Liquidity Creation and Bank Capital Structure in China

Page 1

Global Finance Journal 24 (2013) 188–202

Contents lists available at ScienceDirect

Global Finance Journal

j ourna l homepage: www.e lsev ie r .com/ locate /gf j

Liquidity creation and bank capital structurein China☆

Adrian C.H. Lei⁎, Zhuoyun SongFaculty of Business Administration, University of Macau, Macau

a r t i c l e i n f o

☆ Wewould like to thank an anonymous referee animproved the paper. We are also grateful to the comMelbourne, Australia. We acknowledge the financial⁎ Corresponding author. Tel.: +853 8397 4162.

E-mail address: [email protected] (A.C.H. Lei).1 People's Bank of China was established as a centra

mainly to the four state-owned commercial banks (lending issues of SOCBs, and asset-management coprepare for the privatization of major banks.

2 After the partial privatization, the major banks icorporate entities. See McGuinness and Keasey (201

3 This index is compiled by the Heritage Foundatibanking and finance out of the ten economic freedom

1044-0283/$ – see front matter © 2013 Elsevier Inc.http://dx.doi.org/10.1016/j.gfj.2013.10.004

a b s t r a c t

Article history:Accepted 8 August 2013Available online 28 October 2013

This paper investigates the relationship between liquidity creationand bank capital structure in China. We test the so-called “financialfragility-crowding out” hypothesis and the “risk absorption” hypoth-esis on Chinese banks and find that bank capital is negatively relatedto liquidity creation, which supports the financial fragility-crowdingout hypothesis. In contrast, we find that foreign banks in China have aweaker relationship between liquidity creation and bank capital,which is consistent with the risk absorption hypothesis and findingsin prior studies.

© 2013 Elsevier Inc. All rights reserved.

JEL classification:G21G28

Keywords:Liquidity creationBank capital structureFinancial fragility-crowding-outRisk absorptionChina

1. Introduction

The Chinese banking system has undergone rapid privatization for the transition from a plannedeconomy to a socialist market economy.1 Despite this transition, the government still controls majorChinese banks,2 and, according to the Index of Economic Freedom,3 the Chinese banking system operates

d the editor (Manuchehr Shahrokhi) for their valuable suggestions that greatlyments made by M.H. Liu, Frank Song, and the participants of AFAANZ 2012 atsupport from the University of Macau.

l bank in 1979 and transferred its commercial activities and treasury functionsSOCBs). In the 1990s, policy banks were established to take over the policympanies managed the transfer of the non-performing loans of the SOCBs to

n China are still controlled by state-based regulatory entities or state-related0).on of the US. Here we refer to Financial Freedom, which is the one related toindicators.

All rights reserved.

Page 2

189A.C.H. Lei, Z. Song / Global Finance Journal 24 (2013) 188–202

under extensive government influence. Berger, Hasan, and Zhou (2009) suggest that China's high growthrate cannot be sustained if banking reform cannot reach the necessary efficiency. Lin and Zhang (2009)show that the “Big Four” state-owned banks are less efficient and less profitable than other types of banks.China's institutional settings thus may impair banks' qualitative asset transformation function, which isessential for the existence of financial intermediaries (Bhattacharya & Thakor, 1993). It is unclear,however, how banks create liquidity under the government's strong influence.

Banks create liquidity in two ways: through transforming illiquid assets to liquid liabilities (Diamond &Dybvig, 1983) or through off-balance-sheet activities, such as loan commitments and lines of credit(Kashyap, Rajan, & Stein, 2002). Banks' role as risk transformers has been well studied in the past, andrecent studies focus on the role of liquidity creators. Deep and Schaefer (2004) construct liquidity creationmeasures that focus purely on maturity transformation and include on‐balance sheet activities only. Later,Berger and Bouwman (2009) further investigate the measurement that is based on categories of assets,liabilities, equity, and off-balance-sheet issues and analyze US banks' liquidity creation. Their new liquiditycreation measures enable the quantitative examination of liquidity creation in China.

The so-called “financial fragility-crowding out” hypothesis (Diamond & Rajan, 2000, 2001; Gorton &Winton, 2000) and the “risk absorption” hypothesis (Berger & Bouwman, 2009) explain the relationshipsbetween the bank capital ratio and liquidity creation. Banks' financial fragility-crowding out hypothesisargues that under a fragile bank structure (i.e., with lower bank capital), banks expend more effort toprovide funds and therefore create more liquidity. Also, higher capital ratios reduce liquidity creation byshifting investors' funds from liquid deposits to relatively illiquid bank capital. The risk absorptionhypothesis argues that higher capital ratios expand banks' risk-bearing ability, and thus banks can createmore liquidity (e.g., Bhattacharya & Thakor, 1993; Coval & Thakor, 2005; Repullo, 2004; Von Thadden,2004). Banks' liquidity creation exposes banks to risk, and thus it is associated with greater likelihood andseverity of losses (e.g., Allen & Gale, 2004; Allen & Santomero, 1997; Diamond & Dybvig, 1983). The effectsof these hypotheses, however, are unknown in the setting of the Chinese banking system, which consistsmainly of marketized state-owned banks. If bank capital ratios affect liquidity creation, then what wouldbe the potential impact of liquidity creation on these banks?

To explore the effects of the financial fragility hypothesis and the risk absorption hypothesis in China'sbanking market, we use annual bank data from China over the 1988–2009 period and estimate liquiditycreation, following Berger and Bouwman (2009). Our results show that the capital ratio and liquiditycreation are negatively related, supporting the fragility crowding-out hypothesis for banks in China. Also,consistent with Berger and Bouwman (2009), we find that foreign banks exhibit a less negative relationbetween bank capital and liquidity creation. These results suggest that the effect of risk absorption is stillimportant in foreign banks of China, nullifying the effect of financial fragility.

This study contributes in several ways to the banking literature and policy implementation in China. Tothe best of our knowledge, we are the first to estimate a detailed quantitative measure of the liquiditycreation of all Chinese banks. We show that the effects of liquidity creation on the policy-oriented bankingsystem are different from those on the market-based banking system. Raising the capital reserve ratiowould negatively affect liquidity creation, because our results are tilted towards predictions of thefinancial fragility-crowding out hypothesis. And consistent with Berger and Bouwman (2009), we findthat the risk absorption hypothesis still significantly influences foreign banks in China.

The remainder of the paper is organized as follows. Section 2 presents the related literature anddevelops the hypotheses. Section 3 describes the data and methodology. Section 4 discusses the empiricalresults. Section 5 presents the conclusion.

