Does high-quality auditing decrease the use of collateral?Analysis from the perspective of lenders’self-protection

Yang Yang

Accounting Department,Zhongnan University of Econom ics and Law,China

Does high-quality auditing decrease the use of collateral?Analysis from the perspective of lenders’self-protection

Yang Yang

Accounting Department,Zhongnan University of Econom ics and Law,China

A R T IC L E IN F O

Article history:

Received 15 M ay 2013

Accep ted 11 August 2014

Availableonline18September 2014

Credit risk

W eexam ine theassociation between auditquality and theuseo f co llateral in a samp le o f Chinese f rm s from 2005 to 2011.Using the fu ll sam ple,we document a negative relationship between audit quality and the use o f co llateral that is consistentw ith lenders’interests.W e also show that audit quality and co llateral are regarded as alternativemeans o f reducing debt credit risk.Our conclusions are robust after using an auditor-sw itching test,the Heckman two-stagemodeland a propensity-scorematchingmodel to addressendogeneity issues.China’s institu tional background is also considered.First,w e f nd that in the group of f rm s in which large shareho lders are ab le to contro l borrowers’activities,the substitution efects between collateral and audit quality are reduced when the degree of separation between large shareholders’contro l and ownership is high.Second,these substitution efects are greater when the borrowers’ultimate contro ller is a state-owned enterp rise(SOE)rather than a non-state-ow ned en terp rise(NSOE).Third,the dif erences in substitu tion ef ects between NSOEs and SOEs are smaller in areas w ith a high marketdevelopment index.W e conclude that the substitution ef ects are smaller in high credit risk f rm s than in low credit risk f rm s.

?2014 Production and hosting by Elsevier B.V.on behalf o f China Journalo f Accounting Research.Founded by Sun Yat-sen University and City U niversity o f H ong K ong.

1.Introduction

Bank loans p lay an important role in borrowers’external fnancing.Unlike developed foreignm arkets,China is undergoing econom ic transition,as its em erging equitymarket emerged as recently as1990.In addition,its publicbondm arket rem ainsundeveloped,long-term bond fnancingmakesup only a sm allproportion(less than 5%)of its long-term debt.Therefore,bank loansarestillvital to corporate funding in China.Chen etal.(2010)and Chen and Xiao(2012)fnd thatdespite thesignif cantgrowth ofChina’sequitymarket,borrowersstill rely overwhelminglyon banksto satisfy theirneed for capital.1Chen etal.(2012)show that the capital raised from bankseach year ranged from RMB 1250 to RM B 3100 billion between 2001 and 2006,whereas the capital raised from the stock market during this period wasonly between RMB 30 and RMB 250 billion a year.Chen and X iao(2012)calculate that the ratios o f bank loans and capital raised from the equitym arket to GDP,and show that the proportion constituted by bank loans is 75 tim es greater than that o f capital raised from the equity m arket.Allen etal.(2005)show that theproportion ofGDPcomprised bybank loans ishigher in China than in other developing countries.M oreover,as important intermediaries,banksarevital to the adjustmento f theallocation ofcapital.They providenotonlycapitalloans,butmacroeconom ic regulation.Chenand Li(2011)show that theindustriesinvolved in China’s5-yearsupportplan havepreferentialaccessto capitaland exhibitbetter performance than industriesnot involved in thep lan.Weknow thatbank loansaresignif cantnotonly to borrowersbut to policymakers,as theymakea large contribution to thedevelopmentofChina’seconomy.

Collateral,togetherw ith interest rates,deb tm aturity and o ther covenants,isw idespread in debt con tracts. It is used to so lve them o ral-hazard and adverse-selection p rob lem s caused by in form ation asymm etry(Bester, 1985;Chan and K anatas,1985;Boot et al.,1991).La Porta et al.(1998)argue that collateral is cen tral to the relationship between lenders and borrowers.Collateral requirements arise from the agency p rob lem that af icts f nancial relationships,especially w ith regard to debt f nancing.Collateral isnotonly an ex-antemechanism o f interest alignment but an ex-postmechanism o f control allocation,used to supervise borrowersand to m inim ize losseswhen a borrower defau lts on a loan repayment(Aghion and Bolton,1992).

Externalauditing isa usefulmechanism o f both corporategovernanceand externalsupervision.H igh-quality auditing can improve the quality of fnancial in formation(Becker and Defond,1998;Francis et al.,1999; Teoh and W ong,1993;Qi,2004;W u and Li,2006),increase the transparency of accounting information (Chen and W ang,2006)and decrease a company’s capital costs(Khurana and Raman,2004;Pittman and Fo rtin,2004).In add ition,m any em pirical stud ies show that when m ak ing decisions,lenders consider the quality of accoun ting in form ation as well as several key f nancial m easures(Chen et al.,2010;Rao and Hu,2005;Goncharov and Zimm erman,2007).The quality o f accounting information also af ects the usefulness o f a debt contract(Lu et al.,2008).As important stakeholders,lenders can use high-quality external auditing to decrease credit risk.

However,there is a trade-of between the use o f co llateral and the necessity o f high-quality auditing. A lthough collateral is determ ined ex-ante,itmay incur the costs of screening and monitoring the p ledged assets,aswell as disposalexpensesand losses due to the sale o f specialized assets(Chen et al.,2012).External audit qualitymay be afected by many factors,such as the purchasing o f audit opinion and the collusion o f auditorsw ith borrowers,which severely depreciateexternalauditquality.This isnot in the interestso f lenders. The question arises o f whether lenders use the optim alm eans to control cred it risk when designing deb t contracts?2W e regard collateral as a debt contract issue determ ined ex-ante,as borrowers use their assets or properties to p ledge collateral. Compared w ith externalauditing,collateral isa betterway to prevent credit risk.However,neither collateralnor externalaudit is free of costs,and both have som e problems.Theaim o four study is to exam ine the degree to which lenders rely on audit quality,which is ref ected in the reduction of collateral.W e focus on collateral in this study,on the grounds that the totalnum ber of bank loans is incapable of ref ecting bank-loan restrictions,as bank loans can be d ivided into co llateral,guaranto r and credit loans according to their degree of restriction.3Bank loans can be divided into the follow ing categories,according to their level of restriction:collateral loans,guarantor loans and credit loans.Co llateral involves two kindsof loans,namely p ledged loansandmortgage loans.A guarantor loan isa loan forwhich a third party has joint o r guaranteed liab ility,if the loan defaults.M ortgage loans are bank loans secured on the borrower’s property or the property of the third party.Pledge loans are secured on the borrower’s or third party’sm ovable property or rights of claim.If the borrower defaultson the loan,the lender is permitted to dispose of the pledged ormortgaged properties to cover their lossesand prevent credit risk.Therefore,webelieve thatcollateraland guarantor loans,especially collateral loansaremoreefective in helping lender to avoid cred it risk.W e use the p ropo rtion of co llateral loans in our main tests,and the proportion o f collateral and guarantor loans in ou r sensitivity tests.W e believe that the use o f collateral and guarantor loansmore accurately ref ectsour expected relationshipsbetween lendersand borrowers,leading tomore reliable results.Chen (2011)uses credit loans to exam ine the relationship between auditor reputation and borrowers’bank loans.

W e aim to answer two m ain questions.First,from the perspective of self-pro tection,how do lenders use collateral and external auditing to m inim ize credit risks,and are these strategies generally substitutable or comp lem entary?Black et al.(2004),N ikolaev(2010)and Chen et al.(2012)exam ine this prob lem,but fail to reach the same conclusions.Second,how does China’s institutional background afect the relationship between audit quality and collateral?The country’s f rms are dom inated by large shareho lders,shareho lders are high ly concentrated and lendersare inefectively p rotected(Xiao,2007).Therefore,we consider the extent o f large shareholders’control and the degree o f separation of the controlling owner’s control and ownership. W e also add ress the nature of the ultimate contro ller and themarket-development index,aswew ish to determ ine how China’s institutional background afects lenders’decisions,specif cally w ith regard to the relationship between the use of co llateral and external auditing.

