Risk Issues in Islamic Banks

Overall risks faced by Islamic banks   The first factor to be examined is the respondents' perceptions about the severity of risk facing IFIs. Descriptive statistics for Question 7, in Chapter 7, showed that Islamic and conventional bankers share similar views about the top risks facing IFIs, unlike non-bankers who adopted a more theoretical approach in their views. This section will investigate further to examine the difference in perceptions among different subgroups of respondents. For this purpose, the researcher has employed the Kruskall-Wallis (K-W) test for region, country, respondent's position, nature of Financial Institution, nature of activities and accounting standards

The first control variable is ‘Region’. The results from the K-W test for the entire research sample in Table 8.1 indicate that there is no statistically significant difference among various regions in risk perception (p-value > 0.05) except for corporate governance risk (p-value = 0.002), which is also evident from the mean ranking. With a ‘relaxation’ of the confidence level to 0.06, we can accept displaced commercial risk as significant as well.

Repeating the K-W test with ‘Region’ as the control variable for different samples of data, in terms of the institutional setting of respondents, gives consistent results, as illustrated by Table 8.2, which confirms that there is a difference in risk perception of corporate governance risk among regions for fundamental market reasons. In other words, there is a significant difference between regions when institutional settings were also considered.

In addition, examining the mean rankings across different regions for corporate governance risk confirms the existence of a structural pattern. As apparent from Table 8.3, the rankings do not change much when conducting K-W with different samples identifying different institutional settings. The inclusion of conventional banks and non-bankers in the test sample gives similar results. ‘Americas’ disappear when conventional banks are excluded from the test sample as there were no respondents from IFIs in the ‘Americas’ in this research sample. Also, the difference in values between the highest and the lowest mean rankings is noticeable, which confirms that the distribution of corporate governance risk is significantly different across regions.

There is a pattern regardless of the nature of the respondents included in the sample, which implies that there are structural issues determined by the nature of the market, which can be explained by fundamental market reasons. Although corporate governance practices have material impacts on a bank's risk profile, IFIs do not generally have robust corporate governance frameworks in place particularly in the Gulf Cooperation Council (GCC), Middle East and Southeast Asia.

The same pattern could be identified, although to a lesser extent, when examining concentration risk, one of the main risks identified by respondents, as explained in the previous chapter. Table 8.4 confirms that there are fundamental market reasons for the difference in mean ranking among different regions. The mean ranking for the K-W test for the full sample ranks ‘Southeast Asia’ first (51.25), followed by ‘Other’ (51), then ‘Americas’ (41), while ‘Other Middle East’ comes last with mean rank of 28.21. This ranking changes little when conducting the K-W test for different samples using different institutional settings, which confirms that for concentration risk there is a significant difference between regions when institutional settings are also applied.

Furthermore, examining the mean rankings across different raw data for other significant risks like credit and liquidity risks (as identified by the respondents in Chapter 7) shows that rankings remain very similar between fully fledged Islamic banks and fully fledged Islamic banks combined with Islamic subsidiaries of conventional banks. However, adding conventional banks with no Islamic activities to the sample changes the rankings slightly as summarised in Tables 8.5 and 8.6. Under credit risk, for instance, when only fully fledged Islamic banks are included in the sample, ‘Southeast Asia’ ranks first (15.63), followed by ‘Other Middle East’ (13.2), ‘Europe’ (13.1), ‘GCC’ (13) and ‘Other’ (7). Also, the difference in values between the mean rankings is minimal, reflecting the close perception among different regions. When the institutional sample settings change to include Islamic subsidiaries as well, this pattern of mean rankings remains very similar. However, changing the institutional sample settings to include conventional banks changes the rankings and the gap between mean values becomes wider. Of note is the existence of the same pattern when non-bankers are also included in the sample. This shows that for credit risk there is a difference between regions when conventional banks and other non-banking respondents are also considered. Islamic and conventional bankers have different risk perceptions about credit risk across various regions.

Table 8.6 shows that the same trend exists for liquidity risk. K-W test results for different institutional samples indicate a similar pattern between samples of fully fledged Islamic banks and fully fledged Islamic banks combined with Islamic subsidiaries. Also, there is another similar pattern between the full sample and a sample comprising fully fledged Islamic banks, Islamic subsidiaries and conventional banks. This emphasises that Islamic and conventional bankers have different risk perceptions about liquidity risk across various regions, while the perceptions of Islamic subsidiaries is the same as that of fully fledged Islamic banks.

