The effect of ROA and the leverage ratio (Basel III) on the
default risk of banks in China
By Hao Liu
Student ID: 11377267
July 2017
Presented to the Faculty of Economics and Business University of Amsterdam
In Partial Fulfillment of the Requirements for the Degree of Master of Science in
Economics Master of Science in Economics
Specialization: Monetary Policy, Banking and Regulation
Supervisor: Dr. S. J. G van Wijnbergen
Co-reader: Dr. W.E. Romp
Faculty of Economics and Business
Abstract
In this study I do some research on the factors that have effect on the bank default risk
(Z-score), particularly focus on the return-on-assets ratio (ROA) and the leverage
ratio (CAR). I employ a dynamic panel dataset of 17 Chinese commercial banks that
extends from 2011 to 2015. Several variables are used as alternative elements that
influence the bank default risk, including the bank size (TA), change in short-term
interest rate and so on. My conclusion relies on the results from the
system-Generalized Method of Moments (GMM) estimator. The results show that the lagged
ROA has positive and statistically significant effect on the Z-score while the lagged
CAR has the opposite effect.
Statement of Originality
This document is written by Hao Liu who declares to take full responsibility for the
contents of this document.
I declare that the text and the work presented in this document is original and that no
sources other than those mentioned in the text and its references have been used in
creating it.
The Faculty of Economics and Business is responsible solely for the supervision of
completion of the work, not for the contents.
Contents
1.Introduction ... 4
2.Review of the literature ... 5
2.1Return on assets ratio (ROA) ... 6
2.2Capital to assets ratio(CAR)-leverage ratio ... 7
2.3Bank’s default risk (Z-score) ... 8
3.Methodology and Data description ... 9
3.1 Data description ... 9 3.2Variables definitions ... 10 3.2.1Dependent variables ... 10 3.3 Identification problem ... 12 3.4 Models ... 13 3.4.1 Assumptions ... 13
3.4.2 Unit root test ... 13
3.4.3 ROA model ... 14
3.4.4 CAR model ... 14
3.4.5 Z-score model ... 15
4.Empirical Analysis ... 16
4.1 the Leverage ratio (CAR) ... 16
4.2 the Z-score ... 17
5.Conclusion ... 18
Reference ... 20
1.Introduction
The 2008-2009 Financial Crisis tested the global financial system both of the stability and the resilience. This crisis originated in the United States of America, caused by collapsing valuations of sub-prime mortgage. It indicated malfunctions of risk management in the banking sector.
The role of the banking sector remains central in financial system in general. IMF estimated of total writedowns during the 2008-2009 financial crisis by banks and other financial institutions on assets originated in the United States, Europe and Japan. The result shows that not only banks but also other financial institutions suffered giant losses, about $4.05 trillion in total as a result of this crisis. (IMF,2009)
In order to deal with problems raised in the banking sector, the 2008 global financial crisis motivated Basel Committee on Banking Supervision (BCBS) to come up with the New Basel Capital Accord, known as Basel III. In November 2010, regulators around the world arranged a summit conference in Seoul to discuss risk management in the banking sector and agreed on the new regulation, called the Basel III.
Although has not joined the Basel committee yet, China pays attention to the regulations published by the Basel committee and try to make our own standard that both follow the global criteria and is suitable to the situation in our own country. On 3 May 2011, the China Banking Regulatory Commission (CBRC) published the “Guidelines for Implementing New Regulatory Standards in the PRC Banking Industry” (known as “the New Standards"). The New Standards are China's response to the new Basel III regulation and are sometimes called the "China's Basel III." The new regulation would be put into practice from 1 January, 2013. (Sekine, E., 2011)
Basel III is made to create a stable and resilient banking sector that can better absorb shocks and avoid giant losses in crisis in the future. In order to find out how it works in China, I combine the Basel III regulation to the default risk of banks, particularly the regulation of the leverage ratio. In the rest parts, I follow the measurement used by Giordana, G. and
Schumacher, I. (2012) that measure the banks’ distance to default, defined as the Z-score. Main targets of this thesis are to find out whether the new regulations in Basel III and the
return-on-assets ratio as well as other elements have effect on the default risk in banks of China.
In this study, the bank default risk, Z-Score, is defined as the sum of the return-on- assets ratio plus the capital-to-assets ratio normalized by the standard deviation of the return on assets (Giordana, G. et al, 2012). I use banks’ profits and leverage decisions to analysis how chosen variables impact on the bank default risk.
Because of the endogeneity issue, my main analysis is based on the results from the system-Generalized Method of Moments (GMM) estimator instead of the Ordinary Least Square (OLS)estimator.
I employ data of 17 commercial banks in the Chinese banking sector over a 5-year period (2011-2015). In specifying the model I account for three equations, including the return-on- assets ratio, the leverage ratio and the default risk, by using a dynamic panel data estimation procedure.
I can broadly summarize my findings as follows. The lagged ROA has positive and statistically significant effect on the Z-score while the lagged CAR has the opposite effect.
Besides, the relationship between the bank size (TA) of banks in last year and the bank default risk (Z-score) is positive and statistically significant. While the off-balance sheet activities over total assets (OBSR) of last year has negative and statistically significant effect on the Z-score.
