0.5027 0.283796 1
PLL -0.22639 -0.22978 -0.24062 -0.10862 -0.31998 0.170607 0.42707 1
TE -0.12941 -0.13238 -0.10658 0.007434 0.181587 0.202008 0.198728 -0.01949 1
GAP -0.40242 -0.40166 -0.2956 -0.37621 -0.10192 0.276006 0.225239 0.039181 0.154484 1
INTEXP -0.02171 -0.03568 -0.04124 -0.05267 -0.09454 0.105635 0.281476 0.16713 -0.10523 0.013656 1
INV 0.147917 0.212359 0.296275 0.097691 0.325054 -0.14374 -0.47923 -0.13074 0.103644 -0.09596 -0.16105 1
LTA -0.0837 -0.15145 -0.13 0.066635 -0.47137 0.0782 0.079058 0.231115 -0.37404 -0.20711 -0.06285 -0.34689 1 NONII 0.151185 0.112906 0.264912 0.257457 0.245086 -0.43017 -0.17649 -0.07626 0.294738 -0.1736 0.033058 0.376613 - 0.18556
1
EA -0.30271 -0.16181 -0.25077 -0.17496 -0.22515 0.109455 0.350318 0.172557 0.141076 0.15328 0.234798 0.108347 - 0.10431
0.061665 1
14 Table 3: Result for fixed effect model. (Dependent variable is the Zrisk index)
Model 1(a)
Model
1(b) Model 1(c) Model 2 Model 3 Model 4
C
-0.17824 (- 0.09652)
2.190864 (1.333)
3.135337 (1.519234)
4.603214
**
(2.071691 )
1.933311 (1.313727 )
1.486281 (0.122759 )
BPS
3.171841
***
(3.42174 4)
RE
1.904869
***
(2.96303 7)
Risky
1.597425*
* (2.390668)
LCC
0.04302 (0.102607 )
SPEC
2.399466
***
(3.826118 ) VART
-37.9217 (-1.2111)
TL
0.503194 (0.87751 8)
0.046674 (0.08922 5)
0.137402 (0.217369)
-0.48189 (- 0.84851)
-0.20877 (- 0.31482)
0.073433 (0.037169 )
PLL
27.03244 (1.25581 4)
32.70282
* (1.95067 7)
30.82505 (1.62262)
38.43044
***
(2.679275 )
41.55427
***
(2.902458 )
-28.5132 (- 0.47099)
TE
111.5781
***
(22.6047 9)
106.9709
***
(27.5941 9)
108.1387*
**
(24.39377)
105.0098
***
(20.74422 )
107.4865
***
(29.78733 )
130.2652
***
(16.85821 )
GAP
1.034152
* (1.78949 6)
0.857046 (1.49518 5)
0.610922 (0.992487)
0.250907 (0.351312 )
0.22754 (0.378621 )
1.82174 (1.654041 )
INTEXP
6.941337 (0.60576 4)
9.306749 (0.98191 4)
8.457365 (0.81505)
5.083599 (0.462531 )
5.672099 (0.56441)
27.35011
* (1.885182 ) INV
- 2.33247*
- 2.23312*
-2.6797***
(-3.22371)
- 2.65914*
- 2.8897**
-2.82661 (-
15
**
(- 3.25297)
**
(- 3.01486)
**
(- 2.90317)
* (- 3.23188)
1.44645)
LTA
0.015155 (0.06186 3)
-0.20003 (- 0.83188)
-0.3545 (-1.18466)
-0.45337 (- 1.48636)
-0.21922 (- 1.02646)
0.268996 (0.141047 )
NONII
52.12042
***
(5.81082 4)
43.38507
***
(5.13360 2)
37.60172*
**
(3.90378)
46.75419
***
(4.176816 )
46.77667
***
(4.190864 )
52.63393
**
(2.091674 )
EA
-0.1422 (- 0.12986)
0.282703 (0.33959)
0.322441 (0.323746)
0.936792 (0.976251 )
0.778442 (0.896707 )
- 3.23249*
**
(- 2.83096) AR(1) 0.292681
***
(3.06597 4)
0.339116
***
(3.54164 1)
0.330328*
**
(3.645797)
0.371848
***
(4.246387 )
0.300425
***
(3.06329)
0.395188
***
(2.770079 ) R2 0.960494 0.966827 0.961544 0.964675 0.968708 0.999972 Adj R2 0.951106 0.958945 0.952406 0.956281 0.961272 0.999959 S.E. reg 2.450388 2.581173 2.560345 2.584523 2.565988 2.246942 F-stats 101.4533 117.9968 107.0357 107.1675 119.9008 945.0669
Prob (F) 0 0 0 0 0 0
D.W stat 1.644601 1.722141 1.680373 1.686569 1.685719 2.093364 Note:
1.Figures in parentheses are t-statistics
2.***, **, * denotes significant at 1 %, 5% and 10% confidence level, respectively.