The findings indicate that there is an observed pattern which can be generalised to most of the risk categories. This can be explained only by market realities.

The K-W test was conducted in a similar manner according to ‘country’ as control variable; the results confirm those produced by the test conducted according to ‘Region’.

In addition, an attempt was made to test the impacts of ‘respondent's position’ and ‘accounting standards’ on risk perception; however, the results show that there are no significant differences, as summarised in Table 8.7.

Finally, conducting the K-W test to examine the significance of perceived differences among various risk groups for the entire research sample according to the ‘Nature of Financial Institution’ provided dispersed results. Table 8.8 shows that liquidity, ALM, Shari'ah-non-compliance, concentration, reputation and displaced commercial risks have significant p-values, while the remaining risks do not.

Further examination of the mean rankings for risks with significant p-value, as summarised in Table 8.9, confirms the dispersion of data as no trend could be established. In general, fully fledged Islamic banks and conventional banks with Islamic activities have higher mean values than conventional banks alone and ‘Others', particularly for liquidity, Asset-Liability Management (ALM) and displaced commercial risks. This trend, nonetheless, slightly changes for concentration and reputation risks. Also of note is the proximity of mean value among fully fledged Islamic banks and Islamic subsidiaries, which reflects the similar perception of risks in Islamic banking. One possible reason for this is the similar knowledge and awareness of Islamic banking products and structures among those professionals with hands-on experience in Islamic banking. This confirms the findings of the section ‘Locating Risk Perception’ in Chapter 7.

Based on the above results, it can be concluded that three control variables (region, country and nature of FI) contribute to some significant differences about risk perception among respondents, but not for all risks. In addition, this can also be supported by the fact that there is no significant difference in perception levels between respondents from stand-alone Islamic banks and Islamic subsidiaries. Initially it was expected that respondents from stand-alone Islamic banks would have stronger perceptions compared to those from Islamic subsidiaries for two reasons: firstly, stand-alone Islamic banks have been in existence for much longer than Islamic subsidiaries, and, secondly, respondents from stand-alone Islamic banks have the advantage of dealing with only Islamic banking products and services, whereas Islamic subsidiaries still need to operate side-by-side with their respective conventional counterpart in sharing the same operating platforms and buildings. Nevertheless, the results have indicated otherwise. Differences could be spotted between perceptions of conventional banks and stand-alone Islamic banks, and more noticeably between the perceptions of bankers and non-bankers, represented by ‘Others’. This could be because bankers, whether Islamic or non-Islamic, have hands-on experience and better understanding of the Islamic banking model and its risk architecture than non-bankers, who tend to be more theoretical in their approach.

Islamic finance contracts   Questions 9 and 10 seek respondents' views on various Islamic modes of financing. Question 9 targets institutions that use Islamic finance contracts only. Therefore, when conducting the K-W test for Question 9, only stand-alone Islamic banks and Islamic subsidiaries were included in the data analysis.

Intensity of use of different Islamic finance contracts   Table 8.10 shows that regardless of the respondent's position or the nature of activities, banks use Islamic finance contracts in similar patterns; all products had p-value > 0.05. However, K-W test results according to ‘Region’ indicate that there is significant difference in the use of mudarabah across different regions. Moreover, there is significant difference in the use of wakala and salaam according to the nature of financial institution.

The Friedman test is used to find a tendency for some variables to receive higher ranks than others, i.e. to test whether the ranking is significant or not. The results of the test reflect that ranking for this question is significant.

As can be seen in Table 8.11, ‘Other Middle East’ and ‘Southeast Asia’ use mudarabah the most, with mean values of 32.5 and 27.13 respectively, while ‘Europe’ (17.71) and the ‘GCC’ (16.74) rank lower on the use of mudarabah as financial institutions in these regions tend to rely more on murabahah, wakala and ijarah. This should be explained by the economies of the regions in question, as the lack of financial depth may necessitate greater use of equity financing.

As depicted by Table 8.12, Islamic subsidiaries (25.21) tend to use wakala to a greater extent than do fully fledged Islamic banks (17.08), while the picture is reversed for the use of salaam, where the disparity between mean values of the two groups is wide

In addition, Question 9 was re-tested, excluding Islamic subsidiaries from the sample. However, there were no significant differences between the use of different contracts across the different control variables: region, respondent's position, nature of activities and accounting standards.

