Impact of Risk Factors Used to Construct BSFIs on the

approximately 390 billion dollars were extended in 1997, the beginning of the crisis. Most of the loans extended by foreign banks consist of short-term loans. The fact that this capital was made available by banks in local currency as loans for sectors that do not create value added and thus, the non-repayment of these loans contributed to the growth of the crisis (Chang and Velasco, 1998; Kaplan, 2002).

Note that, even the BSFIs defined in the EWS models are based on the same risk factors, the proxy variables that are used to measure these risk factors differ in each index. Indeed, our results show even if the same risk factors are utilized, the predictive powers of the

liquidity risk factor (DEP proxy) from the Model 14. In addition, the models such as Model 1 and Model 2, Model 3 and Model 5, Model 4 and Model 6, Model 7 and Model 8, Model 9 and Model 10, Model 11 and Model 12, Model 13 and Model 14, Model 15 and Model 16, Model 18 and Model 21, Model 19 and Model 20 and; Model 22 and Model 25 are designed in a similar manner to evaluate the impact of liquidity risk factor on the predictive performance of the EWS models. According to our results, omitting DEP variable for the liquidity risk from the BSFI increases the predictive power of the EWS models. For instance, while Model 7 (constructed by using NPF, DEP, TIER and ROE variables) is able to predict the fragility and tranquil episodes of Islamic banking about 74% accuracy, the predictive performance of the Model 8 (constructed by omitting DEP variable from Model 7) increases to 80%. In fact, the case is valid for all of the pairwise compared models that are mentioned above. The results are in line with the Kibritçioğlu (2003) and Glick and Hutchison (1999) studies on conventional banking system suggesting that the liquidity risk do not play a major role in explaining banking crises.

Credit risk is one of the main risks that seriously causes financial instability and effects banks’ vulnerability. It is considered as the main reason of bank failures (van Greuning and Iqbal, 2007). Considering the fact that the amount of bad debt in Islamic banking has been growing in the last decade⁵⁸we examine whether the credit risk is an essential risk factor for Islamic banking in forecasting the crisis episodes for Islamic banks. In terms of the credit risk factor, our results show that the it plays a crucial role in predicting the fragility episodes for Islamic banks. For instance, in Model 22, the BSFI is defined employing proxies for liquidity risk, market risk and profitability risk factors omitting the credit risk factor. The BSFIs of the Model 7, Model 9, Model 11 and Model 14, on the other hand, are constructed with including credit risk factor into the Model 22 employing different combinations of the proxies. In other words, Model 22 and the Models 7, 9, 11 and 14 differ in terms of the credit risk factor. The results show that, including credit risk proxy enhances the predictive performance of the EWS models. For instance, when we include BC as a proxy for credit risk in Model 9, the predictive performance of the model

58 See Sarker (1999).

increases to 71.2%. Additionally, including NPF also enhances the predictive performance to 73.5% in Model 7.

Islamic banks collect and distribute funds on the basis of profit and loss sharing (PLS).

While conventional banks use interest as a tool against credit risk exposure, Islamic banks do not use interest since it is prohibited. While conventional banks provide debt-based products relying on interest, Islamic banks use funds on the basis of PLS where they provide funds with mainly sale and lease-based products such as Murabahah, Ijarah, Salam, Istisna’, Musharakah and Mudarabah. Thus, it is argued that they are exposed to higher credit risk compared to conventional banks since they have limited risk sharing practices (Chong and Liu, 2009; Abdul-Rahman et al., 2014; Kabir et al., 2015). For this reason, the fragility episodes of the Islamic banks are highly dependent on the credit risk.

As expected, our results indicate that the credit risk is an important factor for the BSFI of Islamic banks as it is in conventional banking (see Kibritçioğlu, 2003). We find that no matter which market risk, liquidity risk or profit risk variable is used, the BSFI for Islamic banks should include the credit risk factor. If the BSF index is constructed considering the credit risk, the EWS captures the fragility and tranquil episodes more successfully and reveals better forecasting results.

In this thesis, apart from the existing studies, we explore whether the profitability risk factor has a significant impact on the predictive power of EWS models for Islamic banks.

In the related literature, the BSFI is constructed by using credit risk factor, market risk factor and liquidity risk factor (or by omitting liquidity risk proxy) and these risk factors are also employed to BSFIs for Islamic banks (Kibritçioğlu, 2003; Ahmad and Mazlan, 2015; Kusuma and Duasa, 2016). However, due to their distinctive nature, Islamic banks are also vulnerable to profitability risk different from the conventional banks.