The f rst em p irical study of the u ltim ate controller and the agency problem between con trolling owners and m inority shareholderswas conducted by La Porta et al.(1999).Researchersexam ining the ro le of large shareholders tend to hold two distinct views:one set of researchers emphasize alignment ef ects(Kanga and Shivdasani,1995;Kahn and W inton,1998),the other highlights entrenchment ef ects(Claessens et al., 2000).Xiao(2007)and La Porta et al.(1999)fnd that borrowershave a highly concentrated,rather than dispersed ownership structure,especially in developing countries.A large shareholder leverages contro l through stock pyram idsor cross-shareho ldingswhile keeping hisor her levelo f ownership low,which resu lts in a high levelof separation between contro land ownership(Luo and Tang,2008;La Porta etal.,1999).Themotivation for separating control and ownership has two further dimensions,depending on the role o f the large shareholder.First,alignm ent and risk-diversif cation(Pan and Yu,2012);and second,the agency problem and entrenchm en t(C laessens et al.,2002;Johnson,2000).A lthough the separation o f contro l and ownership has a considerable inf uence on the agency problem between large shareho lders and m inority shareho lders(Su and Zhu,2003;Yu and Xia,2003a,2003b),weshou ld also consider thebasicp rem ise that the largeshareholder hasa high levelof contro l(Pan and Yu,2012;Fan and Wong,2002;M orck etal.,2005).Ifwedirectly exam ine the relationship between theagency p rob lem and theseparation o f controland ownership,w ithout considering this important prem ise,wemay d raw unreliab le conclusions.Therefore,weneed to determ inewhether the large shareholderhasa high levelo f contro l,beforeexam ining the relationship between theagency p rob lem sand the separation of contro land ownership.Xiao(2007)and Chen andW ang(2010)note that the ultimate controller m ay be either a state-owned or non-state-owned borrower.SOEsand NSOEs havemany diferent characteristics(Cu lland Xu,2005;Chen etal.,2010;Fan et al.,2007;Lin and Li,2004).For instance,SOEshavemore opportunities than NSOEs to ob tain capital,but face a lower credit risk(A llen et al.,2005;Guariglia et al., 2011).Un like the developed m arkets in W estern coun tries,China’s regional econom ic developm ent is unbalanced,w ith vast diferences between regions in term s o f legal developm ent and m arketization(Yu and Pan, 2008;Fan andW ang,2009).W ith greater legaldevelopmentandmarketization,decisions regarding bank loans aremoremarketized,and more independent of government power(Jiang and Li,2006;Fang,2007;Yu and Pan,2008a,2008b;Weiand Shen,2009).As a resu lt,SOEs face greater credit risk.

We fnd that the use o f co llateral and high-quality external auditing are regarded as substitutes by lenders m aking decisionsabout bank loans.We notice that in the group o f f rmswhose large shareho ldersare capable o f contro lling borrowers,these substitution efects are reduced when the degree o f separation of control and ownership is high.In the other group,however,the substitution ef ects are not found to increase.W e show that thesubstitution efectsbetween collateraland high-quality auditing are greater in SOEs,butwealso p rove that a higher m arket developm en t index can reduce the d if erences between SOEs and NSOEs.

This study m akes several potential contribu tions to the literatu re.First,we take a novel perspective in exam ining how lenders p rotect them selves by in fuencing bo rrowers’activities through high-quality aud iting and collateral requirements.M ost prior studieshave addressed this problem from the borrower’s perspective, and conclude that high-quality auditing helpsborrowers to obtain bank loansand reduce the cost o f borrowing.For example,Hu and Tang(2007)notes the relationship between interest rates and auditing.Liao et al. (2010)use going-concern opinions to exam ine the role of auditors.Our results show that lenders use highquality auditing and co llateral requirements to inf uence borrowers’actions;high-quality auditing and collateral are regarded as substitutes,and their substitution ef ects are ad justed according to the credit risk of the borrower,which depends on the institutional environm ent.Second,our study enriches the existing literature on the determ inantsof co llateral.Prior researchersnote that collateral isafected by thedegree of informationasymm etry(Brick and Palia,2007;Chakrabo rty and Hu,2006;Jim enez et al.,2006),the relationship between lendersand borrowers(Sharpe,1990;Rajan,1992),the level o f competition in the loan market(Besanko and Thakor,1987),and the costs(benef ts)of screening borrowers(M anove and Padilla,1999;M anove et al., 2001).We note that high-quality external auditing decreases the use of co llateral and that the extent o f this decrease depends on the level o f credit risk.Third,our study contributes to existing research on large shareholders’separation of control and ownership.A lthough Prior researchershave used the separation of control and ownership to exam ine entrenchment,agency or alignment ef ects,few studieshave addressed its f nancial consequences,such as the cost of borrow ing.Lin et al.(2011)samp le f rms from a number o f foreign countries’to exam ine the relationship between the degree o f separation of control and ownership and the cost o f borrow ing,as refected in interest-rate spreads.W e d ivide our sam p le into two groups to exam ine the inf uence of the degree o f separation of con tro l and ownership on the relationship between co llateral and high-quality aud iting.Our results ind icate that in high contro l group,the substitution ef ects between collateral and highquality auditing are on ly weakened when thedegree o f separation o f contro land ownership ishigh.However, we do not fnd these efects to be stronger in the low-contro l group,when the degree o f separation o f control and ownership ishigh.Our evidence ofers insight into the incentives for and consequenceso f large shareho lders’separation o f contro l and ownership.Finally,follow ing A rm strong et al.(2010)and Skinner(2011),we exam ine the interaction between various aspects of debt contracts to determ ine,specif cally,whether they aresubstitutab leor complementary.W eseek to ascertain whether variousmechanismsareused asalternatives or comp lem ents.Prior researchers have noticed the relationships between collateral and other contracting issues and characteristics and the quality o f accounting(Zhang,2008;N ikolaev,2010;Chen et al.,2012; Black et al.,2004).The focus o f our study is the use of an externalm onito ring m echanism,nam ely,external aud iting.W e exam ine the relationship between externalauditing and the use o f co llateral,and then investigate the efects on this relationship of dif erent levels of credit risk.

The rest of the paper is structured as follows.W e develop our hypotheses in Section 2,and p resent the research design,sample selection and data in Section 3.We present the resultso f ourmain tests in Section 4 and results of our sensitivity tests in Section 5.Section 6 concludes the study.

2.Development of hypotheses

A ccord ing to contract theo ry and agency theory,an enterp rise is fo rm ed from the com bination of various kinds of contracts.Du ring the design and im p lem entation process,properm onitoring m echanism s should be used to align stakeholders’interests(Jensen and M eckling,1976).The focusof our study is the agency problem between lenders and borrowers,whose con f ictso f interestmain ly concern asset substitution(Jensen and M eck ling,1976),under-investment(M yers,1977)and equity delusion(Sm ith and Warner,1979),specif cally the fear that borrowersw ill exp rop riate lenders.The aim o f our study is to determ inewhether a lender w ill adjust the use o f collateral when he or she knows the borrower’s audit quality,and whether the degree o f adjustment is always the same.