Risk perception for different Islamic finance contracts   Unlike Question 9, which targeted financial institutions using Islamic finance contracts, Question 10 seeks risk perceptions for these contracts. The feedback of all respondents is valuable; therefore the K-W test is conducted on the entire research sample.

As can be seen in Table 8.13, the results of the Friedman test reflect that ranking for this question is significant, indicating that there is a significant difference between risk perceptions of Islamic contracts.

K-W test results according to ‘region’, as illustrated in Table 8.13, indicate that murabahah is the only contract that reflects significant results across regions. This is expected because murabahah is extensively used globally. Moreover, mean rankings for murabahah, shown in Table 8.14, show that ‘Other’ regions, like Turkey and Pakistan, have a higher ranking (54.0) than the ‘GCC’ (43.13) and Europe (38.63), while the remaining regions follow. This can be attributed to two main reasons. First, the European and GCC markets are more sophisticated in their financial awareness of risk management, product structures and the use of risk-hedging techniques than are Turkey and Pakistan, a fact which has a direct impact on risk perception among those markets. Second, at the time of conducting this questionnaire, European and GCC markets enjoyed stable political environments and ‘relatively’ less volatile business cycles compared to ‘Others’.

Repeating the K-W test with ‘Region’ as the control variable for different institutional samples of data gives consistent results, as depicted by Table 8.15, which confirms that there is a difference in the risk perception of murabahah among regions in comparing according to institutional nature for fundamental market reasons.

In addition, examining the mean rankings across different regions for murabahah confirms the existence of a structural pattern. As apparent from Table 8.16, the rankings are similar when conducting the K-W with different raw data. The inclusion of conventional banks and non-bankers in the test sample gave similar results. The region ‘Americas’ disappears when conventional banks are excluded from the test sample as there were no respondents from Islamic banks in the Americas in this research sample.

As depicted by Table 8.16, there is a general pattern in terms of perception of murabahah-related issues. Such regional and institutional differences can be attributed to market conditions prevailing in each region.

Furthermore, using the entire research sample, attempts were made to test the impact of the respondents' positions, nature of financial institution, nature of activities and accounting standards on risk perception. However, the results, as depicted in Table 8.17, show that there are no significant differences except for murabahah contracts, which had significant risk perception according to accounting standards (p = 0.028), and nature of financial institution (0.03).

Additional risk issues facing IFIs   Question 11 aimed at exploring the perceptions of the participants in relation to a number of risks-related statements. For this, the K-W test was employed to determine if there were any statistically significant differences across the categories of respondent profiles.

Table 8.18 shows the K-W test results for the ‘Nature of financial institution’ variable. Statements 1, 2, 3, 4, 8, 10 and 11 are statistically significant, which reflects that there are significant differences in risk perception among respondents according to the nature of their financial institution. It should be noted that insignificant categories are eliminated and hence are not depicted in the table. Table 8.18 also breaks down the mean rankings for these statements.

Statements:

1. Risks for Islamic banks should be managed using the same techniques used in conventional banking.
2. Islamic banking is more risky by nature than conventional banking.
3. Risk management for Islamic banks is more challenging than it is for conventional banks.
4. There is naturally inherent conservatism in the principles of Islamic finance.
5. In an Islamic bank, a low rate of return on deposits will lead to withdrawal of funds.
6. Depositors would hold the bank responsible for a lower rate of return on their deposits.
7. Variation among Shari'ah scholars' opinions represents a major risk to Islamic banking.
8. Shari'ah-non-compliance could severely damage the reputation of an Islamic bank.
9. AAOIFI and IFSB standards should be made mandatory for Islamic banks.
10. Corporate governance is generally weak in Islamic banks.
11. Islamic banking in its current state is a safer option than conventional banking.

Studying the mean ranking for each statement does not reveal a certain pattern governing the data; the data is widely dispersed with no clear trend of ranking according to nature of financial institution.

Repeating the K-W test for the entire research sample using other control variables (such as region, position of respondent, nature of activities and accounting standards) gives similar results; as seen in Table 8.19, which confirms that there is a significant difference in risk perception among various groups. With a ‘relaxation’ of the significance level to 0.06, more statements can be considered significant. An attempt was made to study the mean ranking for each statement within each test; however, the results did not reveal a certain pattern, and the data is widely dispersed with no clear ranking trend.