The structure of the paper is as follows: Section 2 briefly reviews the existing literature on the effect of the Basel III requirements on Chinese banks. Section 3 describes the methodology that I adopt and also the description of my data. Section 4 presents the main empirical results. And Section 5 summarizes and concludes and followed by the reference and tables.
2.Review of the literature
Since the purpose of this study is analyzing which and how chosen variables influence the bank default risk. There are a lot of literature that do research on the elements that affect the bank risk. In this thesis I choose the Z-score as the bank default risk. And the return-on-assets
ratio (ROA) and the capital-to-assets ratio (CAR) are two components of it. As a result, the review of literature shows the history of applying the Z-score, how the return-on-assets ratio (ROA) and the capital-to-assets ratio (CAR) affect the bank default risk. Besides, there are also some literature about which elements that have effect on the ROA and CAR. The detail literature shows in the next three subsections.
2.1Return on assets ratio (ROA)
A profitable and solid banking sector can hold out against negative shocks and contribute to a stable economy in a much better way. Generally, people choose to use the return on
assets(ROA) or the return on equity (ROE) to measure the profitability of banks. In this study, I use the return-on-assets ratio (defined as ROA) to measure the bank profitability. ROA represents the ability of management of banks to generate profits from their assets.
There are a lot of literature that do research on the relationship between profitability and bank risk. Some studies support the bank profitability contributes to the bank risk, while there are also other literature that agree the return on assets make banks less risky
Moussa, M. (2015) finds out that return on assets has positive effect on the risk of banks. In contrast, Berger, A. and Bouwman, C. (2011) do research on the how does capital affect bank performance, especially during financial crises. In the results of the effect of the bank’s pre-crisis capital ratio on its ability to survive, the return-on-assets ratio has positive and statistic significant effect on the banks’ ability to survive. In other words, there is negative
relationship between the ROA and the risk of banks.
In addition, there are also some literature study on the factors that have effect on the bank profitability. Elements that determine bank profitability can be divided into two mainly parts, internal elements and external elements.
External elements that affect bank profitability are elements that reflect the economic
environment which influence the performance of banks. When it comes to the external factors that affect bank profitability, industry factors and macroeconomic factors are always taken into account. There are a lot of literature that studied the industry determinants related to the market concentration and industry size. Among the macroeconomic factors, the interest rate, GDP growth rate, change in GDP growth rate as well as unemployment are considered as the mainly parts that influence bank profitability (Slovik, P. and Cournède, B. ,2011)
Internal determinants of bank profitability come from bank balance sheets and therefore could be called bank-specific determinants. They are influenced by not only the bank’s management decisions but also policy objectives. Studies that investigate on internal determinants include variables such as capital, credit risk and size.
Smirlock, M. (1985) and Akhavein, J. et al. (1997) find that there is a positive and significant relationship between bank size and profitability. Similarly, Short, B. (1979) also link bank size to capital ratios, which is believed to be positively related to bank size.
Athanasoglou, P. et al. (2004) do research in Geek banking sector based on the annual data from 1985 to 2001. They find that ‘productivity growth, the coefficient of the capital variable (equity/assets) and expected inflation positively and significantly affects profitability.’
2.2Capital to assets ratio(CAR)-leverage ratio
Leverage allows banks to increase the potential gains on a circumstance what would be possible through a direct investment of its own funds (Hulster, K. 2009). Financial Services Authority in U.K (2009) finds that excessive leverage by banks has contributed to a much severe global financial crisis. In general, banks engage in leverage through borrowing from outside. In this way can banks increase their return on equity. As a result, it leads to more and more fragility in not only the individual bank but also the whole banking system.
There are several ways to measure the leverage, the most common used one is the leverage ratio. The leverage ratio is generally expressed as Tier 1 capital over total assets. In 2010. Basel committee on Banking Supervision proposed New Basel Capital Accord, known as Basel III. The leverage ratio is presented as a new capital regulation for banks. It acts as a supplement to risk-based measure. Because the leverage ratio is relatively easy to calculate comparing to the risk-based measures, it can be adopted quickly and less costly, which makes it easy to use both for banks and for their supervisors. The restriction of the minimum
leverage ratio, acting as a backstop to risk-based capital requirements, resulting the leverage ratio to be a good tool to stop banks from being too leveraged. (Hildebrand, P. 2008)
Even before it was raised by the Basel committee, the Unite States had already chose to use the leverage ratio to regulate its banks, at least 4% for all the banks and 3% for “strong banks”. Besides, Canadian and Swiss also include the minimum leverage ratio in their
useful tool when combined to other Basel capital requirements, in this way can banks avoid building up excessive leverage and reduce default risk.
As mentioned before, the leverage ratio is a main part of the Z-score index, and the Z-score index is used to measure the overall bank risk. There are a lot of literatures that do research on the impact of leverage on the bank’s default risk.
Papanikolaou, N.and Wolff, C. (2010) do research on the leverage and risk in U.S
commercial banks in 2010, their evidence indicates that leverage contributes to both systemic risk and banks’ overall risk.