2. Literature review and hypothesis development

Based on liquidity creation theory, Berger and Bouwman (2009) suggest that when banks transformilliquid assets into liquid liabilities, or finance illiquid assets with liquid liabilities, they function as liquiditycreators. The intuition is that banks create liquidity when they hold illiquid items for the nonbank publicand provide the public with liquid liabilities. For example, when banks absorb corporate long-term loansusing savings deposits, they transform illiquid items (i.e., long-term corporate investments) into liquidones (i.e., saving deposits) for the nonbank public, which creates liquidity. Alternatively, when banks issuelong-term subordinated debt to hold marketable securities, they transform liquid items (i.e., marketable

Page 3

Table 1Definitions and sources of the regression variables.This table provides the definitions and sources for all the regression variables, including dependent and independent variables. They are all calculated in nominal values for each year. LC / TA cat fatequals to liquidity-creation (LC) over total assets (TA), where the LC is based on the activities categories (cat) including off-balance-sheet activities (fat) and LC / TA cat nonfat excludesoff-balance-sheet activities (cat nonfat).

Variables Definition Sources

Dependent variablesLC / TA cat fat Measure of liquidity creation under the “cat fat” method divided by total assets BankScopeLC / TA cat nonfat Measure of liquidity creation under the “cat nonfat” method divided by total assets BankScope

Bank capital ratioEQ / TA The ratio of total equity to total assets BankScope

Bank management efficiencyCO/INC Cost to income ratio BankScope

Bank sizeLn(TA) Natural log of total assets to represent bank sizes BankScope

Bank riskY-STD The standard deviation of annual return on assets over previous 5 years. If there is no such long previous

years' period, it is calculated based on previous available years' data.BankScope

CREDITRISK The sum of the risk-weighted assets and off-balance-sheet activities divided by total assets BankScopeZSCORE The sum of return on assets and equity / TA ratio divided by the standard deviation of annual return on assets BankScope

Banking governanceDPS / TA Total deposits divided by total assets BankScopeNL / TA Net loans divided by total assets BankScopeGL-GROWTH Gross loans growth rate BankScope

190A.C.H

.Lei,Z.Song/GlobalFinance

Journal24(2013)

188–202

Page 4

Local market competitionHHI Bank-level Herfindahl index based on the deposits share BankScopeML-SHARE The share of medium and large banks' deposits to total deposits BankScope

Bank typesState-owned banks Dummy variable that is used to check whether the bank is a state-owned commercial bank; if yes = 1, otherwise = 0 Almanac of China's Finance and BankingForeign banks Dummy variable that is used to check whether the bank is a foreign bank; if yes = 1, otherwise = 0 Almanac of China's Finance and BankingDomestic banks Dummy variable that is equal to one if the bank is not a state-owned, foreign or policy bank Almanac of China's Finance and BankingD-M&A Dummy variable that is used to check whether the bank has a merger and acquisition history over the previous 3 years;

if yes = 1, otherwise = 0BankScope

D-LISTED Dummy variable that is used to check whether the bank is a bank listed in the stock market; if yes = 1, otherwise = 0 Almanac of China's Finance and Banking

Financial market opennessBOR Benefit of openness ratio, measured as the share of income inflow to the share to income outflow in the country's global trades (IMF) Balance of Payments (BOP) Statistics

General local economyGNI-GROWTH Annual nominal growth rate of gross national income GNI (IMF) International Financial Statistics (IFS)Ln(POP) Natural log of population World Economic OutlookGDP-GROWTH Annual nominal growth rate of gross domestic product GDP (IMF) International Financial Statistics (IFS)Ln(M1) Natural log of that sum of legal tender notes and coins held by the public, plus customers demand deposits placed

with licensed banksNational Bureau of Statistics of China

DPSR Deposit rate at the end of each year (IMF) International Financial Statistics (IFS)

Banking reformD1995 Dummy variable that is used to indicate the year of the time period of the observation; if in between

year 1995 and 2001 = 1, otherwise = 0D2001 Dummy variable that is used to indicate the year of the time period of the data; if in or after

year 2001 = 1, otherwise = 0

Fixed and random effectsTime fixed effects Dummies for all time but oneBank fixed effects Dummies for all banks but one

191A.C.H

.Lei,Z.Song/GlobalFinance

Journal24(2013)

188–202

Page 5

192 A.C.H. Lei, Z. Song / Global Finance Journal 24 (2013) 188–202

securities) into illiquid ones (i.e., subordinated debt for the public), which destroys the liquidity. Finally,when banks use savings deposits to buy securities, they transform liquid assets into liquid liabilities, andliquidity remains unchanged. Banks create economic value through this transformation of illiquid assetsinto liquid liabilities.

There are two mainstream hypotheses related to bank capital and liquidity creation: the “financialfragility-crowding out hypothesis” and the “risk absorption hypothesis.” The financial fragility-crowdingout hypothesis suggests that when a bank's capital structure is fragile, it will cause the bank to monitor itsborrowers more, and then the bank can extend loans and create more liquidity. Diamond and Rajan (2000,2001) point out that having more bank capital will make the bank's capital structure less fragile and thusinhibit liquidity creation. Gorton and Winton (2000) argue that as capital and deposits are crowding out,more capital will reduce deposits and cause a decrease of liquidity creation. Some empirical studies alsofind that leverage requirements lead to a decrease in bank loans, which suggests a negative relationshipbetween bank capital and liquidity creation (Berger & Udell, 1994; Hancock, Laing, & Wilcox, 1995; Peek &Rosengren, 1995). Berger and Bouwman (2009) apply this hypothesis to test small banks. They argue thatsmall banks need to monitor borrowers more, because their borrowers are usually smaller with higherrisks, which is consistent with Diamond and Rajan (2000, 2001), and small banks' sources of funds aremainly local residents and corporations; thus the crowding-out effect between capital and deposits maybe significant, as in Gorton and Winton (2000).

The risk absorption hypothesis suggests that banks are able to absorb more risk with more capital, andthus they create more liquidity when they have higher risk tolerance (Allen & Gale, 2004; Allen &Santomero, 1997; Bhattacharya & Thakor, 1993; Coval & Thakor, 2005; Repullo, 2004; Von Thadden,2004). Several empirical studies find that the decrease in bank capital ratios incurred from loan lossesreduces lending (e.g. Peek & Rosengren, 1995). This hypothesis suggests a positive relationship betweencapital and liquidity creation. Berger and Bouwman (2009) find supporting evidence for this hypothesis inlarge US banks. Large Chinese banks, however, even though privatized, are backed by the government,because the Chinese government is still the largest shareholder of these banks. Therefore, the role ofcapital to absorb risk and create liquidity may not apply to China's large banks. We expect that thefinancial fragility crowding-out effect dominates China's banks.