As the conventional banks operate based on interest, on the asset side of their balance sheet, they have fixed income securities and also the return on their deposits are predetermined which means that the conventional banks have fixed rate of returns.

However, since the Islamic banks operate based on PLS, the rate of returns are not certain.

The investments are based on mark-up and equity implying that there is no fixed rate of

return. And, since there is no pre-agreed return on deposits the uncertainties of the rate of return on investments is higher (van Greuning and Iqbal, 2007). For this reason, Islamic banks are also supposed to be exposed to profitability risk. Therefore, we further incorporate profitability risk factor in order to test whether it has an impact on the predictive power of EWS models. To this aim we use ROE, since ROE is accepted as the most important indicator of a bank’s profitability and widely-used also for Islamic banking (see Moin, 2008; Bilal et al., 2016; Ekinci and Poyraz, 2019).

To be able to investigate if profitability risk has any impact on the predictive power results of EWS for Islamic banks, we construct different BSFIs that only differ in terms of the profitability risk proxy. For instance, while the BSFI of Model 2 is comprised of credit risk (NPF) and market risk (Tier) factors, Model 7 is constructed by considering credit risk (NPF), market risk (Tier) and profitability risk factors. Correspondingly, we are able to compare Model 1 and Model 7, Model 2 and Model 8, Model 3 and Model 10, Model 4 and Model 11, Model 5 and Model 9, Model 6 and Model 12, Model 13 and Model 16, Model 14 and Model 15 as well as Model 17 and Model 18.

Our findings indicate ROE-the proxy for the profitability risk-improves the predictive power performances of the EWS models for Islamic banks as expected. Put differently, involving the profitability risk factor into a BSFI in defining the fragility of Islamic banks to crises, expands the ability of the EWS model to make consistent predictions. In all of the models, it is observed that profitability proxy enhances the predictive performance of the system. For instance, while Model 2 correctly predicts the fragility and tranquil episodes by 73%; including profitability risk factor into the BSFI increases the predictive performance of the EWS model by 7%. Furthermore, this outcome is valid for all of the pairwise compared EWS models. Therefore, the results suggest that, as an important risk factor for Islamic banking, profitability risk increases the correctly called fragility and tranquil episodes of banking crises. That is, in constructing BSFIs for Islamic banking system profitability risk factor should not be omitted.

As opposed to conventional banks, since the financial instruments of the Islamic banks are asset-based rather than debt-based, Islamic banks are exposed to higher market risk

than conventional banks. The market risk in Islamic banking arises mainly due to mark-up rates and price fluctuations. For instance, within the context of Salam, market risk arises due to price differences in the period between the delivery and sale of the goods.

In Murabahah, on the other hand, even though the benchmark rate may vary, the mark-up rate is fixed during the contract. Therefore, when prevailing mark-mark-up rate exceeds the rate that is agreed in the contract, then the bank cannot benefit from this price change (van Greuning and Iqbal, 2007). For this reason, it is important to investigate whether including the market risk factor into the BSFI enhances the predictive power performances of the EWS of Islamic banks.

To investigate the impact of market risk proxy on the predictive power rate of the EWS models of Islamic banks, we compare the models: Model 1 and Model 23, Model 5 and Model 17, Model 7 and Model 19, Model 8 and Model 20, Model 9 and Model 18, Model 10 and Model 21, Model 12 and Model 20, Model 11 and Model 19, Model 13 and Model 21, Model 14 and Model 18 and; Model 15 and Model 17. Although the BSFIs of these models are built by using credit risk, liquidity risk and profitability risk, we omit market risk from some of them. For instance, while the BSFI of the Model 12 is built by credit risk (NPF), market risk (FL) and profitability risk (ROE), the BSFI of the Model 20 is constructed by omitting the market risk proxy and constructed by considering the credit risk (NPF) and profitability risk (ROE) only. Therefore, the models differ only in terms of the market risk factor. We find that, as in the case for conventional banking, market risk is an important risk factor for Islamic banks (see Kibrirçioğlu, 2003). Once we include market risk factor by using FL proxy into the models, the predictive power of the EWS models increases. However, we observe a weak evidence for the models where the BSFIs are constructed by tier proxy as a market risk factor. As we investigate in the Section 5.3.1, the FL proxy reveals better predictive power results than the tier proxy. We believe that our interpretations about the FL and tier proxy are also binding at this point.

More precisely, in term of the predictive power performances, the foreign currency liabilities of Islamic banks reflect the market risk of Islamic banks better than the minimum capital requirement to cover the market risk of banks. Therefore, we observe constructing the BSFI with market risk factor increases the predictive ability of the EWS model even if the risk factor is measured by the FL proxy.