Factor analysis for Question 11   In order to provide further statistical robustness to the analysis, factor analysis was conducted. Factor analysis seeks to discover if the observed variables can be explained largely or entirely in terms of a much smaller number of variables called ‘factors’.

As there are 11 statements for Question 11, all analysing the respondents' perceptions of different risk issues in Islamic banking, the researcher felt that reducing these statements into a more manageable number would enhance the analysis and would tell more about how respondents perceived these issues. Hence, factor analysis is deemed to be relevant in this respect as the main task of factor analysis is to cluster the related group of variables through their common variance (Pallant, 2007).

In order to test the factorability of the data in terms of sampling adequacy, there are two statistical measures available in the SPSS software that can be used: Bartlett's test of sphericity and the Kaiser-Meyer-Olkin (KMO) test. As laid down in Pallant (2007), for the factor analysis to be considered appropriate, the Bartlett's test of sphericity value should be significant (p < 0.05), while for the KMO test, the suggested minimum outcome must be at least 0.6 (KMO score ranging from 0 to 1). The KMO test's benchmarks are as follows: for a KMO measure in the 0.90s the sampling is considered marvellous. If the outcome is in the 0.80s, then the sampling is considered meritorious, if it is in 0.70s then the sample is middling, if it is in the 0.60s then the sample is mediocre, if it is in 0.50s then the sample is deemed miserable and lastly if it is below 0.50 then the sample is unacceptable (Pallant, 2007).

Table 8.20 presents the results of KMO and also Bartlett's test for this factor analysis.

The outcome of the KMO measure for all 11 items combined, related to risk perception, produced the value of 0.760, which is higher than 0.60, therefore the factor analysis is appropriate for this study. In addition, the significant p-value as presented in the table of 0.000 is significantly lower than critical p-value of 0.05. Therefore, the identity matrix can be rejected. Based on the very encouraging results from both tests, factor analysis may be performed.

The second step is to choose the most suitable method of data extraction. As discussed in Chapter 6, the researcher selected principal component analysis (PCA) as it is deemed the most suitable method for the data at hand. PCA involves determining the patterns with the objective of studying the similarities and the differences among the components of the data set.

After determining the factors, the next step in order to facilitate the interpretation selection of rotation method is very important. In this regard, orthogonal (uncorrelated) and oblique (correlated) approaches are the two main techniques to rotation (Pallant, 2007). The results of the orthogonal rotation are easier to interpret, describe and report (Field, 2009). There are various rotational approaches in SPSS within both the orthogonal and oblique categories. Varimax, Quartimax and Equamax are the typical orthogonal approaches of rotation, whereas Direct Oblimin, Quartimin and Promax are the oblique methods. Varimax is the most commonly used orthogonal technique in order to reduce the number of variables whereas the Direct Oblimin technique is generally used for the oblique method. The researcher opted for Varimax rotation with Kaiser Normalization, as Table 8.21 suggests.

Table 8.21 presents the output of the number of factors that are retained according to Kaiser's criterion, in which all the eigenvalues are more than 1.0. In this situation, there are three factors that will be retained, since the eigenvalues are 3.732, 1.887 and 1.439 respectively. The screen plot, which is basically a graph of the eigenvalues, shows that the 11 variables could be reduced to only three as the graph slopes down steeply before becoming parallel to the horizontal line. It is therefore clear from the plot that there is only a three-factor solution to this question. Hence it was decided to retain the three factors.

According to Pallant (2007), the eigenvalue has to be greater than 1.0 to be regarded as significant and to be used in determining the factors. The assumption here is that the eigenvalues stand for the amount of total variation represented by the factors and this means that an eigenvalue of 1.0 or above indicates a high level of variation. Table 8.21 shows that there are three factors with an eigenvalue greater than 1.0. This means that the original 11 items can be simply reduced to three factors. The three-component solution explained 64.2% of the variance with component 1 contributing 33.9%, component 2 contributing 17.1% and component 3 contributing 13.1%. The explanatory power of the first factor is very high.

Table 8.22 further provides Rotated Component Matrix by distributing all variables to the identified three components. The factors in each component have some common characteristics and measure the same phenomenon, and therefore each component is named with a general description of the factors or variables it includes. For instance, factors in component 1 deal with the respondents' risk perception. The factors in component 2 deal with Shari'ah principles and their impact on the risk profile of an IFI, while the factors in component 3 deal with the rate of return paid by an IFI and the effect of this on depositors' behaviour and perception of how safe the IFI is. Thus the results indicate that all these statements can be explained with three main components. Figure 8.1 provides a screen plot of the factor analysis results for Question 11.