Bordeleau, E. et al (2009) study the experience in Canada, they review leverage constraints in Canada from 1992 to 2009 and find out that a minimum leverage ceiling would be a useful tool to complete the risk-weighted measures. Besides, a leverage ceiling also helps to mitigate procyclical tendencies in the whole financial system.
However, there are also evidences that show opposite views. ‘The current Tier 1 capital leverage ratio generally is least descriptive of credit risk and, in some instances, even has a negative relationship with credit risk.’ indicated by Blankespoor, E. et al (2012). They also conclude ‘the current adjustments being made to arrive at regulatory capital may be hampering bank regulators’ ability to discover credit risk problems.’
In the end, the impact of the leverage ratio (CAR) on the bank’s default risk is not clearly yet. What is more, there may also mutual effect between the leverage ratio and the risk of banks. The specific analysis is shown in the empirical analysis part. Just like what I do to analysis the elements that effect return on asset, I also do the same thing the find factors that have effect on the leverage ratio, both in internal and external ways. The specific elements that I choose will be shown in the methodology part.
2.3Bank’s default risk (Z-score)
The banking sector serves as a major part in the whole economy. As a result, the instability of the banking sector may be transferred easily to rest the economy through contagion. What is more, increasing competition among banks may also result financial system become more fragile. As a result, regulators are motivated to focus on developing banking rules that can
help to maintain stability in the banking sector (Berger, A.et al 2008). In this case, I choose to focus on the analyzing the bank’s default risk, defined as the Z-score.
A large economic and finance literature provides several methods to measure the bank risk, such as the VaR model (Duffie, D. and Singleton, K., 2004) and DD (distance to default) (Campbell, J. et al,2005). In this case, I choose to use the Z-score index, which is an indicator of financial stability, as the measurement of the distance-to-default for banks in China.
The score has been used in a lot of academic literature. Berger, A.et al (2008) use the Z-score to measure the overall bank risk. They find that banks with a greater market power have less overall risk. Maechler, A.et al (2007) to also choose use the Z-score as a measure of insolvency risk to analysis the decomposing financial risks and vulnerabilities in Eastern Europe.
Berger and Bouwman
(2011),
grounded on existing theories, believe that capital improves probability to survive for banks. Their results show capital improve the survival probability for banks of all sizes, both during crisis and normal times,The Z-score is an appropriate measurement to express bank’s default risk. Several elements have effect on it, I will choose some variables to analysis their effects on the Z-score. The specific results will be shown below.
3.Methodology and Data description
In this section I turn to describe the data source, econometric models and the choice
of estimators.
3.1 Data description
My empirical analysis is based on a dynamic panel dataset that consists of the 17 Chinese commercial banks in the period of 2011 to 2015. Among the 17 Chinese commercial banks, there are 5 state-owned commercial banks, 7 joint-stock commercial banks and 5 city commercial bank. And 16 of them are listed banks. The list of banks’names is shown in the Table 1. Among all the banks that I choose, the 5 state-owned commercial banks have strong
market power because they possess more than 40% of the total assets in the banking system in China.
When it comes to my data source, all my macroeconomic variables come from the Word Bank. While the bank-specific variables are mostly obtained from the Bank Scope database, such as the return on assets ratio (ROA) and total assets (TA). In addition, the leverage ratio (CAR) is generally expressed as Tier 1 capital over total assets. These two variables (tier 1 capital and total assets) are also taken from the Bank Scope database. Besides, I also take parts of my data from bank’s annual reports. For example, I need the data of the return on assets ratio (ROA) from 2007-2010 to get the standard deviation of ROA, these data resource from each bank’s annual reports. And the data of the short-term interest rate (IR) are
published by the People's Bank of China (PBOC).
To be specific, table 2 describes definitions of all the variables that I use in this study, while table 3 shows summary statistics of these variables. And I will explain them in detail in the next subsection.
3.2Variables definitions
Now I turn to describe the variables employed in the econometric analysis in detail.
3.2.1Dependent variables
The main target of this study is to find out which elements that I choose have effect on the bank default risk. So the first step is to define the bank default risk.
As discussed before, I choose to use the Z-score index to measure the bank risk.
The Z-score measures the distance to default for banks. It combines profitability, leverage, and return volatility in a single measure (Berger, A. et al,2008). It relies on Alman’s Z-score and is calculated as follows:
i =1,2…17; t=1,2…5;
Zit=
Eit / Ait+ ROAit
where !"#$% is the period-average return on total assets for bank i in time t, &$% is the bank
i’s Tier 1 capital at time t. #$% are its total assets, then &$%/#$% is the capital assets ratio (defined as CAR ratio). In this study, )#!$% represents the Tier 1 capital to total assets ratio
for bank i in time t, and *!"#$% is the standard deviation of return on assets. The larger the
Z-score index is, the longer distance for banks to default.
To get the standard deviation of return on assets, I use the ROA data of five years for each bank to compute the average ROA. For example, to compute the standard deviation of return on assets for Industrial & Commercial Bank of China (ICBC) in 2011, I use data collected from 2007 to 2011:
2007
2008
2009
2010
2011
Industrial & Commercial Bank of
China (The) - ICBC
1.02
1.21
1.2
1.32
1.35
And follow the method below:
In the end, the average return-on-assets ratio is for 2011 is 1.22, and then the standard deviation of return-on-assets ratio is 0.13 in this year.
