University of Huddersfield Repository Stability and Profitability in the Chinese Banking Industry: evidence from an auto-regressive-distributed linear specification

The important role played by the Chinese commercial banks in the development of China’s economy makes the government and banking regulatory authority give great concern about the performance of Chinese commercial banks, while the stability in the banking sector, without any doubt, has attracted greater attention since the financial crisis during 2007 and 2009. The principal objective of this study is to investigate the inter-relationships between profitability and stability in the Chinese banking industry. Using a sample of Chinese commercial banks over the period 2003-2013, the current study examines the inter-relationships under an auto-regressive-distributed linear model. With regard to the measurement of stability, both Z-score and stability inefficiency were used, while Return on Assets (ROA) was used as the indicator of profitability. In terms of the econometric methods, the current study used different types of Generalized Method of Moments (GMM) estimators including difference GMM, one-step system GMM, two-step system GMM as well as two-step robust GMM. In order to the check the robustness of the results, alternative econometric techniques were used such as ordinary least square (OLS) estimator, between effect estimator as well as fixed effect estimator. The results show that higher insolvency risk/lower bank stability leads to higher profitability of Chinese commercial banks and also that higher profitability leads to higher bank fragility.


Introduction
Profitability is central to the operation of commercial banks, while stability is also of concern to the banking regulatory authorities, particularly since the financial crisis of [2008][2009]. The empirical literature not only investigates risk-taking behaviour but links the risk to the efficiency of commercial banks (Abedifar et al., 2013;Tan and Floros, 2013).
A few studies test the impact of risk-taking behaviour on bank profitability; however, they mainly focus on credit risk (Tennant and Sutherland, 2014;Dietrich and Wanzenried, 2011).
Although Tan (2016) investigates the impacts of credit risk as well as insolvency risk on Chinese bank profitability, the impact of profitability on the risk-taking behaviour in the Chinse banking industry has not obtained enough attention from scholars. There is only one study investigating the inter-relationship between insolvency risk and profitability in the Chinese banking industry under a Seemingly Unrelated Regression (Tan and Floros, 2014).
The current study focuses on the analysis of insolvency risk and its relationship with bank profitability in China due to the fact that several rounds of banking reforms in China have aimed to reduce risk-taking behaviour and improve bank performance, whether the reforms can achieve these two goals at the same time is greatly concerned by the Chinese government as well as banking regulatory authorities. In other words, investigating the inter-relationship between insolvency risk and bank profitability in China, in particular the impact of profitability on insolvency risk, will provide important insights for policy makers and will also significantly contribute to the banking literature.
This study extends the work of Tan and Floros (2014) by testing the inter-relationships between profitability and insolvency risk using a number of econometric techniques under an auto-aggressive-distributed linear specification 1 . However, rather than using the Z-score as the insolvency risk indicator, the current study uses stability inefficiency as well. Different econometric techniques together with a more precise stability indicator (stability inefficiency) will provide more robust results compared to Tan and Floros (2014). We use an unbalanced panel dataset of 100 Chinese commercial banks over the period 2003-2013. The sample includes three different ownership types (state-owned commercial banks (SOCBs), jointstock commercial banks (JSCBs) and city commercial banks (CCBs)).
The findings of our paper shows that Chinese commercial banks with higher levels of insolvency risk have higher profitability, while higher insolvency risk leads to higher profitability of Chinese commercial banks.
The paper is structured as follows: section 2 reviews the empirical literature on bank profitability in China. Section 3 presents the data and methodology. Section 4 discusses the results, while Section 5 provides a summary of the whole paper.