After conducting factor analysis between groups, a MANOVA test was computed in order to investigate if there is any significant difference between the three component groups in relation to the same control variables. This will help to locate the impact or significance of each control variable on the established distribution.

The outputs of the relevant tests are presented in Tables 8.23 to 8.25 in terms of data conforming to the assumptions before the main MANOVA analysis. In this sense, the sig. value of Box's Test of Equality of Covariance Matrices should not be lower than 0.001 in terms of not violating the assumption (Tabachnick and Fidell, 2006). In this example, the output of Box's Test shows that there is no violation of assumption of homogeneity of variances of variance-covariance matrices since the sig. value of 0.248 is higher than the critical value of 0.001.

Additionally, the output of the Levene's Test of Equality of Error Variances is explored. The results in the Sig. column show that sig. values of ‘Risk Perception’ (0.806), ‘Shari'ah Compliance’ (0.121) and ‘Rate of Return’ (0.112) are higher than 0.05. Thus, there is no violation of the assumption of equality of variances for these three factors.

After performing Box's Test of Equality of Covariance Matrices and Levene's test, the set of multivariate tests was employed. Pallant (2007) states that multivariate tests of significance demonstrate if there are any significant differences among groups; the sig. value should be lower than 0.05 in order to find a statistically significant result. There are several statistics which are also used in the SPSS such as Pillai's Trace, Wilks' Lambda, Hotelling's Trace and Roy's Largest Root. In this research Wilks' Lambda result is taken into account since it is one of the most commonly used statistics (Tabachnick and Fidell, 2006). The results of the Wilks' Lambda show that there is a statistically significant difference between regions in relation to the perceptions of the three components since the sig. value of 0.00 is quite a bit lower than the critical level of 0.05.

Since multivariate tests suggest that there is a statistically significant difference, a further investigation is conducted. This is in order to reveal if there is a difference in terms of region on ‘Risk Perception’, ‘Shari'ah Compliance’ and ‘Rate of Return’, or only to some extent. Tests of Between-Subjects Effects provide this information. Bonferroni adjustment, which is one of the most commonly employed methods, gives this information when the alpha level of 0.05 is divided by the number of dependent variables (Pallant, 2007). In this example, there are three dependent variables, therefore 0.05 is divided by three and the new alpha level is 0.0167. As can be seen in the Tests of Between-Subjects Effects in Table 8.26, the results indicate that the dependent variables ‘Risk Perception’ and ‘Shari'ah Compliance’ have significant values of 0.000, while ‘Rate of Return’ has a sig. value of 0.671, which is higher than the critical value of 0.0167 for this example.

Furthermore, Tests of Between-Subjects Effects provide the effect size. Partial Eta Squared is used to determine the impact of independent variables on dependent variables, and it signifies the percentage of the variance in the dependent variable which is explained by the independent variable (Pallant, 2007). In this question, the effect of ‘Region’ (independent variable) on ‘Risk Perception’ and ‘Shari'ah Compliance’ (dependent variables) can be evaluated by the Partial Eta Squared, which is depicted in the Tests of Between-Subjects Effects in Table 8.26. The importance of this impact is explored using the effect size values. Cohen (2005) categorises an effect size of 0.01 as a small effect and 0.06 as a medium effect whereas 0.14 is a large effect.

The effect size values for this case are 0.459 and 0.344, which are deemed large effect sizes using Cohen's. These results signify that 45.9% and 34.4% of the variances in ‘Risk Perception’ and ‘Shari'ah Compliance’ scores are explained respectively by the region.

MANOVA test according to nature of financial institution for Question 11   After conducting a MANOVA test with ‘Region’ as the independent variable, another MANOVA test was computed with ‘nature of financial institution’ as the independent variable in order to investigate if there is any significant difference between the three dependent factors identified by the factor analysis.

In this case, the output of Box's Test, as shown in Table 8.27, shows that there is no violation of the assumption of homogeneity of variances of variance-covariance matrices since the sig. value of 0.080 is higher than the critical value of 0.001.