From the equation of the Z-score, one may easily conclude that higher return-on-assets ratio leads to lower default risk. Similarly, we may also say that banks with high leverage ratio has less risk to default. What is more, the smaller the standard deviation of return on assets ( *!"#$%) is, the lager the distance to default is. However, once we take the simultaneity of
CAR and ROA in to account, there may be opposite conclusion. That is to say, the level of return on assets influences the leverage ratio and vice-verse, which makes it confused to identify the relationship between Z-score and its two components. For instance, if one bank increase its return on assets by increasing the leverage ratio, even though both ROA and CAR raised, the total risk for this bank also increase, which brings it much closer to default.
σ
xit = 1 T−1 (xi,t− j− xit) 2 j=0 T∑
In order to find out whether the hypothesis is correct, I run simple OLS regression in the Stata about the relationship between the ROA and the Z-score as well as the CAR and the Z-score. The results show there is positive and significant relationship between the CAR and the Z-score. However, the ROA has negative effect on the Z-score even though it is not significant. Besides, the relationship between the standard deviation of return on assets and the Z-score negative and significant. As a result, the relation among these element is not as clear as I hypothesis before.
It makes it difficult to identify the impact of CAR and ROA on the bank default risk. To be clear, I decide to analysis the two components of the Z-score first, and then get the model of the Z-score by combining these two parts together.
The next step is which regressors I choose to use for the models of ROA and CAR, and finally for the Z-score.
3.3 Identification problem
Before I estimate these three equations, the first and foremost task is to find out whether there are identification problems in my models. All equations should satisfy mainly two conditions, the rank conditions and order conditions (Bond, S. 2002).
In my multiple equations models above, it is clear that the three equations as a system is identified. For the whole system, there is only three endogenous variable (g) and six predetermined variables (k). In the equation 1, +, is equal to 1 while -, is equal to 6,
k--,=0=+,-1. Hence, equation 1 is just identified. In addition, for the equation 2, +. is equal to 2 while -. is equal to 4, k--.=2=+.-1. Therefore equation 2 is just identified. The equation 3 is the sum of equation 1 and 2. It also satisfy the identification condition. As a result, the whole multiple equations model system is identified, which means I can do the empirical analysis on the whole model system.
The next step is describing the whole model system that I use, which contains three equations, one for the return-on-assets ratio, one for the capital-to-assets ratio and the final one for the Z-score.
3.4 Models
The main econometric models that I cite in this study are the method used by Gaston
Giordana and Ingmar Schumacher (2012). They do a similar research on the effects on bank’s default risk based on the data of Luxembourg banking sector.
In the former subsector, I conclude that there may be opposite conclusions when taking the simultaneity of CAR and ROA in to account to analysis the effect of ROA and CAR on the Z-score. In order to make is clear to analysis the variables that have influence on the bank default risk, I decide to establish three equations, one for ROA, one for CAR, and finally sum the first two equations together to get the last equation for the Z-score.
Just as what I explain in the Z-score part, return on asset ratio (ROA) and the leverage ratio (CAR) are the main components of the Z-score.
The first step is to set up an equation for return-on-asset ratio (ROA), then for the leverage ratio (CAR). Combing these two can I get the final equation for the Z-score. In order to make it easier to get the equation for the Z-score, I use the ROA and CAR normalized by the standard deviation of ROA (*!"#$%) as dependent variables in the models of ROA and CAR.
3.4.1 Assumptions
Since I follow the method used by Gaston Giordana and Ingmar Schumacher, I also follow the assumption in their paper that banks choose their leverage ratio in present to improve future profitability. That is to say, leverage have effect on the future profits. What is more, the more profits banks earn today, the higher Tier 1 capital can they retain in the same year. Thus, leverage is related to the return on assets in the same period.
As a result of the assumption, in the econometric models, CAR is predetermined by the ROA, while ROA is an endogenous regressor in the CAR model.
3.4.2 Unit root test
Before building the models for the ROA, CAR and the Z-core, I first do the unit root test to test whether those variables that I include in my models are stationary. I choose the Fisher type unit root test in Stata to find out if the ROA/sd(ROA), CAR/sd(ROA) and the Z-score
panels contain unit roots. The results of the Fisher type unit root test show that all these three variables are stationary.
3.4.3 ROA model
The return-on-assets ratio is the first component of Z-score. Just as I describe in the literature part, the return-on-assets ratio (ROA) embodies the generating profits ability of banks from their assets. It is influenced mainly by two dimensions, the internal dimension and the external dimension.
I build a dynamic model for ROA by including one lag of dependent variable (L1.ROA) as a regressor. Since there are mainly two parts that effect the bank profitability, the internal elements and the external elements. I include some bank-specific variables as well as macroeconomics indicators.
In the end, I choose the change in GDP growth rate and also the change of short-term interest rate as macroeconomic variables. And as what I assume before, I also include one lag of capital-to-assets ratio in the ROA model. In addition, the bank sheet size (measured by total assets) is also used as one of the bank-specific variable.