Hypothesis 1. The financial fragility-crowding out hypothesis dominates Chinese banks' liquiditycreation. Bank capital is negatively related to liquidity creation for banks in China.

Foreign banks in China are mainly global banks, such as JP Morgan Chase, Citibank, and HSBC. Othersare major banks from Hong Kong. These banks may not be backed by the Chinese government if they runinto financial distress. Furthermore, providing loans in China is generally considered higher risk than otherdeveloped markets. Also, Berger and Bouwman (2009) find that large banks' liquidity creation isdominated by the risk absorption effect. Therefore, we expect that the risk absorption effect should havegreater influence on the liquidity creation of foreign banks.

Hypothesis 2. For foreign banks, the negative relationship of liquidity creation and bank capital isreduced because of the “risk absorption” effect.

3. Data and methodology

3.1. Data

In our study, we use the annual bank data of China over the 1988–2009 period from BankScope. Allvariables are winsorized at the 1st and 99th percentiles to reduce the impact of outliers. We dropobservations that do not have time-series to run the fixed-effects model (i.e., banks that have only asingle-year observation). Our sample contains 4 state-owned commercial banks (SOCBs) with 64observations, 113 domestic banks with 694 observations, and 18 foreign banks (FB) with 78 observations.Note that the decrease in the number of banks for 2008 and 2009 is because of unavailable data, especiallyfor the small banks — city commercial banks, and not because of mergers or defaults. There are 836bank-year observations in our full sample.

Page 6

193A.C.H. Lei, Z. Song / Global Finance Journal 24 (2013) 188–202

We use unconsolidated statements whenever they are available. Otherwise, we refer to the consolidatedstatements.4 Unconsolidated data are preferred because they have more detailed categories than consolidateddata. Also, we can analyze based on banks instead of groups. Macroeconomic variables are obtained fromInternational Monetary Funds' (IMF) World Economic Outlook (WEO), International Financial Statistics (IFS),and Balance of Payments Statistics (BOPS) databases, including gross domestic product (GDP) growth, grossnational income (GNI) growth, inflation, population, exchange rate, and balance of national payments. Table 1lists the main regression variables, sources of data, and brief descriptions.

3.2. Construction of liquidity-creation measures

We construct two alternative liquidity-creation measures, according to Berger and Bouwman (2009).First, bank-balance-sheet and off-balance-sheet activities are classified as liquid, semi-liquid, or illiquid.Second, activities are classified and weighted, following the liquidity-creation intuition. Third, activitiesare combined as classified and weighted to construct two liquidity-creation measures: One includesoff-balance-sheet activities (cat fat), and the other excludes them (cat nonfat). Because of data availability,we calculate only the two measures that are based on the activities categories.5 Nevertheless, of themeasures based on the activities category, including the off-balance-sheet item, the “cat fat” measure ispreferred. This is because how costly, timely, and easy it is to dispose of obligations on the asset side ismore important than the time until self-liquidation (maturity). Table 2 Panel A illustrates how bankactivities are classified and weighted. Panel B illustrates the calculations of “cat fat” and “cat nonfat.”

3.3. Liquidity creation over time and in the cross section for different bank sizes

We sort the total assets of all banks and define the banks with total assets (TA) under $1 billion assmall banks, over $3 billion as large banks, and those in between as medium banks.6 We split the sampleby bank type to test the financial fragility crowding-off hypothesis and the risk absorption hypothesis,similar to Berger and Bouwman (2009).

After calculating the liquidity creation (LC) by “cat fat” and “cat nonfat”methods, we explore the change ofliquidity creation andhow it varies for the subsamples of different bank sizes and types. Table 3 summarizes theentire bank liquidity creation in China based on the cat fat method, using the 1988, 1998, and 2008 data. In1988, only 7 commercial banks are available. This restricts the feasibility of analyzing liquidity over the longhorizon from the 1980s. Following rapid growth in 1998, there are 27 banks, and this number increases to 94 in2008. Liquidity creation increases from 22 billion RMB in 1988 to 2463 billion RMB in 1998 to 11,404 billionRMB in 2008. Fig. 1 shows the rapid increase of liquidity creation over the two decades. Also around the majorbanking reform, there are substantial changes of liquidity creation, suggesting the possible effect of bankingreform on the financial system. With the first bank reform, liquidity creation begins to grow slowly starting in1995, which is brought about mainly by large banks. After the second reform, liquidity creation grows rapidly.In 2008, large banks still maintain over 80% of the share of liquidity creation.

The trends for liquidity creation / total assets (LC / TA) and liquidity creation / total equity (LC / EQ)are different, however. The ratios increase in the first 10 years (e.g., LC / TA increases from 0.24 to 0.32)and decrease in the second 10 years (e.g., LC / TA decreases from 0.32 to 0.28), which indicates that theliquidity creation per asset or equity, or the efficiency of liquidity creation, is stagnant in the recent decade,compared with the expanding total liquidity creation.

For the bank-size subsamples, we can observe that the number of banks for each size increasessubstantially, especially the number of small banks, nowadays 11 times as many as that in 1998. Liquiditycreation of “cat fat” and “cat nonfat” seems to be consistent across different bank sizes. It seems that

4 Among the total 145 banks, 17 lack unconsolidated data. They are: Agricultural Bank of China Limited, Bank of Chongqing, Bankof Dalian, Bank of Jilin Co., Ltd., China CITIC Bank, Corporation Limited, China State Bank, Ltd., Xiamen International Bank, Bank ofCommunications Co., Ltd., Bank of Luoyang Co., Ltd., and Bank of Weifang Co., Ltd.

5 Berger and Bouwman (2009) develop four measures, the two are based on the activities category and the others are based onactivities maturity.

6 The $3 billion and $1 billion cutoffs are determined with reference to Berger and Bouwman's (2009) findings. The cutoffs arebased on 1987 real dollar values; then we adjust the cutoff points with the nominal GDP growth of China per year and the exchangerate for US dollars per RMB. In each year, we separate the banks into small, medium, and large banks in our sample.