2.1.Collateral and audit quality

From the perspective of self-p rotection,agency costs can be decreased by using alignmentmechanisms to solve the agency problem between lendersand borrower.Collateral and external auditing are two suchmechanisms.The use of collateral isnecessitated by in formation asymmetry between lendersand borrowers,and is a common componento f bank loans(Yang and Qian,2008).Bester(1985)and Chan and Kanatas(1985)f nd that in conditions of information asymm etry,co llateral can reduce interest rates and may be regarded as a signal of better credit quality.M any researchers have focused on the role of external audit.Becker and Defond(1998)and Francis(1999)note that the clientso f Big-four audit f rmshave lower discretionalaccruals than the clientso f non-Big-4 f rm s.Qiet al.(2004)and W u and Li(2006)test the Big-4m easure of audit quality and fnd that the Big-4 audit f rm s do p rovide a high aud it quality.Teoh and W ong(1993)note that clien ts of the Big-4 have a higher earnings-response coef cient than clients of other audit f rm s.K hurana and Ram an (2004)report that the use o f a Big-4 f rm can reduce the cost of both equity capital and debt(Pittman andFo rtin,2004).These researchers argue that external aud iting can im p rove the quality of bo rrowers’f nancial in formation and decrease in formation risks,helping lenders to make decisions.

In addition,lenders that pay seriousattention to borrowers’fnancial in formation are found tomakebetter decisions(Goncharov and Zimm erman,2007),and thusalleviate credit risk.Chen etal.(2010)f nd thatm any debt contractsare availab le,based on dif erent accountingmeasures.Rao and Hu(2005)document the roles o f fnancialmeasures and accounting in formation in lenders’decision making,and show that lenders frequently consider borrowers’information.G rama et al.(2008)fnd that lenders are likely to givemore severe debt contracts to borrowers invo lved in fnancial restatements.Lu et al.(2008)report that earningsmanagem ent impairs the efectiveness of debt contracts.

In add ition,Chen et al.(2012)show that the use of collateral incurs the costs of screening and m onitoring the p ledged assets,as well as disposal expenses;it also brings lenders’losses due to the sale of specialized assets.M anove and Pad illa(1999)observe that lenders generally seek to give cap ital to high-quality borrowers,butwhen costs,benef tsand the level o f competition are taken into consideration,lenders’decisionsare unlikely to beso p rudential;even in an undeveloped bank-loansmarket,lenders regard theuseof collateralas a substitute for inef ectivemonitoring(M anoveetal.,2001).W hen designing debt contracts,do lendersuse the optimalmeans to contro l credit risk?Does the tradeo f between costsand benef ts inf uence lenders’decisions?

Finally,the relationship between debt contract issuesand othermechanisms rem ainsan empiricalproblem. Hu and Tang(2007)documents that higher-quality audit opinions and larger auditor o f ces reduce interest ratesand increase thematurity of debt.Chen(2011)fnds thatauditors’reputation thenumber and p roportion o f borrowers’credit bank loans increasesas the auditor’s reputation increases.Black et al.(2004)use a bankindustry sam p le to p rove that external regu lation and deb t covenants are substitutable,whereas,N ikolaev (2010)fnds that conservatism and deb t covenants are com p lem entary.Chen et al.(2012)show that lenders generally regard collateral and conservatism as substitutab le,butwhen the borrower has a low credit quality or a high proportion of intangib le assets,the lenderw ill reverse thisassumption,regarding co llateraland conservatism as complementary.As the studies above are not in comp lete accordance,we rely on our empirical analysis to identify themore accurate of the fo llowing two alternatives.

H 1a.Co llateral requirem en ts are positively related to high-quality external aud iting.

H 1b.Co llateral requirem ents are negatively related to high-quality external auditing.

2.2.Collateral,separation of control and ownership and audit quality

The f rst empirical studies of the u ltimate contro ller and the agency problem between contro lling owners and m inority shareholders were conducted by La Ports et al.(1999)and Shleifer and Vishny(1997,1999). According to La Porta et al.(1999)the agency p roblem that arises between large shareho lders and m inority shareholdershas becomemore severe than that between shareho lders and managers.Researchers exam ining the role of large stockho lders tend to em phasize either alignm ent ef ects o r entrenchm en t efects(C laessens et al.,2000).K anga and Shivdasani(1995)note that large stockho lders can alleviate the“free-rider”prob lem associated w ith m inority shareho lders,and large shareho lders have an incentive to m onito rm anagers(K ahn and W inton,1998).Theses scholars conclude that the presence of large shareho lders can imp rove corporate governance by preventing managers from acting in their self-interest.However,these studies are based on theownership structure o f U.S.f rms,which arehighly dispersed.Samplesof f rms from other countries show that ownership structures in these areas are not dispersed but highly concentrated,especially in developing countries(La Porta et al.,1999;Franks and M ayer,1998;Cronqvist and N ilsson,2003).Xiao(2007)notes that Chinese f rmshave a high ly concentrated ownership structure,usually dom inated by a single large stockholder.This large shareholder has the incentive to leverage contro l through stock pyram ids or cross-shareholdings whilem inim izing his or her ownership,which resu lts in a high level of separation of contro l and ownership(Luo and Tang,2008;La Po rta et al.,1999).This separation of con tro land ownership has received considerab le attention from scho lars.The em p irical research pertaining to the incentives for and consequences o f separating con tro l and ownership can be divided in to two groups.The researchers in the f rst group agree that a high levelo f separation o f contro land ownership can resu lt in an alignmento f interestsby encouragingthem onitoring ofm anagers.They show that under these cond itions,a sm all am ount o f cap ital is required to supervisea largenumber ofmanagers,due to diversif cation.This reduces investment risk(Pan and Yu,2012). The researchers in the other group document that a large degree o f separation of control and ownership can help large shareholders to expropriate from m inority shareholders through tunneling activities,leading to agency or entrenchment ef ects(Johnson et al.,2000).Su and Zhu(2003)observe that large stockho lders can cause severe agency p rob lem s for borrowers,as large shareholdersmay expropriate from m inority shareholdersby increasing their separation of control and ownership.This increasesboth credit risk and the likelihood of remaining in f nancial distress(Lin et al.,2011).

However,directly exam ining the relationship between the agency p rob lem and the degree of separation o f con tro l and ownership w ithou t considering the con trol righ ts o f the large shareholderm ay lead to un reliable results.Fan and W ong(2002)and M orck et al.(2005)docum ent that the large shareholder can only engage in tunneling activities to gain p rivate benef ts and thereby reduce com pany value if he o r she has considerable contro l rights.M ajority contro lm ay even be required.Pan and Yu(2012)concur that substantial control rightsare required for a large shareho lder to engage in tunneling;also they show that the degree of entrenchment and alignment difersaccording to the levelo f contro l.Sh leifer and Vishny(1997)f nd thatwhen a large shareholder is capable of contro lling the borrower,he or she ismore likely to force a manager to engage in asset substitution and expropriate from the lender.However,the authors also show that this would result in a high credit risk and increase the likelihood o f remaining in fnancial distress(Lin et al.,2011).We thus predict that in thehigh-contro lgroup,the entrenchmentand agency efectsw illbestronger than the alignment efectswhen the degree of separation o f control and ownership is high,resu lting in higher credit risk.In contrast,we p red ict the alignm en t efectsw illbe stronger than the en trenchm ent and agency efects in the low-contro l group when the degree o f separation of contro l and ownership is high.W e thus expect a high degree o f separation of controland ownership to improve corporategovernance,as itmay encourage large shareho lders to monitormanagers and improve the ef ciency o f investment.

From the perspective o f lenders’self-protection and considering the tradeof between costs and benef ts,a lender ismore likely to regard external auditing and collateral as substitutes for a borrowerw ith lower credit risk when a borrower choosesa high-quality auditor.In thisway,the lender canm inim ize risk.Even when the borrower’s credit risk is extremely high,the lender is able to regard external auditing and collateral as complementary.We thus test the fo llow ing hypotheses empirically.