Additionally, the output of the Levene's Test of Equality of Error Variances is explored in Table 8.28. The results in the Sig. column show that sig. values of ‘Risk Perception’ (0.753), ‘Shari'ah Compliance’ (0.427) and ‘Rate of Return’ (0.077) are higher than 0.05. Thus, there is no violation of the assumption of equality of variances for these three factors.

The results of the Wilks' Lambda in Table 8.29 show that there is a statistically significant difference according to nature of financial institution since the sig. value of 0.00 is quite a bit lower than the critical level of 0.05.

Since the multivariate test suggests that there is a statistically significant difference, a further investigation is conducted. Tests of Between-Subjects Effects provide this information. In this case, there are three dependent variables, therefore 0.05 is divided by three and the new alpha level is 0.0167. As can be seen in the Tests of Between-Subjects Effects in Table 8.30, the results indicate that the dependent variables ‘Risk Perception’ and ‘Shari'ah Compliance’ have significant values of 0.000, while ‘Rate of Return’ has a sig. value of 0.234, which is higher than the critical value of 0.0167 for this example. Furthermore, the effect size values as evaluated by the Partial Eta Squared for this case are 0.301 and 0.336, which are deemed large-effect sizes using Cohen's. These results signify that 30.1% and 33.6% of the variances in ‘Risk Perception’ and ‘Shari'ah Compliance’ scores are explained respectively by the nature of financial institution.

Conducting the MANOVA test according to ‘Region’ and ‘Nature of Financial Institution’ as independent variables provided consistent results. It can be concluded that ‘Risk Perception’ and ‘Shari'ah Compliance’ are significant dependent variables and have strong explanatory power, while ‘Rate of Return’ does not follow the pattern.

Capital Adequacy for Islamic Banks

This section addresses capital adequacy challenges facing IFIs. It tackles the controversial issues regarding the applicability of the Basel II and Basel III Accords to IFIs, and the appropriate capital requirement levels for Islamic banks.

The results of the K-W test in Table 8.31 show that all statements are statistically significant (p-value < 0.05) except for Statement 5 (p-value = 0.358) implying that regional differences in relation to capital adequacy are significant.

Statements as depicted in the following tables and their coding are:

1. Basel II standards should be equally applied to Islamic banks without modification.
2. IFSB standards on Capital Adequacy should be used by Islamic banks rather than Basel II.
3. Basel II standards should be reviewed after failing to prevent the financial crisis.
4. The proposed Basel III rules would be easily applicable to Islamic banks.
5. Stricter capital, leverage and liquidity rules, as proposed under Basel III, are likely to prevent another financial crisis.

Conducting the K-W test with ‘Nature of Financial Institution’ as the control variable for the entire research sample gives different results, as illustrated by Table 8.32. All statements are statistically insignificant except Statement 5, which shows different views between bankers (whether Islamic or conventional) and non-bankers, which is also evident from the mean ranking. This implies that the nature of financial institution is not a statistically determining factor; and that the opinions of the respondents are rather similar.

Furthermore, repeating the K-W test with ‘Nature of Activities’ and ‘Respondent's Position’ as control variables for the entire research sample gives results consistent with those of K-W according to ‘Nature of Financial Institution’, as illustrated by Tables 8.33 and 8.34 respectively. For ‘Nature of Activities’, statements are statistically insignificant except for Statements 1 and 3, while as depicted by Table 8.34 for ‘Respondent's Position’, all statements are statistically insignificant except Statement 5. This reflects the difference in opinions among different groups regarding the newly developed Basel III capital and liquidity standards and their applicability to Islamic banking.

The Credit Crisis and Islamic Banks

This section of the questionnaire seeks respondents' views on different issues relating to the recent global crisis. For this purpose, Question 16 of the survey presented nine statements to respondents. This part applied to all the respondents, which means replies from all institutional samples of data were obtained by asking respondents to answer using a five-point Likert scale (ranking from Strongly Agree = 5 to Strongly Disagree = 1). Table 8.35 employs the K-W test to examine the significant difference among respondents' perceptions according to ‘Region’.

Statements tested in this section and their coding are as follows:

1. Islamic banks are more resilient to economic shocks than their conventional peers.
2. The recent crisis would not have happened under a true Islamic banking system.
3. Islamic finance could have solved the global crisis.
4. Risk management must be embedded institutionally.
5. Banks in general used to rely heavily on rating agencies.
6. Islamic banks rely less on rating agencies than conventional banks.
7. The Islamic finance industry should develop its own rating agencies.
8. Islamic banks will emerge stronger from the crisis.
9. Consolidation is needed among smaller Islamic banks.