When measuring the bank profitability, there will be bias if the off-balance sheet items of banks are ignored, especially when the off-balance sheet items play an important share of bank’s return. As a result, I include the off-balance sheet activities of banks to get rid of the measurement bias for return on assets.
3.4.4 CAR model
The capital-to-assets ratio, also called the leverage ratio (CAR), is another component of Z-score. It is an indicator of the bank risk profile. Similar to what I do to the ROA model, I also include one lag of dependent variable as a regressor in the CAR model (L1.CAR). In addition,
(ROA σROA )it =α1,0+α1,1( ROA σROA )it−1+α1,2( CAR σROA
)it−1+α1,3lnTAit−1+α1,4OBSRit−1+α1,5ΔIRt+ε1,it(1)
(CAR
σ
ROA )it =α
2,0+α
2,1( CARσ
ROA )it−1+α
2,2( ROAσ
ROAI also choose the bank-specific variables and macroeconomics indicators for CAR as determinants.
In order to satisfy the assumption, I include the return-on-assets ratio (ROA) in the CAR model. It is very clear that ROA is an endogenous variable in the CAR model. As a result, the choice of econometric method should be considered carefully. The Ordinary Least Squares estimator is not plausible under this condition. I also use the Fixed Effects and final the GMM estimators for this equation, but the main analysis is based on results of the GMM estimator.
In the GMM estimation of equation (2), I utilize the ROA/sd(ROA) as the endogenous variable. In the list of explanatory variables, I include the change in Gross Domestic Product ( GDP) as the exogenous variable, while the Non-performing loans over total loans of banks (NPLs) and Off-balance sheet activities over total assets (OBSR) as additional instrumental variables.
3.4.5 Z-score model
Here shows the final equation for the Z-score, which is the major one to analysis in this study.
It is sum of equation 1 and 2. And as defined in the data description section, the ROA and CAR are respectively, the return-on-assets ratio and capital to assets ratio. The TA is the total assets of banks, which stands for the bank size. OBSR is the off-balance sheet items over total assets. ∆0!% is the change in the short-term interest rate published by the People's Bank Of China (PBOC).
As the assumption section says, there is endogeneity issue in my models, I prefer to use the system-Generalized Method of Moments (GMM) estimator instead of the Ordinary Least Square (OLS) one.
There are two reasons that I choose to use the Generalized Method of Moments (GMM) (Arellano – Bond,1991) in this study.
On one hand, the econometric models above examine the effect of variables on the bank default risk based on a dynamic panel dataset of 17 banks for 5 years (2011-2015), which has
Zit=
β
0+β
1(CARσ
ROA)it−1+
β
2(ROAσ
ROAa short time dimension and a large individual dimension. In other words, in this panel dataset, the T (time) is small and N (individual) is large, which is suitable to use the
system-Generalized Method of Moments (GMM) (Arellano – Bond,1991). On the other hand, there is endogeneity issue in my models as discussed before in the assumption part. The regressors may be correlated with the error term. By using the GMM estimator can I avoid the bias caused by the endogeneity regressors.
In order to show the difference among different estimators, I decide to use three types of estimators, the Ordinary Least Square (OLS), fixed effects (FE) and the Generalized Method of Moments (GMM). But my main analysis will still be based on the results from the GMM estimator.
In the GMM estimation of equation (3), I choose the ROA/sd(ROA) as the endogenous variable. In the list of explanatory variables, I include the change in Gross Domestic Product ( GDP) as the exogenous variable, while the Non-performing loans over total loans of banks (NPLs) as the additional instrumental variable.
4.Empirical Analysis
This section shows the econometric results of models that I describe above. The regression results are presented in Tables 4 and 5.
Because the equation (3) is the sum of equation (1) and (2). There is linear relationship among these coefficients. It is not necessary to analysis all three models. As a result, I choose two equations, equation (2) and (3), to analysis the econometric results.
4.1 the Leverage ratio (CAR)
Now I turn to analysis the estimation results for the CAR model. To make it easy to describe, I just shorten the name of the model of CAR/sd (ROA) (equation 2) as the CAR model. Table 4 shows the estimated coefficients of equation (2) by using estimators of OLS, FE and GMM. I only focus my analysis on the results of the GMM estimator.
The coefficient estimated on the ROA in the CAR model is positive and statistically significant. An increase in ROA of one percentage point enhance CAR by 6.076 percentage
points. This result proves what I assume in precious section. That is to say, the more profits banks earn today, the higher Tier 1 capital can they retained in the same year.
The lagged dependent variable (234
5678)%:, coefficient in the GMM regression is positive and
statistically significant, which suggests persistence of leverage ratio in Chinese bankingsector. A high leverage ratio at present will contribute to a much higher leverage ratio in the future.
The GMM regression also shows that there are negative and statistically significant
relationship between the lagged bank size (TA) and CAR, also between the change in short-term interest rate and CAR. The negative effect of the bank size on the leverage ratio may due to that the banks with more assets are more likely to retain less capital, due to the scale effect. Brandao-Marques, L. et al (2012) get a similar result that larger banks tend to be more leveraged due to “too-big-to-fail” effect.