H 2.W hen the large shareholder isable to control the borrower and the degree o f separation of control and ownership is higher,the substitution ef ects between collateral and high-quality auditing areweaker.

H 3.W hen the large shareholder isunab le to contro l the borrower and the degree of separation of contro land ownership is higher,the substitu tion ef ects between collateral and high-quality auditing are stronger.

2.3.Collateral,nature of controller,market development and audit quality

There are two d istinct groups of Chinese f rm s:state-owned enterp rises(SOEs)and non-state-owned en terprises(NSOEs).Since 1978,China has sought to convert SOEs from f rm swhose so le proprietor is the state to modernWestern-stylecorporations(Cu lland Xu,2005).A lthoughmany decision rightshavebeen delegated to the corporatized SOEs,the government retains the ultimate rights to make decisions about the disposal o f assets and mergers and acquisitions undertaken by listed f rms,aswell as the right to appoint the CEOs.As SOEs lack the right to dispose o f state assets,they are subsidized by the governmentwhen they face f nancial distress.Nevertheless,theCEOso f SOEshavemu ltip le responsibilities(Chen etal.,2010).Thep romotion and compensation of CEOsof SOEs’are determ inedmore by their success in fulf lling various po litical and social objectives than by their f rms’operating and accounting performance(Fan et al.,2007).Liand Li(2004)f nd that SOEs facemoredivergentpolicy burdens than NSOEs,which lead to soft-budget constraints.The imp licit insurance provided by the governm ent and its con trolover state-owned banksare im portant sou rceso fexisting soft-budget constrain ts.Despite these constrain ts,SOEs have a lower credit risk.First,as the Chinese governm ent hasbeen engaged in the co rporatization or partialp rivatization of these enterp rises,it hasan incentive to ensure that they remain f nancially sound.Second,the competition for listing in China is extremely intense;SOEsm ustperform better than other f rm s to be chosen to list.A num ber of stud ies provide evidence that compared w ith NSOEs,SOEs have p referential access to capital;specif cally,they havemore opportunities to obtain capitaland their capital costsare lower(A llen etal.,2005;Ayyagarietal.,2010;Guariglia etal.,2011).

Unlike thedevelopedmarkets in Western countries,China’s regionaleconom ic developm ent isunbalanced; there are substantial diferences between regions in terms of legal development and marketization(Yu and Pan,2008;Fan and W ang,2009).Thismakes China’s institutional setting a natural forum in which to study the relationship between market developm ent and the dif erences in substitution efects between SOEs and NSOEs(W ang et al.,2008).Jiang and Li(2006)note thatwhen the government has little power over bank loans and fnancing activities arewell developed,the diferences between SOEsand NSOEs in term so f bank loans are reduced.A ccord ing to Fang(2007),decision m aking regard ing bank loans ism orem arketized and independentwhen the institu tionalenvironm ent is im proved,governm en t power over bank loans is decreased and soft-budget constrain ts are relaxed.Under such cond itions,lenders are likely to give SOEsm o re severe debt contracts,alleviating the f nancial discrim ination associated w ith the nature o f ownership.Yu and Pan (2008a)fnd that NSOEs’political relationshipsmay afect bank loansand that these ef ectsarestrongerwhen the levelsof fnancing and legal development are low.SOEs’bank loansare negatively related to the levelso f f nancing and legal developm ent,which contradicts the f ndings o f La Porta et al.(1998)in“Law and Finance.”Weiand Shen(2009)fnd thatwhen the degree o f government intervention is high,SOEs are less able to obtain credit loanswhen lendersarewell protected.W e expect an increase in market development to decrease the supportive ef ects of government intervention in SOEs’bank loans.

When SOEs choose high-quality auditors,lender decrease their co llateral requirements to reduce credit risk;the degree o f ad justm ent is also larger for SOEs than NSOEs.W ith the developm ent of them arket environm ent,the degree o f governm en t intervention in lenders’decisions decreases,which increases the credit risk faced by SOEs.In a region w ith more developed markets,the gap in co llateral requirements between SOEs and NSOEs is smaller than that in less developed regions.W e thus test the fo llow ing hypotheses empirically. H 4.The substitu tion ef ects between co llateral and high-quality aud iting are stronger in SOEs than NSOEs.

H 5.W hen m arket-developm en t index is higher,the enhancem ent ef ects of SOEs are weaker.

3.Research design

3.1.M easurement ofmain variables

3.1.1.M easurement of collateral

Collateral is the proportion o f co llateral loans,that is calcu lated as the ratio o f total loansco llateralized to total loansoutstanding at theend of the year.W edo notusea dummy variab le for collateral.The information on bank loansisobtained from f nancialstatements.In themain tests,weusethep roportion o fco llateralloansto m easurecollateral;in thesensitivity tests,weuse thep roportion of co llateralloansp lus thato fguarantor loans.4Bank loans can be divided into collateral loans,guarantor loans and credit loans,according to their level of restriction.The use of collateral involves two kinds of loans:p ledged loans and m ortgage loans.W e use the propo rtion of collateral loans to m easure the use of collateral.Some scho larsuse the p roportion of collateraland guarantor loans tomeasure collateral(Chen and Xiao,2012),and we have noticed that the information on bank loans disclosed in fnancial reports,sometimes are related to“collateral and guarantee loans.”Therefore,we use the proportion o f collateral loans tom easure collateral in ourm ain tests,and the proportion of collateraland guaranto r loans in our sensitivity tests.Our p redictions are generally well supported.

3.1.2.M easurement of audit quality

Researchershave used a range of variables to measure audit quality,such as the size of the o f ce(reputation)(DeAngelo,1981).C lients o f Big-4 audit f rm s have lower d iscretional accruals and higher earnings response coef cien ts than the clien ts o f non-Big-4 f rm s(Becker and Defond,1998;Francis,1999;Teoh and W ong,1993).Follow ing Hu and Tang(2007)and W u(2006)we use accounting-f rm size as a dummy variab le to measure audit quality.If the f rm is in the Big-4 or Big-10,the dummy variab le is 1;otherw ise, it is 0.However,som e researchers regard accounting-f rm size as a poor proxy for audit quality,and it ism islead ing to hom ogenize the audit quality of the Big-4,due to d iferences in the institu tional environm en t (Liu and Zhou,2007).We use discretional accruals to so lve these p rob lem s.First,we divide the fu ll sample into two groups according to audit quality and then compare the means o f the two groups to determ ine whether the diferences are signif cant.

3.2.M ain testmodels

In H 1aand H 1b,we p redict that collateral and audit quality are negatively and positively related,respectively.We use themodels developed by Rajan and Zingales(1995)and Chen etal.(2012)to test these hypotheses,as follows.5In accordancew ith the reviewers’suggestions,we use the debt-coverage ratio to p roxy for cash f ow,because lenders pay considerable attention to cash fow.We also use the percentage of intangible assets to control for the borrowers’ability to provide collateral.In addition,weadjust the framework ofourmain results in the fo llow ingways.First,we test thesignif canceo fH 1.Second,weuseameasure ofauditor change,Heckm an two-stagemodeland the PSM techniqueas robustness testsand to addressendogeneity issues.Finally,weuse M odel2 and M odel 3 to test H 2,H 3,H 4 and H 5.

