Finx

Bank risk taking, efficiency, and law enforcement: Evidence from Chinese city commercial banks☆

Jianhua ZHANG a, Peng WANG a, Baozhi QU b,⁎ a People's Bank of China, China b Skolkovo Institute for Emerging Market Studies, Moscow School of Management, Beijing

a r t i c l e i n f o a b s t r a c t

Article history: Received 30 May 2011 Received in revised form 7 November 2011 Accepted 20 December 2011 Available online 29 December 2011

We investigate bank risk taking, efficiency and their relation to law enforcement using a unique sample of 133 Chinese city commercial banks across 31 regions for the 1999–2008 period. We find that stronger law enforcement tends to promote greater bank risk taking in the region. Furthermore, employing a stochastic distance function approach, our analysis shows that the performance of Chinese city commercial banks, as measured by bank efficiency, is heavily influenced by the effectiveness of law enforcement in the region. Better legal envi- ronment, higher efficiency in the legal system, and stronger protection of intellectual property right are associated with a higher level of efficiency among these banks.

© 2012 Elsevier Inc. All rights reserved.

JEL classification: G21 G28 K0

Keywords: Bank risk taking Bank efficiency Law enforcement Stochastic frontier analysis China

1. Introduction

The growing law and finance literature (pioneered by La Porta, Lopez-de-Silanes, Shleifer, & Vishny, 1998) has documented that a well-developed legal system enhance the enforcement of financial contracts. Since banks are usually in the center of a country's financial system and their operation and performance depend heavily on the contract enforcement (e.g., a loan contract), legal institutions may have a significant impact on banking. Previous research has reported some evidence of such influences. For instance, Cole and Turk-Ariss (2008) find that banks have higher loan ratios in countries with English legal origin and weaker creditor rights. Houston, Lin, Lin, and Ma (2010) document that stronger creditor rights protection tends to promote greater bank risk taking. The legal system and the effectiveness of law enforcement may also influence the operation and performance of a country's banking sector through other channels, such as by enhancing lending technology (Berger & Udell, 2006), increased availability of loans (Qian & Strahan, 2007), or reduced rigidity of regulations (Demirguc-Kunt, Laeven, & Levine, 2003).

Most of the abovementioned studies use a standard cross-country setting to examine the links between law and bank opera- tions on the basis of country-level institutions. However, such a setting has two limitations. First, cross-country analysis may not sufficiently account for the great heterogeneities across countries, such as the difference in history, culture, natural endowment,

China Economic Review 23 (2012) 284–295

☆ This paper represents the views of the authors and does not necessarily represent the People's Bank of China's views or its policy. The views expressed herein should be attributed to the authors and not to the People's Bank of China. ⁎ Corresponding author. Tel.: +86 10 64981634x217; fax: +86 10 64981634x208.

E-mail address: Baozhi_Qu@skolkovo.ru (B. Qu).

1043-951X/$ – see front matter © 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.chieco.2011.12.001

Contents lists available at SciVerse ScienceDirect

China Economic Review

etc. Second, by assuming uniform legal and institutional environment within a single country, these studies fail to take into consideration the significant degree of variation in the institutions and law enforcement within a particular country, which has been well documented and proved important in affecting economic and financial outcomes.1 Micro data analysis at the bank level that takes into consideration the variation in institutional quality within a country can therefore provide valuable empirical evidence that complements and extends the existing country-level research. As Acemoglu (2005) correctly points out, questions related to the importance of legal institutions “will be almost impossible to answer with cross-country data alone, and micro data investigations, for example, exploiting differences in regulations across markets and regions appear to be a most promising avenue” (p. 1045).

Our study contributes to the literature by providing bank-level evidence of a link between bank risk taking and law enforce- ment using a within-country setting. It further provides a direct examination of the potential effects of the law enforcement on bank performance (as measured by bank efficiency), which is noticeably absent in this recent literature as correctly argued by Berger and Mester (1997) and Hasan, Wang, and Zhou (2009). We take advantage of a unique data set which has extensive information on 133 Chinese city commercial banks (CCBs) from 1999 to 2008. Such a rich data set of Chinese CCBs serves our purpose particularly well because China is a big country with great regional variation. Although China is a centralized country so that the laws on book are highly uniform across regions, the quality of law enforcement varies significantly across provinces. CCBs usually conduct their business only within the boundaries of a city or province, which provides us with a unique opportunity to study how regional institutional environment affect bank risk taking and performance while controlling for a large set of bank- level characteristics and country-level variables.2 In addition, since CCBs are relatively small and they do not have a significant impact on the regional institutional development, the concern for the endogeneity of legal institutions and law enforcement is significantly reduced in this setting (more discussion in Section 4.1).

In China, a number of market-oriented reforms took place during the past decade and these reforms led to important changes in the institutional environment of Chinese banks.3 These changes developed in a very unbalanced way across different provinces, especially in terms of the effectiveness of law enforcement. According to Fan et al.'s (2000, 2004, 2006, 2010) measures, the necessary market intermediaries that are essential for the effective law enforcement are more developed in some provinces (such as Zhejiang) than others (such as Qinghai). Similarly, the efficiency of the regional legal system and the effectiveness of certain laws (such as the protection of the intellectual property rights) also have great variations across provinces. Such difference in law enforcement could affect bank risk taking and the efficiency of banks in the region both directly and indirectly. For instance, more effective law enforcement reduces credit risk by providing more protection to banks when the loan contract is defaulted, thus encourage banks to take more risks and make more loans (Houston et al., 2010). In addition, better legal protection (such as protection of the intellectual rights) creates a better market environment for firms thus may improve the performance of borrowers. This will in turn affect banks indirectly by reducing the credit risk of banks' loan portfolio and encourage banks to take more risk by lending to more risky borrowers such as new startups of IT companies.

Our empirical results show that stronger law enforcement encourages risk taking among Chinese banks. In addition, using a stochastic function approach, our study shows that better legal environment, more efficient law enforcement and better protec- tion of intellectual property rights are significantly related to greater bank efficiency. The findings are in general consistent with the cross-country research by Houston et al. (2010) and with John, Litov, and Yeung (2008)'s argument that stronger legal pro- tection could lead to riskier but value-enhancing investments.