4.2 the Z-score
Table 5 presents the estimated coefficients of equation (3), the Z-score index, by using estimators of OLS, FE and GMM. I also focus my analysis on the results of the GMM estimator only.
The coefficient estimated on the lagged ROA is positive and statistically significant. A one percent increase in the lagged ROA enhance the Z-score by 0.00232 percentage points, which means if banks manage their profits well currently, they are less likely to default in the coming year.
However, the results also show that there is a negative and statistically significant relationship between the lagged CAR and the Z-score. One percent increase in leverage ratio of banks in last year reduce the Z-score by 0.0116 percentage points, which implies that bank leverage ratio of last year increases the bank default risk of this year. This result shows opposite effect from what I expect in the beginning. However, as I say in the literature section, there are literatures that conclude ‘the current Tier 1 capital leverage ratio generally is least descriptive of credit risk and, in some instances, even has a negative relationship with credit risk’
(Blankespoor, E et al ,2012). As a result, the negative effect for lagged CAR on the bank default risk is reasonable in some extent. The negative results are due to the offsetting effect of the lagged CAR on the ROA. Increasing the lagged CAR reduces the ROA, which results in the negative effect of the lagged CAR on the Z-score.
The relationship between the bank size (TA) of banks in last year and the bank default risk (Z-score) is positive and statistically significant. An increase in lagged TA of one percentage point enhance Z-score by 0.115 percentage points, which implies bigger banks in China is less likely to go bankrupt. The reason of this conclusion may be that the biggest 5 commercial banks in China are all owned by the state. It is difficult for them to go bankrupt since the government will always help them, which may lead to moral hazard problem.
In the GMM regression of the Z-score, the effect of the change in short-term interest rate is positive but not significant. While the off-balance sheet activities over total assets (OBSR) of last year has negative and statistically significant effect on the Z-score. An increase in lagged OBSR of one percentage point deduce the Z-score by 2.947 percentage points.
5.Conclusion
In this paper, I study on the factors that have effect on the bank default risk in the commercial banking sector in China. I use the Z-score as an indicator of bank’s distance to default, which defined as the sum of the return on assets plus the capital-to-total assets ratio normalized by the standard deviation of the return on assets. The bigger the Z-score is, the higher the distance to default.
Besides, I also study the two components of the Z-score, the return-on-assets ratio (namely the ROA ratio) and the capital-to-assets ratio (namely the CAR ratio). By employing a dynamic panel dataset of 17 Chinese commercial banks that extends from 2011 to 2015, I empirically investigate whether these chosen variables have effect on bank default risk.
Since there is endogenous relationship between the return-on-assets ratio and the capital-to-assets ratio, I use the system-Generalized Method of Moments (GMM) (Arellano –
Bond,1991) estimator to do the econometric analysis in this study. Several variables are used as alternative elements that influence the bank default risk, including the bank size (TA), change in short-term interest rate and so on.
My econometric results show that the lagged ROA has positive and statistically significant effect on the Z-score, while the lagged CAR has the negative and statistically significant
effect on the bank default risk. These results suggest that the better for banks to manage their profits from assets of last year, the less risky for banks to default of this year.
When it comes to analysis on the effect on the lagged CAR on the Z-score, the GMM estimation shows a negative effect of the lagged CAR on the Z-score.
Before reaching the results, I expected a positive relationship between the lagged CAR and the Z-score. The negative results are due to the offsetting effect of the lagged CAR on the ROA. The equation of the Z-score is sum of equation of ROA and CAR; the lagged CAR also negatively influences the ROA, which apparently is the dominant effect in this study.
Increasing the lagged CAR reduces the ROA, which results in the negative effect of the lagged CAR on the Z-score. Apparently banks with high capital-to-assets ratio this year will not keep the same high capital-to-assets ratio next year. My study suggests that the negative impact on the ROA will reduce the leverage ratio to a lower level to such an extent that it leads to higher risk in the next year, that is to say, a lower Z-score in the next year.
In the beginning of this study, I also want to include other standards in the Basel III, such as the liquidity rules on the net stable funding ratio (NSFR) and the liquidity coverage ratio (LCR). But the relative data has not been published in China. It will be a better analysis study if I can include these factors.
All in all, the banking sector plays an important role in the whole financial system. Paying attention to the bank default risk is always an emphasis for the whole economy all over the world, especially for developing countries like China. A good management in bank risk can help them to develop in a stable way and avoid serious banking crisis in the future.
Reference
[1] Adrian, T. and Shin, H. (2008), Liquidity and Leverage, Federal Reserve Bank of New York Staff Reports, Federal Reserve Bank of New York, No. 328.
[2] Akhavein, J., Berger, A. and Humphrey, D., (1997), The effects of megamergers on efficiency and prices: evidence from a bank profit function, Finance and Economic Discussion Series 9, Board of Governors of the Federal Reserve System.
[3] Athanasoglou, P.and Brissimis, N., (2004), The effect of mergers and acquisitions on bank efficiency in Greece, Bank of Greece Economic Bulletin 22, 7-34.
[4] Athanasoglou, P., Brissimis, S. and Delis, M. (2005), Bank-specific, industry-specific and macroeconomic determinants of bank profitability, MPRA Paper No. 32026.