Literature review on bank profitability in China
The profitability in the Chinese banking sector has been extensively tested by the empirical literature. Shih et al. (2007)  Using a sample of Chinese commercial banks over the period 2000-2005, Sufian and Habibullah (2009 investigate the impact of credit risk on bank profitability. Their results suggest that credit risk has a significant and positive impact on the profitability of Chinese state-owned commercial banks and joint-stock commercial banks. In addition, Sufian (2009) uses four state-owned commercial banks and twelve joint-stock commercial banks during 2000-2007 to examine the determinants of bank profitability in China with a focus on the credit risk and liquidity risk under a fixed effect model. The results show that Chinese commercial banks with higher levels of credit risk and liquidity risk have higher profitability. Heffernan and Fu (2010) analyze the profitability of Chinese commercial banks over the period 1999-2006 using two econometric techniques including a Generalized Method of Moments (GMM) estimator as well as a fixed effect estimator. Their findings report that bank efficiency, bank listing, GDP growth rate and unemployment are significantly related to bank profitability.
More recently, Tan and Floros (2012a, 2012b, 2012c) use a sample of Chinese commercial banks over the period 2003-2009 to examine the determinants of bank profitability with a focus on the impacts of credit risk and competition on bank profitability under a GMM estimator. The competition is measured by 3-bank and 5-bank concentration ratios. To be more specific, Tan and Floros (2012a) use both 3-bank concentration ratio and 5-bank concentration ratio to investigate the joint effects of credit risk and competition on bank profitability in China. They do not find any significant impact. The findings from Tan and Floros (2012b) show that the profitability of Chinese commercial banks is significantly affected by credit risk. Finally, the results from Tan and Floros (2012c) report that Chinese joint-stock commercial banks with higher levels of credit risk have higher profitability. Through reviewing the related literature, it is clearly shown that although there are study investigating the impact of insolvency risk on bank profitability in China, they mainly focus on using either generalized method of moments estimator or seemingly unrelated regression, this study contributes to the empirical literature by firstly using the auto-aggressivedistributed linear specification to test and inter-relationships between insolvency risk and profitability in banking industry, while the causality is tested using a number of different econometric techniques which is supposed to provide more accurate and robust results compared to previous empirical banking literature.

Data and Methodology
We use 5-state-owned commercial banks, 12 joint-stock commercial banks and 83 city commercial banks over the period 2003-2013 to test the inter-relationships between insolvency risk and bank profitability. The insolvency risk is measured by using two alternative indicators which are Z-score and stability inefficiency, while the profitability is measured by Return on Assets. The data is collected from Bankscope database. Table 1 shows the summary statistics of the variables used in the current study. The table shows that although the insolvency risk measured by stability inefficiency in the Chinese banking industry is not as stable as the profitability of Chinese commercial banks, it is still less volatile than the insolvency risk measured by the Z-score. Due to the fact that stability inefficiency focuses on the insolvency condition of the whole banking industry, while Zscore concentrates on the levels of insolvency risk of commercial banks, the findings suggest that Chinese commercial banks have larger difference in insolvency risk across the period examined, while the whole banking industry has less volatility in terms of banking industry stability.
<<Table 1---about here>> Table 2a-2d present the descriptive statistics of the two different profitability measures for the whole Chinese banking industry as well as for different ownership types. The table shows that the profitability of CCBs is the highest over the examined period which is followed by SOCBs while the profitability of JSCBs is the lowest. When looking at the profitability of different ownership types of Chinese commercial banks on a year by year basis, Figure 1 shows that, in general, SOCBs and CCBs have higher profitability than JSCBs. We use an auto-aggressive-distributed linear specification to test the inter-relationships between profitability and insolvency risk following Casu and Girardone (2009). The method can test not only the short term causal relationship between stability and profitability, but more importantly, also the long-run inter-relationship. It can be expressed as follows: where y represents either a profitability indicator (ROA) (Tan, 2016;Daly and Zhang, 2016) or an insolvency risk indicator (Z-score 2 or stability inefficiency 3 ), i and t represent specific banks operating in a specific year, is the intercept, , , and are the coefficients to be estimated, is the time effect, stands for individual bank effect, and is the error term. ROA shows the profits earned per unit of assets and reflects management ability to utilize banks' financial and real investment resources to generate profits. ROA has emerged as the key ratio for the evaluation of bank profitability and has become the most common measure of bank profitability (Athanasoglou et al., 2008;Garcia-Herrero-et al., 2009;Golin, 2011). Before conducting the econometric analysis, an augmented Dickey Fuller test is used to test whether the variables have unit root. The results are reported in Table 3. The test uses two lagged differences for ROA -stability inefficiency and the logarithm of the Z-score -the results show that all the variables are stationary at the 1% significance level. The test has been re-conducted using fewer or more lagged differences and the first and second lag of stability inefficiency and logarithm Z-score generate the same results. The rejection of unit root of the variables guarantees the valid and accurate results of the model.  (Arellano and Bond, 1991), while the system GMM addresses the issue of unit root property which has been ignored by the difference GMM (Bond, 2002). Compared with the one-step GMM system estimator, the two-step system GMM estimator gives consistent estimates in the presence of heteroskedasticity and autocorrelation (Mileva, 2007). Finally, the two-step Robust GMM estimator is windmeijeier corrected, in order to provide the most efficient estimates (Roodman, 2009). The relationship between profitability and insolvency risk is reported as the sum of the joint-significant coefficients, with a positive (negative) and significant sign 2 The Z-score is an indicator to measure the financial health and risk condition of banks. The calculation of Zscore can be expressed as: To control for outliers and skewness of the distribution, the logarithm of Z-score is used (Abedifar et al., 2013). There are few observations with negative Z-score, because we use the logarithm of Z-score, these observations have been dropped out. 3 See appendix for the estimation.
representing a positive (negative) causal relationship. The long-run inter-relationships between profitability and insolvency risk are also checked by testing . If the probability is less than 0.1, it means there is a long-term effect of one variable on the other. In order to check the robustness of the results, alternative econometric techniques are used, including between-effect estimator, ordinary least square estimator (OLS) and fixed effect estimator. Table 4 reports the results with regard to the effect of insolvency risk on bank profitability.

