Journal of Banking & Finance 37 (2013) 4989–5006

Contents lists available at ScienceDirect

Journal of Banking & Finance

journal homepage: www.elsevier .com/locate / jbf

Loan collateral and financial reporting conservatism: Chinese evidence

0378-4266/$ - see front matter � 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.jbankfin.2013.09.003

⇑ Corresponding author. Tel.: +1 713 743 4838.E-mail addresses: zeyun.chen@colorado.edu (J.Z. Chen), gjlobo@uh.edu

(G.J. Lobo), yanyanwang@xmu.edu.cn (Y. Wang), yulisheng@xmu.edu.cn (L. Yu).

Jeff Zeyun Chen a, Gerald J. Lobo b,⇑, Yanyan Wang c, Lisheng Yu c

a University of Colorado at Boulder, United Statesb University of Houston, United Statesc Xiamen University, China

a r t i c l e i n f o

Article history:Received 5 January 2013Accepted 1 September 2013Available online 8 September 2013

JEL classification:G21G32

Keywords:Loan collateralFinancial reporting conservatismDebt contractingDefault riskAsset tangibility

a b s t r a c t

We examine the relation between the use of collateral and financial reporting conservatism for a sampleof Chinese firms. In the absence of flexibility in risk pricing through interest rates and strong contractenforcement in China, we find that lenders reduce collateral requirements from more conservative bor-rowers and that this negative relation is significantly moderated by borrowers’ poor credit quality andlow asset tangibility. Our finding that conservatism can result in a tangible benefit in the form of lowercollateral requirements indicates that lenders value financial reporting conservatism. However, the ben-efit from financial reporting conservatism is muted as lenders become more concerned about borrowers’default risk or ability to pledge tangible assets as collateral against loans.

� 2013 Elsevier B.V. All rights reserved.

1. Introduction et al. (2011b) find that collateral requirements motivated by ex

Collateral requirements arise from standard agency problems infinancial relationships. One set of theoretical studies explainscollateral as a signaling device to address ex ante information gapsbetween borrowers and lenders. These models predict that collat-eral induces borrowers to reveal their default risk. Specifically,lenders provide a menu of contract terms such that high (low)quality borrowers choose secured (unsecured) loans at lower(higher) premiums (Bester, 1985; Chan and Kanatas, 1985;Besanko and Thakor, 1987a,b; Boot et al., 1991).

Another school of thought proposes that frictions other than exante information asymmetry motivate the use of collateral as partof an optimal debt contract. These frictions include ex post moralhazard problems (Myers, 1977; Smith and Warner, 1979; Aghionand Bolton, 1992), difficulties in enforcing contracts (Albuquerqueand Hopenhayn, 2004; Cooley et al., 2004), and costly state verifi-cation (Townsend, 1979; Gale and Hellwig, 1985; Williamson,1986; Boyd and Smith, 1993). The right to repossess collateralserves as a credible threat to ensure that borrowers behave inthe best interest of lenders. Such a disciplinary role of collateralis central to the theory of incomplete financial contracts. Berger

post moral hazard problems are empirically dominant, relative tocollateral requirements used to address ex ante informationasymmetry.

Although there is a sizable body of research on collateral use indebt contracts, the vast majority of that research focuses ondeveloped markets. As a result, much remains to be known aboutcollateral choice in emerging markets. Theory suggests that bankrequests for collateral to deal with agency problems become higherin emerging markets where the information environment isgenerally more opaque (Hainz, 2003; Menkhoff et al., 2006,2012). However, banks could also be concerned about the suitabil-ity of collateral as an efficient contracting tool due to weaker laws,institutions, and enforcement in emerging markets (Qian andStrahan, 2007; Bae and Goyal, 2009). A natural question that arisesis whether there are other mechanisms that lenders can use to im-prove efficiency in setting collateral requirements. In this study, wefocus on borrowers’ financial reporting conservatism (hereafter re-ferred to as conservatism) and study how it relates to the use ofcollateral in addressing the debt agency problem in China, thelargest emerging market.

Basu (1997) interprets conservatism as ‘‘the accountants’ ten-dency to require a higher degree of verification to recognize goodnews as gains than to recognize bad news as losses’’. Researchersadvance four explanations for conservatism: contracting, litigation,regulation, and taxation (Watts, 2003). Under the contracting

4990 J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006

explanation, conservative reporting is a means of addressing moralhazard.1 In the context of risky debt financing, asset substitution andunderinvestment problems drive lenders’ demand for timely infor-mation about the value of the firm’s net assets in the event of liqui-dation. The consequences of moral hazard are of serious concern tolenders when firm value falls and shareholders’ incentive to delaythe recognition of bad news for fear of losing control rights to thefirm’s assets becomes stronger. Many accounting-based debt cove-nants (such as the debt-to-cash flow ratio, interest coverage ratio,and debt-to-equity ratio) restrain shareholders from opportunisti-cally expropriating wealth from lenders when a firm approacheseconomic distress. Covenants that constrain such expropriation onlybecome binding if the financial reporting system recognizes thedeterioration of a firm’s financial position. In that regard, conserva-tism (i.e., timely loss recognition) can improve the efficiency of thesecovenants because they are more likely to be binding in distress andtherefore are more likely to limit wealth expropriation by sharehold-ers. It is the use of accounting-based covenants that essentially leadsto increased demand for conservatism. Clearly, lenders benefit fromborrowers’ conservatism and they can induce it by imposing certaincontracting costs on borrowers.2

Most studies examining the (debt) contracting explanationfocus on conditional conservatism (e.g., Zhang, 2008; Beattyet al., 2008; Nikolaev, 2010; Chen et al., 2010), often referred toas news-dependent or ex post conservatism, which involves firmswriting down the book value of net assets in a timely manner uponreceiving bad news but not writing up net assets as quickly uponreceiving good news. In comparison, unconditional conservatism,often referred to as news-independent or ex ante conservatism,involves the predetermined understatement of the book value ofnet assets. Under unconditional conservatism, firms commit atinception to recognizing book values of net assets that are belowthe expected market values during their lives (Ryan, 2006). Balland Shivakumar (2005) point out that conditional conservatismenhances contracting efficiency, but unconditional conservatismseems inefficient or at best neutral in contracting. Following theprior literature, we examine conditional conservatism and itsrelation to the use of collateral.

Because the use of collateral is costly to lenders and they canbenefit from borrowers’ conservatism, we expect that lenders willreduce collateral requirements from more conservative borrowers.More conservative borrowers are more likely to violate debtcovenants and to violate them sooner, so conservatism benefitslenders at the expense of borrowers (Zhang, 2008). It is reasonableto argue that lenders are willing to share the benefits withconservative borrowers in the form of reduced collateralrequirements.

1 The litigation explanation points out that litigation produces asymmetric payoffsin that overstating the firm’s net assets is more likely to generate litigation costs forthe firm than understating net assets. By understating net assets, conservatismreduces the firm’s expected litigation cost. The taxation explanation posits thatasymmetric recognition of gains and losses enables profitable firms to reduce thepresent value of taxes and thereby increase the value of the firm. Under the regulationhypothesis, financial reporting standard setters and regulators have incentives tofavor conservative reporting because of the asymmetry in regulators’ costs. They arelikely to face more criticism if firms overstate net assets than if they understate netassets. Conservatism reduces the political costs imposed on them. We refer interestedreaders to Watts (2003) for a more detailed discussion on the four explanations forconservatism.

2 Our discussion above focuses on the role of conservatism in enhancing theefficiency of (accounting-based) debt covenants after the loan is issued. When lendersassess a potential loan, they are concerned about the likelihood the borrower willhave enough net assets to cover the loan. Since future values of the firm and net assetsare generally not verifiable at the time when lenders evaluate a loan application, theyobtain verifiable lower bound measures of the current value of net assets (generatedby a conservative financial reporting system) and use those as inputs in the loangranting decision (Watts, 2003).

We also conjecture that the negative relation between the useof collateral and conservatism is moderated as lenders becomemore concerned about default risk and the potential recovery indefault. From lenders’ perspective, the marginal benefit of collat-eral increases in loan default risk. As their exposure to this risk in-creases, lenders become reluctant to reduce collateralrequirements and share the benefits from conservatism. Similarly,when borrowers have only limited ability to pledge tangible assetsas collateral against loans, lenders find it more critical to maintaincollateral requirements as opposed to relaxing them in exchangefor borrowers’ conservatism.

We test our hypotheses using a sample of Chinese firms. Chinais the largest emerging market and Chinese firms overwhelminglyrely on banks to finance their capital needs.3 China also provides aunique opportunity for examining collateral requirements becauseof the government’s tight control over interest rates during oursample period, which severely limits lenders’ use of loan pricing todifferentiate across borrowers with different risks (Podpiera, 2006;Koivu, 2009). In developed markets, lenders not only use collateralbut also price risk to maximize profits. However, the endogenousdecisions of setting interest rates and collateral requirements arelikely to contaminate any observed relation between collateral useand conservatism. This is less of a concern in China.4

We hand-collect information on the sources of bank loans andthe use of collateral disclosed in firms’ annual financial reportsfor 5358 firm-year observations between 2001 and 2006. Wemeasure collateral use at the firm-year level as the ratio of totalcollateral loans to total loans outstanding at the end of the yearfor which the sources of the loans and information on collateralare disclosed in the financial statements. Collateral loans accountfor 26.2% of outstanding loans in our sample. Following Khan andWatts (2009), we construct C_Score to measure conservatism andvalidate this measure in the Chinese setting.5

We first document that there is a negative relation between theuse of collateral and conservatism after controlling for financialperformance, risk and other loan features. The lower collateralrequirements established for more conservative borrowers suggestthat lenders generally value conservatism and are willing to sharethe benefits from conservatism with borrowers. Next, we classifysample firms into high and low observed credit quality groupsbased on whether they had any loans in default in the last year(Jiménez et al., 2006). We find that the negative relation betweenthe use of collateral and conservatism is significantly lesspronounced for firms with low observed credit quality. This isconsistent with our expectation that when lenders offer debtfinancing to riskier borrowers, they perceive the marginal benefitof collateral to be higher and therefore are less likely to relaxcollateral requirements in exchange for conservatism.

Lastly, we classify sample firms into two equal-size groups byindustry and year, according to the level of their asset tangibility(measured as the proportion of fixed assets to total assets at the

3 During our sample period (2001–2006), Chinese listed firms obtained RMB (i.e.,Renminbi, the Chinese currency) 12.5 trillion in new loans from banks, about 20 timesthe amount raised from the stock market (China Securities and Futures StatisticalYearbook, 2007).

4 In a developed loan market such as the US, where interest rates are uncon-strained, research shows that conservatism is negatively related to interest ratesbecause lenders reward borrowers for conservative reporting by lowering interestrates (Zhang, 2008). Addressing our research questions in the US setting is likely tosuffer from an endogeneity bias because we cannot rule out any indirect relationshipbetween collateral use and conservatism through their common effect on interestrates.

5 We note that although it is preferable to examine the relation between collateralrequirements and conservatism at the individual loan level, comprehensive samplesof private loan agreements with specific information on loan characteristics are notavailable in China. Therefore, we examine a firm-year measure of collateral loans anduse proxies for various loan characteristics at the firm-year level.

6 We find that 52% of our sample firms’ loans are guaranteed by a third party.Although interesting in its own right, we do not investigate the relation between loanguarantees and conservatism in this study. Prior studies show that loan guaranteesmay be used as a tunneling tool to expropriate wealth from minority shareholders inChina (see, for example, Berkman et al., 2009). Therefore, we are concerned that the

J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006 4991

end of the last year), and assess the impact of asset tangibility onthe relation between the use of collateral and conservatism. Con-sistent with our hypothesis, we find that low asset tangibility sig-nificantly moderates the negative relation between collateral useand conservatism. This result suggests that when borrowers be-come sufficiently constrained to provide tangible assets to securethe loans, lenders are less willing to relax collateral requirementsand allow borrowers to share in the benefits from conservatism.Our main results are robust to a battery of sensitivity tests.

The results of our study have several important implications.We extend the conservatism literature by directly testing the rela-tion between the use of collateral and conservatism. Zhang (2008)finds that lenders reward conservatism and reduce interest ratesfor conservative borrowers. Our study shows that reduced collat-eral requirements can be another type of reward for conservatism.In addition, we find that such a reward may be diminished whenborrowers’ observed credit quality and asset tangibility are low.

Our study also makes an important contribution to the collat-eral literature. Given weaker enforceability of contracts and lend-ers’ legal rights in reorganization and liquidation procedures inemerging markets, research on how lenders use collateral effi-ciently to manage credit risk is important. Menkhoff et al. (2012)find that Thailand lenders rely on loan guarantees as an importantsubstitute for collateral. We show that, in addition to loan guaran-tees, lenders care about financial reporting conservatism (which isnot an explicit contract term) and adjust collateral requirements toshare the benefits from conservatism with borrowers.

In addition to informing the academic debate on theeconomic demands for conservatism and the role of collateralin efficient debt contracting in emerging markets, the results ofour study may also shed light on Chinese bank lending behaviorand its consequences. Cull and Xu (2000, 2003) find that bankfinancing in China is associated with higher productivity,profitability, and adoption of reforms compared to governmenttransfers. Jia (2009) concludes that lending by Chinese banks ismore prudent as a result of the reforms triggered by China’s en-trance to the WTO. Ayyagari et al. (2010) document that Chinesefirms with bank financing grew faster than similar firms with nobank financing. We contribute to this line of research by offeringa glimpse into how Chinese banks play their monitoring rolesand the tools they can potentially draw onto enhance theirmonitoring efficiency.

The rest of the study is organized as follows. We describe thesalient features of China’s banking industry in Section 2. Followingthe literature review in Section 3, we develop the hypotheses inSection 4, present the research design in Section 5, and describethe sample selection procedures and data in Section 6. We discussthe results of our main tests and additional analyses in Section 7and Section 8, respectively, and present our conclusions inSection 9.

measure of loan guarantees is contaminated because of such opportunistic behavior.In addition, investigation of loan guarantees requires information about the guaran-tors who may be individuals, independent third parties, related parties, or state orlocal governments. The wide range of guarantors and lack of information on theirquality prevent us from performing a thorough test of the relation between loanguarantees and conservatism. Due to these concerns, we do not consider loanguarantees in our hypotheses testing. Nevertheless, in untabulated analyses, we findthat our main conclusions are not affected if the percentage of guaranteed loans(calculated as the ratio of total loans that are guaranteed to total loans outstanding atthe end of the year) is incorporated as a control variable.