By focusing on CCBs in China, our paper also makes a contribution to the efficiency study of Chinese banks. A number of studies have examined bank efficiency in China (e.g., Berger, Hasan, & Zhou, 2009; Chen, Skully, & Brown, 2005; Hasan et al., 2009; Jiang, Yao, & Zhang, 2009). Most of these studies focus on the listed Chinese banks that are relatively large with nationwide business operations, probably because the financial information of these banks is publicly available and thus easily obtained. However, one of the most important changes in the Chinese banking sector in the past decade is the emergence and rapid development of a new breed of dynamic regional banks — CCBs. In contrast to the major state-owned commercial banks, these banks are recent entities, have a lower level of state ownership and operate mainly in the regional market of a city or province. As of 2008, there were more than 136 CCBs in China, with at least one in almost every major city. Most of them, however, are unlisted. This constraint on data availability makes CCBs significantly underrepresented in previous efficiency studies of Chinese banks. The very few studies of CCBs have had to rely mainly on survey data, and their sample size is significantly limited (e.g., Ferri, 20094). The rich data set of Chinese CCBs in our study thus allows us to fill the gap by providing empirical evidence on their risk-taking behavior and the efficiency of CCBs. Our study also differs from the previous research (e.g., Ferri, 2009) in several other ways: we use a distance function approach to study the efficiency of CCBs, which may better capture the quality of banking

1 For instance, corruption in some regions may be more pervasive than that in other regions, and a country may have strong rules and regulations on the books but weak law enforcement in some regions and for some firms. Berkowitz and Clay (2006) show that the quality of state courts varies significantly across U.S. states and is greatly affected by the initial conditions of a state. Laeven and Woodruff (2007) find significant variation in the quality of the Mexican legal system. Acemoglu and Dell (2010)) argue that both de jure and de facto institutions vary greatly within countries. Using World Bank enterprise survey data, Ma, Qu, and Zhang (2010) find that the average within-country variation in judicial quality is much greater than the cross-country variation in judicial quality.

2 This empirical setting distances our paper from similar research (e.g., Hasan et al., 2009) that focuses on Chinese banks that have country-wide banking op- eration. If a bank is allowed to operate across different provinces, then it will be difficult to isolate the effect of regional institutional environment on its efficiency.

3 Given that the banking sector in China has been under heavy regulation of the government and most of the reforms related to this sector followed a “top- down” procedure, we expect that the causality goes from market development and institutional changes to banks' risk-taking and efficiency and not the other way around.

4 In addition, Ferri (2009) did not consider the risk-taking of Chinese CCBs.

285J. Zhang et al. / China Economic Review 23 (2012) 284–295

institutions and their functions in the economy (Hasan et al., 2009). We also pinpoint regional factors that affect the risk-taking and efficiency of CCBs and focus on the difference in law enforcement across regions.

The rest of the paper is organized as follows. The next section explains the institutional background of the Chinese banking sector and the development of CCBs. Section 3 describes the estimation procedure and data. Section 4 reports the empirical results, and Section 5 concludes the paper.

2. The development of CCBs in China

The Chinese banking system has gone through significant transformation since 1978. The all-in-one mono-bank system under the planned economy was replaced by a banking sector dominated by the four state-owned banks (the Big Four). In addition to the Big Four, there are quite a number of joint-stock commercial banks (JSCBs), CCBs and other types of financial institutions operating in the banking sector. Since 2006, a number of foreign banks have been allowed to operate in China with full banking licenses. As of the end of 2008, the total assets of the Chinese banking sector amounted to RMB62.4 trillion. The Big Four, the JSCBs, and the CCBs held 71.8% of the total assets (51% by the Big Four, 13.1% by the JSCBs, and 6.6% by CCBs).

CCBs emerged in China only in 1995. On 7 September 1995, the State Council issued Guidelines of the Establishment of the City Cooperative Bank. City cooperative banks (the precursors of CCBs) were set up in 35 medium- and large-sized cities. The first gen- eration of CCBs were formed through the merging of 2194 urban credit cooperatives, rural credit cooperatives and local financial service institutions. Hence, CCBs were also regarded as JSCBs at the time. Similar banks were later established in other cities. By 2008, the number of CCBs had increased, with at least one CCB in almost every major city. However, these banks are unevenly dis- tributed, with more branches in the economically more developed eastern provinces (such as Guangdong) than in the less developed western provinces (such as Qinghai). By regulation, CCBs can provide financial services only in their own administrative region.5

Therefore, a typical CCB is much smaller than a Big Four bank. For instance, the amount of total loans per CCB was RMB6219 million in 2008 (2000 price level), compared to RMB1,841,047 million for a Big Four bank and RMB173,869 million for a JSCB. Table 1 (the upper panel) presents the summary statistics of the basic financial ratios for CCBs used in our analysis.

In addition, CCBs differ from the Big Four in that the former have more diversified shareholders from different classes in society, including individuals, private enterprises, institutional investors, state-owned enterprises and the treasury of local governments. On average, only about one-fourth of CCB shares are held directly by SOEs. Hence, CCBs are subject to less state intervention and may have relatively better corporate governance compared to the Big Four (Ferri, 2009). Using survey data on 20 CCBs from three provinces, Ferri (2009) studies the performance of these banks, finding that they outperform the Big Four and that geographic location significantly affects CCB performance and business.

3. Data and model

3.1. Sample and measure of law enforcement in China

Our data set was obtained from the PBC, the central bank of China. It includes extensive information on 133 Chinese CCBs, including listed and unlisted banks, over 10 years (1999–2008). After deleting some observations with incomplete data, we have a sample of 1083 observations. Table 2 shows the sample distribution over time.

We rely on the various issues of the Market Development Report of Fan et al. (2000, 2004, 2006 and 2010) to obtain measures of law enforcement in 31 provinces of China. The reports publish annual information about indices that measure the law enforce- ment across different provinces from 1998 to 2007 and have been used by some other studies (e.g., Chen, Schipper, Wang, & Xiao, 2011; Wang, Wong, & Xia, 2008). Three indices from Fan et al. (2000, 2004, 2006, 2010) are used in our study. The first index, “Legal Environment”, is based on the number of lawyers and certified public accountants in relative to the population in the region. Since the effectiveness of law enforcement depends heavily on the expertise of lawyers and accountants, higher ratio of these professionals to the population indicates more effective law enforcement in the region. The second index, “Efficiency of Legal System”, concerns the efficiency of law enforcement, and it is based on survey results of firms. Every year, a number of firms of each province are surveyed and they are asked to evaluate the efficiency level of the local law enforcement. A high value in this variable indicates higher efficiency level of local law authority, as perceived by surveyed firms in the region. The third variable, “Protection of Intellectual Property Rights”, is based on the number of patents applied and granted in a certain year relative to the number of scientific and technical personnel of the region. Since the creation of intellectual property rights depends critically on the legal protection of the innovating firms' rights, a high value in this variable could reflex better enforce- ment of laws that protect such rights in the region.

As an illustration, Table 3 presents the index of the legal environment. The index shows significant variation both across regions and over time.6

5 The branching regulation on CCBs has been relaxed in the recent 2–3 years which allows CCBs to establish branches in other regions. Our sample period is from 1999 to 2008 and therefore our results are not affected by this recent regulatory change.

6 The data of Fan et al. (2000, 2004, and 2006) cover four sub-periods (1997–1999, 1999–2000, 2000–2002, 2001–2005) and cannot be compared across the different periods. In the newest publication in 2010, the authors adjusted the data of the previous years so that they are comparable over time. We used the ad- justed data in the regressions. As a robustness test, we have also tried our own transformation of the data using 2001 as the base year to make them comparable. The results are very similar.

286 J. Zhang et al. / China Economic Review 23 (2012) 284–295

The effects of law enforcement on bank risk taking and efficiency are likely to be lagged. Consequently, we use the value of the measures of law enforcement of the previous year in the analysis. Our results are not sensitive to how the lagged effects are taken into account in the regressions.