M odel 1:

W e expect the sign o fα1to be signif cantly negative or positive.In M odel1,we use Soe asa con trol variable becausewe expect SOEs to be less likely than NSOEs to p rovide co llateral for their debt.W e thusexpect the sign of Soe to benegative.W ealso include leverageand thepercentage of long-term debt in M odel 1,because we expect these variab les to be positively related to co llateral.To add ress f rm performance,operational risk and solvency,we add the fo llow ing variables in M odel 1:return on assets,natural logarithm o f total assets, interest-coverage ratio,grow th o f assets and current ratio.We also use the number o f years since the f rm’s establishment to control for the bank-f rm relationship,because the resu ltso f p revious literature indicate that the bank-f rm relationship can solve the information-asymm etry p rob lem and decrease the use o f collateral. However,others believe that banks’information superiority enables them to ask for high interest,thereby exacerbating the agency prob lem.Therefo re the relationship between collateraland age isexpected to be either negative o r positive.W e p redict that the sign of them arket-developm ent index,a proxy for m arket developm ent,w illbe positive,because the higher the levelof com petition,them ore likely the use o f collateral.W e use the percentage o f in tangible assets,the volatility of net operating cash fow,the deb t coverage and the sales cash fow ratio to control for the ability to provide collateraland the operating cash f ow.Finally,we use year and industry dummy variab les to control for year and industry factors.

M odel 2:

We useM odel2 to exam ineH 2 and H 3.We predict that if the large shareholder hasa high levelo f contro l, α3w ill be signif cantly positive;if the large shareholder has a low level o f contro l,we expectα3to be significantly negative.

M odel 3:

W e use M odel 3 to exam ine H 4 and H 5,in which we pred ict thatα4w ill be signif cantly negative andα5signif cantly positive,respectively.

3.3.Sample selection and data

We use a sample of listed A-share Chinese f rms from 2005 to 2011.Our sample selection process isas follows.First,we om it the ST,*ST and PT f rms,as they are fail to p rovide the necessary information to exam ine the relationship between collateral and high audit quality.Next,we om it f rms from the f nance industry, because these do not f t our aims of exam ining bank loans.We then om it f rms w ith listings later than 2006 and f rm s w ith fewer than th ree f rm-year observations.In addition,we om it f rm s whose u ltim ate contro lling shareho lder cannot be identif ed.Finally,we elim inate observations that p rovide insuf cient data to calculate aud it quality or Z-scores,f rm sw ith zero bank loans,and f rm s for which we are unab le to iden tify m ajor loan sources or the number o f co llateral loans at the end o f the year.A fter elim inating these observations,wehave4877 observations from 828 f rms,ofwhich SOEscomp rise 3148 observationsand NSOEs provide1729.6The samp le used in the draft paper com prised 4919 observations.Following the reviewers’recomm endations,we have added several variables to ourmodel,and thushave lost42 observations(due tom issing data)to give4877 observations,ofwhich 3148 pertain to SOEs and 1729 to Non-SOEs.W ew insorize severalof the contro lvariablesat the 1%level.Our data on co llateral and guarantor loans,audit quality,ultimate controlling shareholders and other f nancial detailsare obtained from the Chinese Stock M arket and Accounting Research(CSMAR)database(see Table 1).

4.Empirical results

4.1.Descriptive statistics for main variables

Table 2 p rovides the descriptive statistics for themain variables.

Table2 reveals that themean o f proportion o f collateral in the fu llsamp le is0.420,which dif ers from Chen etal.’s(2012)resultsof 0.262 fora samp leo f f rms from 2001 to 2006.Themean of proportion o f co llateraland guarantor loans together is0.74,which isconsistentw ith Chen and Xiao’s(2012)fndings.This resu lt indicates that co llateraland/or guarantor loansarew idesp read in debt contracts.Firmsw ith a high auditquality comp rise30%of our samp le;that is,appropriately one third of the f rms sampled have chosen international Big-4 or domestic Big-10 auditors.SOEscomprise64.5%of thesamp le.Thereare largediferencesbetween themaximum and m inimum values for the other contro l variables,such as,interest coverage,market developm ent, num ber o f years since the f rm was estab lished and separation o f con tro land ownership.This result ind icates that these characteristics difer between f rm s,which are consistentw ith the literatu re.7These resultsarenotw inso rized,becauseweuse dumm y variables form arket development and separation of ownership and con tro l in them ain tests.The reviewers also note the problem s w ith the interest-coverage ratio and them easure o f separation of ownership and control.In the descriptive statistics,we state that a continuous variable is used to represent the degree of separation of ownership and control,w ith amean of 6.783.Themean degreeofseparation of ownership and controlis thus6.783%.Thedata are collected directly from the CSM AR database and our m ethod is as described by La Porta(1997).The interest-coverage ratio is calculated as=EBIT/in terest expenses.Our descriptive statistics for them ain variab les are generally consistent w ith those presen ted by Chen et al.(2012).

Table 3 lists the descrip tive statistics for f rms divided according to audit quality.

Table3 displays thedif erences inmeansbetween the groupsand the resu ltsof t-tests.Themean discretionary accruals dif er signif cantly between the groups,which suggest that our choice o f criteria to build the high audit quality variab le isappropriate for our samp le.The proportion o f collateral is signif cantly higher in the low auditquality group than in thehigh-quality group.This isprelim inary p roof of Hypothesis1,in which we p redict that co llateral and audit quality are regarded as alternatives.We also fnd thatw ith the exception o f the current ratio,proportion of guarantor loans,leverage and separation of contro land ownership,themean o f the variab les are signif cantly higher in the high audit quality group,than in the low-quality group.

Table 4 lists descrip tive statistics fo r the dumm y variab les for separation o f contro l and owner-ship,large shareholders’contro l righ ts and m arket developm ent.

Table 1 Def nitions of variables.

Table 2 Descriptive statistics.

Table 3 D escriptive statistics by audit quality.

We use various criteria to build our dummy variab les.W ith regard to the degree of separation of control and ownership,themean and 2/3 values do not dif er signif cantly between the groups.Neither themean nor the2/3 valuesaresignif cantly dif erent formarket development.When 25%ischosen asa criterion,high-contro l f rms represent 73.7%of the sample,whereas the30%criterion yields60.2%o f high-contro l f rms.Therefore,we use 30%as the criterion in themain tests,and 25%in the sensitivity tests.

4.2.Correlation coef cients ofmain variables

Table 5 lists the correlation coef cients of themain variab les.

The results disp layed in Tab le 5 show that the correlation between collateral and audit quality isnegative (-0.1638),which is consistent w ith our p rediction.SOEs are less likely than NSOEs to p ledge collateral (-0.2491),which makes them better able to obtain credit loans.M oreover,SOEs have longer debtmaturity, which is ref ected in the high p roportion of long-term debt(0.0593).Our results support the conclusion that long-term debt ismore dif cult to monitor than short-term debt.W e fnd the level o fmarket developm ent to be negatively related to co llateral(-0.0995),which indicates thatwhen lenders are better pro tected,borrowers are less likely to p ledge collateral.Deb tm aturity is signif cantly longer in the high audit quality group than in the low-quality group(0.0922),which p roves that external auditing m ay afect bo rrowers’debt m aturity. Despite the high correlation between collateral and guarantor loans(0.5628),the other data are reasonable; there is no evidence ofmu lticollinearity.

4.3.M ain results

A fter contro lling for year and industry factors,we use M odel 1,2 and 3 to exam ine H 1,H 2,H 3,H 4 and H 5.First,we useM odel 1 to test H 1;next,we use auditor sw itching,theHeckman two-stagemodel and propensity-scorematching(PSM)as robustness tests;fnally,we use M odel 2 and 3 to test H 2 to H 5.