3.As Zrisk index is a ‘safety index’, a high index means a low bank insolvency risk exposure; thus the relationship between independent variables and bank insolvency risk exposure is reversed from the sign in this table.
CONCLUSION
The fact that lending structure to some extent conveys a signal capable of revealing bank risk exposures is important for policy makers and practitioners. Therefore, a few recommendations with caution can be drawn. For the policy makers, three suggestions are highlighted. First, as lending to real estate sectors could reduce Islamic banks insolvency risk, the Bank Negara Malaysia should introduce special guidelines for capital adequacy standard that taking into consideration the lower impact of real estate lending on bank insolvency risk for the Islamic banks. This regulatory change should be motivated by the desire to lessen the capital constraint for the Islamic banks. In order to achieve the target set by Bank Negara Malaysia for the Islamic banks to command a 20 percent market share by year 2010, urgent action is pronounced as the Islamic banks have to compete hand in hand with the conventional banks. By incorporating a lower risk factor for real estate lending, the risk-weighted-asset (RWA) for capital adequacy standard for the Islamic banks can be reduced. Then, more loans can be distributed based on a limited capital, hence promoting the growth of the Islamic banking market share. Second, the effect of bank lending strategy may be amplified by endogeneous changes in the sector itself. For instance, if the government wishes to promote agricultural sector, several incentives such as a lower funding rate or a looser loan approval are given in that particular sector. This scenario
16 could motivate banks to change their lending portfolio composition by moving towards higher risk lending portfolio as a reaction to the erosion of bank profit resulted from a low return policy for the desired sector. Thus, policy makers must take into consideration this endogeneous factor in the decision making process. Third, as the increasing price for real estate could affect the amount of loan default, this study suggests that the issue is tackled from its roots. In minimizing the nonperforming loan, the cost for fund plays a significant role. This study advances that a separate rate should be given to: 1) the first and second onwards property buyers and 2) the local and foreign purchasers. Not only to curtail an excessive demand for real estate prices, this different rate could also help segregating the real estate speculators and borrowers who purchase properties for self-accommodation. When a lower rate is given to the first timer property buyers, presumably they want to hold the properties for good, the probability for them to default the loan would be lower, hence lower the bank insolvency risk exposures. Besides, a higher rate for more than first timer property buyers and foreigners would indirectly filter the borrowers for speculative purposes, thus reducing bank credit risk exposures.
For the practitioners, by knowing our findings, the Islamic banks should be less likely to fall victim to the generalized panic of its conventional counterparts. Specialization on real estate sector should be encouraged for the Islamic banks as it will not jeopardize the bank safety and soundness. Accordingly, the calculation formula for cost of funding real estate sectors should be revised by taking into account this lower risk impact. Instead of simply benchmarking the profit rates based on the conventional banks as well as the prevailing interest rate, the Islamic banks should be able to come out with their own benchmark. From social point of view, the objectives of Islamic banking to fulfill the goal of Maqasid Al-Syariah can be achieved when the Islamic banks are capable to reduce the profit rate charged especially to the real estate sector.
REFERENCES
Amin Gutierrez de Pineres, S. & Ferrantino, M. 1997. Export Sector Dynamics and Domestic Growth: The Case of Colombia. Review of Development Economics 3(3). 268-80 Amin Gutierrez de Pineres, S. & Ferrantino, M. 1999. Export Diversification and Structural
Dymanics in the Growth Process: The Case of Chile. Journal of Development Economics. 52. 375-91
Anderson, Ronald D, and Fraser, Donald R. Corporate Control, Bank Risk Taking, and the Health of The Banking Industry. Journal of Banking and finance 24 (8) 2000. 1383-1398.