Page 7

Table 2Liquidity classification of bank activities and construction of two liquidity creation measures.Step 1: Bank activities are classified as liquid, semiliquid, and illiquid, based on the activities category in Panel A.Step 2: We assign weights to all bank activities classified in Step 1. Based on the liquidity-creation theory, banks create liquidity onthe balance sheet when they transform illiquid assets into liquid liabilities. According to Berger and Bouwman (2009), when bankstransform $1 of illiquid assets into $1 of liquid liabilities, that $1 of liquidity is created. Thus, they assign a weight of 1/2 for illiquidassets and liquid liabilities, since the amount of liquid created is only “half” by the use of the funds alone — both are needed to create“one” liquidity. Similarly, they apply a weight of −1/2 to liquid assets, illiquid liabilities, and equity. By transferring $1 of liquidassets into $1 of illiquid liabilities, $1 of liquidity is destroyed. The semi-liquid items are weighted to 0, and 1/2 to illiquidoff-balance-sheet activities.Step 3: We combine the bank activities classification in Step 1 with weights in Step 2 in two ways to construct liquidity-creationmeasures by using the “cat” based on the activities category, and by alternatively including off-balance-sheet activities (fat) orexcluding these activities (nonfat), which is in Panel B.

Panel A: Liquidity classification of bank activities

Illiquid assets (weight = 1/2) Semiliquid assets (weight = 0) Liquid assets (weight = −1/2)

AssetsCorporate & commercial loans Residential mortgage loans Cash and due from banksInvestments in property Other mortgage loans Trading securities and at fv

through incomeForeclosed real estate Other consumer/retail loans Tradable derivativesFixed assets Loans and advances to banks Available-for-sale securitiesGoodwill Reverse repos and cash collateral Held to maturity securitiesOther intangibles At-equity investments in associatesOther assets Other securities

Liabilities plus equityLiquid liabilities (weight = 1/2) Semiliquid liabilities (weight = 0) Illiquid liabilities plus equity

(weight = −1/2)Customer deposits — current Customer deposits — term Senior debt maturing after 1 yearCustomer deposits — savings Deposits from banks Subordinated borrowingTradable derivatives Repos and cash collateral Other fundingTrading liabilities Other deposits and short-term

borrowingsCredit impairment reserves

Fair value portion of debt Reserves for pensions and otherCurrent tax liabilitiesDeferred tax liabilitiesOther deferred liabilitiesOther liabilitiesTotal equity

Off-balance-sheet activitiesIlliquid activities (weight = 1/2) Semiliquid activities (weight = 0) Liquid activities (weight = −1/2)Acceptances and documentary creditsreported off-balance-sheet

Managed securitized assets reportedoff-balance-sheet

Committed credit lines Other off-balance-sheet exposureto securitizations

Other contingent liabilities Guarantees

Panel B: “cat fat” and “cat nonfat” formulas

Cat fat= +1/2 * illiquid assets +0 * semiliquid assets −1/2 * liquid assets+1/2 * liquid liabilities +0 * semiliquid liabilities −1/2 * illiquid liabilities

−1/2 * equity+1/2 * illiquid activities +0 * semiliquid activities −1/2 * liquid activities

Cat nonfat= +1/2 * illiquid assets +0 * semiliquid assets −1/2 * liquid assets+1/2 * liquid liabilities +0 * semiliquid liabilities −1/2 * illiquid liabilities

−1/2 * equity

194 A.C.H. Lei, Z. Song / Global Finance Journal 24 (2013) 188–202

mid-size banks are the most efficient in creating liquidity, as their LC / TA is the largest. For the bank-typesubsamples, as expected, the state-owned banks, even though there are only four of them, are the majorcreators of liquidity in China.

Page 8

Table 3Summary of bank liquidity creation in China.This table shows liquidity creation (LC) in RMB million and divided by total assets (LC / TA) and total equity (LC / EQ) for 1988,1998, and 2008. The LC RMBmil presented are the total volumes for that sample, respectively, and the LC ratios including LC / TA andLC / EQ are the mean values of that sample. We sort the gross total assets of all banks and define the banks with total assets (TA)under $1 billion as small banks. All banks with over $3 billion are defined as large banks, and the others are medium banks. Thebank-types are dummy variables for Domestic banks, State-owned banks, and Foreign banks. “Cat fat” and “cat nonfat” methods aretwo methods that measure liquidity creation, which classify all bank activities based on product category. The “cat nonfat” liquiditycreation measures include (exclude) off-balance-sheet activities. The $3 billion and $1 billion cutoffs are estimated with reference toBerger and Bouwman (2009).

Liquidity creation measure 1988 liquidity creation 1998 liquidity creation 2008 liquidity creation

N LC RMBbil

LC / TA LC / EQ N LC RMBbil

LC / TA LC / EQ N LC RMBbil

LC / TA LC / EQ

“Cat fat” All banks 7 22 0.24 3.17 27 2463 0.32 4.45 94 11403 0.28 4.27Large 3 13 0.24 3.79 12 2317 0.32 5.25 18 10366 0.28 5.89Medium 4 8 0.24 2.70 10 137 0.33 4.41 20 683 0.34 5.25Small 5 9 0.31 2.60 56 355 0.25 3.50Foreign 17 261 0.32 3.22Domestic 7 22 0.24 3.17 23 405 0.33 4.60 73 3706 0.31 5.66State-owned 4 2059 0.37 5.95 4 7408 0.24 4.07

“Cat nonfat” All banks 7 14 0.15 2.03 27 2101 0.29 4.11 94 8732 0.21 3.30Large 3 8 0.15 2.48 12 1969 0.28 4.75 18 7903 0.21 4.29Medium 4 6 0.15 1.70 10 124 0.30 4.12 20 545 0.28 4.11Small 5 8 0.29 2.54 56 283 0.19 2.72Foreign 17 219 0.27 2.69Domestic 7 14 0.15 2.03 23 364 0.31 4.26 73 2585 0.22 3.95State-owned 4 1737 0.32 5.02 4 5900 0.19 3.24

195A.C.H. Lei, Z. Song / Global Finance Journal 24 (2013) 188–202

Comparing the figures of liquidity creation in China with Berger and Bouwman's (2009) US results, theLC / TAs are around 0.3 in China, which is relatively smaller than the United States' 0.4.7 The LC / EQs,which are around 4 to 6 in China, however, are much larger than the US banks' 3 to 5 in 2008, but verysimilar to the US banks' LC / EQ in 1993. These figures indicate that the base of total assets for Chinesebanks is a bit larger relative to liquidity creation, while the total equity is smaller, compared with those ofUS banks. This is consistent with the argument that bank equity is abnormally small for Chinese banks(García-Herrero, Gavilá, & Santabárbara, 2009), and it seems that the liquidity-creation speed cannotmatch the growth of bank sizes.

