The Index-Based Approach

Within the context of the index-based banking crisis definitions, a banking crisis is defined by considering on various banking sector risk factors such as credit risk, liquidity risk and market risk and/or macroeconomic variables. After constructing a BSFI, an arbitrarily determined threshold level, 𝜑, is defined which identifies the crisis and non-crisis episodes.

𝐵𝑎𝑛𝑘𝑖𝑛𝑔 𝐶𝑟𝑖𝑠𝑖𝑠 (𝐵𝐶)_𝑖,𝑡= {1, 𝑖𝑓 𝐵𝑆𝐹𝐼_𝑖,𝑡 < 𝜑

0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 (1)

According to the equation 1, if the fragility index exceeds a certain threshold value in a specific time period, this is considered as a crisis period. More precisely, the BSFI is transformed into a binary variable and takes the value 1 if BSFI is less than 𝜑. This states that there occurs a banking crisis in county i at time t. On the other hand, an episode is classified as a non-crisis episode if the BSFI exceeds 𝜑.

It is often stated in the literature that determining an index value to define the banking crisis is difficult due to lack of reliable data on the financial activities of the banking sector such as NPL (Hawkins and Klou, 2011; Kibritçioğlu, 2003). Despite that, efforts to identify the banking crises by determining an index value have increased especially in the resent years. Related to this, one of the main studies is done by Kibritçioğlu (2003) pioneering the development of the studies in this field. In this prominent study, a banking crisis is defined by constructing a banking sector fragility (BSF) index. The BSF index is defined as the average standardized values of credit risk proxy, exchange rate risk proxy and liquidity risk proxy. Bank deposits is considered as a proxy for the liquidity risk;

Domestic credit to private sector by banks is considered as a proxy for the credit risk and foreign liabilities of banks is considered as a proxy for the market/exchange rate risk. The author defines two fragility episodes of the banking system as medium fragility and high

fragility episodes. According to this, if BSF index is between 0 and -0.5, then the sector experiences a medium fragility episode. Moreover, if the index is equal or lower than -0.5, then the banking system is highly fragile to the systemic crisis. In order to indicate whether the liquidity risk has any impact on the crisis identification, Kibritçioğlu (2003) also constructs two alternative BSF indices. The first alternative index is comprised of the same proxies as in the main BSF index except the bank deposit variable. The second alternative BSF index is constructed subtracting the bank deposit variable from the main index. The results reveal that while BSF index is highly beneficial in the context of monitoring and determining the banking crises, bank runs generally does not play a crucial role in triggering the crisis.

Ahmad and Mazlan (2015) develop an annual BSF index for Malaysian local-based and foreign-based commercial banks to investigate the fragility of these banks. Although, the BSF index is obtained following Kibritçioğlu (2003), the authors prefer to use different proxies in order to measure credit risk and market risk. On this basis, NPL variable is used to measure credit risk and time interest earned ratio (Tier) proxy is chosen for the market risk.

To identify the episodes of Islamic banking crises in Indonesia, Kusuma and Asif (2016) construct an EWS by using an Islamic banking sector fragility index (IBSFI). Following Kibritçioğlu (2003), the authors construct the IBSFI based on liquidity risk and credit risk. While Islamic bank deposits is considered as proxy for the liquidity risk, domestic credit is used as the credit risk proxy. The results show that Islamic banking in Indonesia experienced high fragility episodes between 2005 and 2006. Moreover, their EWS model predicts 80% of the banking crisis periods correctly.

Another study that covers EWS of Islamic banking crisis is done by Wiranatakusuma and Duasa (2017). The scholars construct an EWS in order to identify the signaling indicators of Islamic banking resilience in Indonesia between 2004 and 2016. For this aim, following Kibritçioğlu (2003), the Islamic Banking Resilience Index (IBRI) is obtained regarding the liquidity risk and credit risk proxies. However, different from Kibritçioğlu

(2003), the authors prefer using financing of Islamic banks variable in order to measure the credit risk.

Van Hogen and Ho (2007) define the banking crises by constructing an index of money market pressure (IMP) following the index-based currency crisis definition of Eichengreen et al. (1996). The main motivation for defining the banking crises based on a money market pressure index is due to the link between the banking crisis and the aggregate demand of the banking sector for central bank reserves. In other words, the authors explain that any banking crisis is connected with increasing non-performing assets, deposit withdrawals and decreasing inter-bank lending. Accordingly, they create the IMP by considering the weighted average of the changes in the ratio of total reserves of banking system to total non-bank deposits and short-term interest rate. The scholars accept the presence of the banking crisis if the index exceeds 98% and, if the index increases more than 5% with respect to the previous year.

Davis and Karim (2008) use two separate depended variables following the Demirgüç-Kunt and Detragiache (1998) and Caprio and Klingebiel (1996) banking crisis definitions.

Their results reveal that, different banking crisis definitions reveal different results. For instance, according to the dependent variable which is constructed with Demirgüç-Kunt and Detragiache’s (1998) banking crisis definition, real interest rate is a significant indicator of banking crisis. However, it becomes insignificant when the dependent variable is constructed based on Caprio and Klingebiel’s (1996) banking crisis definition.

Singh (2011) constructs two monthly BSFIs in order to identify the fragility episodes of Indian banks. The first BSFI is constructed considering weighted averages of the annual growth of real time deposits, real non-food credits, real investments, real foreign currency assets and liabilities and, real net reserves. The alternative index, on the other hand, is created by using the same proxies as in the main index, except for the real time deposits variable. Singh distinguishes the fragility episodes as high fragility and medium fragility episodes. Accordingly, the banking sector is in a high fragility episode if BSFI is lower than the negative standard deviation of the index. The sector is medium fragile if the BSF index is between zero and negative standard deviation of the index. The results reveal that

two BSFIs show similar movement patterns, thus the bank runs do not play an important role in the fragility of the Indian banking sector.

Jing et al. (2015) use the money market pressure index of Von Hagen and Ho (2007) however they modify the index by changing the weights of the variables. Moreover, they create alternative indices by using nominal interest rate variable instead of real interest rate in order to detect the stress in money market better.