3.2. Bank risk taking

As a measure of bank risk taking, we use the natural logarithm of the Z-score which is the number of standard deviations that a bank's rate return of assets has to fall for the bank to become insolvent. The Z-score has been widely used in the recent literature (e.g., Demirguc-Kunt & Huizinga, 2010; Houston et al., 2010; Laeven & Levine, 2009). Specifically, Z-score=(ROA+CAR)/ σ(ROA), where ROA is the mean rate of return on assets and CAR is the mean equity-to-assets ratio. σ(ROA) is the standard deviation of ROA. Higher Z-score implies more stability. We calculate a Z-score for a bank over time based on annual accounting

Table 1 Summary statistics of bank characteristics (CCBs, 1999–2008) and regional variables.

Mean Std. dev. Min. Max. Obs.

Net interest income 323.68 644.86 −101.04 8418.23 1083 Non-interest income 15.23 52.66 −14.54 973.10 1083 Total loans 6219.47 12,269.37 76.55 130,384.40 1083 Deposits 10,220.21 20,319.47 387.73 211,077.70 1083 Other earning assets 5691.75 12,991.87 124.70 156,363.50 1083 Interest expenses 158.89 346.65 1.90 4027.82 1083 Overhead 182.32 302.00 3.80 2907.11 1083 Net fixed assets 165.61 205.08 9.08 2064.14 1083 Equity/assets 0.0435 0.0361 −0.1273 0.3741 1083 NPL ratio 0.1588 0.1735 0.0000 0.8177 1083 Loan loss reserves/total loans 0.0139 0.0154 −0.0115 0.1147 1083 Loans/deposits 0.6143 0.1679 0.0670 1.2543 1083 Interbank funds/(interbank funds+deposits) 0.1015 0.0947 0.0000 0.4959 1083 Long-term loans/total loans 0.3007 0.1886 0.0000 0.8431 1083 Securities/earning assets 0.1701 0.1118 0.0000 0.5185 1083 Ln(Z score) 2.6491 1.2299 0.0000 4.9231 550 Operating income growth rate 0.5386 1.1499 −3.1741 9.8409 550 Non-interest income/operating income 0.0415 0.0797 −0.1140 0.7828 550 Dynamic governance 0.0691 0.2538 0 1 550 Regional variables:

Legal environment 4.76 2.42 1.15 16.61 1083 Efficiency of the legal system 4.13 1.98 −0.46 10.00 1083 Protection of intellectual property right 4.53 5.70 −0.24 40.47 1083 Loans/GDP 0.9447 0.2241 0.5913 2.2587 1083 Log(GDP) 13.2536 0.7509 10.0159 14.7825 1083 GDP growth rate 0.1167 0.0245 0.0510 0.2380 1083

Note: Input and output factors are measured in million RMB at the 2000 price level.

Table 2 Sample distribution.

Year Number of banks

1999 84 2000 89 2001 102 2002 107 2003 108 2004 108 2005 112 2006 120 2007 120 2008 133 Total 1083

Note: In January 2007, ten CCBs (including the Wuxi Commercial Bank) merged to form the Jiangsu Commercial Bank; therefore, the number of CCBs did not greatly increase from 2006.

287J. Zhang et al. / China Economic Review 23 (2012) 284–295

data for the past five years. Following Laeven and Levine (2009) and Houston et al. (2010), we use the natural logarithm of the Z-score in the regressions to smooth out the skewness of the Z-score.

To examine the impact that legal environment has on bank risk taking, we implement the following fixed-effects estimation model:

Zijt ¼ α þ β1Law Enforcement Measuresjt þ β2Macro Controlsjt þ β3Bank Controlsit þ β4Bank Fixed Effecti þ β5Time Variablet þ ε:

ð1Þ

The dependent variable is the log Z-score, and the key independent variable is the measure for the law enforcement in the region. We also control for a number of other macro variables at the provincial level to reduce the omitted variable bias: the degree of financial deepening, measured by the ratio of total loans of all banks in the province to GDP of the province; size of the economy, measured by the natural logarithm of gross domestic product (GDP) at the beginning of the year (the GDP value is adjusted to the 2000 price level using the GDP deflator); and economic growth, measured by the annual GDP growth rate. Following previous literature (e.g., Houston et al., 2010), we include a large set of bank level control variables in the regressions: bank size, the quadratic form of bank size, growth rate of the operating income, non-interest income to operating income ratio, equity to assets ratio, interbank funds to interbank funds plus deposits ratio, non-performing loans (NPLs) ratio, ratio of securities to total earnings assets, and a dummy variable for dynamic governance change in the bank (IPO or the introduction of foreign investors). In addition, we control for firm fixed-effects and time effect in the regressions. The measures of law enforcement and other macro variables are both time (t) and region (j) variables while the bank-level control variables are bank specific variables (i) that vary over time.

3.3. Bank efficiency: the distance function approach

A stochastic frontier approach is commonly employed for efficiency analysis. Fries and Taci (2005) argue that such an approach is appropriate in efficiency studies of transition economies, in which problems related to measurement error and an uncertain economic environment are more likely to prevail. Following previous studies (e.g., Jiang et al., 2009; Lovell,

Table 3 Legal environment of different regions in China (1998–2007).