Angbazo, Lazarus. Commercial bank net interest margin, default risk, interest rate risk, and off- balance sheet banking, 1997, Journal of Banking and finance 21. 55-87
Beaver, W. Eger C., Ryan, S. and Wolfson, M. 1989 Financial Reporting. Supplemental Disclosures and bank share Prices. Journal of Accounting Research 27. 157-178
Blasko, Matej. and Sinkey Jr, Joseph F. Bank asset Structure, real-estate lending, and risk taking.
Article in Press in The quarterly review of economics and finance. XXX. 2005XXX- XXX.
Brewer, E. 1989. Relationship between bank holding company risk and nonblank activity.
Journal of Economics and Business. 41. 337-353
17 Brewer III, Elijah, Jackson III, Willian E, and Mondschean, Thomas S. 1996 Risk, Regulation,
and S&L Diversification into Nontraditional Assets. Journal of Banking and finance 20.
723-744.
Cebeyonan, A Sinan, and Strahan, Philip E. 2004 Risk Management, Capital Structure and Lending at Banks. Journal of Banking and finance 28. 19-43.
Eisenbeis, R.A., and Kwast, M.L. 1991. Are real estate depositories viable? Evidence from commercial banks. Journal of financial services research. 5-24
Gallo, John G, Apilado, Vincent P, and Kolari, James W. 1996. Commercial Bank Mutual Fund Activities: Implications for Bank Risk and Profitability. Journal of Banking and finance 20. 1775-1791
Gonzales, Francisco. 2004. Bank Regulation and Risk-taking Incentives: An International Comparison of Bank Risk. Journal of Banking and finance 29. 1153-1184
Greene, W.H. 2003. Econometric Analysis 5thEdition (International edition). McGraw-Hill. Inc.
New York
Gujarati, Damodar N. 2003. Basic Econometrics 4thEdition. McGraw-Hill Higher Education, Singapore.
Hannan, T.H. and Hanweck, G.A. 1988. Bank insolvency risk and the market for large certificates of deposit. Journal of Money, Credit and Banking. 203-211
Hanson, Samuel, Pesaran, M Hashem, and Schuermann, Til. Firm Heterogeneity and credit diversification. Working paper in Wharton Financial Institutions Center. (accessed in June 2005).
Hsio, C. 2002. Analysis of panel data. 2ndEdition. New York: Cambridge University Press.
Konishi and Yasuda and Yasuda, Masaru, and Yasuda Yukihiro. 2004 Factors Affecting Bank Risk taking: Evidence From Japan. Journal of Banking and finance. 28.215-232
Liang, J.N. and Savage, D.T. 1990. The Nonbank activities of bank holding companies. Federal Reserve Bulletin. 280-292
Madura, Jeff, Martin, Anna D, Taylor, Don A. 1994. Determinants of implied risk of depository institutions. Applied financial economics 4. 363-370
Madura , Jeff, and Zarruk, Emillio R. Spring 1995. Bank exposure to Interest rate risk: A global Perspoective. The journal of Financial research. Vol XVIII. No 1, 1-13.
Mansor H. Ibrahim, and Ruzita Mohd amin 2004. Export Expansion, Export Structure, and Economic Performance in Malaysia. Asia Pasific Journal of Economics and Business.
Vol 7(2). 89-103.
M. Kabir Hassan. 1993. Capital Market Tests of Risk Exposure of Loan Sales Activities of Large
18 U.S Commercial Banks. Quarterly Journal of Business and Economics. 27-49.
Nash, R.C. and Sinkey, J.F. 1997. On competition, Risk, and the Hidden Assets in the Market for Bank Credit Cards. Journal of Banking andFfinance 21. 979-1013
Nor Hayati Ahmad, and M. Ariff. 2003. What factors Determine The Total Risk of Deposit- Taking Institutions? In MFA’s 5thAnnual Symposium. 888-902.
Roza Hazli Zakaria. 2007. Banks’ Securitization Involvement, Bank Lending and Bank Stability:
A Panel Data Analysis. Phd. Diss. Universiti Kebangsaan Malaysia.
Rubi, Ahmad, M. Ariff, Skully, Micheal. 2006. The Determinants of Capital ratios at Banking Institutions: Evidence From Malaysia. In the Malaysian Finance association’s 8thAnnual conference, Sabah. 575-591.
Saunders, Anthony, Strock, Elizabeth, and Travlos, Nickolaos G. June 1990. Ownership structure, deregulation and bank risk taking. Journal of finance. Vol XLV (2) 643-54 Sayrs, L. 1989. Pooled Time Series Analysis. Sage Publications. Newbury Park, Canada.