4.3.1.H 1 and endogeneity issues

Tables6 and 7 list the resu lts of testing H 1 and the sensitivity tests.The f rst colum n disp lays the results fo r M odel 1.W e f nd the relationship between co llateral and aud it quality to be signif cantly negative at the 1% level(-0.0649,t-stat.=-8.18).Thus,H 1bis supported,because collateral and high-quality auditing areregarded assubstitutesby lenders.Apart from interest coverage and grow th of assets,the controlvariab lesare allsignif cantat the1%level.In addition,we f nd that thenumber o f yearssinceestablishmentand the current ratio are positively related to co llateral,which isnot the expected resu lt.SOEsare likely to pledge less collateral than NSOEs.H igh leverage,low return on assets,smallsize,a low levelofmarket developmentand a high proportion o f long-term debt all increase the likelihood of a f rm using co llateral.Interestingly,the degree o f separation o f controland ownership isnot signif canteven at the10%level,but thisoutcome isconsistentw ith our p rediction.Prior researchersnote that the incentives to separate contro land ownership arem ixed.Such a separation m ay im prove alignm ent and reduce risk by increasing the ef ciency of m onito ring(Pan and Yu, 2012),or cause agency and en trenchm ent efects th rough tunneling activities;the relationship between these two sets of ef ects canno t be sim p ly negative or positive.

Table 4 Descriptive statistics for the dummy variables for separation o f control and ownership,large shareholders’control rights and market development.

Table 5 Correlation coef cients ofmain variables.

The self-selection p rob lem is a type o f endogeneity problem regu larly encountered in studies o f auditors, especially when the sizeo f an audit f rm isused tomeasure auditquality,as in our study.Borrowersare likely to have certain preferenceswhen choosing auditors,which may result in biased samples,in turn afecting the accuracy and validity o f the resu lts.An auditor-sw itching sensitivity testof auditor change can partly solve the self-selection problem.Follow ing previous literature,wealso use the two-stageHeckman(1979)m odel to provide a fuller so lution to the prob lem.The second column of Tab le 6 provides the resu ltso f the Heckman twostagemethod.W hen the inverse M ills ratio is added to the equation,co llateral and audit quality are again found to be signif cantly negatively related.In the f rst stage,we use a probit regression to test not only the con tro l variab les in M odel1,bu t other variab les such as the ratio of receivables and inven tory to total assets, and a dumm y variab le fo r audit opinion in the previous year.However,we report only the resu lts o f the sec-ond stage,which support our p rediction that collateral is negatively related to audit quality(-0.0788, t=-9.16).

Table 6 Results fo r H 1.

As docum en ted in them ain tests,substitu tion efects are found to exist between collateral and high-quality auditing.W e use an auditor-sw itching test to further exam ine the relationship between co llateral and highquality auditing.The auditor’s size(reputation)isused to build the auditor dummy.W e predict that the proportion o f co llateralw ill increase if the borrower sw itches from an international Big-4 or domestic Big-10 to another auditor,and the p roportion of co llateralw ill decrease if the borrower sw itches from a non-international Big-4 and non-domestic Big-10 to an international big-4 or domestic big-10.We do not exam ine thediference between high-quality auditorsand low-quality auditors,because according to our criteria,this kind of sw itch does not af ect lenders’decisions.Our p redictions are fully conf rmed by the resu lts of the regressions.There are251 caseso f auditor-sw itching in our samp le,o fwhich 28 f rmssw itch from high-quality auditors to low quality auditors and 223 f rms sw itch from low quality auditors to high-quality auditors.8M odel 4 is designed to calculate the changes between the current year and the previous year.We do not use the data for 2004 as a benchmark;instead,wegenerate the lagged variablesdirectly from the sam p le from 2005 to 2011.Consequently,our samp le is reduced by 828 observations,from 4877 to 4049 observations.The results shown in the third colum n support our p redictions.The values o f Change1(-0.0219,t=-1.86) and Change2(0.102,t=3.14)indicate thatwhen the f rm moves from a low-quality auditor to a high-quality auditor,the useo f co llateraldecreases;when the high-quality auditor is rep laced by a low-quality auditor,the use of co llateral increases.

Table 7 Propensity-sco rematch ing.

The results of the p ropensity-sco rem atching are listed in Table 7.There are two m ain stages.During the f rst stage,the observations are graded accord ing to the contro l variables and their scores are then m atched. During the second stage,we calcu late the diferences between the control group and the treatment group, namely the low audit quality group and the high audit quality group.W e use fourm ethods to calculate these diferences.The resu lts of the second stage are listed in Tab le 7,and support our p rediction that high audit quality can reduce the use o f collateral.

4.3.2.Results for H 2 to H 5

The f rst three columns of Tab le 8 p rovide the resu lts o f testing hypotheses 2 and 3.W e f rst determ ine whether the large shareholder hasa high or low level of control rights,then we divide the sam ple accordingly in to two groups.W e pred ict that in the high-con trol group,the high level o f separation of con trol and ownership righ tsw illweaken the substitution efectsbetween co llateraland high-quality aud iting,as itw ill increase the oppo rtunity for large shareholders to expropriate from m ino rity shareho lders,resulting in a higher credit risk.In the low-control group,however,a high level o f separation of control and ownership may strengthen the substitution ef ectsbetween collateral and high-quality auditing by helping large shareholders to monitor managers and reducing investment risk through diversif cation.Thisw ill reduce credit risk and increase the ef ciency of investment.The results disp layed in the second co lumn of Table 8 show that in the high-control group,co llateraland high-quality auditing are always signif cantly negatively related at the 1%level(-0.103, t-stat=-6.97),and that the auditor*wedge interaction is signif cantly positive at the 5%level(0.0439,tstat=2.12).This con f rms H ypothesis 2,in which we p redict that the substitution ef ects areweaker in the high-contro lgroup when the degree of separation of contro land ownership ishigh.However,there isno evidence to suppo rt H ypo thesis 3,which relates to the low-con trol group,as Aud ito r*wedge is no t signif can tly negative(-0.0251,t-stat=-0.94).This resu lt m ay be attribu ted to the non-signif cant d if erences between entrenchm en t efects and alignm en t efects in the low-con trol group.M oreover,we fnd W edge to be signifcantly positive at the 5%level(0.0327,t-stat=2.06).

The fourth colum n d isp lays the resu ltsofM odel3,which wasdesigned to testH ypo theses4 and 5.W e expect thesubstitution ef ectsbetween collateraland high-quality auditing to bestrongerwhen theborrower isan SOE rather than an NSOE,because SOEshavea lower credit risk,for several reasons.However,wealso expect the enhancem entef ectso f SOE to beweakerwhen the levelofmarket development ishigh,because amoredevelopedmarket is likely to reduce thegovernment’sintervention in bank loans.Thisw illincreasethecredit risk faced by SOEs;thereby reducing thedif erence in credit risk between SOEsand NSOEs.The resultsof the regression fu lly con f rm our prediction.AsAuditor*Soeissignif cantlynegativeat the1%level(-0.0951,t-stat=-4.72), the substitution efectsare stronger in SOEs than in NSOEs,so Hypothesis4 issupported.In addition,Auditor*Soe*Geo is signif cantly positive at the5%level(0.0412,t-stat=2.24)which con f rmsour prediction in H ypo thesis5 that the enhancem en t efects of SOEsare reduced when the levelo fm arket developm en t ishigh.

5.Sensitivity tests

5.1.Using Z-scoremodel to evaluate credit risk

H ypotheses2,3,4 and 5 are developed based on the grounds that SOEsare likely to have a lower credit risk and thatborrowers face a higher credit risk when largestockholdershavesubstantial control rightsand a high degree of separation of control and ownership;and the latter predictions does not app ly to the low control rights group.In this section,we use A ltman’s(1968)Z-scoremodel to validate our reasoning by exam ining the dif erences in Z-scores between the two groups,as a lower Z-score indicates a higher credit risk.

The Z-scorem odel is as follows.