7 According to the World Bank–People’s Bank of China (2006) Report on ‘‘SecuredTransactions Reform and Credit Market Development in China’’, only 4% of loans usemovable assets as collateral.

8 Yang and Qian (2008) identify the following five potential risks of over-relianceon collateral in loan contracting: (1) banks tend to underweight borrowers’fundamental performance; (2) collateralized asset markets are not efficient; (3) someborrowers have unclear property rights over collateralized assets; (4) it takes a longtime to liquidate collateralized assets, resulting in smaller net realizable value; and(5) collateral value decreases due to impairment of collateralized assets.

2. Institutional background

Banks grant three primary types of loans in China: credit,guaranteed and collateral loans. Until 1990, capital flows were alldirected and controlled by the government (through the banks).There was little room or need for risk management by banks; loanswere granted based on ‘‘credit’’. With sufficiently low interestrates, these ‘‘credit’’ loans were essentially free gifts from thegovernment. During the early stage of China’s transition toward amarket economy in the early 1990s, banks started to exercise riskawareness and require guarantees or collateral to manage loanrisk. However, ‘‘credit’’ loans (i.e., policy lending) still overwhelm-ingly predominated the loan market, resulting in sizable nonper-forming loans. Not until the mid-1990s were loan guarantees and

collateral widely used by banks as important risk managementtools.6 According to a survey by Yang and Qian (2008) of 13 majordomestic banks, collateral loans increased from 22% to 32% of allloans granted between 2000 and 2005. In China, the only type ofcollateral acceptable to most banks is land or buildings; moveableassets are rarely used as collateral.7 In addition, under ChineseSecurities Law, inventory and receivables are not allowed to be usedas collateral. Although collateral has become a common contractterm in China and helped banks substantially reduce their riskexposures, loan risk cannot be fully eliminated by relying only oncollateral (Yang and Qian, 2008).8

In theory, banks can price credit risk through interest rates.However, the Chinese government maintains tight control overlending interest rates, which prevents interest rates from effi-ciently reflecting the risks of borrowers. Since 1996, when loansto all firms had to range between 0.9 times to 1.1 times the offi-cial benchmark rate, the government has been gradually deregu-lating lending interest rates. As of October 2004, there is no upperlimit on interest rates, only a floor of 0.9 times the benchmarkrate. However, despite the increased flexibility provided by theliberalization of interest rates in recent years, loan pricing(through interest rates) remains largely undifferentiated (Podpi-era, 2006). Koivu (2009) argues that despite the expanded free-dom of setting interest rates above the benchmark rate, thehigh level of liquidity in the credit market in recent years has re-stricted Chinese banks’ ability to raise interest rates. She observesthat the deviations from benchmark are still modest and ques-tions the capability of Chinese banks to efficiently price theircredit through interest rates.

3. Review of related research

Collateral is a borrower’s pledge of specific assets as security fora loan. A collateralized lender’s claim to pledged assets reduces itslosses if the borrower becomes bankrupt. Collateral requirementsare widely used in loan contracts, together with interest rate,maturity, size and other covenants, to address adverse selectionand/or moral hazard problems between borrowers and lenders(Bester, 1985; Chan and Kanatas, 1985; Besanko and Thakor,1987a,b; Boot et al., 1991). Most studies find that observed bor-rower credit risk is a key driver for the use of collateral, and theincidence and degree of collateral are higher for young and smallfirms, consistent with the risk reduction and monitoring role ofcollateral (Berger et al., 2011a,b).

9 Note that lenders can also reduce the interest rate charged to more conservativeborrowers in exchange for receiving more timely signals of default risk and controlrights from borrowers (see, for example, Zhang, 2008). We focus on collateralrequirements rather than interest rates because of the Chinese government’s tightcontrol over loan pricing through interest rates.

4992 J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006

The majority of studies on collateral focus on developed mar-kets such as the U.S. and Europe. However, the use of collateralin emerging markets could be different from that in developedmarkets. One school of thought is that collateral should play a lar-ger role in emerging markets because of higher information asym-metry, lower liquidation payoff, or lower competition in thebanking sector (Hainz, 2003; La Porta et al., 2003; Menkhoffet al., 2006, 2012). Allen et al. (2005), based on survey evidencefrom start-up, private firms in China, find that collateral value ex-ceeds 80% of loan value on average. They conclude that it is con-sistent with Chinese banks understanding the risk of start-upfirms and ‘‘pricing’’ the risk in loan contracts. However, manyresearchers believe that collateral requirements should be lessprevalent in emerging markets given weaker laws, institutions,and enforcement of debt contracts (Qian and Strahan, 2007; Baeand Goyal, 2009). Lin (2011) finds that Chinese firms with highervalue of potential collateral have no incremental increase in long-term bank loans after foreign bank loans become available, indi-cating that collateral may play only a limited role in mitigatingthe problem of information asymmetry in China, where creditorrights are not well protected.

Lacking strong creditor rights protection in emerging mar-kets, lenders may choose to use various other tools to monitorand manage credit risk. Other loan terms such as interest rate,loan duration, size, and restrictive covenants are natural candi-dates, but are all subject to the same concern about contractenforceability. Prior research shows that relationship lendingand loan guarantees are used as substitutes for collateral inemerging markets. Menkhoff et al. (2012) find that third partyguarantee is an important substitute for collateral in Thailand.The role of lender-borrower relationship in dealing withinformation asymmetry is discussed in detail by Boot (2000).Relationship lending relies on soft or private information aboutborrower risk obtained through a close lender-borrower rela-tionship. However, Ayyagari et al. (2010) find that informalfinancing through relationship provides little support for thegrowth of the private sector in China and is unlikely to substi-tute for formal financing.

In contrast to relationship lending, collateral-based lendingrelies on hard or public information (Brick and Palia, 2007).We focus on conservatism, an important attribute of borrowers’financial information quality, and study how the use of collateralis related to the degree of borrowers’ conservatism.

The accounting literature explains conservatism as an impor-tant feature of financial reporting to address agency problemscaused by lenders having asymmetric information and asymmet-ric payoffs (Watts, 2003; Ball et al., 2008b). The empirical liter-ature largely supports theoretical predictions on the role ofconservatism in monitoring default risk and mitigating debtagency cost. For example, Zhang (2008) finds that conservatismbenefits lenders ex post through the timely signaling of defaultrisk, and benefits borrowers ex ante through lower initial interestrates. Focusing on syndicated loan deals, Ball et al. (2008a) findthat when borrowers’ accounting information is more conserva-tive (in terms of capturing credit quality deterioration in atimely fashion), information asymmetry between the lead arran-ger and other syndicate participants is lower, resulting in leadarrangers holding a smaller proportion of new loan deals.

Evidence on lenders’ demand for conservatism to protectinvestment is documented in emerging markets as well. Chenet al. (2010) find that Chinese lenders demand more conserva-tism from non-state-owned firms than from state-ownedfirms, because the default risk is substantially lower for firmswith government ownership and support. Using data fromIndia, Gormley et al. (2012) show that foreign bank entry duringthe 1990s is associated with higher demand for borrowers’

conservatism, which in turn is positively related to subsequentdebt levels.

4. Hypotheses

4.1. Relation between financial reporting conservatism and use ofcollateral

Given their information disadvantage, lenders use collateral tolimit potential losses in instances of default. The use of collateralis, however, not free of cost for the lender as the lender incurs costsof screening and monitoring the pledged assets and, in the case ofrepossession, pays for any enforcement and disposal expenses.Selling specialized assets may be difficult and may result in a lossrelative to the true value of the assets because of a possible lemonsproblem. To the extent that lenders can benefit from borrowers’conservatism, they are likely to reduce the use of collateral.

Two important benefits of conservative reporting are valued bylenders. First, conservatism accelerates (accounting-based) cove-nant violations, resulting in more timely transfer of control rightsfrom borrowers to lenders (Watts, 2003; Ball et al., 2008b). Second,to the extent that the book value of net assets is understated, con-servatism also provides lenders with a measure of the lower boundof the collateral’s value. This is important because understated col-lateral likely is associated with a higher recovery rate in default(Watts, 2003; Zhang, 2008). Lenders can share these benefits withconservative borrowers in the form of reduced collateral require-ments which, in turn, induces borrowers’ conservatism.9 The moreconservative the borrower, the greater are the benefits to the lender,and the greater is the reduction in collateral requirements. This leadsto our first hypothesis:

H1. Financial reporting conservatism is negatively related to theuse of collateral in debt contracts.

Note that conservatism and collateral serve related but stillsomewhat different purposes in loan contracting. Conservatismtriggers a timely violation of financial covenants and transfersdecision rights to the lender quickly following economic losses.By contrast, loan collateral reduces the lender’s losses when theborrower actually defaults. As a result, under certain circum-stances, the lender may find it more beneficial and critical to relyon collateral in monitoring default risk and mitigating losses fromloan defaults despite borrowers’ conservative reporting. We con-jecture that the hypothesized tangible benefit from conservatismin the form of reduced collateral requirements is less pronouncedas the lender becomes sufficiently concerned about default riskand the potential recovery in default. We use the borrower’s ob-served credit quality and its ability to pledge assets in instancesof default to reflect such concerns.

4.2. Collateral, conservatism, and observed borrower credit quality

As the lender’s exposure to the risk of loan default increases, themarginal benefit of collateral becomes higher. It could be appealingto the lender to draw on multiple precautions to safeguard its loaninvestment. In other words, as the lender becomes sufficiently con-cerned about default risk, it may be less willing to relax collateral

J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006 4993

requirements and allow the more risky borrower to share in thebenefits from conservatism.

Boot et al. (1991) show that when the lender can observe theborrower’s credit quality, a low quality borrower obtains loanswith collateral and a high quality borrower obtains loans withouthaving to pledge collateral. Several empirical studies find evidencein support of this theory and document a positive associationbetween the use of collateral and past observed repayment prob-lems (Chakraborty and Hu, 2006; Jiménez et al., 2006; Brick andPalia, 2007; Berger et al., 2011b). Following this line of research,we also assume that a borrower with loans in default in the pastis observably riskier; consequently, the lender finds the use ofcollateral more critical in managing risk. Thus, we expect thenegative relation between the use of collateral and conservatismamong borrowers with past loan default experience to be lesspronounced. The above discussion leads to our second hypothesis:

H2. The negative relation between financial reporting conserva-tism and the use of collateral in debt contracts is less pronouncedwhen the borrower has lower observed credit quality.

10 We provide a detailed discussion of the C_Score measure in Appendix A.

4.3. Collateral, conservatism, and access to tangible assets

Many factors other than credit quality or risk, such as theborrower’s access to tangible assets which can be pledged, alsoaffect the collateral decision (Berger and Udell, 1990). Theories ofcapital structure suggest that if a large fraction of the firm’s assetsare tangible, then assets should serve as collateral to reduce therisk of the lender suffering agency cost of debt (Harris and Raviv,1991; Rajan and Zingales, 1995). Tangible assets are likely to havea greater value than intangible assets in case of bankruptcy. Inaddition, intangible assets generate more asymmetric informationthan tangible assets and the agency cost of debt increases in theborrower’s informational opacity. Indeed, previous studies gener-ally find that leverage increases in the proportion of tangible assetson the balance sheet, because the lender is more willing to supplyloans (Titman and Wessels, 1988; Rajan and Zingales, 1995).

Given that tangible assets have higher value in liquidation thanintangible assets, the use of collateral in debt contracts for aborrower with more tangible assets is presumably less costly fromthe lender’s perspective. With greater availability of less costlycollateral, the lender may be more willing to reduce collateralrequirements given its access to more conservative financial infor-mation. However, when the borrower’s ability to provide tangibleassets as collateral against loans is limited, the lender may find itmore costly to relax collateral requirements and share the benefitsfrom conservatism with the borrower. This will potentially moder-ate the negative relation between the use of collateral and conser-vatism. Based upon the above reasoning, we formally state our lasthypothesis as follows:

H3. The negative relation between financial reporting conserva-tism and the use of collateral in debt contracts is less pronouncedwhen the borrower has lower asset tangibility.

11 All Chinese firms use a December 31 fiscal year-end. Listed firms are required torelease their audited annual reports no later than April 30 of the next calendar year.We calculate buy-and-hold annual returns that end four months after the fiscal year-end to ensure that the market response to the previous year’s earnings is excluded(Basu 1997). However, one could argue that the extent to which financial reportsincorporate bad news in a timely manner depends on auditors having the newsinformation (captured by stock returns) available to them at the time of the audit, inwhich case fiscal year returns may be more appropriate. When we recalculate thereturns variable (R) using buy-and-hold returns over the fiscal year, our main resultsremain qualitatively the same (untabulated).

12 Khan and Watts (2009) caution that C_Score ‘‘. . .may not be an appropriateconservatism measure in studies using data from countries where the institutionalfeatures differ from US institutional features in important ways.’’

5. Research design

5.1. Measure of conservatism

Our empirical analyses call for a metric that can capture bothcross-sectional and time-series variation in conservatism. Khanand Watts (2009) develop a firm-year conservatism measure,C_Score, and provide evidence that this measure reflects the empir-ical properties of conservatism documented in the prior literature.

Following Khan and Watts (2009), we calculate and use C_Score inour analyses. Specifically, we first estimate the following annualcross-sectional model:10

Ei;t=Pi;t�1 ¼ ðk0 þ k1SIZEi;t þ k2LEVi;t þ k3MBi;tÞþ DRi;tðj0 þ j1SIZEi;t þ j2LEVi;t þ j3MBi;tÞþ Ri;tðl0 þ l1SIZEi;t þ l2LEVi;t þ l3MBi;tÞþ DRi;t � Ri;tðm0 þ m1SIZEi;t þ m2LEVi;t þ m3MBi;tÞ þ ei;t ð1Þ

where E is earnings per share; P is year-end stock price per share; Ris 12-month, buy-and-hold, annual return from May of year t toApril of year t + 111; DR is a dummy variable that equals 1 if R is neg-ative and 0 otherwise; SIZE is the natural logarithm of market valueof equity; LEV is leverage ratio, defined as the sum of long-term andshort-term debt divided by market value of equity; MB is market-to-book ratio. Annual cross-sectional estimation of model (1) yields m0

to m3 that are constant across firms but vary over time. We then cal-culate C_Score for each firm-year as: C_Scorei,t = m0 + m1SIZEi,t

+ m2LEVi,t + m3MBi,t.There are a few noteworthy observations about this measure.

