The dynamic relationship between earnings volatility, concentration, stability and size in the Turkish banking sector

Abstract

This article investigates the causal relationship between earnings volatility, concentration, stability and bank size in the Turkish banking sector in the period 2002 to 2011. A relatively new empirical methodology, dynamic panel Granger-causality test, is used to analyse the causal relationship between these variables. The empirical result shows that bank size and concentration negatively Granger-cause earnings volatility, suggesting that larger banks and more concentrated banking market decrease earnings volatility. Moreover, the result also indicates that concentration in the banking sector increases bank stability and supports the ‘concentration-stability’ hypothesis.

Keywords:

• earnings volatility
• concentration
• stability
• bank size
• Turkish banking

JEL Classification:

• G21
• G32

Acknowledgements

We would like to acknowledge the financial support provided by the Turkish Scientific and Technological Research Council under the project no. SOBAG-112K039.

Notes

¹ The introduction of a lagged dependent variable among the right hand side variables in Equations 1(1) (2) (3)– 4(4) creates an endogeneity problem since the lagged dependent variable is correlated with the disturbance, . To solve this problem, Arellano and Bond (1991) developed a difference GMM estimator for the coefficients in the above-mentioned equations where the lagged levels of the regressors are the instruments for the equation in first differences. However, Arellano and Bover (1995) and Blundell and Bond (1998) suggest to difference the instruments instead of the regressors in order to make them exogenous to the fixed effects. This leads from the difference GMM to the system GMM estimator, which is a joint estimation of the equation in levels and in first differences. Hence, we use the two-step system GMM estimators with Windmeijer (2005) corrected standard error, along with the fixed effects estimators, to conduct our analysis.

² We also use the SD of returns on assets (ROA) as a proxy for earnings volatility for bank i for a robustness check.

³ To calculate earnings volatility (ROE or ROA), we used quarterly data 2002:Q1–2011Q:4, and used four-quarter rolling time windows to compute the SD of ROE (or ROA) specified in Equation 5.

⁴ In accordance to the Hausman test, the random effects model was rejected.

⁵ The Sargan test is a test on whether the instruments are uncorrelated with the error term. Moreover, the Arellano–Bond test results also require significant AR(1) serial correlation and lack of AR(2) serial correlation.

⁶ To check whether our results are robust, we also used the SD of ROA as earnings volatility and HHI (total assets as inputs). The results are similar to those obtained in Table 2. Although not reported, they are available upon request from the authors.