3.4. Regression framework

To test the effect of bank capital on liquidity creation, we employ the following model, similar to that ofBerger and Bouwman (2009). The model is as follows:

7 Det

Lit ¼ cþ β1 EQ=TAð Þit−1 þ ΣλΠit−1 þ ε; ð1Þ

Lit is liquidity creation divided by total assets of ith bank at time t, with i = 1,…, N, and t = 1,…, T.

whereWe use lagged variables for all of the independent variables to mitigate endogeneity problems, yet we donot establish a causal claim and interpret the results with care. (EQ / TA)it − 1 is the lagged one-periodtotal equity divided by total assets, andΠit − 1s are the lagged one-period control variables, including banksize Ln(TA), bank risk Y-STD, CREDITRISK, ZSCORE, merger and acquisition history D-M&A, local marketcompetition HHI, ML-SHARE, and local market economic environment, such as population Ln(POP) andGNI growth GNI-GROWTH. Detailed definitions are displayed in Table 1.

ails about liquidity creation in the United States are available in Page: 11Berger and Bouwman (2009, table 2).

Page 9

A: Cat fat liquidity-creation measure with and without equity effects10

B: Cat nonfat liquidity creation measure with and without equity effects

Fig. 1. Liquidity creation under “cat fat” and “cat nonfat”measures over the sample period. The sample contains all commercial banksin China from 1988 to 2009. Panel A shows the first liquidity-creation measure using the cat fat method, where “cat fat” defines loanswith category, not maturity, and includes loan commitments and off-balance-sheet activities in China. The measure in the left graphcontains equity effects, and the right one excludes equity.8 Panel B shows the second liquidity-creation measure calculated by the catnonfat method, where “cat nonfat” also differs according to loans based on category, while excluding off-balance-sheet activities. Thevertical lines represent the start of a major banking reform. All calculating values are expressed in nominal terms.

196 A.C.H. Lei, Z. Song / Global Finance Journal 24 (2013) 188–202

3.5. Descriptive statistics

Table 4 shows the summary statistics of all variables in our regression models, before and afterwinsorization. Table 5 is the correlation matrix of key variables. Correlations larger than an absolute valueof 0.50 are in bold in Table 5.

HHI, ML-SHARE, and Ln(POP) are highly correlated, because they are all related to market competition.Following the model described in Berger and Bouwman (2009), we include all three variables in ourmodels. In addition, we test these models by dropping one or both of the variables that are collinear, andthe coefficients on our main variables in all the following models are mostly similar. So even though

8 The equity effect includes means conferring the weight of −1/2 to equity to calculate liquidity creation. The right-graphliquidity-creation measure is without setting −1/2 for equity.

Page 10

Table 4Summary statistics of the regression variables.This table provides summary statistics of all the regression variables. We drop the observations that do not have time-series to run the fixed-effects model (i.e., banks that have only a single-yearobservation). The remaining observations in our sample are 836 over the 1988–2009 period. All variables are winsorized at the 1st and 99th percentiles. We display the unwinsorized andwinsorized summary statistics, and the medians for winsorized variables are omitted as they are of the same value.Liquidity creation measures are as follows: (1) LC cat fat and LC cat nonfat, theliquidity creation volumes measured by cat fat and cat nonfat methods, respectively; and (2) LC / TA cat fat and LC / TA cat nonfat, the liquidity creation divided by total assets. Performancemeasures are: (1) NIM, net interest margin, and (2) ROA, return on average assets.EQ / TA is the capital ratio, measured as total equity divided by total assets. Ln(TA) is bank size, measured as natural log of total assets. Y-STD is the standard deviation of return on assets.CREDITRISK is a credit risk measure, calculated as the bank's risk-weighted assets and off-balance-sheet activities divided by total assets. ZSCORE is the sum of return on assets and the equity / TAratio divided by the standard deviation of annual return on assets. DPS / TA is total deposits divided by total assets. NL / TA is net loans divided by total assets. GL-GROWTH is growth of gross loans.D-M&A is a dummy that equals to 1 if the bank was involved in mergers and acquisitions over the past 3 years. D-LISTED is a dummy that equals to 1 if the bank is listed on the stock market. HHI is abank-level Herfindahl index based on deposits. ML-SHARE is the share of large and medium banks' deposits to the total deposits in the market. BOR is the benefit-of-openness ratio, measured as theshare of income inflow to the share to income outflow in the country's global trades. GNI-GROWTH is the growth of gross national income in China. Ln(POP) is the natural log of population.GDP-GROWTH is the growth of the gross domestic product. Ln(M1) is the natural log of M1, the sum of legal tender notes and coins held by the public plus customers' demand deposits placed withlicensed banks. DPSR is the annual deposit rate at the end of each year. PB, SOCB, JSCB, and FB are the dummies that equal to 1 if the banks are Policy banks, State-owned commercial banks, Jointstock commercial banks, or Foreign banks; if the above 4 dummies are all equal to 0, the bank is a city commercial bank. All variables are measured in nominal values, and currency units are in termsof RMB.

Variables Observations Unwinsorized data Winsorized data

Mean Median Maximum Minimum Std. dev. Mean Maximum Minimum Std. dev.

LC cat fat 836 122339 10760 3238000 −99062 398234 118506 2394000 −7349 371746LC cat nonfat 836 102607 8833 2723000 −99062 344685 98637 2058000 −8352 317050LC / TA cat fat 836 0.369 0.377 0.948 −0.413 0.2 0.369 0.804 −0.253 0.195LC / TA cat nonfat 836 0.314 0.325 0.921 −0.413 0.194 0.314 0.743 −0.27 0.189NIM 836 2.713 2.56 9.41 0.17 1.141 2.699 6.68 0.73 1.064ROA 836 0.00774 0.00715 0.0314 −0.0653 0.00633 0.00786 0.0235 −0.0049 0.00553EQ / TA 836 0.07 0.0565 0.591 −0.137 0.0606 0.0695 0.363 0.0121 0.0529CO / INC 830 48.06 42.94 350.8 2.03 21.22 47.47 95.32 17.62 16.8Ln(TA) 836 10.52 10.36 16.28 0.923 2.206 10.52 15.76 2.286 2.174Y-STD 836 0.303 0.239 2.729 0.00707 0.296 0.295 1.201 0.01 0.248CREDITRISK 836 0.271 0.119 6.624 0 0.388 0.264 1.07 0 0.316ZSCORE 836 0.64 0.309 11.41 −0.372 1.101 0.625 6.194 0.0344 0.993DPS / TA 834 0.871 0.906 1.101 0.00339 0.109 0.873 0.971 0.466 0.0915NL / TA 836 0.524 0.524 0.966 0.0732 0.109 0.524 0.838 0.258 0.106GL-GROWTH 725 0.222 0.198 0.99 −0.482 0.211 0.223 0.864 −0.203 0.206D-M&A 836 0.311 0 1 0 0.463 0.311 1 0 0.463D-LISTED 836 0.0478 0 1 0 0.214 0.0478 1 0 0.214HHI 836 0.178 0.151 0.339 0.134 0.055 0.178 0.339 0.134 0.055ML-SHARE 836 0.976 0.971 1 0.965 0.0116 0.976 1 0.965 0.0116BOR 836 2.509 1.986 9.302 0.315 2.202 2.509 9.302 0.315 2.202GNI-GROWTH 836 0.159 0.171 0.364 0.0636 0.0601 0.159 0.364 0.0636 0.0601