First, a potential concern with the C_Score measure is that it maysimply capture systematic risk as opposed to conservatism. Khanand Watts (2009) rule out this alternative explanation as theyfind a negative correlation between CAPM beta and C_Score(i.e., the sign of the correlation is the opposite of that predicted).Second, C_Score is motivated from the theory of conservatism indeveloped markets.12 To address the concern about the constructvalidity of C_Score in our setting, we assess its effectiveness as ametric of conservatism by examining whether its empirical prop-erties are consistent with other conservatism measures identifiedin the prior literature. We discuss these results in Section 7. Third,Khan and Watts (2009) recognize that many studies using C_Scoreas an independent variable in a regression may also require con-trolling for SIZE, LEV, and/or MB. Failing to control for these firmcharacteristics may result in finding an association between con-servatism and the variable of interest where there is no associa-tion. They recommend that one option is to directly control forSIZE, LEV, and/or MB, so C_Score loads after controlling for thesevariables. As collateral requirements are likely to vary with firmsize, leverage, and/or market-to-book ratio, we follow their sug-gestion and include these control variables in our regressionanalyses.

5.2. Model relating collateral and conservatism

We measure the use of collateral as the percentage of loansthat are collateralized, rather than as a dichotomous variablesimply indicating whether or not the firm has collateral loans.Doing so allows us to exploit any information contained in theamount of the collateral loans. We use the following model toestimate the relation between the use of collateral andconservatism:

4994 J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006

COLLATERALi;t ¼ a0 þ a1C Scorei;t þ a2SOEi;t þ a3LSBi;t

þ a4LEVi;t þ a5SIZEi;t þ a6ROAi;t

þ a7CFOVOLi;t þ a8INTCOVi;t þ a9GROWTHi;t

þ a10MBi;t þ a11LLOSSi;t þ a12CURRENTi;t

þ a13AGEi;t þ a14GEOi;t þ a15LTDEBTi;t

þ aj

X

j

YEARi;j þ ak

X

k

INDi;k þ ei;t ð2Þ

where COLLATERAL is the fraction of collateral loans, calculated asthe ratio of total loans that are collateralized to total loansoutstanding at the end of the year for which the sources of the loansand information on collateral are disclosed in the financialstatements;13,14 SOE is a dummy variable that equals 1 if the firmis an SOE and 0 if it is an NSOE. We classify borrowers as SOEs andNSOEs based on the ownership type of their ultimate controllingshareholders.15 LSB is measured as the ratio of total loans from SBsto total loans outstanding at the end of the year for which thesources of the loans and information on collateral are disclosed inthe financial statements; LEV is leverage ratio, defined as the sumof long-term and short-term debt divided by market value of equity;SIZE is the natural logarithm of market value of equity; ROA is returnon assets; CFOVOL is volatility of operating cash flows over the pastthree years; INTCOV is interest coverage ratio, calculated as incomebefore interest and tax expense divided by interest expense;GROWTH is growth in total assets, calculated as ending total assetsdivided by beginning total assets; MB is market-to-book ratio; LLOSSis a dummy variable that equals 1 if the firm reports a loss in theprevious year and 0 otherwise; CURRENT is current ratio, calculatedas current assets divided by current liabilities; AGE is the naturallogarithm of the number of years since the firm was established;GEO is the natural logarithm of the index of financial market compet-itiveness for each province or provincial level region; LTDEBT is thepercentage of long-term debt, calculated as the ratio of long-termdebt to total assets at the end of the year; 16 YEAR and IND are yearand industry dummies, respectively.

In Eq. (2), the coefficient of interest is a1. A negative value of a1

implies that lenders relax collateral requirements in exchange forborrower conservatism and is consistent with H1. We controlfor borrower ownership type (SOE) because SOEs, with the

13 Chinese Accounting Standards require that if the firm has significant outstandingloans or the collateral items are of significant value, it should disclose relevantinformation in the annual financial statements. We find that, in total, our samplefirms disclosed details of 92% of their loans outstanding. The disclosed loanspresumably have a significant impact on firms’ financial status. Since 92% of the loansoutstanding are disclosed by our sample firms, we are less concerned about apotential self-selection issue regarding firms’ disclosure choices.

14 A potential concern about the COLLATERAL measure is that it fails to incorporatethe information on collateral value. We thank an anonymous referee for pointing thisout. It is possible that two firms have the same percentage of collateralized loans (i.e.,the same amount of COLLATERAL), but their collateral value may differ. Unfortunately,because information on the exact collateral value is not available in our setting, weare unable to find the exact factor that lenders use to discount collateral value.

15 SOEs are defined as those borrowers directly or indirectly owned or controlled bystate asset management bureaus or other state-owned enterprises controlled by thecentral government or local governments. NSOEs are defined as those borrowerscontrolled by private investors. We exclude township-village enterprises from ouranalyses. A township-village enterprise refers to a business unit that belongs to allresidents of a rural community where it is also usually located. According to Che andQian (1998), it is neither an SOE nor an NSOE. They argue that it is best characterizedas a community enterprise with a governance structure in which the communitygovernment has control.

16 There may be an endogeneity problem as COLLATERAL, LEV, and LTDEBT all containelements of total loans in the numerator and/or denominator. If so, they are likely tobe jointly and endogenously determined, which may bias our results. We check therobustness of our results by estimating equation (2) and comparing the results whenLEV and LTDEBT are included in and excluded from the regression. Untabulated resultsindicate that our conclusions are qualitatively the same when we exclude thesepotentially endogenous variables.

government’s implicit insurance, are less likely to default ontheir loans (Chen et al., 2010). Therefore, lenders are less concernedabout the safety of the loans granted to SOEs and require lesspledged collateral. Accordingly, we expect a2 to be negative. Wecontrol for the percentage of loans borrowed from state banks(LSB) because SBs, with their less shareholder-wealth-orientedobjective functions, are less prudent in making loan granting deci-sions than their non-state counterparts (Jia, 2009). As a result, SBswill require less pledged collateral from their clients. We expect a3

to be negative. We also include various performance and risk mea-sures commonly used in the literature as control variables. The gen-eral expectation is that the use of collateral increases as theborrower becomes riskier and its financial performance deterio-rates. Note that we investigate the relation between collateral useand conservatism at the firm-year level. Accordingly, we controlfor loan size and maturity in Eq. (2) using LEV and LTDEBT.

We include an index for each province or provincial level region(GEO) to capture regional financial market disparity in China.17 Thelarger the index, the more advanced the region’s market-orientedinstitutional transformation and financial market development.Presumably, there is greater competition among creditors in regionswhere the financial market is more developed. According to Besankoand Thakor (1987b), the use of collateral is more likely in a competi-tive rather than in a monopoly credit market. We therefore expectthe coefficient on GEO to be positive. We include year (YEAR) andindustry (IND) dummies to control for year and industry effects onthe use of collateral in loan contracts.

H2 investigates whether the borrower’s low observed creditquality moderates the negative relation between use of collateraland conservatism. To test this hypothesis, we classify each samplefirm into one of two groups based on whether the firm had anyloan defaults in the past year (i.e., t � 1) and augment Eq. (2) topermit the relation between use of collateral and conservatism tovary across the two groups. Specifically, we create a dummy vari-able, Defaultt�1, which equals 1 for firms that had loan defaults inthe past year, and 0 otherwise.18 We include Defaultt�1 and its inter-action with C_Scoret in Eq. (2). A positive coefficient on the interac-tion term is consistent with the prediction of H2 that as theborrower’s observed credit quality decreases, the lender is less will-ing to relax collateral requirement and share the benefits from con-servatism with the borrower.

H3 examines the effect of the borrower’s asset tangibility on therelation between the use of collateral and conservatism. We mea-sure asset tangibility as the fraction of tangible assets on the balancesheet, i.e., fixed assets divided by total assets at the end of year t � 1.Based on their asset tangibility, we classify sample firms into twoequal-size groups within each industry and year, and create a dum-my variable LowTGt�1 that equals 1 for firms with asset tangibilitybelow the sample median, and 0 otherwise. To test H3, we augmentEq. (2) by including LowTGt�1 and its interaction with C_Scoret. If theborrower’s low asset tangibility suppresses the lender’s willingnessto reduce collateral requirements in exchange for more conserva-tism, we should find a positive coefficient on the interaction term.19

17 The index captures the degree of competitiveness of the financial market at theprovincial level. A higher index indicates better financial market development andhigher competition among lenders. See Fan and Wang (2007) for a more detaileddiscussion on the construction of the index.

18 As a sensitivity check, we expand the time horizon for measuring Default from thepast one year (i.e., t � 1) to the past two years (i.e., t � 1 to t � 2) and the past threeyears (i.e., t � 1 to t � 3) to assess the firm’s past repayment problems. We continue tofind results that support H2 when we measure Default over the past two years.However, the results no longer support H2 when we expand the time horizon to thepast three years (untabulated).

19 As sensitivity checks, we control for asset tangibility to test H2, and control forpast loan default experience to test H3. Our results (untabulated) remain qualitativelythe same.

22 In contrast, NSOEs are more represented in the Information and Technology, RealEstate, and Conglomerates industries.

23 The Pearson (Spearman) correlations between C_Score and its three componentsare 0.51(0.75) for LEV, �0.29 (�0.46) for SIZE, and 0.08 (�0.40) for MB.

24 Specifically, t statistics are defined as the mean of the coefficients from the sixannual regressions (2001–2006) divided by its standard error.

25 To construct C_Score, it is necessary to estimate Eq. (1) by year so that the

J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006 4995

6. Data and descriptive analysis

Our sample period ranges from 2001, when China entered theWTO, to 2006. We focus on the post-WTO period, when the lendingindustry became substantially more market-oriented and began tovalue borrower conservatism to a larger degree.20 Our initial sampleconsists of 7595 firm-year observations that are included in the ChinaSecurities Markets and Accounting Research Database (CSMAR). Thescreening of firm ownership type results in an elimination of 359observations related to township–village enterprises and firms whoseultimate controlling shareholder cannot be identified. We further de-lete 6 observations from the financial industry, 373 observations withno outstanding loans, and 657 observations without sufficient data tocalculate C_Score and various control variables. Of the remaining 6200observations, we eliminate an additional 842 observations becausewe are unable to identify the sources of their major loans and theamount of collateral loans from their financial statement disclosures.The final sample comprises 5358 firm-year observations.

Panel A of Table 1 details the distribution of sample firms acrossindustries. The industry composition of our sample is similar to thatof the population of firms in CSMAR, with Manufacturing comprising59.96% of sample observations. To divide manufacturing firms intomore homogenous groups, we classify them into ten subcategoriesaccording to the China Securities Regulatory Commission’s SICcodes. Panel B summarizes descriptive statistics of the main vari-ables for the full sample. On average, 26.2% of our sample firms’ loansare collateralized.21 C_Score has a mean of 0.176 and median of 0.117,indicating that its distribution is not very skewed. The first quartile(Q1) is positive (0.060), indicating that conservatism is a common fea-ture of financial reporting for more than 75% of the sample firms. SOEscomprise 77.3% of the sample. Not surprisingly, the majority of loansare issued by SBs (93.1%), indicating that SBs dominate the loan mar-ket when measured by the total value of loans. However, further anal-ysis reveals that 12.5% of the sample firms obtain at least some loansfrom NSBs. This suggests that the NSB loan market measured by thenumber of clients is still economically significant. Various perfor-mance and risk measures indicate that on average our sample firmsare financially healthy.

To shed more light on the differences between firms with highand low observed credit quality, we report descriptive statisticsfor each group in Panel C. We identify 630 observations (12% ofthe sample) that have at least one loan in default in year t � 1. Thesefirms pledge collateral significantly more frequently than the firmswithout any loan defaults (36.1% versus 24.9%), consistent withbanks viewing firms with past observed repayment problems asmore risky loan clients. We find that firms with past loan defaultshave larger C_Score than firms without past loan defaults (0.223 ver-sus 0.169), consistent with firms with past loan defaults facingstronger demand for conservatism as well. The results reported inPanel C clearly indicate that firms without past loan defaults havebetter performance and lower risk than firms that experienced pastdefaults. They are more likely to be SOEs and obtain a larger percent-age of their loans from SBs. They are also less leveraged, larger, moreprofitable, have lower cash flow volatility, fewer losses, and higherinterest coverage ratio and current ratio. These findings lend supportto our classification of a firm as high or low borrower observed creditquality based on whether it had any loans in default in the past year.

20 The consequent reforms of the banking industry following China’s entry into theWTO have led to substantial bank ownership changes with important implications forbank efficiency and lending behavior. For example, Jia (2009) shows that SBs havebecome more prudent, measured by bank excess reserve, loan-to-asset ratio anddeposit-to-loan ratio, as a result of the reforms.

21 Recall that the denominator of this ratio is total loans outstanding at the end ofthe year for which the sources of the loans and information on collateral are disclosedin the financial statements. The ratio is 25% if we calculate the ratio using total loansoutstanding reported on the balance sheet as the denominator.

Panel D shows the comparison between firms with high and lowasset tangibility. Not surprisingly, firms with high asset tangibilitycan pledge more collateral relative to firms with low asset tangibility(27.7% versus 24.8%). We find a larger percentage of SOEs amongfirms with high asset tangibility, presumably because SOEs typicallyoperate in the more asset-intensive industries such as Mining,Utilities and Transportation.22 Banks’ willingness to lend more whenborrowers can pledge more collateral translates to a higher leverageratio for firms with more tangible assets (Titman and Wessels, 1988;Rajan and Zingales, 1995). In addition, given the ability to supply morecollateral, firms with high asset tangibility obtain more long-termdebt than firms with low asset tangibility (9.3% versus 6.4%). Interest-ingly, we observe a higher C_Score for firms with a larger fraction of to-tal assets that are tangible. Recall that by construction, C_Scoreincreases in leverage. Therefore, a higher leverage ratio for firms withhigh asset tangibility can manifest in a higher C_Score. We control forleverage in our multiple regression analyses.