Regions 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Beijing 9.20 4.52 5.12 7.37 7.62 7.63 8.10 7.78 7.87 8.41 Tianjin 3.97 4.05 4.35 5.86 6.50 6.95 7.49 8.51 8.98 9.92 Hebei 1.59 2.95 3.15 3.26 3.39 3.48 3.90 5.11 5.13 5.27 Shanxi 3.11 3.24 3.94 2.91 3.08 3.20 3.61 4.38 4.51 4.78 Inner Mongolia 2.30 2.87 3.42 3.30 3.38 3.56 3.96 4.47 4.43 4.50 Liaoning 3.24 3.16 3.54 4.28 4.51 5.15 5.46 6.35 6.55 7.23 Jilin 2.79 3.35 3.36 3.65 3.69 3.83 3.89 4.79 4.84 5.37 Heilongjiang 2.65 3.13 3.51 4.07 4.24 4.44 4.56 5.30 5.15 5.46 Shanghai 5.32 3.92 5.51 9.42 10.52 12.15 11.06 12.84 13.87 16.61 Jiangsu 2.17 3.66 3.76 5.20 5.69 6.18 6.61 8.18 9.07 11.50 Zhejiang 2.32 3.49 3.99 6.19 7.16 8.09 8.39 10.59 11.97 13.89 Anhui 1.34 3.15 3.05 2.53 2.67 2.63 3.15 4.99 5.53 5.99 Fujian 2.36 3.66 4.31 4.68 5.05 5.23 5.30 6.41 6.61 6.92 Jiangxi 1.55 2.76 2.59 2.23 2.61 3.01 3.38 4.32 4.28 4.75 Shandong 2.22 3.15 3.11 3.80 4.18 4.67 5.13 6.14 6.71 7.37 Henan 1.50 2.85 3.00 3.06 3.12 3.07 3.38 4.52 4.66 4.99 Hubei 2.04 2.87 3.04 2.96 3.24 3.74 3.81 4.87 5.02 5.79 Hunan 1.68 1.35 1.50 2.29 2.64 3.13 3.75 4.29 4.20 4.32 Guangdong 3.85 4.17 6.02 7.11 7.69 8.45 8.86 10.64 11.47 12.59 Guangxi 1.90 2.89 2.99 2.92 3.10 3.20 3.17 3.80 3.70 4.23 Hainan 2.09 3.73 3.93 3.86 3.80 3.64 3.57 3.63 3.74 3.87 Chongqing 2.19 2.13 1.72 2.37 2.74 3.29 3.95 4.89 5.20 5.61 Sichuan 1.96 2.65 2.43 3.49 3.81 4.03 4.11 5.04 5.24 5.96 Guizhou 1.27 2.47 2.32 1.82 1.88 1.96 2.16 3.12 3.20 3.76 Yunnan 1.85 2.26 2.74 2.17 2.33 2.41 2.75 3.91 4.15 4.63 Tibet NA NA 0.00 1.90 1.91 2.25 2.63 3.60 3.78 3.89 Shannxi 2.93 1.73 1.92 1.78 2.29 2.47 2.88 3.96 4.29 4.99 Gansu 1.45 2.27 1.89 1.15 1.36 1.52 2.11 3.34 3.57 3.79 Qinghai 1.88 2.75 3.02 1.47 1.58 1.49 1.53 1.85 2.06 2.79 Ningxia 2.37 2.92 2.58 1.67 1.86 2.24 2.83 3.47 3.52 3.80 Xinjiang 2.83 2.17 2.12 3.28 3.85 4.37 4.48 4.83 4.64 4.56 Country average 2.60 3.01 3.16 3.61 3.92 4.24 4.51 5.48 5.74 6.37

288 J. Zhang et al. / China Economic Review 23 (2012) 284–295

Richardson, Travers, & Wood, 1994; Wang & Schmidt, 2002), we adopt the estimation procedure proposed by Battese and Coelli (1995). The output distance function7 for a firm producing m outputs using n inputs is defined as:

ln Do x t ; yt; t

� � ¼ α0 þ

Xn

k¼1 αk ln x

t k þ

Xm

j¼1 βj ln y

t j þ

1 2

Xn

k¼1

Xn

h¼1 αkh ln x

t k ln x

t h þ

1 2

Xm

j¼1

Xm

l¼1 βjl ln y

t j ln y

t l

þ Xn

k¼1

Xm

j¼1 γkj ln x

t k ln y

t j þ φtt þ

1 2 φttt

2 þ Xn

k¼1 ξkt ln x

t kt þ

Xm

j¼1 τjt ln y

t j t;

ð2Þ

where x is input, y is output (multiple inputs, multiple outputs) and t is time. Do(x t,yt,t) is homogeneous of degree 1 in y. For

individual firm i, the following equation can be derived:

ln Dt oi − ln ytmi ¼ α0 þ

Xn

k¼1 αk ln x

t ki þ

Xm−1

j¼1 βj ln y

t ji

� �� þ 1 2

Xn

k¼1

Xn

h¼1 αkh ln x

t ki ln x

t hi þ

1 2

Xm−1

j¼1

Xm−1

l¼1 βjl ln y

t ji

� �� ln ytli � ��

þ Xn

k¼1

Xm−1

j¼1 γkj ln x

t ki ln y

t ji

� �� þ φtt þ 1 2 φttt

2 þ Xn

k¼1 ξkt ln x

t kit þ

Xm−1

j¼1 τjt ln y

t ji

� �� t þ vti ;

ð3Þ

where (yji t)*=yji

t/ymi t (j=1,2,…m−1). We further define uit=−ln Doi

t: i.e., ui t follows a non-negative truncated normal distribu-

tion. In addition, vi t~N(0,σv

2): i.e., vi t follows a standard normal distribution. ui

t and vi t are independent.8 Following Battese

and Coelli (1995), the inefficiencies (ui t) can be decomposed as ui

t=zi tδ+ei

t, where zi t denotes the exogenous factors that affect

the technical inefficiency term, δ denotes the coefficients of these factors and ei t is a random variable, which is independently

distributed as truncations of a normal distribution N(0,σu 2) (here, ei

t≥−zitδ). How to define and measure the inputs and outputs of a bank is an important issue in the bank efficiency research. The two

basic methods are the production and the intermediation approach. Following the previous literature (e.g., Jiang et al., 2009), we use the latter9 and implement two different models: the first is an income-based model and the second is an earning assets-based model. In both models, input factors include interest expenses, non-interest expenses (operating expenses), and net value of fixed assets. However, the output factors differ in the two models. In the first model, the output variables are the banks' income factors which include net interest income and non-interest income. In the second model, the output factors are based on the banks' earning assets, including total loans, total deposits, other earning assets, and non-interest income. The definitions of the input and output factors included in these two models are presented in Table 4.

Following previous studies (e.g., Fu & Heffernan, 2007; Jiang et al., 2009), factors in the analysis of technical efficiency are divided into macro factors and bank-level characteristics as listed below. The main macro factor of interest is the measure of law enforcement of the region. Other province-level macro factors include: the degree of financial deepening, size of the economy, and economic growth. In addition, we consider a large set of bank characteristics in the inefficiency analysis.

3.3.1. Governance variables First, we include a dummy variable that measures the selection governance of a bank (Berger, Clarke, Cull, Klapper, & Udell,

2005), which takes the value of one if the bank has foreign investors as big shareholders or is listed in a financial market (including both domestic and oversea stock exchanges) during the sample period, and zero otherwise. The coefficient on this variable indicates whether the change in governance resulting from the introduction of foreign investors or being listed has an impact on bank efficiency. In our sample, 9.8% of banks (13 among 133 CCBs) have been selected for such governance change by 2008. Second, we include a dummy variable that measures the dynamic governance of a bank. The dynamic governance variable takes the value “zero” prior to the aforementioned governance changes in a bank and “one” after the governance changes to capture short-term effects of such changes. If a bank does not experience these governance changes during the sample period, the value of this variable will always be zero. The coefficient on this variable measures whether a bank's efficiency is significantly different before and after these governance changes. The mean value of this variable in our sample is 0.06.

7 A major advantage of the distance function approach is that it can be applied in the case of multiple inputs, multiple outputs or absence of price information when the traditional dual approach is inapplicable (Jiang et al., 2009).

8 Simultaneous equation bias may exist when both inputs and outputs are included in the distance function as regressors. After the normalization procedure, output ratios may be treated as exogenous (Coelli & Perelman, 1996).