(continued on next page)

197A.C.H

.Lei,Z.Song/GlobalFinance

Journal24(2013)

188–202

Page 11

Table 4 (continued)

Variables Observations Unwinsorized data Winsorized data

Mean Median Maximum Minimum Std. dev. Mean Maximum Minimum Std. dev.

Ln(POP) 836 7.165 7.179 7.205 7.009 0.0395 7.166 7.205 7.026 0.0391GDP-GROWTH 836 10.55 10.03 14.24 3.837 2.193 10.55 14.16 4.066 2.186Ln(M1) 836 9.061 9.281 9.999 6.308 0.793 9.061 9.999 6.368 0.792DPSR 836 0.0343 0.0225 0.113 0.0198 0.0234 0.0343 0.11 0.0198 0.0233FB 836 0.123 0 1 0 0.329 0.123 1 0 0.329JSCB 836 0.207 0 1 0 0.405 0.207 1 0 0.405SOCB 836 0.0766 0 1 0 0.266 0.0766 1 0 0.266PB 836 0.00598 0 1 0 0.0772 0.00598 1 0 0.0772

198A.C.H

.Lei,Z.Song/GlobalFinance

Journal24(2013)

188–202

Page 12

Table 5Correlation matrix of the regression variables.This table provides correlations of all the regression variables. Some observations that display 1.00 are not perfectly collinear, but instead are a result of rounding.Figures in bold are correlations large than 0.5.

LC / TA cat fat LC / TA cat nonfat EQ / TA CO / INC Y-STD CREDITRISK ZSCORE Ln(TA) D-M&A HHI ML-SHARE GDP-GROWTH

LC / TA cat fat 1LC / TA cat nonfat 0.94 1EQ / TA −0.27 −0.30 1CO / INC −0.06 −0.01 0.39 1Y-STD −0.07 −0.08 0.46 0.33 1CREDITRISK 0.15 −0.03 0.00 −0.16 0.05 1ZSCORE −0.11 −0.09 0.04 0.04 −0.35 −0.09 1Ln(TA) 0.01 0.00 −0.28 −0.14 −0.21 0.09 −0.05 1D-M&A 0.07 0.02 −0.14 −0.23 −0.08 0.23 −0.10 0.25 1HHI −0.15 −0.10 0.11 0.23 −0.03 −0.32 0.08 0.04 −0.44 1ML-SHARE −0.30 −0.27 0.16 0.21 0.02 −0.27 0.06 0.06 −0.38 0.81 1GDP-GROWTH 0.19 0.15 −0.15 −0.33 −0.13 0.20 0.01 −0.04 0.16 −0.33 −0.50 1BOR −0.01 −0.04 −0.06 −0.05 −0.13 0.02 0.06 0.00 0.08 −0.07 0.05 0.07GNI-GROWTH 0.07 0.03 −0.11 −0.31 −0.13 0.12 0.05 −0.02 0.08 −0.13 −0.28 0.79LN(POP) 0.13 0.10 −0.04 −0.23 0.15 0.31 −0.14 0.00 0.42 −0.83 −0.78 0.23NIM 0.02 0.00 0.00 −0.27 0.17 0.22 −0.12 −0.09 0.03 −0.09 −0.10 0.17ROA −0.12 −0.14 −0.12 −0.67 −0.07 0.10 −0.07 −0.02 −0.01 0.10 0.17 0.15NL / TA 0.28 0.24 0.06 −0.10 0.10 0.24 −0.01 0.05 0.07 −0.17 −0.24 0.10GL-GROWTH −0.01 −0.01 0.08 −0.33 0.11 −0.03 0.03 −0.06 −0.05 0.06 0.04 0.01DPS / TA 0.39 0.37 −0.48 −0.18 −0.10 0.13 −0.12 0.06 0.22 −0.33 −0.37 0.15D-LISTED −0.06 −0.11 −0.05 −0.12 −0.04 0.22 −0.07 0.36 0.26 −0.17 −0.10 0.03DPSR −0.17 −0.18 0.03 −0.17 −0.12 −0.12 0.12 0.02 −0.26 0.58 0.57 0.26Ln(M1) 0.10 0.07 −0.03 −0.24 0.16 0.32 −0.14 0.01 0.42 −0.82 −0.75 .23

BOR GNI-GROWTH Ln(POP) NIM ROA NL-TA GL-GROWTH D-TA D-LISTED DPSR Ln(M1)

BOR 1GNI-GROWTH 0.35 1Ln(POP) −0.14 −0.08 1NIM −0.05 0.16 0.19 1ROA 0.08 0.25 −0.11 0.41 1NL / TA −0.02 0.01 0.29 0.23 −0.08 1GL-GROWTH 0.08 0.05 −0.08 0.07 0.07 0.10 1DPS / TA −0.01 0.00 0.35 0.12 0.10 −0.14 −0.05 1D-LISTED −0.01 0.02 0.19 0.03 0.08 −0.01 −0.03 0.07 1DPSR 0.33 0.57 −0.73 0.03 0.35 −0.18 0.10 −0.30 −0.08 1Ln(M1) −0.11 −0.07 1.00 0.20 −0.08 0.29 −0.07 0.33 0.20 −0.68 1 199

A.C.H

.Lei,Z.Song/GlobalFinance

Journal24(2013)

188–202

Page 13

200 A.C.H. Lei, Z. Song / Global Finance Journal 24 (2013) 188–202

multicollinearity does exist in our regression models, the overall results are consistent. Additional controlvariables are chosen carefully to avoid multicollinearity.