We report Pearson pairwise correlations among the variables ofinterest for the full sample in Table 2. As expected, SOEs and firmsborrowing more from SBs pledge fewer assets as collateral. Lendersrequire more collateral for riskier borrowers that are highly lever-aged and suffer operating losses. The use of collateral is less for largerborrowers and borrowers with stronger financial performance.Somewhat surprisingly, we observe that COLLATERAL is negativelyrelated to GROWTH and positively related to AGE, suggesting thatyounger/higher growth firms are unable to provide as much collat-eral as older/more mature firms. On a univariate level, COLLATERALand C_Score are positively correlated. Note that C_Score, by construc-tion, is a function of LEV, SIZE, and MB.23 The positive univariate cor-relation between COLLATERAL and C_Score is largely due to the factthat they both increase in LEV and decrease in SIZE. We suggest thatcaution be exercised when interpreting the univariate correlation be-tween COLLATERAL and C_Score.

7. Main results

Panel A of Table 3 reports the results of estimating Eq. (1) for eachyear of the 2001–2006 period. We show the mean coefficients overthe six years and t statistics for the means (Fama and Macbeth,1973).24 We calculate the C_Score for a firm-year as given in Eq. (1), usingthe coefficient estimates for each year. The significantly positive coeffi-cient on D� Ret indicates that our sample firms are on average conser-vative (i.e., earnings reflect bad news more quickly than good news).As expected, the coefficient on D� Ret� SIZE is significantly negative,indicating that larger firms have a richer information environment andface less demand for conservatism. The coefficient on D� Ret� LEV issignificantly positive, consistent with more levered firms facing higherdemand for conservatism. Similar to Khan and Watts (2009), we fail todetect a significant coefficient on D � Ret �MB.25

measure of conservatism can vary over time. However, the Fama–Macbeth standarderrors could be biased in the presence of a firm fixed effect (Petersen 2009). To assesswhether the Fama–Macbeth procedure is robust to a firm fixed effect in our sample,we re-estimate Eq. (1) using a pooled regression and calculate t statistics based ontwo-way, cluster-robust standard errors to control for both cross-sectional and time-series dependence (Petersen, 2009). We continue to detect a significantly positivecoefficient on D � Ret (t = 1.84), a significantly negative coefficient on D � Ret � SIZE(t = -2.02), and a significantly positive coefficient on D � Ret � LEV (t = 5.41). Theinsignificant coefficient on D � Ret �MB under the Fama-Macbeth procedurebecomes significantly positive (t = 2.87), which supports Khan and Watts (2009)’sprediction.

Table 1Sample industry distribution and firm characteristics.

Panel A: Sample composition by industry

Industry group No. of obs. % of obs.

Manufacturing–Food and beverages 265 4.95–Textile, clothing and fur 175 3.27–Wooden and furniture 21 0.39–Papermaking and printing 104 1.94–Petroleum, chemicals and plastic products 632 11.80–Electronics 164 3.06–Metal & non-mental 536 10.00–Machinery, equipment and instruments 916 17.10–Medical and bio-product 376 7.02

–Others 23 0.43Subtotal 3212 59.96Agriculture, forestry and fishing 112 2.09Mining 82 1.53Utilities 235 4.39Construction 72 1.34Transportation 218 4.07Information and technology 305 5.69Wholesale trade 431 8.04Real estate 289 5.39Services 120 2.24Entertainment 28 0.52

Conglomerates 254 4.74Total 5358 100.00

Panel B: Descriptive statistics of the full sample (N = 5358)

Variable Mean Std. Dev. Q1 Median Q3

COLLATERAL 0.262 0.313 0 0.131 0.444C_Score 0.176 0.212 0.060 0.117 0.226SOE 0.773 0.419 1 1 1LSB 0.931 0.198 0.990 1 1LEV 0.757 0.932 0.259 0.541 0.990SIZE 21.405 0.824 20.852 21.367 21.875ROA 0.021 0.064 0.004 0.025 0.052CFOVOL 0.054 0.039 0.026 0.044 0.070INTCOV 11.244 32.329 1.604 3.177 7.462GROWTH 1.094 0.205 0.982 1.070 1.181MB 3.575 15.065 1.707 2.533 4.062LLOSS 0.113 0.317 0 0 0CURRENT 1.370 0.817 0.871 1.195 1.619AGE 2.483 0.397 2.197 2.565 2.639GEO 1.892 0.418 1.723 1.923 2.216LTDEBT 0.078 0.096 0.000 0.043 0.118

Panel C: Descriptive statistics by borrowers’ observed credit quality

Variable Low Credit Quality High Credit Quality High–Low

Mean Median Std. Mean Median Std. Mean Median

COLLATERAL 0.361 0.260 0.353 0.249 0.118 0.305 �7.60*** �7.92***

C_Score 0.223 0.154 0.259 0.169 0.113 0.204 �4.99*** �7.34***

LSB 0.804 0.994 0.307 0.948 1 0.172 11.57*** 18.73***

SOE 0.708 1 0.455 0.782 1 0.413 3.87*** 4.17***

LEV 1.123 0.701 1.617 0.709 0.521 0.785 �6.34*** �7.44***

SIZE 21.160 21.149 0.881 21.437 21.387 0.811 7.48*** 7.44***

ROA �0.022 0.005 0.083 0.026 0.028 0.058 14.17*** 15.72***

CFOVOL 0.058 0.045 0.044 0.053 0.044 0.039 �2.46** �1.68*

INTCOV 3.766 1.631 20.69 12.241 3.479 33.452 8.85*** 15.64***

GROWTH �2.019 �0.499 7.301 �0.363 �0.267 4.140 8.22*** 12.20***

MB 5.087 3.184 32.547 3.373 2.483 10.765 �1.31 �6.77***

LLOSS 0.300 0 0.459 0.088 0 0.284 �11.30*** �15.74***

CURRENT 1.049 0.959 0.614 1.413 1.222 0.831 13.35*** 12.47***

AGE 2.584 2.639 0.296 2.470 2.565 0.406 �8.73*** �9.94***

GEO 1.815 1.901 0.450 1.902 1.934 0.412 4.64*** 4.87***

LTDEBT 0.090 0.057 0.102 0.077 0.042 0.095 �1.51 �0.02Observations 630 4728

4996 J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006

Table 1 (continued)

Panel D: Descriptive statistics by borrowers’ asset tangibility

Variable Low Asset Tangibility High Asset Tangibility High–Low

Mean Median Std. Mean Median Std. Mean Median

COLLATERAL 0.248 0.121 0.301 0.277 0.140 0.323 3.38*** 12.16***

C_Score 0.163 0.111 0.176 0.188 0.128 0.242 4.22*** 3.93***

LSB 0.938 1 0.191 0.924 1 0.204 �2.66*** �3.66***

SOE 0.760 1 0.427 0.786 1 0.410 2.32** 2.32**

LEV 0.734 0.524 0.907 0.781 0.557 0.956 1.85* 2.67***

SIZE 21.378 21.346 0.754 21.432 21.384 0.888 2.43** 1.15ROA 0.017 0.022 0.063 0.024 0.028 0.064 4.27*** 5.04***

CFOVOL 0.058 0.048 0.043 0.049 0.041 0.035 �8.93*** �7.69***

INTCOV 12.273 3.309 34.924 10.215 3.071 29.478 �2.33** �1.32GROWTH 1.096 1.073 0.214 1.092 1.069 0.196 �0.78 �0.72MB 3.715 2.683 14.658 3.434 2.404 15.462 �0.68 �5.51***

LLOSS 0.105 0 0.306 0.122 0 0.327 2.03** 2.03**

CURRENT 1.536 1.323 0.866 1.204 1.058 0.728 �15.19*** �18.15***

AGE 2.466 2.565 0.339 2.500 2.565 0.447 3.18*** 2.46**

GEO 1.904 1.934 0.422 1.880 1.921 0.413 �2.06** �2.90***

LTDEBT 0.064 0.029 0.085 0.093 0.060 0.103 11.24*** 11.28***

Observations 2679 2679

Panel A reports the industry distribution of the full sample. Industry groups are based on the China Securities Regulatory Commission’s classification. Panel B presentsdescriptive statistics of firm characteristics for the full sample. Panel C presents subsample descriptive statistics for firms with high and low observed credit quality. Firmswith low (high) observed credit quality are those with (without) loan defaults in year t � 1. Panel D presents subsample descriptive statistics for firms with high and low assettangibility. Asset tangibility is measured as the proportion of tangible assets on the balance sheet, i.e., fixed assets divided by total assets at the end of year t � 1. Sample firmsare classified as high (low) if their asset tangibility is above (below) the median for each industry and year. t tests are used to test differences between means. Wilcoxon two-sample tests are used to test differences between medians.COLLATERAL is the percentage of collateral loans, calculated as the ratio of collateralized loans to total loans outstanding at the end of the year for which the sources of theloans and information on collateral are disclosed in the financial statements; C_Score is the Khan and Watts (2009) conservatism score measure; SOE is a dummy variable thatequals 1 if the firm is an SOE and 0 if it is an NSOE; LSB is measured as the ratio of total loans from SBs to total loans outstanding at the end of the year for which the sources ofthe loans and information on collateral are disclosed in the financial statements; LEV is leverage ratio, defined as the sum of long-term and short-term debt divided by totalequity; SIZE is the natural logarithm of market value of equity; ROA is return on assets; CFOVOL is the volatility of operating cash flows over the past three years; INTCOV isinterest coverage ratio, calculated as income before interest and tax expense divided by interest expense; GROWTH is growth in assets, measured as ending total assetsdivided by beginning total assets; MB is market-to-book ratio; LLOSS is a dummy variable that equals 1 if the firm reports a loss in the previous year and 0 otherwise; CURRENTis current ratio, calculated as current assets divided by current liabilities; AGE is the natural logarithm of the number of years since the firm was established; GEO is thenatural logarithm of the index of financial market competitiveness for each province or provincial level region. LTDEBT is the percentage of long-term debt, calculated as theratio of long-term debt to total assets at the end of the year.* Significance at 0.10 level in a two-tailed test.** Significance at 0.05 level in a two-tailed test.*** Significance at 0.01 level in a two-tailed test.

J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006 4997

We next assess the validity of C_Score as a measure of conserva-tism in the Chinese setting. To do so, we examine whether C_Scorecan effectively distinguish firms with different levels of conserva-tism in a manner consistent with other commonly used conserva-tism measures. Specifically, we examine whether the Basu (1997)returns-based and Ball and Shivakumar (2005) accruals-based mea-sures of conservatism estimated for each of three equal-sized groups(low, medium and high) ranked on the basis of C_Score increase withC_Score. These results are presented in Panel B and Panel C for Basu’sand Ball and Shivakumar’s methods, respectively. Note that for theremaining analyses in our paper, we estimate the regression modelswith the pooled sample and report t statistics based on two-way,cluster-robust standard errors in order to control for both cross-sec-tional and time-series dependence (Petersen, 2009).

In Panel B, we find that the Basu measure increases monotonicallyacross the three groups. The difference between the high and lowC_Score groups is 0.111 (t = 3.52). In Panel C, we observe a similar trendfor the Ball and Shivakumar (2005) measure across the three groups(difference between the high and low C_Score groups = 0.347,t = 2.94). In both Panels, we only detect reliable evidence of conserva-tism for the high C_Score group. Under the Basu measure, we find thatearnings of low C_Score firms reflect bad news in a less timely mannerthan they reflect good news. Overall, the consistency in ranking be-tween C_Score and the Basu (1997) and Ball and Shivakumar (2005)measures indicates that C_Score is effective in distinguishing betweenChinese firms with different levels of conservatism.26

26 The implicit assumption is that the Basu and the Ball and Shivakumar asymmetrictimeliness coefficients are appropriate measures of accounting conservatism in theChinese setting. Chen et al. (2010) validate this assumption.

We turn next to the results of our hypotheses testing which arereported in Table 4. The first column shows the results of estimat-ing Eq. (2). We find a significantly negative coefficient on C_Score(a1 = �0.109, t = �6.07), consistent with the prediction of H1 thatlenders reduce collateral requirements for more conservative bor-rowers. We find significantly negative coefficients on SOE and LSB,suggesting that SOE borrowers and borrowers with more loansfrom SBs in general can pledge less collateral because their lendersare less concerned about default risk. The coefficients on most ofthe other control variables exhibit the expected signs. One excep-tion is the negative coefficient on GEO, which indicates lowercollateral requirements for firms located in regions with moredeveloped financial markets.27

The second column reports the results of estimating a modifiedversion of Eq. (2) that allows the relation between COLLATERAL andC_Score to vary between firms with and without loan defaults inthe last year. We continue to find a significantly negativecoefficient on C_Score (coefficient = �0.132, t = �5.80), indicatingthat lenders reduce collateral requirements in exchange for bor-rowers’ conservatism when they observe high credit quality. How-ever, our results show that borrowers’ past loan default experiencemoderates the negative relation between COLLATERAL andC_Score, as indicated by the significantly positive coefficient onC_Score � Defaultt�1 (coefficient = 0.193, t = 3.40). This is consistentwith the prediction of H2 that lenders find the marginal benefit ofcollateral to be higher and therefore are less likely to relax

27 The mean of the variance inflation factor (VIF) for the model is 1.75 and thelargest single VIF is 3.91 indicating that multicollinearity does not appear to be aconcern.

Table 2Correlation matrix.

Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Collateral 1SOE 2 �0.15

(0.00)LSB 3 �0.21 �0.03

(0.00) (0.02)C_Score 4 0.09 �0.09 �0.02

(0.00) (0.00) (0.00)CURRENT 5 �0.07 0.05 0.02 �0.34

(0.00) (0.00) (0.07) (0.00)LEV 6 0.11 �0.08 �0.06 0.51 �0.33

(0.00) (0.00) (0.00) (0.00) (0.00)ROA 7 �0.19 0.07 0.05 �0.24 0.23 �0.24

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00)SIZE 8 �0.25 0.25 �0.07 �0.29 0.15 �0.11 0.37

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)CFOVOL 9 0.03 �0.10 0.02 0.02 �0.02 0.03 �0.10 �0.14

(0.02) (0.00) (0.25) (0.11) (0.24) (0.03) (0.00) (0.00)INTCOV 10 �0.08 0.04 �0.02 �0.18 0.33 �0.19 0.32 0.17 �0.03

(0.00) (0.01) (0.14) (0.00) (0.00) (0.00) (0.00) (0.00) (0.03)GROWTH 11 �0.08 0.03 0.04 �0.12 �0.08 �0.16 0.50 0.12 0.00 0.13

(0.00) (0.10) (0.00) (0.31) (0.12) (0.77) (0.00) (0.00) (0.05) (0.00)MB 12 0.01 �0.03 �0.00 0.08 �0.02 0.32 �0.02 0.02 �0.00 �0.02 �0.01

(0.59) (0.04) (0.99) (0.00) (0.16) (0.00) (0.15) (0.23) (0.96) (0.22) (0.40)LLOSS 13 0.15 �0.07 �0.04 0.15 �0.17 0.15 �0.40 �0.24 0.10 �0.11 �0.21 0.03

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.05)AGE 14 0.11 �0.03 �0.02 0.05 �0.12 0.10 �0.13 �0.05 0.07 �0.06 �0.03 0.03 0.09

(0.00) (0.02) (0.24) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.03) (0.00)GEO 15 �0.05 �0.08 0.02 0.11 �0.08 0.02 0.04 �0.04 0.05 �0.01 0.05 �0.05 �0.01 0.04

(0.00) (0.00) (0.13) (0.00) (0.00) (0.08) (0.00) (0.01) (0.00) (0.69) (0.34) (0.00) (0.39) (0.01)LTDEBT 16 0.06 0.04 �0.07 0.40 �0.19 0.32 �0.03 0.09 �0.06 �0.12 0.09 �0.02 0.02 0.02 �0.08

(0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.06) (0.00) (0.00) (0.00) (0.00) (0.23) (0.24) (0.11) (0.00)

The table contains pairwise Pearson correlations for the full sample (5358 firm-year observations). Two-tailed p values are reported in parentheses.COLLATERAL is the percentage of collateral loans, calculated as the ratio of collateralized loans to total loans outstanding at the end of the year for which the sources of theloans and information on collateral are disclosed in the financial statements; C_Score is the Khan and Watts (2009) conservatism score measure; SOE is a dummy variable thatequals 1 if the firm is an SOE and 0 if it is an NSOE; LSB is measured as the ratio of total loans from SBs to total loans outstanding at the end of the year for which the sources ofthe loans and information on collateral are disclosed in the financial statements; LEV is leverage ratio, defined as the sum of long-term and short-term debt divided by totalequity; SIZE is the natural logarithm of market value of equity; ROA is return on assets; CFOVOL is the volatility of operating cash flows over the past three years; INTCOV isinterest coverage ratio, calculated as income before interest and tax expense divided by interest expense; GROWTH is growth in assets, measured as ending total assetsdivided by beginning total assets; MB is market-to-book ratio; LLOSS is a dummy variable that equals 1 if the firm reports a loss in the previous year and 0 otherwise; CURRENTis current ratio, calculated as current assets divided by current liabilities; AGE is the natural logarithm of the number of years since the firm was established; LTDEBT is thepercentage of long-term debt, calculated as the ratio of long-term debt to total assets at the end of the year.

4998 J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006

collateral requirements despite borrowers’ conservative reportingif they perceive these borrowers to be more risky clients.

The last column examines the effect of asset tangibilityon the relation between the use of collateral and conservatism.Specifically, our modified version of Eq. (2) permits the relation be-tween COLLATERAL and C_Score to vary between firms with highand low asset tangibility. The coefficient on C_Score remains signif-icantly negative (coefficient = �0.126, t = �4.23), consistent withlenders reducing collateral requirements for more conservativeborrowers when they have high asset tangibility. Consistent withthe prediction of H3, the coefficient on C_Score � LowTGt�1 is sig-nificantly positive (coefficient = 0.062, t = 1.69), indicating thatwhen faced with a limited supply of tangible assets that lenderscan rely on as collateral against loans, collateral becomes morecostly and lenders are less willing to share the benefits from con-servatism with those borrowers in the form of reduced collateral.28

28 As a sensitivity check, we separately examine our hypotheses for the pre- andpost-2004 periods. Untabulated results continue to show a significantly negativerelation between COLLATERAL and C_Score in both periods, but the magnitude of thecoefficient on C_Score becomes smaller in the post-2004 period. This could suggestthat lenders start to share the benefits from conservatism with borrowers throughinterest rates in addition to collateral requirements after the Chinese governmentremoved the upper limit on interest rates. We continue to find that borrowers’ pastdefault experience moderates the negative relation between the use of collateral andconservatism in both periods. We find strong evidence that the reward forconservatism is diminished for borrowers with low asset tangibility in the post-2004 period, but the result is much weaker in the pre-2004 period.

8. Additional tests

8.1. Lead-lag relation between use of collateral and conservatism

Our main tests focus on the contemporaneous relation be-tween use of collateral and conservatism. However, one could ar-gue for a potential lead-lag relation between these variables indesigning debt contracts if collateral decisions are partially af-fected by borrowers’ past conservatism. To explore this possibil-ity, we re-run our analyses after replacing C_Scoret with itslagged value. An implicit assumption is that lenders are able toinfer the future level of conservatism from the past, since whatreally matters for lenders is borrowers’ conservatism during thedebt contract period. Zhang (2008) and Nikolaev (2010) arguethat borrowers will adhere to conservative reporting due to con-cerns about reputation, a key factor influencing their future ac-cess to the debt markets, and auditor pressure for conservativecompliance with covenants.

Table 5 summarizes the results. We find a significantlynegative relation between the use of collateral and past conser-vatism in the first column (coefficient = �0.061, t = �5.25). Asshown in the next two columns, when the borrower has lowcredit quality or asset tangibility, the negative relation betweenthe use of collateral and past conservatism becomes lesspronounced. These results are strikingly consistent with thosereported in Table 4.

Table 3Effectiveness of C_Score in distinguishing between Chinese firms with varying degreesof conservatism.

Panel A: Results of estimating Eq. (1)

Variable Pred. sign. Mean coeff. t Value

Intercept �0.297 �3.87***

DR �0.104 �0.59R + �0.170 �1.41R � SIZE + 0.010 1.77*

R �MB � �0.002 �1.49R � LEV � 0.068 1.02DR � Ret + 1.522 1.92*

DR � Ret � SIZE � �0.069 �1.88*

DR � Ret �MB + 0.002 1.31DR � Ret � LEV + 0.282 3.21***

SIZE 0.015 4.00***

MB �0.001 �0.43LEV �0.017 �1.38DR � SIZE 0.005 0.61DR �MB 0.001 0.11DR � LEV 0.016 0.87Adj. R2 11.52%

Panel B: Coefficients from Basu (1997) regressions by high, medium and lowC_Score groups

C_Score group Intercept DR R DR � R

Low 0.025 �0.012 0.030 �0.026(15.71) (�5.93) (5.21) (�2.58)

Medium 0.018 �0.012 0.043 0.001(5.12) (�2.72) (3.05) (0.02)

High 0.021 �0.021 0.052 0.085(2.84) (�2.97) (4.91) (2.84)

High–Low 0.022 0.111(t value) (1.80) (3.52)

Panel C: Coefficients from Ball and Shivakumar (2005) regressions by high,medium and low C_Score groups

C_Score group Intercept DCFO CFO DCFO � CFO

Low 0.016 0.002 �0.605 �0.038(6.48) (0.33) (�30.78) (�0.47)

Medium 0.006 �0.002 �0.726 0.044(2.43) (�0.42) (�41.69) (0.60)

High �0.011 0.013 �0.717 0.309(�4.37) (2.19) (�40.76) (3.64)

High–Low �0.112 0.347(t Value) (�4.25) (2.94)

Panel A reports the results of estimating Eq. (1). We estimate Eq. (1) annually andreport the mean coefficients over the 2001–2006 period (6 years). The t statistic forthe mean is defined as the mean divided by its standard error [the time-seriesstandard deviation of the regression coefficients divided by 61/2] (Fama and Mac-beth, 1973). Panel B reports coefficients from the Basu (1997) model for high,medium and low C_Score groups. Panel C reports coefficients from the Ball andShivakumar (2005) model for high, medium and low C_Score groups. In Panels B andC, t statistics based on two-way, cluster-robust standard errors are reported inparentheses (Petersen, 2009).E is earnings per share; P is year-end stock price per share; R is 12-month, buy-and-hold, annual return from May of year t to April of year t + 1; DR is a dummy variablethat equals 1 if R is negative and 0 otherwise; SIZE is the natural logarithm ofmarket value of equity; MB is market-to-book ratio; LEV is leverage ratio, defined asthe sum of long-term and short-term debt divided by total equity; ACC is accrualsscaled by beginning total assets; accruals are defined as earnings before exceptionalitems and extraordinary items minus cash flows from operations; CFO is cash flowsfrom operations scaled by beginning total assets; cash flows from operations aredefined as earnings before exceptional items and extraordinary items + Deprecia-tion � D(Working capital); D(Working capital) = D(Current assets) � DCash� D(Current liabilities) + D(Current portion of long-term loans); DCFO is 1 if CFO isnegative and 0 otherwise. t values are based on two-way, cluster-robust standarderrors adjusting for cross-sectional and time-series dependence.�� 0.05 Significance level in a two-tailed test.* 0.10 Significance level in a two-tailed test.*** 0.01 Significance level in a two-tailed test.

J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006 4999

8.2. Interest rate and credit risk

As discussed earlier, one advantage of investigating the relationbetween loan collateral and conservatism using the Chinese settingis that the Chinese government’s tight control over interest ratesseverely limits lenders’ use of interest rates to differentiate bor-rowers with differential credit risk. In this section, we assess the(in)effectiveness of interest rates in reflecting credit risk for oursample firms.

We begin with a univariate comparison of interest expense(deflated by average loans outstanding in the year) betweenfirms with different levels of credit risk. We use four metricsto measure credit risk: (1) whether the firm is an SOE; (2)whether the firm has any loan default in the current year; (3)whether the firm reports a loss in the previous year; and (4)whether the firm is designated as an ST/PT/�ST firm.29 For noneof the four risk metrics do our results, reported in Panel A of Ta-ble 6, indicate a significant intra-group difference in interest ex-pense. We also perform the analyses separately for the pre- andpost-2004 periods, and still find no evidence that lenders effi-ciently price risk (based on these metrics) through interest ratesin either period.

In Panel B of Table 6, we re-estimate Eq. (2) using interest ex-pense rather than loan collateral as the dependent variable. Theadjusted R2 decreases substantially to 0.9% compared to 17.8%(reported in the first column of Table 4). None of the explana-tory variables in Eq. (2), except INTCOV, is significantly correlatedwith the dependent variable. This result indicates that, whencompared to loan collateral, a considerably smaller portion ofthe variation in interest expense is explained by the variationin firm performance and credit risk for our sample. Overall, theresults reported inTable 6 provide evidence supporting the argument that interestrates are relatively insensitive to borrowers’ credit risk duringour sample period in China. We note that caution must be exer-cised when interpreting these results as we do not have accessto loan-level data. The use of total interest expense deflated byaverage loans outstanding as a proxy for (average) interest ratesat the firm level is at best a noisy measure of loan pricing orcost of debt.

8.3. Controlling for several non-lender demands for conservatism

The theory of conservatism suggests that conservatism mayalso be driven by non-lender demands such as litigation, taxationand regulation (Watts, 2003). C_Score measures the overall levelof conservatism in a firm’s financial reporting system and may re-flect both lender and non-lender demands for conservatism. Tomore narrowly focus on testing our hypotheses from the lenders’perspective, we repeat our main tests after controlling for severalnon-lender demands for conservatism.

29 ST/PT/�ST firms are firms that report consecutive losses and face substantialdelisting risk. Banks will be especially concerned about the safety of their loansbecause these loss-making firms, which are unlikely to be able to raise more capital inthe stock market, are more likely to lack sufficient net assets to cover their interestand loan repayment. According to the rules introduced by the China SecuritiesRegulatory Commission (CSRC) in 1999, a firm is designated as a special treatment(ST) firm if it incurs losses for two consecutive years and a particular treatment (PT)firm if it continues to report a loss for another year. If a PT firm does not becomeprofitable in one year, it is delisted. PT stocks can be traded only on Fridays and arelimited to a maximum 5% price increase over the previous Friday’s close, but have nodownside limit. Since 2002, the CSRC has ceased the PT designation. Instead, itintroduced a new designation labeled ‘‘⁄ST’’, which is similar to ST but without thetransition period. In other words, if a firm incurs losses for three consecutive years, itis de-listed without a PT period.

Table 4Relation between collateral use and financial reporting conservatism.