Z-score is used as the dependent variable,w ith the nature of the ultimate contro ller,the degree of separation o f con trol and ownership,and several con trol variables on the right-hand side o f the equation.The contro l variables com p rise f rm size,f rm leverage,cu rrent ratio,retu rn on assets,in terest coverage,the num ber o f years since the f rm wasestab lished,a dumm y variable for loss,and variables contro lling fo r year and industry factors.W e f nd that SOEshave higher Z-scores than NSOEs.The Z-score for thehigh control rightsgroup is not signif cantly lower when the separation of contro l and ownership is greater,but the Z-score for the low contro l rightsgroup is signif cantly higherwhen theseparation of controland ownership isgreater.The results are generally consistentw ith our predictions,but are not reported in this paper.

5.2.Alternativemeasures of collateral,large-shareholder control rights,degree of separation of control and ownership,and the level ofmarket development

In ourmain tests,we use the p roportion of collateral as the exp lained variab le and a 30%measure,a 2/3 m easure and themedian as criteria to build variab les for large-shareholder control rights,separation o f contro l and ownership and market development.In this section,however,the proportion of co llateral and guarantor loans is the exp lained variab le.W e use a 25%measure as the criterion to build a dummy variab le for large-shareholder control rights;then themean and median are used as criteria to build a dummy variable fo r separation o f con trol and ownership,and them ean and 2/3 m easu re as the criteria fo r am arket-developm ent dumm y variab le.W e use these alternativem easu res to regress M odels 1,2,and 3 and,reexam ine all the f ve hypotheses.Our predictions are generally supported but the resu lts are not reported in this paper.

5.3.M easurement of audit quality

In them ain tests,we use the international Big-4 and the dom estic Big-10m easures asa com bined p roxy fo r audit quality.Wealso usediscretionalaccruals,calcu lated using themodif ed Jonesmodel to verify thep roxy.A lthough the international Big-4 f rms occupy less than 5%of our sample,we use on ly international Big-4 f rms as a proxy for audit quality to test all o f the hypotheses.Our predictions are generally supported.

Table 8 Results for H 2 to H 5.

6.Conclusion

In this study,we attem p t to ascertain from the perspective o f lenderswhether externalaud iting contribu tes to bo rrowers’bank loans by reducing co llateral requirem en ts,as suggested in the literature.W e also aim todeterm inewhether the relationship between collateraland high-quality auditing isaf ected by the natu re of the u ltimate contro ller and the borrowers’ownership structure.Our results show that lenders regard collateral and high-quality auditing asalternativemeansof p reventing credit risk.In China’s institutionalcontext,single shareholders are dom inant,shareholders are high ly concentrated and lenders are inefectively p rotected.In this setting,the substitution efects between co llateral and high-quality auditing are greater in SOEs than NSOEs,as SOEs have lower credit risk.We also notice thatwhen themarket development level is higher, the supportive efects of government’s intervention in SOEs’bank loans are reduced,the credit risk faced by SOEs is stronger and the enhancement ef ects o f substitution in SOEs are weaker.In addition,when the large shareho lder has considerable control rights,the substitution efects between co llateral and high-quality aud iting areweakened if the borrower hasa high levelo f separation of con tro land ownership.W hen the large shareholder can con tro l the bo rrower and there is a high level of separation of contro l and ownership the entrenchm en t and agency ef ects are stronger than the alignm en t efects,so the bo rrower faces a higher credit risk.In contrast,a greater separation o f control and ownership does not result in stronger alignment ef ects than entrenchment and agency ef ects in the group w ith fewer control rights.Our evidence shows that high-quality auditing and co llateral are regarded asalternativeswhen lendersmake self-p rotection decisions, but that degree o f their substitutability isadjusted by lendersaccording to the borrower’s level of credit risk. To further exam ine our p redictions,we use alternativemeasures,an auditor-sw itching test and A ltman’s Z-scoremodel as sensitivity tests.We also use Heckman’s two-stage regression model to add ress self-selection p rob lemsassociated w ith auditor choice,and p ropensity scorematching to dealw ith endogeneity.Our predictionsare generally well supported by the resu ltso f themain tests and the sensitivity tests.Finally,the lim itations o f our study shou ld be acknow ledged.First,although the in ternational Big-4 and dom estic Big-10 m easures of audit quality generally support our p redictions,and we conduct sensitivity tests to exam ine these m easures further,it possible that our resu lts contain measurement errors.Second,we contro l for debtmaturity,then exam ine the relationship between collateraland externalauditing,whichmay lead to the problem o f endogeneity caused by interest rates.However,p revious literatureshows that interest rates in China are tightly regulated by thegovernment,preventing lenders from using interest rates to distinguish between thehigh-and low-risk borrowers.Due to this lim ited functionality(Chen et al.,2012;Podpiera,2006;Koivu,2009),we do not use interest rates as a contro l variable in our models,nor do we conduct further tests to exam ine this p rob lem.

Aghion,P.,Bolton,P.,1992.An incomp lete contracts approach to f nancial con tracting.Rev.Econ.Stud.59,473–494.

Allen,D.,Qian,J.,Qian,M.J.,2005.Law,fnance and economic growth in China.J.Financ.Econ.77,57–116.

Armstrong,C.S.,Guay,W.R.,Weber,J.P.,2010.The role of information and fnancial reporting in corporate governance and debt contracting.J.A ccount.Econ.50(2–3),179–234.

Ayyagari,M.,D em irgu¨c-Kun t,Asli,M aksim ovic,V.,2010.Formal versus inform al fnance:evidence from China.Rev.Financ.Stud.23 (8),3048–3097.

Becker,C.L.,Defond,E.A.M.L.,1998.The efect of audit quality on earningsmanagement.Contem p.Account.Res.15,1–24.

Besanko,D.,Thakor,A.V.,1987.Competitiveequ ilibria in the creditmarketunder asymmetric information.J.Econ.Theory 42,167–182.

Bester,H.,1985.Screening vs.rationing in creditm arketsw ith imperfect in formation.Am.Econ.Rev.75,850–855.

Black,E.L.,Carnes,T.A.,M osebach,M.,M oyer,S.E.,2004.Regulatorymonitoring asa substitute for debt covenants.J.Account.Econ. 37(3),367–391.

Boot,A.W.,Thakor,A.V.,U dell,G.F.,1991.Secured lending and default risk:equilibrium analysis,policy imp lications and emp irical resu lts.Econ.J.101,458–472.

Brick,I.E.,Palia,D.,2007.Evidence of jointress in the term s of relationship lending.J.Financ.Intermed.16,452–476.

Chakraborty,A.,Hu,C.X.,2006.Lending relationship in line-of-creditand nonline-of-credit loans:evidence from collateraluse in small business.J.Financ.Interm ed.15,86–107.

Chan,Y.S.,K anatas,G.,1985.Asymm etric valuation and the role of collateral in loan agreem ents.J.M oney Cred it Bank.17,85–95.

Chen,H.W.,Wang,Y.Y.,Chen,J.Z.,Lobo,G.J.,2010.Association between borrower and lender state ownership and accounting conservatism.J.Account.Res.48,973–1014.

Chen,J.Z.,Lobo,G.J.,W ang,Y.Y.,Yu,L.S.,2012.Loan Collateraland Accounting Conservatism.W o rking Paper.

Chen,J.,2011.Auditor reputation and bank loans allocation—evidence from bank credit loans decision.Aud it Res.5,90–97.

Claessens,S.,D jankov,S.,Lang,L.H.P.,2000.Theseparation of ownership and control in East Asian corporations.J.Financ.Econ.58, 81–112.

Cronqvist,H.,N ilsson,M.,2003.Agency costso f controllingminority shareholders.J.Financ.Quant.Anal.38,695–719.

Cull,R.,Xu,L.C.,2005.Institutions,ownership and fnance:the determ inants of pro f t reinvestment am ong Chinese Firms.J.Financ. Econ.77,117–146.

DeAngelo,L.,1981.Auditor size and auditor quality.J.Account.Econ.3,183–199.