Shahida Shahimi. 2005. Penentu Margin Pendapatan Bersih bank Islam. PhD diss. University Kebangsaan Malaysia.
Sinkey, J.F.,Jr. and Nash, R.C. 1993. Assessing the Riskiness and Profitability of Credit Card Banks, Journal of Financial Service Research 7(2). 127-150
19 Appendix 1
Table 3(a): Result of None Effect Model. (Dependent variable is the Zrisk Index) Model
1(a)
Model
1(b) Model 1(c) Model 2 Model 3 Model 4
C
2.757273 (0.53284 1)
2.551484 (0.3973)
3.584613 (0.711846)
5.681493 (1.015555 )
3.912718 (0.722252 )
2.296749 (0.100652 )
BPS
1.082249 (1.11318 5)
RE
1.47037 (1.18221 1) Risky
0.563091 (0.87781)
LCC
-0.35816 (- 0.82956) SPEC
0.722722 (1.07619)
VART
-34.887 (- 0.94988)
TL
-0.1001 (- 0.13486)
0.088407 (0.12682 5)
-0.33154 (-0.46604)
-0.50214 (- 0.67117)
-0.42898 (- 0.61817)
-2.68117*
(- 1.77433)
PLL
3.0319 (0.11262 3)
5.918204 (0.22637 6)
7.521404 (0.296894)
5.785053 (0.216792 )
11.24445 (0.434996 )
-36.6639 (- 0.76362)
TE
103.5535
***
(15.7463 2)
103.4843
***
(14.9867 5)
102.9076*
**
(16.48442)
103.1579
***
(15.32675 )
102.8334
***
(16.55571 )
136.6873
***
(17.82115 )
GAP
0.324478 (0.79996)
0.396473 (0.77646 1)
0.41122 (0.918045)
0.560203 (1.133764 )
0.480273 (0.906103 )
0.040298 (0.042402 )
INTEXP
12.3201*
* (2.18575 3)
12.37778
***
(2.14792 5)
12.62426*
* (2.210264)
12.90318
**
(2.216088 )
12.15086
**
(2.087506 )
12.4858 (1.27932)
INV
- 2.4417**
* (- 2.64698)
- 2.36399*
* (- 2.50463)
- 2.63399**
* (-2.79148)
- 2.74031*
**
(-3.1433)
- 2.65136*
**
(- 2.85415)
- 4.3715**
* (- 2.68359)
LTA
-0.05427 (- 0.08964)
-0.00681 (- 0.00901)
-0.19165 (-0.32569)
-0.38638 (- 0.60734)
-0.24853 (- 0.39342)
-1.16502 (- 1.00088) NONII 50.67666 50.99582 48.16027* 50.32563 50.30068 23.67688
20 Note:
1. White cross-section heteroskedasticity-consistent covariance matrix estimators are reported.
2. Figures in parentheses are t-statistics
3. ***, **, * denotes significant at 1 %, 5% and 10% confidence level, respectively.
4. All independent variables are ratios to total asset except for total asset which is expressed in log.
5. As Zrisk index is a ‘safety index’, a high index means a low bank insolvency risk exposure; thus the relationship between independent variables and bank insolvency risk exposure is reversed from the sign in this table.
***
(5.05058 6)
***
(5.13939 8)
**
(4.692439)
***
(4.674299 )
***
(4.656297 )
(1.176571 )
EA
-0.29464 (- 0.22661)
-0.67564 (- 0.54551)
-0.02292 (-0.01775)
-0.00948 (- 0.00698 )
0.012403 (0.009894 )
-1.11362 (- 1.65702) AR(1) 0.898462
***
(27.1899 2)
0.902403
***
(28.8111 8)
0.90136**
* (27.83914)
0.903031
***
(29.87512 )
0.903248
***
(29.07097 )
1.02437*
**
(27.22543 ) R2 0.943281 0.949906 0.945317 0.947613 0.947424 0.972045 Adj R2 0.937808 0.945072 0.940041 0.942558 0.942351 0.967003 S.E. reg 3.072034 3.070015 3.109556 3.100947 3.109924 2.399608 F-stats 150.6233 156.6233 156.3926 153.5111 157.1381 165.0998
Prob (F) 0 0 0 0 0 0
D.W stat 1.874895 1.898489 1.894483 1.884879 1.897109 2.199414
21 Table 3(b):
Result of Random Effect Model.