Variable Pred. H1 H2 (D = Default t�1) H3 (D = LowTGt�1)

Sign Coeff. t Value Coeff. t Value Coeff. t Value

Intercept ? 2.312 11.00*** 2.328 10.67*** 2.427 9.43***

C_Score � �0.109 �6.07*** �0.132 �5.80*** �0.126 �4.23***

D ? �0.588 �0.78 �0.375 �1.56C_Score � D + 0.193 3.40*** 0.062 1.69*

SOE � �0.071 �4.10*** �0.085 �4.36*** �0.078 �3.90***

LSB � �0.333 �6.55*** �0.255 �6.08*** �0.281 �5.23***

LEV + 0.019 3.26*** 0.016 2.61*** 0.034 4.34***

SIZE � �0.071 �7.03*** �0.077 �7.27*** �0.075 �6.37***

ROA � �0.214 �2.38** �0.292 �4.47*** �0.208 �1.59CFOVOL + �0.117 �0.88 �0.073 �0.63 0.010 0.04INTCOV � �0.000 �1.06 �0.000 �0.18 �0.000 �2.25**

GROWTH � �0.076 �3.60*** �0.079 �2.74*** �0.110 �3.29***

MB � �0.000 �0.83 �0.000 �0.28 �0.000 �1.26LLOSS + 0.039 1.99** 0.058 3.09*** 0.036 1.58CURRENT � 0.009 1.33 0.007 1.08 �0.008 �1.06AGE ? 0.053 3.63*** 0.055 3.25*** 0.058 3.38***

GEO + �0.068 �5.22*** �0.055 �3.97*** �0.081 �5.80***

LTDEBT + 0.278 5.26*** 0.379 7.55*** 0.142 1.77*

SOE � D 0.099 3.71*** 0.007 0.40LSB � D �0.230 �3.32*** �0.119 �1.65*

LEV � D �0.002 �0.17 �0.032 �3.21***

SIZE � D 0.045 1.30 0.015 1.38ROA � D 0.179 1.02 �0.034 �0.20CFOVOL � D �0.226 �0.81 �0.140 �0.65INTCOV � D �0.001 �3.74*** 0.000 1.31GROWTH � D 0.022 0.34 0.064 1.96**

MB � D 0.000 �0.58 0.000 0.96LLOSS � D �0.061 �1.79* 0.000 �0.01CURRENT � D 0.009 0.30 0.036 2.97***

AGE � D �0.033 �0.74 �0.014 �1.19GEO � D �0.079 �1.74* 0.031 2.10**

LTDEBT � D �0.564 �3.60*** 0.256 2.96***

YEAR Yes Yes YesIND Yes Yes YesAdj. R2 17.8% 19.1% 18.6%

Column 1 reports the results of estimating the relation between collateral use and financial reporting conservatism for the full sample. Column 2 reports the results ofestimating the relation between collateral use and financial reporting conservatism for firms with high and low observed credit quality. 630 and 4728 observations areclassified as low and high observed credit quality, respectively. Column 3 reports the results of estimating the relation between collateral use and financial reportingconservatism for firms with high and low asset tangibility. Asset tangibility is measured as the proportion of tangible assets on the balance sheet, i.e., fixed assets divided bytotal assets at the end of year t � 1. Sample firms are classified as high (low) if their asset tangibility is above (below) the median for each industry and year.COLLATERAL is the percentage of collateral loans, calculated as the ratio of collateralized loans to total loans outstanding at the end of the year for which the sources of theloans and information on collateral are disclosed in the financial statements; C_Score is the Khan and Watts (2009) conservatism score measure; Defaultt�1 is a dummyvariable that equals 1 if the firm has low observed credit quality and 0 otherwise; LowTGt�1 is a dummy variable that equals 1 if the firm is in the low asset tangibility groupand 0 otherwise; SOE is a dummy variable that equals 1 if the firm is an SOE and 0 if it is an NSOE; LSB is measured as the ratio of total loans from SBs to total loansoutstanding at the end of the year for which the sources of the loans and information on collateral are disclosed in the financial statements; LEV is leverage ratio, defined asthe sum of long-term and short-term debt divided by total equity; SIZE is the natural logarithm of market value of equity; ROA is return on assets; CFOVOL is the volatility ofoperating cash flows over the past three years; INTCOV is interest coverage ratio, calculated as income before interest and tax expense divided by interest expense; GROWTH isgrowth in assets, measured as ending total assets divided by beginning total assets; MB is market-to-book ratio; LLOSS is a dummy variable that equals 1 if the firm reports aloss in the previous year and 0 otherwise; CURRENT is current ratio, calculated as current assets divided by current liabilities; AGE is the natural logarithm of the number ofyears since the firm was established; GEO is the natural logarithm of the index of financial market competitiveness for each province or provincial level region; LTDEBT is thepercentage of long-term debt, calculated as the ratio of long-term debt to total assets at the end of the year; YEAR and IND are year and industry dummies, respectively. tvalues are based on two-way, cluster-robust standard errors adjusting for cross-sectional and time-series dependence.* 0.10 Significance level in a two-tailed test.** 0.05 Significance level in a two-tailed test.*** 0.01 Significance level in a two-tailed test.

5000 J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006

Our proxies for non-lender demands for conservatism arepercentage of controlling shareholders’ ownership (CTRL OWN),natural logarithm of the index of legal environment strengthfor each province or provincial level region (LEGAL IDX), andeffective tax rate (ETR). We include CTRL OWN to capture the ex-tent of equity demand for conservatism. Chinese firms are char-acterized by highly concentrated ownership. As ownershipbecomes more concentrated, it provides controlling shareholderswith incentives and opportunities to expropriate wealth fromminority shareholders. We expect the information asymmetrybetween management/controlling shareholders and minorityshareholders to increase with CTRL OWN. Since conservatism

plays an important role in addressing the demand for reducinginformation asymmetry, we expect the equity demand for con-servatism to be positively related to CTRL OWN. Watts (2003) ar-gues that firms subject to high litigation risk prefer conservativefinancial reporting to reduce their expected litigation costs.Although the overall legal environment of China is still weak,there exists cross-sectional variance in litigation risk exposureby region. We use the regional legal index developed by Fanand Wang (2007) to measure firms’ litigation risk exposureand expect the litigation demand for conservatism to be posi-tively related to LEGAL IDX. Another important factor that influ-ences conservatism is taxation. Watts (2003) argues that firms

Table 5Relation between collateral use and lagged financial reporting conservatism.

Variable Pred. H1 H2 (D = Default t�1) H3 (D = LowTGt�1)

Sign Coeff. t Value Coeff. t Value Coeff. t Value

Intercept ? 2.290 12.11*** 2.292 12.32*** 2.513 10.37***

C_Scoret�1 � �0.061 �5.25*** �0.074 �6.42*** �0.078 �4.72***

D ? �0.314 �0.42 �0.525 �2.01**

C_Scoret�1 � D + 0.113 1.88* 0.065 2.29**

SOE � �0.071 �4.39*** �0.086 �4.60*** �0.072 �3.62***

LSB � �0.333 �6.44*** �0.251 �6.25*** �0.294 �5.60***

LEV + 0.018 3.72*** 0.012 1.50 0.031 4.26***

SIZE � �0.069 �8.20*** �0.074 �8.11*** �0.076 �7.22***

ROA � �0.166 �1.71* �0.223 �2.56*** �0.136 �1.12CFOVOL + �0.112 �0.84 �0.040 �0.34 �0.019 �0.09INTCOV � �0.000 �1.26 �0.000 �0.26 �0.000 �1.62GROWTH � �0.083 �3.60*** �0.092 �3.03*** �0.131 �4.11***

MB � �0.000 �1.43 �0.000 �0.36 �0.001 �1.68*

LLOSS + 0.040 2.02** 0.060 3.18*** 0.035 1.50CURRENT � 0.011 1.44 0.009 1.25 �0.007 �0.78AGE ? 0.053 3.43*** 0.054 3.03*** 0.056 3.11***

GEO + �0.075 �5.84*** �0.060 �3.95*** �0.092 �5.87***

LTDEBT + 0.088 1.51 0.251 4.69*** �0.023 �0.64SOE � D 0.102 3.74*** �0.003 �0.15LSB � D �0.242 �3.53*** �0.092 �1.10LEV � D 0.003 0.23 �0.030 �3.08***

SIZE � D 0.030 0.91 0.018 1.39ROA � D 0.082 0.46 �0.090 �0.56CFOVOL � D �0.337 �1.11 �0.081 �0.52INTCOV � D �0.001 �4.30*** 0.000 0.74GROWTH � D 0.044 0.61 0.088 2.46**

MB � D 0.000 �0.28 0.001 1.14LLOSS � D �0.068 �2.04** 0.003 0.18CURRENT � D 0.012 0.41 0.039 2.64***

AGE � D �0.029 �0.57 �0.011 �0.84GEO � D �0.065 �1.57 0.042 2.61***

LTDEBT � D �0.400 �4.72*** 0.314 6.06***

YEAR Yes Yes YesIND Yes Yes YesAdj. R2 17.6% 19.0% 18.5%

Column 1 reports the results of estimating the relation between collateral use and lagged financial reporting conservatism for the full sample (5,050 firm-year observations).Column 2 reports the results of estimating the relation between collateral use and lagged financial reporting conservatism for firms with high and low observed credit quality.604 and 4446 observations are classified as low and high observed credit quality, respectively. Column 3 reports the results of estimating the relation between collateral useand lagged financial reporting conservatism for firms with high and low asset tangibility. Asset tangibility is measured as the proportion of tangible assets on the balancesheet, i.e., fixed assets divided by total assets at the end of year t � 1. Sample firms are classified as high (low) if their asset tangibility is above (below) the median for eachindustry and year.COLLATERAL is the percentage of collateral loans, calculated as the ratio of collateralized loans to total loans outstanding at the end of the year for which the sources of theloans and information on collateral are disclosed in the financial statements; C_Score is the Khan and Watts (2009) conservatism score measure; Defaultt�1 is a dummyvariable that equals 1 if the firm has low observed credit quality and 0 otherwise; LowTGt�1 is a dummy variable that equals 1 if the firm is in the low asset tangibility groupand 0 otherwise; SOE is a dummy variable that equals 1 if the firm is an SOE and 0 if it is an NSOE; LSB is measured as the ratio of total loans from SBs to total loansoutstanding at the end of the year for which the sources of the loans and information on collateral are disclosed in the financial statements; LEV is leverage ratio, defined asthe sum of long-term and short-term debt divided by total equity; SIZE is the natural logarithm of market value of equity; ROA is return on assets; CFOVOL is the volatility ofoperating cash flows over the past three years; INTCOV is interest coverage ratio, calculated as income before interest and tax expense divided by interest expense; GROWTH isgrowth in assets, measured as ending total assets divided by beginning total assets; MB is market-to-book ratio; LLOSS is a dummy variable that equals 1 if the firm reports aloss in the previous year and 0 otherwise; CURRENT is current ratio, calculated as current assets divided by current liabilities; AGE is the natural logarithm of the number ofyears since the firm was established; GEO is the natural logarithm of the index of financial market competitiveness for each province or provincial level region; LTDEBT is thepercentage of long-term debt, calculated as the ratio of long-term debt to total assets at the end of the year; YEAR and IND are year and industry dummies, respectively. tvalues are based on two-way, cluster-robust standard errors adjusting for cross-sectional and time-series dependence.* 0.10 Significance level in a two-tailed test.** 0.05 Significance level in a two-tailed test.*** 0.01 Significance level in a two-tailed test.

J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006 5001

accelerate loss recognition relative to gain recognition so thatthey can minimize the present value of their tax payments.Accordingly, we expect the taxation demand for conservatismto increase in ETR.

The results of re-testing our three hypotheses are presented inTable 7. Consistent with our main results, we find a significantlynegative association between COLLATERAL and C_Score aftercontrolling for these non-lender demands for conservatism. Thisreinforces H1 that lenders, on average, value conservatism andreduce collateral requirements for more conservative borrowers.The significantly positive coefficients on C_Score � Defaultt�1 is

consistent with H2, suggesting that the relation between the useof collateral and conservatism (after controlling for non-lenderdemands) is moderated as lenders become more concerned aboutdefault risk. The coefficient on C_Score � LowTGt�1 remainspositive, but its statistical significance drops slightly below the10% level (coefficient = 0.059, t = 1.60).

Turning to our proxies for non-lender demands for conservatism,we find that CTRL OWN and ETR are both positively related to the useof collateral (significant at the 1% level), and LEGAL IDX is negativelyrelated to the use of collateral (significant at the 10% level). Thesesignificant relations suggest that either our C_Score measure does

Table 6Relation between interest rate and credit risk.

Panel A: Mean values of interest expense deflated by average loans outstanding between firms with different levels of credit risk

Measure of credit risk 2001–2006 2001–2003 2004–2006

Yes No t Value Yes No t Value Yes No t Value

(1) Is the firm an SOE? 0.062 0.060 1.30 0.061 0.058 1.21 0.064 0.061 0.91(2) Does the firm default in the current year? 0.060 0.060 0.04 0.061 0.060 0.99 0.062 0.063 �0.75(3) Is the firm designated as an ST/PT/⁄ST firm? 0.065 0.061 0.65 0.059 0.061 �0.91 0.080 0.061 1.11(4) Does the firm report a loss in the previous year? 0.060 0.062 �0.84 0.058 0.061 �1.52 0.071 0.061 0.66

Panel B: Re-estimating Eq. (2) using interest expense deflated by average loans outstanding as the dependent variable

Variable Coeff. t Value

Intercept 0.581 1.43C_Score �0.235 �1.16SOE �0.091 �1.06LSB �0.641 �1.05LEV �0.070 �1.47SIZE 0.021 0.96ROA �1.512 �1.03CFOVOL 1.577 1.34INTCOV �0.000 �2.15**

GROWTH �0.331 �1.34MB 0.003 1.09LLOSS 0.010 0.35CURRENT �0.037 �1.18AGE �0.028 �1.24GEO 0.065 1.05LTDEBT �0.076 �0.75YEAR YesIND YesAdj. R2 0.9%

Panel A reports the results of comparing the mean values of interest expense deflated by average loans outstanding between firms with different levels of credit risk for theentire sample period, the pre-2004 sample period, and the post-2004 sample period. t tests are used to test differences between means. Panel B reports the results of re-estimating Eq. (2) using interest expense deflated by average loans outstanding as the dependent variable. t Values are based on two-way, cluster-robust standard errorsadjusting for cross-sectional and time-series dependence.C_Score is the Khan and Watts (2009) conservatism score measure; SOE is a dummy variable that equals 1 if the firm is an SOE and 0 if it is an NSOE; LSB is measured as theratio of total loans from SBs to total loans outstanding at the end of the year for which the sources of the loans and information on collateral are disclosed in the financialstatements; LEV is leverage ratio, defined as the sum of long-term and short-term debt divided by total equity; SIZE is the natural logarithm of market value of equity; ROA isreturn on assets; CFOVOL is the volatility of operating cash flows over the past three years; INTCOV is interest coverage ratio, calculated as income before interest and taxexpense divided by interest expense; GROWTH is growth in assets, measured as ending total assets divided by beginning total assets; MB is market-to-book ratio; LLOSS is adummy variable that equals 1 if the firm reports a loss in the previous year and 0 otherwise; CURRENT is current ratio, calculated as current assets divided by currentliabilities; AGE is the natural logarithm of the number of years since the firm was established; GEO is the natural logarithm of the index of financial market competitiveness foreach province or provincial level region; LTDEBT is the percentage of long-term debt, calculated as the ratio of long-term debt to total assets at the end of the year; YEAR andIND are year and industry dummies, respectively.** 0.05 Significance level in a two-tailed test.

5002 J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006

not fully capture the non-lender demands for conservatism (proxiedby CTRL OWN, LEGAL IND, and ETR) or our proxies for non-lenderdemands can affect the use of collateral through other channels(unrelated to financial reporting conservatism).

8.4. Unconditional conservatism and the use of collateral

So far, our analyses have focused on conditional conservatism.From a contracting perspective, conditional and unconditional con-servatism are substantially different concepts. Researchers gener-ally believe that conditional conservatism can enhancecontracting efficiency, but unconditional conservatism seems inef-ficient or at best neutral in contracting (Ball and Shivakumar,2005).