Results
The findings show that AR (1) is significant for some cases; in other words, the first-order autocorrelation is present for some cases. All the second order (AR(2)) autocorrelations are rejected which guarantees the consistency of the estimation. The findings show that the second lags of profitability as well as of insolvency risk are significant for most of the cases, which indicates that profitability is influenced by previous years' profitability as well as insolvency risk. Two cases show that an increase in the value of Z-score (lower insolvency risk) causes increases in profitability, while an increase in stability inefficiency (higher insolvency risk) is found to precede an improvement in bank profitability for three cases. The different finding reflected by Z-score and stability inefficiency can be explained by the fact that the actual condition of stability/insolvency risk can be more accurately measured by stability inefficiency rather than the Z-score (Tabak et al., 2012). Table 5 shows the impact of profitability on insolvency risk. The results indicate that the insolvency risk at year t is significantly affected by the previous year's insolvency risk. It is found that an increase in ROA causes an increase in Z-score (insolvency risk) only for one case, while the stability inefficiency indicates that an increase in profitability leads to higher insolvency risk in three cases. As discussed previously, we believe that stability inefficiency provides more precise findings with regard to banks' insolvency risk. Thus we believe that in the Chinese banking industry, higher bank profitability will increase insolvency risk and lead to bank fragility. Table 6 and Table 7 show the inter-relationships between profitability and insolvency risk using the between-effect estimator, OLS, as well as a fixed effect estimator. The findings suggest that profitability and insolvency risk in the current year are significantly influenced by the previous years' values, while an increase in insolvency risk (Z-score) leads to an improvement in bank profitability, while this impact is opposite for stability inefficiency.
When stability inefficiency is used as the insolvency risk indicator, it shows that an increase in profitability precedes increases in insolvency risk. All the significant cases from the tables show that there is a long run relationship between insolvency risk and profitability. We believe that the causal and long-run inter-relationship between insolvency risk and profitability in China is significant and positive because of the more accurate estimate of stability generated by stability inefficiency. Two insolvency risk indicators are used namely Z-score and stability inefficiency.
Interestingly, Z-score and stability inefficiency provide different finding with regard to its inter-relationships with bank profitability. Stability inefficiency measures the risk condition in a more concise way. Thus, we believe that the relationship between profitability and insolvency risk in the Chinese banking industry is significant and positive.
to the Chinese banking industry. To be more specific, Chinese commercial banks can allocate long-term credit to different businesses and they can also make full use of their available fund to invest in relatively long-term projects. Although there is a mismatch of liquidity between assets and liabilities, the increase in the insolvency risk can be redeemed by the improvement in bank profitability.

Estimation of stability in the Chinese banking sector-stability inefficiency
Tabak et al. (2012) argue that the potential stability of banks cannot necessarily be reflected by the Z-score 4 . The deviation from the banks' current stability and the maximum stability must be considered. This study provides a measure of the bank's stability inefficiency by estimating a stochastic frontier (Aigner et al., 1977;Meeusen and Van den Broeck, 1977), with the Z-score as the dependent variable of a translog specification.
W represents the input price; this study considers two input prices which are the price of funds (the ratio of interest expenses to total deposits) and the price of capital (the ratio of non-interest expenses to total assets). Y represents four outputs which are total loans, total deposits, other earning assets and non-interest income. The sub-indices i and t represent bank The Z-score reflects the extent to which banks have the ability to absorb losses. Thus, a higher value of Z-score indicates lower risk and greater stability. The Z-score has been widely used in empirical studies to measure the stability of financial institutions (Iannotta et al. 2007;. The Z-score can be expressed as follows: where ROA is banks' Return on Assets, E/A is the ratio of equity to total assets, and  International Financial Markets, Institutions and Money, 40, pp. 85-110. Tennant. D. and Sutherland, R. (2014). What types of bank profit most from fees charged? A cross-country examination of bank-specific and country-specific determinants, Journal of Banking and Finance,49,     •*, **, *** represents significance at 10%, 5% and 1% level, respectively. • the number outside the () represents coefficient, while the number in the () is t-stat, SI represents stability inefficiency.