How Do Bank Features and Global Crises Affect Scale Economies? Evidence from the Banking Sectors of Oil-Rich GCC Emerging Markets

ABSTRACT

This study investigates the types of scale economies (SE) for the exchange-listed banks of the Gulf Cooperation Council (GCC) countries over the 2000–2016 period, using the stochastic frontier for modeling banking technology that explicitly incorporates managerial preferences for the bank risk taking. It explores how the levels of economies of scale (ES) are associated with banks’ features that include the business model, risk, profitability, and capital strength. The results underscore that ES are exhausted over the subperiod 2000–2008, while substantial ES are available over the subsequent 2009–2016 subperiod that followed the global financial crisis. The ES are substantial especially for the small- and middle-sized banks. Based on the geographical location, banks operating in Bahrain and UAE have shown the highest levels of ES, while those in Saudi Arabia and Oman have shown the least. Regarding bank specialization, the investment banks have shown the highest levels of ES, while commercial banks have indicated the least ES. Concerning the bank features, we find that the levels of ES are not strongly correlated with the ratio of securities-to-total assets, the profitability from lending activities and the equity-to-capital ratio. These outcomes underscore that many banks in the GCC countries, mainly in the aftermath of the global financial crisis, have failed to alter their scales of operations to land on the most efficient scale that minimizes the average cost. Thus, the financial reforms that aim to restrict the motives of banks to expand their scale of business to benefit from the too-big-to-fail (TBTF) status, especially during financial crises, are not justifiable as a valued goal.

KEY WORDS:

• economies of scale (ES)
• GCC banking
• risk
• risk preference
• scale economies (SE)
• too-big-to-fail

JEL:

• D20
• D21
• G21.

Notes

1. The study sample comprises the data for all the exchange-listed banks of various specializations over the considered period. Although each category of banks based on specialization is supposed to offer a unique menu of financial services, the reality indicates that the offered services identified on the basis of their asset structure and funding mix are alike. The line separating the services of the categories is blurred due to the intensive competition between these institutions to attain and retain clients.

2. The Hausman test does reject the hypothesis of random effects in favor of the alternative hypothesis of fixed-effect in Eq. (4(4) $\begin{array}{c}S{E}_{i,t}={\alpha }_0+{\beta }_{1}sec-T{A}_{i,t}+{\beta }_{2}NI{M}_{i,t}+{\beta }_{3}L{R}_{i,t}+{\beta }_{4}LRs{q}_{i,t}+{\beta }_{5}LLP-Loan{s}_{i,t}\\ +{\beta }_6{E}_{i,t}+{\beta }_{7}Es{q}_{i,t}+{\epsilon }_{i,t}\end{array}$(4) ). This result is expected as some explanatory variables may be omitted from the model or correlated with each other. Thus, the employed fixed time-invariant model, in Eqs. (4)(4) $\begin{array}{c}S{E}_{i,t}={\alpha }_0+{\beta }_{1}sec-T{A}_{i,t}+{\beta }_{2}NI{M}_{i,t}+{\beta }_{3}L{R}_{i,t}+{\beta }_{4}LRs{q}_{i,t}+{\beta }_{5}LLP-Loan{s}_{i,t}\\ +{\beta }_6{E}_{i,t}+{\beta }_{7}Es{q}_{i,t}+{\epsilon }_{i,t}\end{array}$(4) and (5(5) $\begin{array}{c}S{E}_{i,t}={\alpha }_0+{\beta }_{1}sec-T{A}_{i,t}+{\beta }_{2}NI{M}_{i,t}+{\beta }_{3}L{R}_{i,t}+{\beta }_{4}LRs{q}_{i,t}+{\beta }_{5}LLP-Loan{s}_{i,t}\\ +{\beta }_6{E}_{i,t}+{\beta }_{7}TBT{F}_i+{\beta }_8{E}_{i,t}\ast TBT{F}_i+{\epsilon }_{i,t}\end{array}$(5) ) above, controls for any variables’ bias in such models. On the other hand, the panel fixed-effect (FE) is employed as many researchers argue that the choice of estimators should be guided by the objective and data characteristics rather than by the Hausman test. Furthermore, and in line with the suggestion of many authors, if the results of both FE and RE are consistent, then the use of fixed effects is generally recommended (see, for instance Borenstein et al., 2010).