9 Berger and Mester (1997) argue that under the production approach, financial institutions are thought of as primarily carrying out services for account holders. These institutions perform transactions and process documents for customers, including loan applications, credit reports, checks and other payment ser- vices, and insurance services. Under this approach, output is best measured by the number and type of transactions or documents processed over a given period. Unfortunately, such detailed transaction flow data are typically proprietary and not publically available. As a result, data on the stock of the number of deposit or loan accounts serviced or insurance policies outstanding are used instead. Also, only physical inputs such as physical capital, labor and their cost as well as op- erating expenses (excluding interest expenses) are used. In contrast, under the intermediation approach, financial institutions are thought of as primarily inter- mediating funds between savers and investors. Because service flow data are not usually available, the flows are typically assumed to be proportional to the financial value of the accounts, such as the dollar amount of loans, deposits or insurance policy premiums as well as the value of other earning assets. The input of funds and their interest costs should also be included in the analysis together with physical inputs. As service flows to depositors are proportional to the value of deposits, if we treat deposits as both input and output, then interest expenses are usually used as costs. In addition, the interest expense-to-deposit ratio is used as the price of the input and the value of deposits as the output.

289J. Zhang et al. / China Economic Review 23 (2012) 284–295

3.3.2. Risk factors These include the ratio of non-performing loans (NPLs) to total loans, which measures the credit risk of the loan portfolio; total

loans-to-deposits ratio, which measures the bank's liquidity risk and operating efficiency; and ratio of interbank funds (including borrowing from the central bank, interbank loans from other commercial banks and funds from the repo markets) to the sum of interbank funds and total deposits, which measures the bank's reliance on short-term funds from the interbank loan market and is an indicator of the bank's liquidity position.

3.3.3. Asset structure Asset structure variables include the ratio of securities to total earning assets (including cash and deposits, reserves at the

central bank, deposits at other financial institutions, reverse repurchase agreements and total loans and securities) and the ratio of long-term loans to total loans.

In addition, a time variable is included to control for the time effect, using 1999 as the base year. Table 1 reports the summary statistics of these bank characteristics for CCBs. The variable values are converted to the 2000 price level with the GDP deflator and measured in million RMB.10 Our data set mainly covers the RMB transactions of the CCBs.11

4. Empirical results

4.1. Law enforcement and bank risk taking

Table 5 reports the fixed effect regression results of the Z-score (log value) on the law enforcement variables and various con- trol variables. The calculation of standard variation and t values is clustered at the bank level. We find that measures of the law enforcement are in general negatively associated with the Z-score (again, higher value of Z-score indicates more stability). This shows that better law enforcement in a region encourages banks to take more risk. Specifically, the index of legal environment has a significantly negative coefficient (regressions 1 and 4). The coefficient on the efficiency of the regional legal system has the expected sign but the result is insignificant (regression 2). The coefficient estimate on the protection of intellectual property right has a negative sign as expected and is significant (at the 5% level), indicating that better law enforcement as represented by the stronger protection of intellectual property rights in the region increases bank risk taking (regressions 3 and 5). These results are not sensitive to the inclusion or exclusion of other macro variables in the regressions (regressions 4 and 5).12

10 In our analysis, we need to take the logarithm of both input and output factors. To avoid taking the logarithm of zero or a negative number, it is a common practice in the literature to find the minimum value for each factor (usually a negative number), calculate its absolute value plus one (∣y∣+1) and then add this number to the initial variable value before taking the logarithm. However, if the quantitative level of ∣y∣ is exceptionally large, then the input–output relationship may change significantly for some banks. In this paper, we adopt the following procedure to tackle this issue. We use negative one million as the benchmark and add 1.01 million to the initial value of the variable. If the result is less than 1, then we treat the logarithm value of this observation as zero. For the rest, we directly take the logarithm. Although this approach effectively imposes some “penalty” on negative values and sacrifices some information by smoothing out the variation in some variables with negative and large absolute values, it allows us better control of the potential distortion of the estimation results resulting from exception- ally large negative values. Our main results are not sensitive to the choice of the benchmark value or following the common practice to obtain the logarithm values. 11 Given the difference in domestic and foreign currency transactions among banks and because RMB transactions account for the majority of the banking busi- ness in Chinese domestic financial institutions, we focus on the RMB business of banks in this paper. Foreign banks own only a small share of the total assets in the Chinese banking sector and thus are excluded from our analysis. 12 We have also tried other model specifications with one (or two) of the regional control variables (loans/GDP, ln(GDP), and GDP growth rate) being included in the regressions and our main estimation results remain unchanged. We thank one anonymous referee for raising this point.

Table 4 Comparison of the income-based model and the earning assets-based model.

Income-based model Earning assets-based model

Input factors Interest expenses, non-interest expenses (operating expenses), net value of fixed assets Output factors Net interest income, non-interest income Total loans, total deposits, other earning assets, non-interest income Note 1. Interest expenses: costs on deposits and other earning assets.

2. Non-interest expenses (operating expenses) include employee expenses, business expenses, depreciation and amortization, business taxes and other expenses. 3. Net value of fixed assets measures the input factor of fixed assets. 4. Non-interest income includes net commission income, net exchange income, other operating income, net income from property leasing, investment income and so forth. 5. Non-interest income is included in the earning assets-based model as an output factor to control for the impact of off- balance-sheet activities (see Fu & Heffernan, 2007; Jiang et al., 2009; Orea, 2002). 6. Deposits are regarded as both input and output, because the service provided to the depositors is closely associated with the amount of deposits. 7. Other earning assets are the remaining assets after loans and fixed assets are deducted from the total assets, including cash, deposits in the central bank and other financial institutions, reverse repurchase agreements, securities and so forth (see Fu & Heffernan, 2007; Jiang et al., 2009; Orea, 2002).

290 J. Zhang et al. / China Economic Review 23 (2012) 284–295

While the endogeneity of institutions in the form of omitted variables bias or reverse causality is common in this type of stud- ies, our inclusion of a large set of control variables at both the bank level and the province level ought to minimize the omitted variable bias. As for the concern for reverse causality, given that CCBs are relatively small in size, their influence on regional institutions and law enforcement is mostly likely to be trivial, if any. Therefore the causality should go from law enforcement to CCBs' risk taking, and not the other way around (also see footnote 3).13

Examination of the coefficients on the various control variables shows interesting results that are generally consistent with previous research (e.g. Houston et al., 2010). We find an inverse U-shape relation between bank size and bank risk taking. In addition, banks with higher growth rate of operating income, greater NPL ratio and higher ratio of interbank borrowing engage in more risk taking. In contrast, banks that rely more on non-interest income, hold more security assets, and have a greater capital ratio tend to take less risk.

4.2. Law enforcement and bank efficiency: results of the income-based model

In this section, we investigate how regional law enforcement affects bank efficiency using the income-based model. Table 6 presents the one-step maximum-likelihood estimation results of the inefficiency analysis. All the regression models have a high γ value, indicating a high overall significance level (at the 1% level). The signs of first-order coefficients in all the regression models are as expected and most of the coefficient estimates are significant (at least at the 10% level).14

We find strong effects of regional law enforcement on the efficiency of CCBs. The index of the legal environment, efficiency of the legal system, and the protection of intellectual property right are all negatively related to the technical inefficiency of CCBs as expected, and the coefficients are highly significant (at the 5% level and above). In other words, good law enforcement significantly

Table 5 Z-score and law enforcement: bank level fixed-effects regressions.