(Dependent variable is the Zrisk Index) Model
1(a)
Model
1(b) Model 1(c) Model 2 Model 3 Model 4
C
- 20.351**
* (- 5.11167)
- 12.443**
* (- 3.75944)
- 19.4788**
* (-3.4894)
-7.53896 (- 1.25668)
- 17.4717*
**
(- 2.79895)
-11.4009 (-0.8922)
BPS
7.752755
***
(3.40936 7)
RE
4.989469
* (1.69018 4)
Risky
7.061447*
**
(2.827264)
LCC
1.414379 (0.815635 )
SPEC
6.142205
**
(2.386292 )
VART
-95.04***
(- 2.87725)
TL
1.365432 (0.74616 3)
0.957449 (0.67915 7)
1.577932 (0.962543)
-0.10913 (- 0.06329)
0.52873 (0.304048 )
-0.64571 (- 0.21171)
PLL
17.05255 (0.41376 3)
18.82374 (0.86724 2)
15.84353 (0.451421)
37.56188
**
(2.390535 )
26.64745 (1.306201 )
11.77534 (0.155126 ) TE 111.5887 112.2477 113.9138* 111.7011 110.7969 129.7622
22 Note:
1. White cross-section heteroskedasticity-consistent covariance matrix estimators are reported.
2. Figures in parentheses are t-statistics
3. ***, **, * denotes significant at 1 %, 5% and 10% confidence level, respectively.
4. All independent variables are ratios to total asset except for total asset which is expressed in log.
5. As Zrisk index is a ‘safety index’, a high index means a low bank insolvency risk exposure; thus the relationship between independent variables and bank insolvency risk exposure is reversed from the sign in this table.
***
(8.83307 1)
(9.03720 9)
**
(9.155996)
***
(8.732154 )
***
(8.724232 )
***
(13.12996 )
GAP
0.575774 (0.23155 1)
1.141598 (0.46336 3)
0.284181 (0.118278)
1.796148 (1.125095 )
0.433199 (0.188677 )
-0.66649 (- 0.39927)
INTEXP
-16.4732 (- 0.50354)
-8.85585 (- 0.25729)
-9.58262 (-0.2884 )
-8.63636 (- 0.25344)
-3.16885 (- 0.09515)
0.022567 (0.001305 )
INV
0.382021 (0.19549 6
-0.38368 (- 0.18509)
0.630652 (0.24541)
-1.23291 (- 0.53965)
0.108761 (0.043107 )
-4.75883 (- 1.34537)
LTA
2.885948
***
(6.85990 8)
1.979192 (5.46750 2)
2.673124*
**
(5.121446)
1.261117
**
(2.086494 )
2.538534
***
(3.744918 )
2.945817 (1.352935 )
NONII
116.4862
***
(4.55929 1)
91.87582 (3.04852 5)
75.27675*
**
(5.169271)
73.34205
***
(5.41891)
82.34414
***
(5.100074 )
82.19883
***
(3.604664 )
EA
-1.06324 (0.7413)
-1.13984 (- 0.34939)
-0.72405 (-0.23939)
0.44829 (0.159606 )
-0.09531 (- 0.03284)
-3.20411 (- 1.40948) R2 0.74095 0.771646 0.754172 0.793767 0.782851 0.827561 Adj R2 0.721023 0.75408 0.735262 0.777903 0.766147 0.804872 S.E. reg 4.790347 4.323402 4.569437 4.008702 4.175699 2.707852 F-stat 37.18332 43.92914 39.88252 50.03554 46.86672 36.47357
Prob (F) 0 0 0 0 0 0
D.W stat 0.62208 0.759585 0.744599 0.913495 0.874807 0.827561
24 Table 3(c): Results for Likelihood Ratio Test for Zrisk Index
Model
Statistic
Degree of
freedom Probability
Model 1 (a) Statistic F 32.31715 -13,117 0
Statistic Chi-square 214.8915 13 0
Model 1 (b) Statistic F 30.75129 -13,117 0
Statistic Chi-square 209.4439 13 0
Model 1 (c) Statistic F 30.65666 -13,117 0
Statistic Chi-square 209.1078 13 0
Model 2 Statistic F 30.64516 -13,117 0
Statistic Chi-square 209.0669 13 0
Model 3 Statistic F 30.11815 -13,117 0
Statistic Chi-square 207.18 13 0
Model 4 Statistic F 29.65776 -13,63 0
Statistic Chi-square 170.7712 13 0
Table 3(d): Hausman Test using Wald Coefficient
Model Chi-Sq.