4. Empirical results

4.1. Liquidity creation and bank capital — all banks

Table 6 displays the results of the pooled, fixed-effects, and random-effects regressions of liquiditycreation on bank capital. A Durbin–Hausman test between our models suggests the use of a fixed-effectsmodel. In the fixed-effects and other models, the lagged EQ / TA is negative and significant at the 1% level.This means that Chinese banks with more capital create less liquidity, which is in line with the financialfragility hypothesis and contrast with Berger and Bouwman's (2009) results. There may be several reasons

Table 6Liquidity creation on bank capital (all banks) — pooled, fixed, and random effects.This table presents the regression results using different modeling methods based on data from all banks. The dependent variable isLC / TA cat fat, the liquidity creation divided by total assets according to the cat fat method. The cat fat method measures liquiditycreation by classifying all bank activities based on product category and includes off-balance-sheet activities.EQ / TA is the capital ratio, measured as total equity divided by total assets. Y-STD is the standard deviation of return on assets.CREDITRISK is a credit-risk measure, calculated as the bank's risk-weighted assets and off-balance-sheet activities divided by totalassets. ZSCORE is the sum of return on assets and equity / TA ratio, divided by the standard deviation of annual return on assets.Ln(TA) is bank size, measured as the natural log of total assets. D-M&A is a dummy that equals to 1 if the bank was involved inmergers and acquisitions over the past 3 years. HHI is a bank-level Herfindahl index based on deposits. ML-SHARE is the share oflarge and medium banks' deposits to the total deposits in the market. BOR is the benefit-of-openness ratio, measured as the share ofincome inflow to the share to income outflow in the country's global trades. GNI-GROWTH is the growth of gross national income inChina. Ln(POP) is the natural log of population.The regressions run a pooled panel model in the first column, a fixed-effects model in the second column, a model with bank fixedeffects and time fixed effects in the third column, and a random effects model with bank fixed effects in the fourth column.t-statistics based on robust standard errors clustered by bank are in parentheses. *, **, and *** denote significance at the 10%, 5%, and1% levels, respectively.

LC / TA cat fat

Pooled Fixed effects Random effects

EQ / TA −0.864 −1.068 −0.801 −0.9(−6.29)*** (−5.83)*** (−4.54)*** (−5.75)***

Ln(TA) −0.004 −0.049 −0.007 −0.014(−1.22) (−2.91)*** (−0.43) (−1.97)**

D-MA 0.006 −0.029 −0.024 −0.02(0.42) (−1.84)* (−1.56) (−1.36)

HHI 0.898 0.536 0.652(4.50)*** (3.24)*** (4.26)***

ML-SHARE −10.82 −7.648 −8.568(−9.19)*** (−8.46)*** (−9.92)***

GNI-GROWTH −0.495 −0.4 −0.435(−3.99)*** (−4.37)*** (−4.83)***

Ln(POP) −1.462 −0.709 −1.565(−4.36)*** (−1.49) (−5.04)***

Y-STD 0.006 −0.049 0.019 −0.034(0.20) (−1.66)* (0.65) (−1.21)

CREDITRISK 0.094 0.184 0.181 0.153(4.31)*** (8.03)*** (7.96)*** (7.35)***

ZSCORE −0.049 −0.034 −0.006 −0.039(−2.29)** (−1.76)* (−0.33) (−2.11)**

c 21.407 13.76424 0.58 20.088(6.57)*** (3.65)*** (2.66)*** (7.34)***

F test for no fixed effects (Pr.)Hausman test for random effects (Pr.) 0.001Time fixed effects No Yes NoBank fixed effects Yes Yes YesObservations 836 836 836 836R-square 0.1993 0.7095 0.7501 0.2401

Page 14

201A.C.H. Lei, Z. Song / Global Finance Journal 24 (2013) 188–202

for this finding. First, large bank size and frequent mergers and acquisitions obstruct banks' liquiditycreation efficiency, as the asset base increases rapidly. Also, large state-owned commercial banks andpolicy banks are subject to heavy intervention by the government, and they have lower profitability (Lin &Zhang, 2009). Further investigation of different bank sizes may provide explanations for this result.

4.2. Effects of capital on liquidity creation — interactions of bank capital and bank types

Table 7 shows the regression results for the capital ratio on both “cat fat” and “cat nonfat” liquidityratios, including the interaction terms of bank-type dummies. We create dummy variables to representdomestic, state-owned, and foreign banks. Note that these bank type dummies are omitted because theyare time-invariant. Coefficients of both cat fat and cat nonfat lagged EQ / TA are negative and significant atthe 1% level for all models except in the first two columns. The interaction terms for domestic banks arenegative and significant. Domestic banks are the major type of banks in the sample; therefore, theirinteraction showing negative significance is consistent with the results for the full sample regression.State-owned banks show a similar relationship with other domestic banks, with insignificant coefficientsof the interaction terms EQ / TQ*State-owned banks. These results are in line with the government-backedintuition, such that the government should support Chinese banks if under financial distress. For foreignbanks, we find a positive and significant interaction term with EQ / TA. The interaction terms for the

Table 7The effect of capital on liquidity creation for different bank-type subsamples.This table presents the regression results based on different bank-type subsamples, using both fixed bank effects and time effectsmodels. The sample period is from 1988 to 2009. The dependent variable is LC / TA cat fat or LC / TA cat nonfat, the liquidity creationdivided by total assets according to the “cat fat” and “cat nonfat”methods. “Cat fat” and “cat nonfat” methods are two methods usedto measure liquidity creation, which classify all bank activities based on product category. The fat (nonfat) liquidity-creationmeasures include (exclude) off-balance-sheet activities. Please refer to Table 1 for variable definitions. The bank types are dummyvariables Domestic banks, State-owned banks, and Foreign banks. These dummy variables are omitted in the regression because oftheir time-invariant nature; thus only the interaction terms and the EQ / TA coefficients are shown. The regressions are run withboth bank fixed effects and time fixed effects, which cause variables HHI, ML-SHARE, BOR, GNI-GROWTH, and Ln(POP) to be omitted,because they have only one value in each year. Therefore, they are not displayed in the table.c is the intercept. t-statistics based on robust standard errors clustered by bank are in parentheses. *, **, and *** denote significance atthe 10%, 5%, and 1% levels, respectively.