(1) (2) (3) (4) (5)

Legal environment −0.3199** – – −0.2940** – [−2.16] [−2.26]

Efficiency of the legal system – −0.0640 – – – [−0.35]

Protection of intellectual property right – – −0.1128** – −0.0879** [−2.18] [−2.27]

Loans/GDP −1.3384 −2.1380** −1.3845 – – [−1.38] [−2.13] [−1.52]

Ln(GDP) 2.9945 −0.4204 4.2630 – – [0.97] [−0.17] [1.16]

GDP growth rate −25.2631** −15.2031 −29.2678** – – [−2.04] [−1.40] [−1.99]

Log(size) 2.3130* 2.6863* 2.4319* 1.8820 1.9598 [1.67] [1.95] [1.76] [1.29] [1.33]

Log(size) square −0.2389 −0.2748* −0.2591* −0.2103 −0.2242 [−1.50] [−1.73] [−1.65] [−1.29] [−1.36]

Operating income growth rate −0.5861** −0.6220** −0.5922** −0.5537** −0.5547** [−2.48] [−2.44] [−2.50] [−2.25] [−2.24]

Non-interest income/operating income 2.1699 1.7856 2.0583 2.4061* 2.2932* [1.58] [1.21] [1.53] [1.75] [1.65]

Equity/asset 17.6440*** 17.0764*** 17.5486*** 17.2806*** 17.0586*** 4.38 [4.19] [4.40] [4.10] [4.07]

Interbank funds/(interbank funds+deposits) −2.6707* −2.8076** −2.8491** −2.4886* −2.6274* [−1.92] [−1.99] [−2.02] [−1.78] [−1.85]

NPL ratio −2.7815*** −2.6202** −2.7400*** −2.9128*** −2.8658*** [−2.66] [−2.42] [−2.68] [−2.76] [−2.70]

Securities/earning assets 4.0963** 4.3088** 4.4526** 4.2846** 4.6221*** [2.25] [2.35] [2.43] [2.53] [2.84]

Dynamic governance 0.2974 0.3303 0.2537 0.2939 0.2631 [1.20] [1.31] [0.97] [1.27] [1.11]

Constant −42.4806 −1.3618 −59.7120 −4.9656 −6.0744 [−1.05] [−0.04] [−1.25] [−0.70] [−0.86]

Year Yes Yes Yes Yes Yes Observations 550 550 550 550 550 Cluster 108 108 108 108 108 R-square (within group) 0.3795 0.3705 0.3840 0.3619 0.3629

Note: The dependent variable is the natural logarithm of the Z-score. Z-score=(ROA+CAR)/σ(ROA), where ROA is the return on assets and CAR is the capital- asset ratio, both averaged over the past five years. σ(ROA) is the standard deviation of ROA over the past five years. Higher Z-score implies more stability. Operating income=interest income+non-interest income. Values of macro variables are from the previous year. t-Values are computed by the robust standard errors clus- tered for individual banks and are presented in brackets. ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively.

13 The same argument applies to our bank efficiency analysis, too. 14 The estimations results of the first-order coefficients are not reported to save space. These results are available from the authors upon request.

291J. Zhang et al. / China Economic Review 23 (2012) 284–295

improves the efficiency level of CCBs in the region. These results are not sensitive to whether or how the three regional control variables (loans/GDP, ln(GDP), and GDP growth rate) are included. In addition, GDP of the region is positively correlated with the technical inefficiency of CCBs. This finding is consistent with that of Hasan et al. (2009): before the 1990s, the Chinese banking markets were pretty much monopolized by the Big Four and a few other banks. As the new comers to the Chinese banking market (since mid-1990s), CCBs performed relatively better in regions where the monopoly power of the Big Four was weaker and thus the market room for CCBs' development was bigger. Since regions with more advanced economy and bigger banking market (greater GDP) have always been the focus of and have been occupied by the large banks, the performance of CCBs (as measured by their efficiency in creating profits and expanding the bank size) is relatively poor in these regions, ceteris paribus. In contrast, high growth rate of the size of the local market (as measured by GDP of the region) benefits all banks in the region and results in higher bank efficiency.

The risk factors all have the expected relation with bank efficiency. We find that the technical efficiency level among banks with a high NPL ratio is low. The equity-to-assets ratio has a negative coefficient, indicating that a strict constraint on the capital requirement of banks actually improves bank efficiency. Not surprisingly, the loan-to-deposit ratio has a negative impact on bank inefficiency as a greater ability to offer loans enhances bank profitability. Similarly, the interbank fund-to-total borrowings ratio is positively related to bank inefficiency, showing that banks that rely more on deposits than on short-term borrowing are more efficient probably due to the low funding cost associated with deposits. As for asset structure, the ratio of securities to total assets has a negative relation with bank inefficiency, which shows that the greater allocation of bank assets to securities improves the ability of banks to generate profit. This result can be taken as evidence that the risk-adjusted return on securities is higher than that on other assets such as loans and cash-type assets in China. In addition, we detect a strong effect of selection governance variable on bank efficiency. Banks which have been selected for IPOs and foreign investment are significantly more efficient than other.

The second column of Table 7 presents the time trend of the efficiency score (weighted average using the bank's total assets as weight) based on the income-based model.15 The efficiency scores of Chinese CCBs show a rising trend, especially before 2003.

15 In Table 7, only the results of model 1 in Tables 6 and 8 are reported. Efficiency scores using other models are similar and the untabulated results are available from the authors upon request.

Table 6 Inefficiency analysis: results of the income-based model.

(1) (2) (3) (4) (5) (6)

Legal environment −0.074*** −0.082*** – – – – [−3.27] [−4.66]

Efficiency of the legal system – – −0.189*** −0.149*** – – [−6.05] [−8.69]

Protection of intellectual property right – – – – −0.023** −0.032*** [−2.14] [−3.32]

Loans/GDP 0.003 – −0.402** – −0.466*** – [0.02] [−2.40] [−2.60]

Ln (GDP) 0.424*** 0.529*** 0.697*** 0.601*** 0.427*** 0.538*** [5.79] [6.55] [8.18] [10.51] [6.51] [12.48]

GDP growth rate – −0.025** – −0.031*** – −0.039*** [−2.20] [−3.15] [−3.42]

Selection governance −1.628*** −1.434*** −1.062*** −1.534*** −1.221*** −1.435*** [−8.00] [−10.19] [−3.46] [−8.11] [−8.17] [−8.98]

Dynamic governance 0.326 0.418* 0.517*** 0.365 0.467** 0.467*** [0.94] [1.65] [4.21] [1.22] [2.12] [2.62]

Equity/assets −7.149*** −7.716*** −8.292*** −6.786*** −7.968*** −8.022*** [−8.83] [−9.40] [−9.78] [−10.31] [−9.85] [−7.73]