Statistic
Degree of
freedom Probability
Model 1 (a) 10.28541 10 0.4158
Model 1 (b) 16.28085 10 0.0919
Model 1 (c) 18.00273 10 0.0549
Model 2 14.11564 10 0.1678
Model 3 35.33311 10 0.0001
Model 4 11.71213 10 0.3048
25 Table 3(e): Results for Fixed Effect Model: Using NPL instead of PLL
Model 1(a)
Model
1(b) Model 1(c) Model 2 Model 3 Model 4
C -0.11776 2.1135 3.190669
4.455915
** 1.944565 2.142234 BPS
3.223636
***
RE
1.981747
***
Risky 1.600533*
LCC 0.119595
SPEC
2.438469
***
VART -37.1279
TL 0.442337 -0.04791 -0.02645 -0.5952 -0.21764 -0.59715
NPL 1.663095
2.287429
**
2.519403*
*
3.468761
**
2.793282
** -1.89056
EA -0.28035 0.184859 0.262786 1.094804 0.733952
- 2.60657*
* GAP 0.892288 0.709268 0.450931 0.177119 -0.02594 1.491078 INTEXP 2.453541 4.416441 0.955527 -1.59261 0.578469
28.45446
**
INV
- 2.60305*
**
- 2.58613*
** -3.0312***
- 3.18347*
** -3.185*** -3.11409 LTA 0.080706 -0.11407 -0.2735 -0.38141 -0.15635 0.146334 NONII
53.63003
***
45.07288
***
38.90965*
**
49.15564
***
48.3426*
**
42.3323*
**
TE
113.376*
**
109.199*
**
110.0529*
**
107.0485
***
110.3559
***
131.2198
***
R2 0.957971 0.96313 0.95833 0.960086 0.961176 0.994211
Adj R2 0.947679 0.9541 0.948125 0.950311 0.951668 0.991255 S.E. reg 2.452431 2.58699 2.566659 2.593918 2.564893 2.282605 F-stat 93.07241 106.6656 93.90877 98.22056 101.0914 336.3221
Prob (F) 0 0 0 0 0 0
D.W stat 1.434748 1.530854 1.484983 1.496557 1.484706 1.851967 Table 3(f): Results for Fixed Effect Model: Using lag NPL instead of PLL.
Model 1(a)
Model
1(b) Model 1(c) Model 2 Model 3 Model 4
C 2.641702 5.848545 6.267801 6.225302 3.314767 1.056762
BPS
4.654146
***
RE
2.807191
***
Risky
3.04028**
*
LCC 0.007752
SPEC
3.296893
***
26
VART -46.7959
TL 0.607046 0.499675 0.905026 0.008871 0.362788 0.029603 NPL (-
1) -6.40823 -4.11695 -4.1874 -4.80097 -4.66408
- 12.4606*
**
EA
120.7119
***
120.2188
***
119.6884*
**
120.4423
***
124.811*
**
129.9788
***
GAP
1.835082
*** 1.393487 0.96465 0.761011 0.763534
1.997896
* INTEXP 19.9866
24.16145
* 13.68579
21.79263
*
22.11627
*
33.0479*
*
INV
- 2.86512*
**
- 2.60028*
* -3.1417*** -2.74403*
- 2.97121*
** -2.62246 LTA -0.59545 -0.9115 -0.99687 -0.79874 -0.56381 0.349469 NONII
42.44091
***
30.41655
* 25.33281
31.16859
*
28.88237
**
40.50143
*
TE -1.13069 -1.00945 -1.05511 -0.22716 -0.8786
- 2.57274*
* R2 0.965018 0.970763 0.965646 0.971174 0.966432 0.989912 Adj R2 0.953357 0.961018 0.954194 0.961566 0.955243 0.984868 S.E. reg 2.393709 2.571675 2.495786 2.619676 2.486708 2.257466 F-stat 82.75855 99.61132 84.32529 101.0744 86.37165 196.2533
Prob (F) 0 0 0 0 0 0
D.W stat 1.456355 1.482627 1.473627 1.464029 1.511668 2.070598