Nevertheless, in our context, lenders may perceive uncondi-tional conservatism as helpful in making collateral decisionsbecause of the systematic understatement of the book value ofnet assets over time (even though it does not reflect any new infor-mation about changes in net assets). To examine whether lendersvalue unconditional conservatism, we replace C_Score withUcdConsv, a proxy for unconditional conservatism, and re-estimateEq. (2). Following Beaver and Ryan (2000), we measure uncondi-

tional conservatism using a regression of the book-to-market ratioon the current and six lagged annual stock returns with bothfirm-fixed and time-fixed effects:

BTMit ¼ ai þ at þX6

j¼0

bjRETi;t�j þ eit ð3Þ

where BTM denotes the book value of common equity divided bythe end of year market value of common equity, RET is annual stockreturns, and at is the time intercept, which captures the year-by-year variation in the BTM common to all sample firms. Thefirm-specific variation in BTM can be decomposed into the firmeffect (a consequence of financial reporting with unconditionalconservatism) and the portion associated with current and laggedreturns. The firm effect, ai (our measure of UcdConsv), is expectedto capture the cumulative effect of unconditional conservatism onthe book value of net assets.

Table 8 reports the results of re-examining our three hypothesesusing UcdConsv in place of C_Score. We detect a significantly nega-tive relation between COLLATERAL and UcdConsv, suggesting thatlenders reduce collateral requirements for borrowers with moreunconditional conservatism. However, we fail to find a significantdifference in the effect of unconditional conservatism on collateral

Table 7Relation between collateral use and financial reporting conservatism after controlling for non-lender demands for conservatism.

Variable Pred. H1 H2 (D = Default t�1) H3 (D = LowTGt�1)

Sign Coeff. t Value Coeff. t Value Coeff. t Value

Intercept ? 2.242 10.10*** 2.262 9.81*** 2.339 9.32***

C_Score � �0.113 �6.16*** �0.135 �5.82*** �0.130 �4.42***

D ? �0.635 �0.83 �0.342 �1.31C_Score � D + 0.181 3.01*** 0.059 1.60SOE � �0.073 �4.25*** �0.087 �4.55*** �0.081 �4.13***

LSB � �0.333 �6.58*** �0.256 �6.11*** �0.280 �5.18***

LEV + 0.019 3.14*** 0.016 2.62*** 0.034 4.30***

SIZE � �0.068 �6.15*** �0.073 �6.39*** �0.071 �6.03***

ROA � �0.226 �2.59*** �0.312 �4.65*** �0.232 �1.75*

CFOVOL + �0.092 �0.67 �0.052 �0.46 0.050 0.20INTCOV � �0.000 �1.05 �0.000 �0.19 �0.001 �2.39**

GROWTH � �0.074 �3.39*** �0.078 �2.63*** �0.110 �3.18***

MB � �0.000 �0.88 �0.000 �0.31 �0.000 �1.26LLOSS + 0.019 0.85 0.043 1.77* 0.019 0.73CURRENT � 0.010 1.35 0.007 1.11 �0.008 �0.99AGE ? 0.055 3.71*** 0.056 3.32*** 0.059 3.39***

GEO + �0.047 �3.52*** �0.034 �3.12*** �0.050 �3.19***

LTDEBT + 0.260 4.86*** 0.362 6.99*** 0.124 1.54CTRL OWN ? 0.002 5.62*** 0.002 8.35*** 0.002 2.22***

LEGAL IDX ? �0.029 �1.87* �0.027 �1.89* �0.040 �2.23**

ETR ? 0.030 2.64*** 0.022 1.66* 0.022 1.46***

SOE � D 0.098 3.96*** 0.007 0.39LSB � D �0.228 �3.53*** �0.121 �1.69*

LEV � D �0.001 �0.06 �0.032 �3.20***

SIZE � D 0.047 1.34 0.013 1.14ROA � D 0.229 1.26 �0.009 �0.05CFOVOL � D �0.165 �0.58 �0.175 �0.74INTCOV � D �0.001 �2.96*** 0.000 1.37GROWTH � D 0.018 0.29 0.068 2.00**

MB � D 0.000 �0.59 0.000 0.89LLOSS � D �0.066 �1.90* �0.007 �0.27CURRENT � D 0.006 0.20 0.035 2.95***

AGE � D �0.027 �0.68 �0.013 �0.99GEO � D �0.063 �0.87 0.010 0.41LTDEBT � D �0.579 �3.67*** 0.254 3.07***

CTRL OWN � D 0.000 0.09 0.000 0.58LEGAL IDX � D �0.025 �0.43 0.023 1.16ETR � D 0.016 0.75 0.019 0.55YEAR Yes Yes YesIND Yes Yes YesAdj. R2 18.1% 19.4% 18.9%

This table reports the results of the relation between collateral use and financial reporting conservatism after we control for conservatism associated with non-lenderdemands.COLLATERAL is the percentage of collateral loans, calculated as the ratio of collateralized loans to total loans outstanding at the end of the year for which the sources of theloans and information on collateral are disclosed in the financial statements; C_Score is the Khan and Watts (2009) conservatism score measure; SOE is a dummy variable thatequals 1 if the firm is an SOE and 0 if it is an NSOE; LSB is measured as the ratio of total loans from SBs to total loans outstanding at the end of the year for which the sources ofthe loans and information on collateral are disclosed in the financial statements; LEV is leverage ratio, defined as the sum of long-term and short-term debt divided by totalequity; SIZE is the natural logarithm of market value of equity; ROA is return on assets; CFOVOL is the volatility of operating cash flows over the past three years; INTCOV isinterest coverage ratio, calculated as income before interest and tax expense divided by interest expense; GROWTH is growth in assets, measured as ending total assetsdivided by beginning total assets; MB is market-to-book ratio; LLOSS is a dummy variable that equals 1 if the firm reports a loss in the previous year and 0 otherwise; CURRENTis current ratio, calculated as current assets divided by current liabilities; AGE is the natural logarithm of the number of years since the firm was established; DEFAULT is adummy variable that equals 1 if the firm has a loan in default in the current year and 0 otherwise; GEO is the natural logarithm of the index of financial marketcompetitiveness for each province or provincial level region; LTDEBT is the percentage of long-term debt, calculated as the ratio of long-term debt to total assets at the end ofthe year; CTRL OWN is the percentage of controlling shareholder’s ownership; LEGAL IDX is the natural logarithm of the index of legal environment strength for each provinceor provincial level region; ETR is the effective tax rate, calculated as total tax expense divided by pretax financial income. We set ETR to zero for firms with tax refunds and100% for firms with positive tax expense and negative (or zero) pretax financial income; YEAR and IND are year and industry dummies, respectively. t values are based on two-way, cluster-robust standard errors adjusting for cross-sectional and time-series dependence.* 0.10 Significance level in a two-tailed test.** 0.05 Significance level in a two-tailed test.*** 0.01 Significance level in a two-tailed test.

J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006 5003

requirements between firms with and without past defaultexperience or between firms with high and low asset tangibility.Overall, our results suggest that lenders may take into account bor-rowers’ unconditional conservatism when making collateralrequirements. However, since unconditional conservatism has noinformation role in the debt contracting process, lenders are unlikelyto treat it as an effective tool to monitor moral hazard. As a result, wedo not find the relation between unconditional conservatism and

the use of collateral to vary with either borrowers’ defaultrisk or borrowers’ ability to pledge tangible assets as security forthe loans.

9. Conclusions

The agency problem of debt in emerging markets can be moresevere due to the more opaque information environments in such

Table 8Relation between collateral use and unconditional financial reporting conservatism.

Variable Pred. H1 H2 (D = Default t�1) H3 (D = LowTGt�1)

Sign Coeff. t Value Coeff. t Value Coeff. t Value

Intercept ? 2.318 10.86*** 2.312 10.16*** 2.404 8.93***

UcdConsv ? �0.022 �1.90* �0.021 �1.72* �0.025 �1.46D ? �0.314 �0.45 �0.268 �1.04UcdConsv � D ? �0.000 �0.01 0.002 0.12SOE � �0.075 �3.94*** �0.089 �4.15*** �0.080 �3.70***

LSB � �0.334 �6.49*** �0.252 �5.88*** �0.281 �5.01***

LEV + 0.009 1.76* 0.002 0.34 0.020 2.87***

SIZE � �0.070 �7.10*** �0.074 �7.20*** �0.072 �6.31***

ROA � �0.174 �1.91* �0.231 �3.26*** �0.171 �1.37CFOVOL + �0.109 �0.81 �0.058 �0.48 �0.007 �0.03INTCOV � �0.000 �0.97 0.000 �0.11 �0.000 �2.23**

GROWTH � �0.076 �3.50*** �0.080 �2.69*** �0.115 �3.44***

LLOSS + 0.039 1.97** 0.059 3.03*** 0.036 1.54CURRENT � 0.012 1.57 0.010 1.42 �0.007 �0.91AGE ? 0.043 3.32*** 0.045 3.18*** 0.049 2.99***

GEO + �0.068 �5.43*** �0.056 �4.15*** �0.082 �6.10***

LTDEBT + 0.220 4.78*** 0.315 6.68*** 0.070 1.13SOE � D 0.099 3.55*** 0.005 0.27LSB � D �0.233 �3.33*** �0.122 �1.66*

LEV � D 0.010 0.90 �0.022 �2.95***

SIZE � D 0.033 1.04 0.010 0.93ROA � D 0.096 0.53 �0.037 �0.24CFOVOL � D �0.193 �0.66 �0.110 �0.55INTCOV � D �0.001 �3.75*** 0.000 1.36GROWTH � D 0.028 0.42 0.072 2.27**

LLOSS � D �0.062 �1.73* 0.000 0.01CURRENT � D 0.001 0.02 0.037 3.12***

AGE � D �0.038 �0.96 �0.021 �1.57GEO � D �0.069 �1.61 0.033 2.19**

LTDEBT � D �0.446 �3.44*** 0.297 4.55***

YEAR Yes Yes YesIND Yes Yes YesAdj. R2 17.7% 18.9% 18.5%

Column 1 reports the results of estimating the relation between collateral use and unconditional conservatism for the full sample (5,358 firm-year observations). Column 2reports the results of estimating the relation between collateral use and unconditional conservatism for firms with high and low observed credit quality. 630 and 4728observations are classified as low and high observed credit quality, respectively. Column 3 reports the results of estimating the relation between collateral use andunconditional conservatism for firms with high and low asset tangibility. Asset tangibility is measured as the proportion of tangible assets on the balance sheet, i.e., fixedassets divided by total assets at the end of year t � 1. Sample firms are classified as high (low) if their asset tangibility is above (below) the median for each industry and year.COLLATERAL is the percentage of collateral loans, calculated as the ratio of collateralized loans to total loans outstanding at the end of the year for which the sources of theloans and information on collateral are disclosed in the financial statements; UcdConsv is the Beaver and Ryan (2000) measure of unconditional conservatism; Defaultt�1 is adummy variable that equals 1 if the firm has low observed credit quality and 0 otherwise; LowTGt�1 is a dummy variable that equals 1 if the firm is in the low asset tangibilitygroup and 0 otherwise; SOE is a dummy variable that equals 1 if the firm is an SOE and 0 if it is an NSOE; LSB is measured as the ratio of total loans from SBs to total loansoutstanding at the end of the year for which the sources of the loans and information on collateral are disclosed in the financial statements; LEV is leverage ratio, defined asthe sum of long-term and short-term debt divided by total equity; SIZE is the natural logarithm of market value of equity; ROA is return on assets; CFOVOL is the volatility ofoperating cash flows over the past three years; INTCOV is interest coverage ratio, calculated as income before interest and tax expense divided by interest expense; GROWTH isgrowth in assets, measured as ending total assets divided by beginning total assets; MB is market-to-book ratio; LLOSS is a dummy variable that equals 1 if the firm reports aloss in the previous year and 0 otherwise; CURRENT is current ratio, calculated as current assets divided by current liabilities; AGE is the natural logarithm of the number ofyears since the firm was established; GEO is the natural logarithm of the index of financial market competitiveness for each province or provincial level region; LTDEBT is thepercentage of long-term debt, calculated as the ratio of long-term debt to total loans outstanding at the end of the year; YEAR and IND are year and industry dummies,respectively. t Values are based on two-way, cluster-robust standard errors adjusting for cross-sectional and time-series dependence.* 0.10 Significance level in a two-tailed test.** 0.05 Significance level in a two-tailed test.*** 0.01 Significance level in a two-tailed test.

5004 J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006

markets. As a result, lenders’ use of collateral to safeguard theirinvestment may become more critical. However, weaker legalsystems and contract enforceability could undermine the effi-ciency of loan collateral as a contracting tool. Accordingly, lendersmay strategically resort to other mechanisms (beyond commoncontract terms) in addition to collateral in addressing moral haz-ard. One potential tool is the extent to which the borrower dis-closes bad news in a more timely fashion than good news in itsfinancial reports, i.e., financial reporting conservatism.

Our study examines the relation between the use of collateraland conservatism for a sample of Chinese firms from 2001 to2006. We find that lenders reduce collateral requirements for moreconservative borrowers. This result suggests that lenders valueconservatism and share the benefits from conservatism with bor-

rowers. However, borrowers’ poor observed credit quality andlow asset tangibility can moderate the negative relation betweenthe use of collateral and conservatism. It appears that lenders areless willing to relax collateral requirements despite borrowers’conservative reporting if they are sufficiently concerned about de-fault risk and potential recovery in default.

One caveat of our study is that collateral and conservatism canbe endogenous terms of a firm’s financial profile that arise from theinteraction of firm characteristics and bank policy. Therefore, it ispossible that collateral and conservatism are simultaneously deter-mined but we are unable to identify and control all their determi-nants in our analyses. To take into account the possibleendogeneity issue, one needs to use a simultaneous equations ap-proach with well-identified instruments. Given the potential risk of

J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006 5005

endogeneity and the difficulty in finding legitimate instruments forboth collateral and conservatism, our study should be viewed onlyas documenting an association between these two variables. Weleave it for future research to further explore and establish thecausality.