The results in Table 8.35 show that most statements are statistically significant (p-value < 0.05). With a ‘relaxation’ of the confidence level to 0.06, we can accept all statements except Statements 6 and 9. Mean rankings reveal that, although there is no clear pattern that could be traced, the ‘GCC’ and ‘Other’ categories are usually ranked at the top for most statements. This emphasises the fact that respondents from these two regions are more aggressive than those from other regions in their views about the credit crunch and Islamic finance. Thus the findings indicate that there are statistically different and significant opinions among the respondents coming from different regions.

In addition, attempts were made to test the impacts of ‘Nature of Financial Institution’, ‘Nature of Activities’, ‘Accounting Standards’ and ‘Respondent's Position’ on the responses; the results are summarised in Tables 8.36 to 8.39.

With the exception of Statement 3 (p = 0.005), there is no statistically significant difference among all other statements. Mean ranking for Statement 3, as seen in Table 8.36, shows that fully fledged Islamic banks are far more aggressive in their belief that Islamic finance could have solved the global crisis than other categories (46.2), followed by Islamic subsidiaries (40.25), then by Others and Conventional Banks. This is consistent with the K-W test result according to ‘Region’ as the control variable (Table 8.35) because respondents from the ‘GCC’ and ‘Other’ regions in this research sample are mainly fully fledged Islamic banks and Islamic subsidiaries.

The same statements are further investigated in relation to the nature of activities. As depicted by Table 8.37, only Statement 6 is statistically significant. With a relaxation of the confidence level to 0.06, one can also accept Statement 1. For both Statements 1 and 6, ‘Retail Banking’ is the most aggressive category according to mean rankings, and ‘Private Equity Houses’ is by far the least aggressive. Other categories fall in between, although no particular trend can be established. These results confirm the K-W test results according to ‘Region’ (Table 8.35), as out of the 10 retail banks included in the research sample, five are located in the ‘GCC’ and one is located in ‘Other’. It should be noted that the sole Private Equity House in this research sample is also located in the ‘GCC’. Furthermore, results from Table 8.37 are also consistent with the K-W test results according to ‘Nature of Financial Institution’ (Table 8.36) because 8 out of the 10 retail banks included in the research sample are fully fledged Islamic banks, and the other two are Islamic subsidiaries. The one Private Equity House is also a fully fledged Islamic bank.

Table 8.38 shows that only Statements 3, 5 and 7 are statistically significant according to accounting standards. Mean rankings reflect that for these three statements, ‘AAOIFI’ and ‘International & AAOIFI standards’ are always top ranked, followed by other criteria. These results confirm the K-W results for the previous control variables in Tables 8.35 to 8.37; therefore the results indicate that the perceived views in relation to accounting standards are statistically significant for these three statements.

The potential impact of the respondents' positions on the same statements is also investigated. The p-values in Table 8.39 show that there are no significant differences according to respondent's position. By relaxing the confidence level to 0.06, one can also accept Statements 5 and 7. No pattern could be concluded by studying the mean ranking. The only obvious conclusion is that Shari'ah scholars ranked the highest mean and solicitors had the lowest mean values for most, but not all, statements. This is expected because Shari'ah scholars tend to be more conservative in their views about Islamic banking and Shari'ah compliance, while solicitors usually focus more on legal structures rather than the Shari'ah side of transactions.

Factor analysis for Question 16 (credit crisis and Islamic finance)   To locate the perception of the participants regarding the credit crisis in relation to a number of issues related to Islamic finance, they were provided with a number of statements. The opinions are analysed through factor analysis.

As previously explained, factor analysis seeks to discover if the observed variables can be explained largely or entirely in terms of a much smaller number of variables called the factors.

As there are nine statements for Question 16, analysing the respondents' perceptions toward Islamic banking and the global credit crisis, the researcher felt that reducing these statements to a more manageable number would enhance the analysis and tell more about how respondents perceived these issues. Hence, factor analysis is deemed to be relevant in this respect as the main task of factor analysis is to cluster the related group of variables through their common variance.

In order to test the factorability of the data in terms of sampling adequacy, Table 8.40 presents the results of KMO and also Bartlett's Test for this factor analysis.