Dependent variable Domestic banks State-owned banks Foreign banks

LC / TAcat fat

LC / TAcat nonfat

LC / TAcat fat

LC / TAcat nonfat

LC / TAcat fat

LC / TAcat nonfat

EQ / TA −0.677 −0.575 −1.14 −1.285 −1.294 −1.44(−1.56) (−1.43) (−6.07)*** (−7.38)*** (−6.52)*** (−7.87)***

EQ / TA*Domestic banks −0.592 −0.863(−1.24) (−1.96)**

EQ / TA*State-owned banks −0.534 −0.073(−0.59) (−0.09)

EQ / TA*Foreign banks 0.951 1.135(1.76)* (2.28)**

Y-STD −0.049 −0.019 −0.054 −0.026 −0.047 −0.019(−1.38) (−0.59) (−1.53) (−0.79) (−1.34) (−0.58)

CREDITRISK 0.127 0.036 0.128 0.037 0.126 0.035(7.91)*** (2.42)** (7.96)*** (2.5)** (7.87)*** (2.38)**

ZSCORE −0.033 −0.026 −0.038 −0.033 −0.029 −0.023(−1.62) (−1.38) (−1.86)* (−1.77)* (−1.44) (−1.22)

Ln(TA) −0.031 −0.037 −0.031 −0.036 −0.032 −0.037(−3.30)*** (−4.21)*** (−3.25)*** (−4.11)*** (−3.35)*** (−4.26)***

D-MA −0.032 −0.041 −0.032 −0.042 −0.030 −0.039(−2.04)** (−2.88)*** (−2.03)** (−2.92)*** (−1.94)* (−2.76)***

c 0.782 0.809 0.782 0.808 0.784 0.812(7.59)*** (8.52)*** (7.59)*** (8.49)*** (7.63)*** (8.56)***

Time fixed effects Yes Yes Yes Yes Yes YesBank fixed effects Yes Yes Yes Yes Yes YesObservations 836 836 836 836 836 836R-square 0.4038 0.3875 0.4203 0.4026 0.4021 0.3871

Page 15

202 A.C.H. Lei, Z. Song / Global Finance Journal 24 (2013) 188–202

foreign bank dummy and the capital ratio are significantly positive at 10% and 5% levels for LC / TA cat fatand LC / TA cat nonfat, respectively (t-statistics are 1.76 and 2.28, respectively). Consistent with Bergerand Bouwman (2009), where large banks are dominated by the risk absorption effect, liquidity creation offoreign banks in China is still affected by risk absorption effect. The overall effects (financial fragility vs riskabsorption) seem to be offset, however, when foreign banks are operating in China. These results suggestthat even though Chinese banks' liquidity creation is dominated by the financial fragility hypothesis,foreign banks are more affected by the risk absorption effect. This result supports the notion thatgovernment-backed banks do not require additional capital to absorb the risk of providing loans.

5. Conclusion

This paper explores liquidity creation in China and the relationship between liquidity creation andbank capital. Our results support the “financial fragility-crowding out” hypothesis for Chinese banks. Wefind that bank capital is negatively related to liquidity creation in general. For foreign banks, however, the“risk absorption” hypothesis nullifies the financial fragility effect as the negative relationship of bankcapital and liquidity creation is reduced. These results are in line with Berger and Bouwman's (2009)findings for large US banks. The information asymmetry in the recent years between banks and investorshas declined substantially, especially for listed large banks, thus reducing the financial fragility that leadsto lower liquidity creation (Diamond & Rajan, 2000, 2001).

There are several potential implications from this study. State-owned or government-backed banks,which are financially less fragile, may have lower liquidity creation. The recent European debt crisis andsub-prime crisis lead to the temporary nationalization of some major banks.9 Most banks in developedmarkets have been operating under laissez faire for decades. Shifting to complete government controlwould decrease their ability to create liquidity, since according to our findings of government-controlledbanks, the financial fragility-crowding out hypothesis dominates. As a result, the government can thendecrease bank capital to increase liquidity creation, yet resources are prone to misallocation. Thegovernment then “absorbs” the risk of these nationalized banks, putting the public interest in jeopardy.

References

Allen, F., & Gale, D. (2004). Financial intermediaries and markets. Econometrica, 72(4), 1023–1061.Allen, F., & Santomero, A. M. (1997). The theory of financial intermediation. Journal of Banking & Finance, 21(11), 1461–1485.Berger, A. N., & Bouwman, C. H. (2009). Bank liquidity creation, monetary policy, and financial crises. Review of Financial Studies,

22(9), 3779–3837.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 & Finance, 33(1), 113–130.Berger, A. N., & Udell, G. F. (1994). Did risk-based capital allocate bank credit and cause a “credit crunch” in the United States? Journal

of Money, Credit, and Banking, 26(3), 585–628.Bhattacharya, S., & Thakor, A. V. (1993). Contemporary banking theory. Journal of Financial Intermediation, 3(1), 2–50.Coval, J. D., & Thakor, A. V. (2005). Financial intermediation as a beliefs-bridge between optimists and pessimists. Journal of Financial

Economics, 75(3), 535–569.Deep, A., & Schaefer, G. (2004). Are banks liquidity transformers? KSG Working Paper, No.RWP04-022.Diamond, D. W., & Dybvig, P. H. (1983). Bank runs, deposit insurance, and liquidity. The Journal of Political Economy, 91, 401–419.Diamond, D. W., & Rajan, R. G. (2000). A theory of bank capital. The Journal of Finance, 55(6), 2431–2465.Diamond, D. W., & Rajan, R. G. (2001). Liquidity risk, liquidity creation, and financial fragility: A theory of banking. Journal of Political

Economy, 109, 287–327.García-Herrero, A., Gavilá, S., & Santabárbara, D. (2009). What explains the low profitability of Chinese banks? Journal of Banking &

Finance, 33(11), 2080–2092.Gorton, G., & Winton, A. (2000). Liquidity provision, bank capital, and the macroeconomy. Working Paper (October 9, 2000).Hancock, D., Laing, A. J., & Wilcox, J. A. (1995). Bank capital shocks: Dynamic effects on securities, loans, and capital. Journal of

Banking and Finance, 19(3), 661–677.Kashyap, A. K., Rajan, R., & Stein, J. C. (2002). Banks as liquidity providers: An explanation for the coexistence of lending and

deposit-taking. The Journal of Finance, 57(1), 33–73.Lin, X., & Zhang, Y. (2009). Bank ownership reform and bank performance in China. Journal of Banking & Finance, 33(1), 20–29.McGuinness, P. B., & Keasey, K. (2010). The listing of Chinese state-owned banks and their path to banking and ownership reform.

China Quarterly, 201(1), 125.Peek, J., & Rosengren, E. (1995). The capital crunch: Neither a borrower nor a lender be. Journal of Money, Credit, and Banking, 27(3), 625–638.Repullo, R. (2004). Capital requirements, market power, and risk-taking in banking. Journal of Financial Intermediation, 13(2), 156–182.Von Thadden, E. L. (2004). Bank capital adequacy regulation under the new Basel Accord. Journal of Financial Intermediation, 13(2), 90–95.

9 There are previous crises that induce nationalization of banks, for example, in Japan during the 1990s.