Fan,G.,Wang,X.,2009.The Report on the Relative Processo fM arketization of Each Region in China.The Econom ic Science Press.

Fan,J.P.H.,W ong,T.J.,Zhang,T.,2007.Politically connected CEOs,corporate governance and post-ipo perform ance of china’s new ly partially p rivatized f rms.J.Financ.Econ.84,330–357.

Fan,J.P.,Wong,T.J.,2002.Corporate ownership structure and the informativeness of accounting earnings in East Asia.J.Account. Econ.33,401–425.

Fang,J.X.,2007.Ownersh ip,institutionalenvironm ent and bank loans allocation.Econ.Res.12,82–92.

Francis,J.R.,M aydew,E.L.,Sparks,H.C.,1999.The ro le of big 6 audito rs in the cred ible reporting o f accruals.Audit.J.Pract.Theory 18,17–34.

Franks,J.,M ayer,C.,1998.Bank control takeovers and corporate governance in Germany.J.Bank.Finance 22,1385–1403.

Guariglia,A.,Liu,X.,Song,L.,2011.Internal fnance and grow th:m icroeconom etric evidenceon Ch inese f rms.J.D ev.Econ.96(1),79–94.

Hu,Y.M.,Tang,L.L.,2007.Audit,information transparency and bank loans interest rates.Audit Res.6(1),2–8.

Jensen,M.,M eckling,W.,1976.Theory of the f rm:managerialbehavior,agency costs,and ownership structure.J.Financ.Econ.3(4),5–50.

Jim enez,G.,Salas,V.,Saurina,J.,2006.Determ inants of co llateral.J.Financ.Econ.81,255–281.

Johnson,S.,La Porta,R.,Lopez-de-Silanes,F.,Shleifer,A.,2000.Tunnelling.NBERWorking Paper Series 7523.

Kanga,Jun.-Koo.,Shivdasani,A.,1995.Firm performance,corporate governance,and top executive turnover in Japan.J.Financ.Econ. 38(1),29–58.

Kahn,C.,W inton,A.,1998.Ownersh ip structure,speculation and shareholder in terven tion.J.Finance 53(1),99–129.

Khurana,I.K.,Raman,K.K.,2004.A re Big-4Audits in ASEAN countrieso f higher quality than non-Big-4Audits?Asia-Pac.J.Account. Econ.11(2),139–165.

Koivu,T.,2009.Has the Chinese econom y becomemore sensitive to interest rates?China Econ.Rev.20,455–470.

La Porta,R.,Lopez-de-Silanes,F.,Shleifer,A.,1999.Corporate ownership around the wo rld.J.Finance 54(2),471–517.

La Porta,R.,Lopez-de-Slianes,F.,Shleifer,A.,Vishny,R.W.,1998.Law and f nance.J.Polit.Econ.106,1113–1155.

Liao,Y.G.,Zhang,L.,Xie,2010.Institutionalenvironment,independent auditing and bank loans-evidence from Chinese listed fnancial distress f rm s.Audit Res.2.

Lin,Y.F.,Li,2004.Politicalburdens,m oral hazard and soft budget constrains.Econ.Res.1(2).

Liu,F.,Zhou,F.,2007.Dose the big-4mean high audit quality?Analysis from accounting conservatism.Account.Res.03,79–87.

Lu,Z.F.,Sun,B.X.,Zhu,J.G.,2008.Earningmanagement,accounting information and bank loanscontracts.M anage.World 3(1),1–2.

Luo,D.L.,Tang,Q.Q.,2008.Pyramid structure,ownership and m inority shareholder p rotection-evidence from chinese listed f rm s. Finance Econ.Res.34(9),132–143.

M anove,M.,Pad illa,A.J.,Pagano,M.,2001.Co llateral versus project screening:am odel of lazy banks.J.Econ.32,726–744.

M anove,M.,Padilla,A.J.,1999.Banking(conservatively)w ith optim ists.J.Econ.30,324–350.

M orck,R.,Wolfenson,D.,Yeung,B.,2005.Corporategovernance,econom ic entrenchmentand grow th.J.Econ.Literature43,655–720.

M yers,S.C.,1977.Determ inants of corporate borrow ing.J.Financ.Econ.5,147–175.

N ikolaev,V.,2010.D ebt covenan ts and accounting conservatism.J.Account.Res.48,137–175.

Pan,H.B.,Yu,M.G.,2012.The Separation of controland ownership o f Large Shareholder:Agency Ef ectsor Risk D iversif cation Efects. W orking Paper.

Pittm an,J.A.,Fo rtin,S.,2004.Auditor choice and the cost of debt capital for new ly pub lic f rms.J.A ccount.Econ.37(1),113–136.

Podpiera,R.,2006.Progress in China’s Banking Sector Reform:H as Bank Behavior Changed?W orking Paper,International M onetary Fund.

Qi,J.N.,2004.Auditor size,brand efects,price and audit quality-audit fees and quality of international big-four.Audit Res.3.

Rajan,G.R.,1992.Insiders and ou tsiders:the choice between inform ed and arm’s-length deb t.J.Finance 47,1367–1399.

Rajan,R.G.,Zingales,L.,1995.W hat do we know about capital structure?Som e evidence from in ternational data.J.Finance 50,1421–1460.

Rao,Y.C.,Hu,Y.M.,2005.Investigation and analysis of accounting information using in bank loans.Account.Res.4,36–41.

Sharpe,S.,1990.Asymm etric inform ation,bank lending and imp licits:a stylized m odel of customer relationships.J.Finance 45,1069–1087.

Skinner,J.D.,2011.D iscussion ofaccounting standardsand debtcovenants:has the“balancesheetapproach”led to a decline in the useof balance sheet covenants?J.Account.Econ.52,203–208.

Sm ith,C.,W arner,J.,1979.On fnancial con tracting:an analysis of bond covenants.J.Financ.Econ.7,117–161.

Su,Q.L.,Zhu,W.,2003.Fam ily controlled listed f rms and f rm value.Econ.Res.2.

Teoh,S.H.,Wong,T.J.,1993.Perceived auditor quality and the earnings response coef cient.Account.Rev.68,346–366.

Wang,Q.,W ong,T.J.,Xia,L.J.,2008.State ownership,the institutional environment,and auditor choice:evidence from China.J. Account.Econ.46,112–134.

W ei,F.,Shen,2009.Ownership,lender protection and bank credit loans.Financ.Res.9,26–39.

Wu,S.P.,Li,Q.F.,2006.Audit quality o f international big-four,domestic big-ten and nondomestic big-ten.Contem p.Finance Econ.2.

Xiao,Z.P.,2007.Large shareholder,lender p rotection and the debtmaturity choice––evidence from chinese listed f rm s.M anage.W orld 10,99–113.

Xiao,Z.Z.,Chen,D.Q.,2012.Fam ily Enterp rises’Preference for Contro l R ights and Bank Loans Con tracts.W orking Paper.

Yu,M.G.,Pan,H.B.,2008a.Political relationsh ip,institutionalenvironment and bank loans of p rivateenterprises.M anage.W orld 8,9–21.

Yu,M.G.,Pan,H.B.,2008b.Government intervention,legal and fnancing development and bank loans of state-owned enterp rises. Financ.Res.9,1–22.

Zhang,J.,2008.The contracting benef ts of accounting conservatism to lenders and borrowers.J.A ccoun t.Econ.45,27–54.

E-mail address:kobeyang827@163.com

http://dx.doi.org/10.1016/j.cjar.2014.08.003

1755-3091/?2014 Production and hosting by Elsevier B.V.on behalf of China Journalof Accounting Research.Founded by Sun Yat-sen U niversity and City University of H ong K ong.

Loan collateral

Audit quality

Ownership structure