NPL ratio 4.734*** 5.237*** 5.255*** 3.968*** 5.283*** 5.289*** [10.41] [12.36] [12.68] [12.80] [11.92] [12.40]

Loans/deposits −2.917*** −3.198*** −3.114*** −2.465*** −3.268*** −3.199*** [−10.83] [−12.10] [−10.08] [−11.27] [−9.48] [−9.93]

Interbank funds/(interbank funds+deposits) 1.246*** 1.117*** 1.052*** 0.869*** 1.180*** 1.147*** [5.15] [3.94] [3.52] [3.98] [3.67] [3.46]

Long-term loans/loans 0.153 −0.007 −0.287* 0.023 −0.083 0.038 [1.11] [−0.05] [−1.95] [0.23] [−0.50] [0.22]

Securities/earning assets −3.732*** −3.985*** −4.436*** −3.637*** −4.001*** −3.972*** [−11.18] [−11.38] [−10.87] [−12.86] [−10.42] [−8.47]

Constant −5.396*** −4.095*** −8.117*** −3.606*** −5.248*** −2.884** [−5.24] [−3.59] [−7.46] [−3.33] [−5.48] [−2.29]

Gamma 0.911*** 0.923*** 0.924*** 0.898*** 0.919*** 0.924*** [71.42] [78.17] [94.29] [70.57] [66.50] [95.37]

Log likelihood −100.59 −98.52 −92.70 −96.58 −98.63 −98.83 LR test 567.18 571.32 582.94 575.19 571.10 570.68 Observations 1083 1083 1083 1083 1083 1083

Note: ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively; t-values are in square brackets.

292 J. Zhang et al. / China Economic Review 23 (2012) 284–295

4.3. Law enforcement and bank efficiency: results of the earnings assets-based model

An alternative model to study bank efficiency is based on the earning assets of a bank. Table 4 lists the main differences between this model and the income-based model. Besides the different input and output factors included, there are a few minor differences in the variable definitions. For example, in the income-based model, net interest income is ym in Eq. (3), whereas in the earning assets-based model, it is measured by total deposits.

Table 8 presents the one-step maximum-likelihood estimation results using the full sample. Only the inefficiency analysis results are reported to save space. The earning assets-based model has a γ value smaller than that of the income-based model, indicating the better model fit of the latter. Their estimation results, however, are generally similar. For example, all of the coef- ficients on the law enforcement variables are negative (regressions 1–6) and most of them are significant (except the efficiency

Table 8 Inefficiency analysis: results of the earning assets-based model.

(1) (2) (3) (4) (5) (6)

Legal environment −0.048*** −0.038*** – – – – [−5.28] [−4.38]

Efficiency of the legal system – – −0.002 0.002 – – [−0.47] [0.47]

Protection of intellectual property right – – – – −0.019*** −0.013*** [−4.46] [−3.68]

Loans/GDP 0.150*** – 0.080* – 0.162*** – [2.63] [1.74] [2.98]

Ln (GDP) 0.130*** 0.106*** 0.065*** 0.067*** 0.111*** 0.090*** [5.99] [5.44] [3.70] [3.77] [5.61] [4.99]

GDP growth rate – −1.016 – −1.609* – −1.293* [−1.28] [−1.94] [−1.65]

Selection governance −0.098*** −0.088** −0.050 −0.047 −0.082** −0.074** [−2.66] [−2.32] [−1.57] [−1.40] [−2.38] [−2.08]

Dynamic governance 0.079 0.076 0.112** 0.121** 0.061 0.054 [1.15] [1.06] [2.10] [2.29] [1.05] [0.86]

Equity/assets 0.302 0.263 −0.086 −0.268 0.270 0.226 [1.11] [0.97] [−0.32] [−0.99] [1.00] [0.84]

NPL ratio 0.041 0.065 0.060 0.079 0.064 0.078 [0.57] [0.91] [0.87] [1.17] [0.92] [1.11]

Loans/deposits −1.257*** −1.233*** −1.039*** −0.826*** −1.411*** −1.357*** [−5.50] [−5.52] [−6.23] [−4.63] [−7.87] [−6.37]

Interbank funds/(interbank funds+deposits) 0.407*** 0.422*** 0.289*** 0.168 0.478*** 0.488*** [3.07] [3.25] [2.62] [1.60] [4.02] [3.81]

Long-term loans/loans −0.266*** −0.280*** −0.111** −0.097** −0.207*** −0.246*** [−4.65] [−4.86] [−2.33] [−2.00] [−3.45] [−4.01]

Securities/earning assets −0.104 −0.130 0.089 −0.018 −0.111 −0.141 [−1.12] [−1.44] [0.94] [−0.20] [−1.22] [−1.48]

Constant −0.297 0.139 0.240 −0.074 −0.091 0.339 [−0.92] [0.49] [0.82] [−0.35] [−0.31] [1.28]

Gamma 0.325*** 0.327*** 0.279*** 0.267*** 0.286*** 0.304*** [4.15] [4.36] [3.60] [3.43] [3.52] [3.98]

Log likelihood 393.85 391.35 390.27 384.06 390.17 388.87 LR test 143.90 138.89 136.73 124.32 139.04 133.94 Observations 1083 1083 1083 1083 1083 1083

Note: ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively; t-values are in parentheses.

Table 7 Technical efficiency of CCBs over time.

Year Income-based model Earning assets-based model

1999 0.6990 0.4510 2000 0.7195 0.5023 2001 0.7618 0.6699 2002 0.8049 0.8264 2003 0.8308 0.8926 2004 0.8427 0.9164 2005 0.8638 0.9215 2006 0.8788 0.9310 2007 0.8828 0.9392 2008 0.8772 0.9423 Average 0.8161 0.7993

293J. Zhang et al. / China Economic Review 23 (2012) 284–295

of the legal system), which indicates that good law enforcement enhances bank efficiency. The size of the regional market also has a negative impact on bank efficiency while its growth rate has a positive effect. The loan-to-GDP ratio has a significantly positive coefficient in the inefficiency analysis. The possible explanation is that in regions with a high level of financial deepening, CCBs face more competition from the JSCBs and Big Four, which constrains the ability of the former to expand their business. As for bank characteristics, the selection governance variable is negatively related to technical inefficiency, a result that is similar to that of the income-based model. Unsurprisingly, the loan to deposit ratio has a negative coefficient as a greater ability to offer loans enhances the bank's ability to expand business. In addition, more reliance on the interbank borrowing reduces a bank's efficiency.

There are a few differences in the estimation results for bank characteristics. The coefficient on NPL ratio is insignificant in the earning assets-based model, indicating that NPLs have a greater effect on the profitability of banks than on the ability of banks to expand their business. The coefficients on equity-to-assets ratio and the ratio of securities to total assets also become insignificant in the earning assets-based model. The ratio of long-term to total loans has no significant impact on the efficiency of banks in the income-based model but is positively related to bank efficiency in the earning assets-based model.

Table 7 (column 3) presents the efficiency scores of banks using the earning assets-based model. The time trend of the CCBs' efficiency using the earning assets-based model is similar to that of the income-based model.