[5] Arellano, M, and Bond, S. (1991), Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations, The Review of Economic Studies, Vol. 58, No. 2 (Apr., 1991), pp. 277-297.
[6] Berger, A. and Bouwman, S. (2013), How Does Capital Affect Bank Performance During Financial Crises, Journal of Financial Economics Volume 109, Issue 1, July 2013, Pages 146-176.
[7] Berger, A., Klapper, L. and Turk-Ariss, A. (2008), Bank Competition and Financial Stability, the World Bank policy research working paper WPS4696.
[8] Blankespoor, E., Linsmeier, T., Petroni, K. and Shakespeare, C. (2012), Fair value accounting for financial instruments: Does it improve the association between bank leverage and credit risk? Rock Center for Corporate Governance at Stanford University Working Paper No. 121.
[9] Bordeleau, E., Crawford, A. and Graham, C. (2009), Regulatory Constraints on Bank Leverage: Issues and Lessons from the Canadian Experience, Bank of Canada Discussion Paper 2009-15 December 2009.
[10] Bond, S. (2002), Dynamic panel data models: a guide to micro data methods and practice, Port Econ J (2002) 1: 141–162.
[11] Bourke, P. (1989), Concentration and Other Determinations of Bank Profitability in Europe, North America and Australia, Journal of Banking and Finance 13 (1989) 65-79, North-Holland.
[12] Brandao-Marques, L., Correa, R. and Sapriza, H. (2012), International evidence on government support and risk-taking in the banking sector, IMF Working Paper WP 13/94.
[13] Campbell, J., Hilscher, J. and Szilagyi, J. (2005), In Search of Distress Risk, Harvard Institute of Economic, Research Discussion Paper Number 2081
[14] Chan-Lau, J. and Sy, A. (2006), Distance-to-Default in Banking: A Bridge Too Far? International Monetary Fund WP/06/215.
[15] Demirgüç-Kunt, A. and Detragiache, E. (2010), Basel Core Principles and Bank Risk: Does Compliance Matter? IMF working paper WP/10/81.
[16] Duffie, D. and Singleton, K. (2004), Credit Risk: Pricing, Measurement, and
Management, Canada Research Chair in Risk Management and Finance Department Working Paper 04-06.
[17] Gennotte, G. and Pyle, D (1990), Capital Control and Bank Risk, University of California at Berkeley, Finance Working Paper 197 Revised.
[18] Giordana, G. and Schumacher, I. (2012), An Empirical Study on the Impact of Base III Standards on Bank’s Default Risk: The Case of Luxembourg, Banque centrale du
Luxembourg working paper.
[19] Giordana, G. and Schumacher, I. (2013), Bank Liquidity Risk and Monetary Policy Empirical Evidence on the Impact of Basel III Liquidity Standards, International Review of Applied Economics.
[20] Green, W. (2012), Econometric Analysis, chapter 10,290-343, New York University, ISBN:0-13-139538.
[21] Hildebrand, P. 2008 Is Basel II Enough? The Benefits of a Leverage Ratio Financial Markets Group Lecture, London School of Economics, London, December 15.
[23] Maechler, A., Mitra, S. and Worrell, D. 2007 , Decomposing Financial Risks and Vulnerabilities in Eastern Europe, IMF working paper WP/07/248.
[24] Mileva, E. (2007), Using Arellano – Bond Dynamic Panel GMM Estimators in Stata, Economics Department, Fordham University.
[25] Moussa, M. (2015), The Relationship between Capital and Bank Risk: Evidence from Tunisia, International Journal of Economics and Finance; Vol. 7, No. 4; 2015.
[26] Molyneux, P. and Wilson, J. (2004), The Profitability of European Banks: A Cross-Sectional and Dynamic Panel Analysis, The Manchester School Vol 72 No. 3 June 2004 1463–6786, 363–381.
[27] Nickell, S. (1981), Biases in Dynamic with Fixed Effects, Econometrica Vol.49, No.6, 1417-1426.
[28] Nicolo, G. (2000), Size, Charter and Risk in Banking: An International Perspective, Board of Governors of the Federal Reserve System, International Finance Discussion Papers, No.689, December 2000.
[29] Papanikolaou, N.and Wolff, C. (2010), Leverage and risk in US commercial banking in the light of the current financial crisis, CEPR Discussion Paper No. DP10890.
[30] Roodman, D. (2006), How to Do xtabond2: An Introduction to “Difference” and “System” GMM in Stata, the Center for Global Development, Working Paper Number 103 December 2006.
[31] Sekine, E .(2011) China's Own Version of Basel III and Its Likely Impact on China's Banking Sector, Nomura Journal of Capital Markets Autumn 2011 Vol.3 No.2.
[32] Short, B.,(1979), The relation between commercial bank profit rates and banking concentration in Canada, Western Europe and Japan. Journal of Banking and Finance 3, 209-219.
[33] Slovik, P. and Cournède, B. (2011), Macroeconomic Impact of Basel III, OECD Economics Department Working Papers No. 844.
[34] Smirlock, M., 1985. Evidence on the (non) relationship between concentration and profitability in banking. Journal of Money, Credit, and Banking 17, 69-83.