As China has gradually liberalized lending interest rates inrecent years, potential avenues for future research includeexamining the difference in the contracting benefits of conser-vatism as reflected in lowering interest rates and collateralrequirements. Studies on the effects of conservatism on multi-ple loan contract terms should provide a more thorough andcomplete view of the role of conservatism in addressing theagency problem of debt.

Acknowledgements

We thank Sudipta Basu, Bjorn Jorgensen, Jeff Yu, JerryZimmerman, two anonymous reviewers, and workshop partici-pants at Chinese University of Hong Kong, Fudan University,McMaster University, Peking University, Renmin University ofChina, Rutgers University, Shanghai National Accounting Institute,Singapore Management University, University of Colorado atBoulder, the 2011 American Accounting Association Annual Meet-ing, and the 2011 Annual Congress of the European AccountingAssociation for helpful comments. Yanyan Wang acknowledgesfinancial support from the National Nature Science Foundation ofChina (NSFC-71002043, NSFC-71372074), New Century ExcellentTalents of Ministry of Education (NCET-11-0297), Fok Ying TungEducation Foundation (131083), and Research Foundation for Cen-tral Universities (2013221012). Lisheng Yu acknowledges financialsupport from the National Nature Science Foundation of China(NSFC-70972114), New Century Excellent Talents of FujianProvince, and Research Foundation for Central Universities(2012221010).

30 These firm attributes are correlated to the demand for conservatism. First, largerfirms, for example, have lower information asymmetry (Easley et al. 2002), andaccordingly, have lower contracting demand for conservatism (Khan and Watts 2009).Second, a firm’s higher market-to-book ratio indicates that firms have more growthoptions. These growth options increase information asymmetry between insiders andoutside investors and accordingly, increase investors’ demand for accountingconservatism (Watts 2003; LaFond and Watts 2008). Third, a higher level of financialleverage causes agency conflicts between debtholders and shareholders, conse-quently increasing debt contracting demand for conservatism (Watts 2003; Khan andWatts 2009).

Appendix A. Appendix: Khan and Watts (2009) firm-yearmeasure of conservatism

Khan and Watts (2009) draw on the Basu (1997) measure ofasymmetric timeliness to estimate a firm-year measure of conser-vatism (denoted as C_Score). To test for asymmetric timeliness,Basu (1997) uses positive and negative returns as proxies for goodand bad news, and estimates a cross-sectional, reverse regressionof earnings on current returns. Specifically, his model is:

Xit ¼ b0 þ b1Dit þ b2Rit þ b3Dit � Rit þ eit ð1Þ

where X is the Net income before extraordinary items, divided bybeginning market value; R is the Buy-and-hold returns during theperiod and D is an indicator variable that takes the value of one ifR is negative, and zero otherwise.

The measure of incremental timeliness for bad news over goodnews, or conservatism, is b3. A significantly positive b3 indicatesthat financial reporting requires a lower degree of verification inrecognizing economic losses relative to economic gains. Basu’s ap-proach works well under the assumption that returns summarizenews from sources other than earnings that becomes available tothe market during the period and this news in principle could berecognized in earnings in that period (Ryan, 2006). He finds thatthe slope coefficient on returns is about five times more positivewhen returns are negative than when they are positive.

Basu’s model is estimated cross-sectionally by year, so his mea-sure of conservatism has intertemporal variation but forces allfirms to have the same level of conservatism in each year. Khanand Watts (2009) build upon Basu (1997) and propose a method

to estimate conservatism at the firm-year level. They first specifyb2 and b3 as a linear function of firm-specific characteristics:

b2 ¼ l0 þ l1MVi;t þ l2M=Bi;t þ l3Lev i;t ð2Þb3 ¼ k0 þ k1MVi;t þ k2M=Bi;t þ k3Lev i;t ð3Þ

where MV is the natural logarithm of market value of equity; M/B isthe Market-to-book ratio and Lev the Long-term debt plus short-term debt divided by market value of equity.

Next, they substitute Eqs. (2) and (3) into (1) and run the fol-lowing annual cross-sectional regression model:

Xit ¼b0 þ b2Dit þ Ritðl0 þ l1MVi;t þ l2M=Bi;t þ l3Lev i;tÞþ Dit � Ritðk0 þ k1MVi;t þ k2M=Bi;t þ k3Lev i;tÞþ ðd0 þ d1MVi;t þ d2M=Bi;t þ d3DitLev i;t þ d4DitMVi;t

þ d5DitM=Bi;t þ d6Lev i;tÞ þ eit ð4Þ

Khan and Watts (2009) include the additional terms in the lastparenthesis because model (4) includes interaction terms betweenreturns and firm characteristics. Therefore, they control for the firmcharacteristics separately (the ‘‘main effects’’).

The coefficients, li and ki, i = 0–3, are constant across firms,but vary over years since they are estimated from annualcross-sectional regression models. C_Score is defined as k0 + k1MVi,t +k2M/Bi,t + k3Levi,t, which varies across firms through cross-sectionalvariation in the firm-year characteristics (Size, M/B, and Lev), andover time through intertemporal variation in ki,t and the firm-yearcharacteristics.30 Conservatism is increasing in the C_Score.

References

Aghion, P., Bolton, P., 1992. An incomplete contracts approach to financialcontracting. Review of Economic Studies 59, 473–494.

Albuquerque, R., Hopenhayn, H.A., 2004. Optimal lending contracts and firmdynamics. Review of Economic Studies 71, 285–315.

Allen, F., Qian, J., Qian, M., 2005. Finance, law, and economic growth in China.Journal of Financial Economics 77, 57–116.

Ayyagari, M., Demirguc-Kunt, A., Maksimovic, V., 2010. Formal versus informalfinance: evidence from China. Review of Financial Studies 23, 3048–3097.

Bae, K.H., Goyal, V.K., 2009. Creditor rights, enforcement, and bank loans. Journal ofFinance 64, 823–860.

Ball, R., Shivakumar, L., 2005. Earnings quality in UK private firms: comparative lossrecognition timeliness. Journal of Accounting and Economics 39, 83–128.

Ball, R., Bushman, R.M., Vasvari, F.P., 2008a. The debt-contracting value ofaccounting information and loan syndicate structure. Journal of AccountingResearch 46, 247–287.

Ball, R., Robin, A., Sadka, G., 2008b. Is financial reporting shaped by equity marketsor by debt markets? An international study of timeliness and conservatism.Review of Accounting Studies 13, 168–205.

Basu, S., 1997. The conservatism principle and the asymmetric timeliness ofearnings. Journal of Accounting and Economics 24, 3–37.

Beatty, A., Weber, J., Yu, J.J., 2008. Conservatism and debt. Journal of Accounting andEconomics 45, 154–174.

Beaver, W.H., Ryan, S.G., 2000. Biases and lags in book value and their effects on theability of the book-to-market ratio to predict book return on equity. Journal ofAccounting Research 38, 127–148.

Berger, A.N., Udell, G.F., 1990. Collateral, loan quality, and bank risk. Journal ofMonetary Economics 25, 21–42.

Berger, A.N., Espinosa-Vega, M.A., Frame, W.S., Miller, N.H., 2011a. Why doborrowers pledge collateral? New empirical evidence on the role ofasymmetric information. Journal of Financial Intermediation 20, 55–70.

Berger, A.N., Frame, W.S., Ioannidou, V., 2011b. Tests of ex ante versus ex posttheories of collateral using private and public information. Journal of FinancialEconomics 100, 85–97.

5006 J.Z. Chen et al. / Journal of Banking & Finance 37 (2013) 4989–5006

Berkman, H., Cole, R.A., Fu, L.J., 2009. Expropriation through loan guarantees to relatedparties: evidence from China. Journal of Banking and Finance 33, 141–156.

Besanko, D., Thakor, A.V., 1987a. Collateral and rationing: sorting equilibria inmonopolistic and competitive credit markets. International Economic Review28, 671–689.

Besanko, D., Thakor, A.V., 1987b. Competitive equilibria in the credit market underasymmetric information. Journal of Economic Theory 42, 167–182.

Bester, H., 1985. Screening vs. rationing in credit markets with imperfectinformation. American Economic Review 75, 850–855.

Boot, A.W.A., Thakor, A.V., Udell, G.F., 1991. Secured lending and default risk:equilibrium analysis, policy implications and empirical results. EconomicJournal 101, 458–472.

Boot, A.W.A., 2000. Relationship banking: what do we know? Journal of FinancialIntermediation 9, 7–25.

Boyd, J.H., Smith, B.D., 1993. The equilibrium allocation of investment capital in thepresence of adverse selection and costly state verification. Economic Theory 3,427–451.

Brick, I.E., Palia, D., 2007. Evidence of jointness in the terms of relationship lending.Journal of Financial Intermediation 16, 452–476.

Chakraborty, A., Hu, C.X., 2006. Lending relationships in line-of-credit and nonline-of-credit loans: evidence from collateral use in small business. Journal ofFinancial Intermediation 15, 86–107.

Chan, Y.S., Kanatas, G., 1985. Asymmetric valuation and the role of collateral in loanagreements. Journal of Money, Credit and Banking 17, 85–95.

Che, J., Qian, Y., 1998. Institutional environment, community government, andcorporate governance: understanding China’s township-village enterprises.Journal of Law, Economics, and Organization 14, 1–23.

Chen, H., Chen, J.Z., Lobo, G.J., Wang, Y., 2010. Association between borrower andlender state ownership and accounting conservatism. Journal of AccountingResearch 48, 973–1014.

China Securities Regulatory Commission, 2007. China Securities and FuturesStatistical Yearbook (2007). Academia Press, Shanghai, China.

Cooley, T., Marimon, R., Quadrini, V., 2004. Aggregate consequences of limitedcontract enforceability. Journal of Political Economy 112, 817–847.

Cull, R., Xu, L.C., 2000. Bureaucrats, state banks, and the efficiency of creditallocation: the experience of Chinese state-owned enterprises. Journal ofComparative Economics 28, 1–31.

Cull, R., Xu, L.C., 2003. Who gets credit? The behavior of bureaucrats and state banksin allocating credit to Chinese state-owned enterprises. Journal of DevelopmentEconomics 71, 533–559.

Easley, D., Hvidkjaer, S., O’Hara, M., 2002. Is information risk a determinant of assetreturns? Journal of Finance 57, 2185–2221.

Fan, G., Wang, X.L., 2007. NERI Index of Marketization of China’s Provinces.Economics Science Press, Beijing, China.

Fama, E., Macbeth, J., 1973. Risk, return, and equilibrium: empirical tests. Journal ofPolitical Economy 81, 607–636.

Gale, D., Hellwig, M., 1985. Incentive-compatible debt contracts: the one-periodproblem. Review of Economic Studies 52, 647–663.

Gormley, T.A., Kim, B.H., Martin, X., 2012. Do firms adjust their timely lossrecognition in response to changes in the banking industry? Journal ofAccounting Research 50, 159–196.

Hainz, C., 2003. Bank competition and credit markets in transition economies.Journal of Comparative Economics 31, 223–245.

Harris, M., Raviv, A., 1991. The theory of capital structure. Journal of Finance 46,297–355.

Jia, C., 2009. The effect of ownership on the prudential behavior of banks – the caseof China. Journal of Banking and Finance 33, 77–87.

Jiménez, G., Salas, V., Saurina, J., 2006. Determinants of collateral. Journal ofFinancial Economics 81, 255–281.

Khan, M., Watts, R.L., 2009. Estimation and empirical properties of a firm-yearmeasure of accounting conservatism. Journal of Accounting and Economics 48,132–150.

Koivu, T., 2009. Has the Chinese economy become more sensitive to interest rates?Studying credit demand in China. China Economic Review 20, 455–470.

La Porta, R., Lopez-de-Silanes, F., Zamarripa, G., 2003. Related lending. QuarterlyJournal of Economics 118, 231–268.

LaFond, R., Watts, R.L., 2008. The information role of conservatism. The AccountingReview 83, 447–478.

Lin, H., 2011. Foreign bank entry and firms’ access to bank credit: evidence fromChina. Journal of Banking and Finance 35, 1000–1010.

Menkhoff, L., Neuberger, D., Suwanaporn, C., 2006. Collateral-based lending inemerging markets: evidence from Thailand. Journal of Banking and Finance 30,1–21.

Menkhoff, L., Neuberger, D., Rungruxsirivorn, O., 2012. Collateral and its substitutesin emerging markets’ lending. Journal of Banking and Finance 36, 817–834.

Myers, S.C., 1977. Determinants of corporate borrowing. Journal of FinancialEconomics 5, 145–175.

Nikolaev, V.V., 2010. Debt covenants and accounting conservatism. Journal ofAccounting Research 48, 137–175.

Petersen, M.A., 2009. Estimating standard errors in finance panel data sets:comparing approaches. Review of Financial Studies 22, 435–480.

Podpiera, R., 2006. Progress in China’s Banking Sector Reform: Has Bank BehaviorChanged?. Working Paper, International Monetary Fund.

Qian, J., Strahan, P.E., 2007. How laws and institutions shape financial contracts: thecase of bank loans. Journal of Finance 62, 2803–2834.

Rajan, R.G., Zingales, L., 1995. What do we know about capital structure? Someevidence from international data. Journal of Finance 50, 1421–1460.

Ryan, S.G., 2006. Identifying conditional conservatism. European AccountingReview 15, 511–525.

Smith, C., Warner, J., 1979. On financial contracting: an analysis of bond covenants.Journal of Financial Economics 7, 117–161.

Titman, S., Wessels, R., 1988. The determinants of capital structure choice. Journal ofFinance 43, 1–19.

Townsend, R.M., 1979. Optimal contracts and competitive markets with costly stateverification. Journal of Economic Theory 21, 265–293.

Watts, R.L., 2003. Conservatism in accounting part I: explanations and implications.Accounting Horizons 17, 207–221.

World Bank – People’s Bank of China (WB-PBOC), 2006. Secured Transactions Reformand Credit Market Development in China. China CITIC Press, Beijing, China.

Williamson, S.D., 1986. Costly monitoring, financial intermediation and equilibriumcredit rationing. Journal of Monetary Economics 18, 159–179.

Yang, J.Y., Qian, H., 2008. Survey evidence on commercial bank mortgage loans.Journal of Financial Research 333, 79–87 (in Chinese).

Zhang, J., 2008. The contracting benefits of accounting conservatism to lenders andborrowers. Journal of Accounting and Economics 45, 27–54.