5. Conclusion

The recent financial crisis has generated renewed interest into how the institutional environment and regulatory environment influence bank risk taking and performance (Houston et al., 2010). In this paper, we focus on within-country effects of regional institutions on bank risk taking and performance by taking into account the striking heterogeneity of bank-level characteristics. Using a unique sample of 133 Chinese city commercial banks for the 1999–2008 period, we find that stronger law enforcement tends to promote greater bank risk taking in the region. Employing a stochastic function approach, our analysis further shows that the performance of Chinese banks, as measured by bank efficiency, is heavily influenced by the law enforcement in the region. Better legal environment, higher efficiency in the legal system, and stronger protection of intellectual property right are associated with a higher level of efficiency among these banks.

Our research has important policy implications to the banking reform in China. The Chinese government just released a policy memo in 2011 that emphasizes the critical role played by the Chinese CCBs in the regional financial market and encourages CCBs to get listed on domestic and/or international stock exchanges. To policy makers, it has become an important issue to evaluate the performance and insolvency risk of these banks properly. For the first time in the literature, our study covers the majority of Chinese CCBs and it provides some important insights about this group of local banks in China. Our findings on factors that affect risk-taking and efficiency of CCBs may also be used as a guideline by the regulators to further enhance the efficiency of the Chinese local banks while maintaining the financial stability in the regional banking market.

References

Acemoglu, D. (2005). Constitutions, politics and economics: A review essay on Persson and Tabellini's “The Economic effects of constitutions”. Journal of Economic Literature, 43, 1025–1048.

Acemoglu, D., & Dell, M. (2010). Productivity differences between and within countries. American Economic Journal: Macroeconomics, 2(1), 169–188. Battese, G. E., & Coelli, T. J. (1995). A model for technical inefficiency effects in a stochastic frontier production function for panel data. Empirical Economics, 120,

325–332. Berger, A. N., Clarke, G. R. G., Cull, R., Klapper, L., & Udell, G. F. (2005). Corporate governance and bank performance: A joint analysis of the static, selection, and

dynamic effects of domestic, foreign, and state ownership. Journal of Banking and Finance, 29, 2179–2221. Berger, A. N., Hasan, I., & Zhou, M. (2009). Bank ownership and efficiency in China: What will happen in the world's largest nation? Journal of Banking and Finance,

33, 113–130. Berger, A. N., & Mester, L. J. (1997). Inside the black box: What explains differences in the efficiencies of financial institutions? Journal of Banking and Finance, 21,

895–947. Berger, A. N., & Udell, G. F. (2006). A more complete conceptual framework for SME finance. Journal of Banking and Finance, 30, 2945–2966. Berkowitz, D., & Clay, K. (2006). The effect of judicial independence on courts: Evidence from the American states. The Journal of Legal Studies, 35(2), 399–400. Chen, Q., Schipper, K., Wang, K., & Xiao, X. (2011). The effects of market development on controlling shareholders' participation in rights offerings. Working Paper,

Duke University. Chen, X., Skully, M., & Brown, K. (2005). Banking efficiency in China: Application of DEA to pre- and post-deregulations era: 1993–2000. China Economic Review,

16, 229–245. Coelli, T., & Perelman, S. (1996). Efficiency measurement, multi-output technologies and distance functions: With application to European railways. CREPP Working

Paper 96 (05), University of Liege. Cole, R., Turk-Ariss, R., 2008. Legal origin, creditor protection and bank risk-taking: Evidence from emerging markets. Unpublished working paper, DePaul

University, Lebanese American University. Demirguc-Kunt, A., & Huizinga, H. (2010). Bank activity and funding strategies: The impact on risk and returns. Journal of Financial Economics, 98(3), 626–650. Demirguc-Kunt, A., Laeven, L., & Levine, R. (2003). Regulations, market structure, institutions, and the cost of financial intermediation. Journal of Money, Credit, and

Banking, 36, 593–622. Fan, G., Wang, X., Zhu, H. (ed.), 2000, 2004, 2006, 2010 (in Chinese). Zhongguo shi chang hua zhi shu: ge di qu shi chang hua xiang dui jin cheng du bao gao.

Beijing: Jing ji ke xue chu ban she. Ferri, G. (2009). Are new tigers supplanting old mammoths in China's banking system? Evidence from a sample of city commercial banks. Journal of Banking and

Finance, 33, 131–140. Fries, S., & Taci, A. (2005). Cost efficiency of banks in transition: Evidence from 289 banks in 15 post-communist countries. Journal of Banking and Finance, 29,

55–81. Fu, X., & Heffernan, S. A. (2007). Cost X-efficiency in China's banking sector. China Economic Review, 18, 35–53. Hasan, I., Wang, H., & Zhou, M. (2009). Do better institutions improve bank efficiency? Evidence from a transitional economy. Managerial Finance, 35(2), 107–127.

294 J. Zhang et al. / China Economic Review 23 (2012) 284–295

Houston, J., Lin, C., Lin, P., & Ma, Y. (2010). Creditor rights, information sharing, and bank risk taking. Journal of Financial Economics, 96, 485–512. Jiang, C., Yao, S., & Zhang, Z. (2009). The effects of governance changes on bank efficiency in China: A stochastic distance function approach. China Economic

Review, 20, 717–731. John, K., Litov, L., & Yeung, B. (2008). Corporate governance and risk-taking. Journal of Finance, 63(4), 1679–1728. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. (1998). Law and finance. Journal of Political Economy, 106, 1113–1155. Laeven, L., & Levine, R. (2009). Bank governance, regulation and risk taking. Journal of Financial Economics, 93, 259–275. Laeven, L., & Woodruff, C. (2007). The quality of the legal system, firm ownership, and firm size. The Review of Economics and Statistics, 89(4), 601–614. Lovell, C. A. K., Richardson, S., Travers, P., & Wood, L. L. (1994). Resources and functionings: A new view of inequality in Australia. In W. Eichhorn (Ed.), Models and

measurement of welfare and inequality (pp. 787–807). Berlin: Springer Verlag. Ma, Y., Qu, B., & Zhang, Y. (2010). Judicial quality, contract intensity and trade: Firm-level evidence from developing and transition countries. Journal of Comparative

Economics, 38(2), 146–159. Orea, L. (2002). Parametric decomposition of a generalized Malmquist productivity index. Journal of Productivity Analysis, 18, 5–22. Qian, J., & Strahan, P. E. (2007). How laws and institutions shape financial contracts: The case of bank loans. Journal of Finance, 62, 2803–2834. Wang, H., & Schmidt, P. (2002). One-step and two-step estimation of the effects of exogenous variables on technical efficiency levels. Journal of Productivity Analysis,

18, 129–144. Wang, Q., Wong, T. J., & Xia, L. (2008). State ownership, the institutional environment, and auditor choice: Evidence from China. Journal of Accounting and Economics,

46(1), 112–134.

295J. Zhang et al. / China Economic Review 23 (2012) 284–295