[35] Staikouras, C. and Wood, G. (2004), The Determinants of European Bank Profitability, International Business & Economics Research Journal, volume 3, Number 6.
[36] U.K. Financial Services Authority 2009 The Turner Review: A Regulatory Response to the Global Banking Crisis. London.
[37] Vazquez, F. and Federico, P. (2012), Bank Funding Structures and Risk: Evidence from the Global Financial Crisis, IMF Working Paper WP/12/29.
Tables
Table1 Sample of Banks
1 Industrial & Commercial Bank of China (The) – ICBC 2 China Construction Bank Corporation Joint Stock Company 3 Agricultural Bank of China Limited
4 Bank of China Limited
5 Bank of Communications Co. Ltd
6 China Merchants Bank Co Ltd
7 Industrial Bank Co Ltd
8 China CITIC Bank Corporation Limited 9 Shanghai Pudong Development Bank 10 China Minsheng Banking Corporation
11 China Everbright Bank Co Ltd
12 Hua Xia Bank co., Limited
13 Bank of Beijing Co Ltd
14 Bank of Nanjing
15 Bank of Chongqing
16 Huishang Bank Co Ltd
Table 2
Variables definitions
Variable Abbreviation Definition Data source
Banks distance to default
Z-Score (expressed in logarithm)
The sum of return on assets and the Tier 1 capital divided by the standard
deviation of return on assets caculate Return on assets ROA (expressed as
a percentage) Return-on-assets ratio times 100
Bank Scope database
Standard deviation
of ROA sd(ROA) Standard deviation of ROA using 5 years ahead data
Bank Scope database and banks' annual
reports Capital to assets CAR (expressed as
a percentage) Capital-to-assets ratio times 100
Bank Scope database Total assets TA (expressed in
logarithm) Total assets
Bank Scope database Off-balance sheet
activity OBSR Off-balance sheet activities over total assets times 100
Bank Scope database
Change in Gross
domestic product GDP Change in Gross Domestic Product the Word Bank
Non-performing
loans NPLs Non-performing loans over total loans of banks
Banks' annual reports Change Short-term
interest rate IR Change Short-term interest rate
the People's Bank of China (PBOC)
Table 3
Summary statistics
Variable Obs
Mean
Std. Dev.
Min
Max
Z-score
84
1.856374
0.2205122
1.05
2.306627
ROA
85
1.177835
0.1514113
0.741
1.482
CAR
85
6.330241
0.7669038
5.016813
8.753685
Sd(ROA)
84
0.1126495
0.0482691
0.0424
0.2396648
lnTA
85
19.61285
1.227219
16.82168
21.73149
OBSR
85
0.2108805
0.0878107
0.083395
0.5211006
IR
85
2.616
0.4437347
1.77
2.98
(This table reports the summary statistics for all regression variables used in the present paper. This Panel relies on data from 2011 to2015.)
Table 4
Estimation results: CAR/sd(R0A)
---
(1)OLS (2)FE (3)GMM CARsdROA CARsdROA CARsdROA --- ROA/sdROA 5.406*** 6.029*** 6.076*** (19.27) (23.16) (24.02) L.CAR/sdROA 0.119* 0.0793 0.0859* (2.44) (1.42) (2.10) IRchange -13.72*** -15.47** -16.63*** (-3.70) (-2.84) (-5.86) L.lnTA 0.415 -1.963 -0.887*** (0.45) (-0.24) (-4.48) _cons -18.19 22.28 (-1.04) (0.14) --- N 67 67 67 ---
Regression results for the period (2011-2015). The dependent variable is capital-to-assets ratio divided by the standard deviation of return on assets. As independent variables I include the return-on-assets ratio (ROA/sdROA), lagged capital-to-assets ratio(CAR/sdROA) , change in short-term interest rate (IR change) and lagged total assets(TA).
The number of total observations is 67. ***, **, * correspond to 1%, 5%, and 10% level of significance respectively for a two-tailed distribution.
Table 5
Estimation results: Z-score
---
(1)OLS 2)FE (3)GMM Zscore Zscore Zscore --- L.ROA/sdROA 0.0263* 0.00232 0.0841* (2.37) (0.13) (2.20) L.CAR/sdROA -0.000219 -0.000446 -0.0116* (-0.12) (-0.16) (-2.00) IRchange -0.00524 -0.0354 0.00900 (-0.09) (-0.30) (0.10) L.lnTA 0.00735 0.307 0.115*** (0.49) (1.76) (10.89) L.OBSR -0.116 0.399 -2.947** (-0.57) (0.42) (-2.69) _cons 1.456*** -4.205 (5.25) (-1.24) --- N 67 67 67 ---
Regression results for the period (2011-2015). The dependent variable is the bank distance to default (Z-score). As independent variables, I include the lagged return-on-assets ratio (ROA/sdROA), lagged capital-to-assets ratio(CAR/sdROA), change in short-term interest rate (IR change), lagged total assets(TA) and lagged off-balance sheet activities-to-total assets ratio (OBSR).
The number of total observations is 67. ***, **, * correspond to 1%, 5%, and 10% level of significance respectively for a two-tailed distribution.
