Bank Relationships and the Value Relevance of the Income...

22
Journal of Business Finance & Accounting, 00, 1–22, XXX/XXX 2007, 0306-686X doi: 10.1111/j.1468-5957.2007.02023.x Bank Relationships and the Value Relevance of the Income Statement: Evidence from Income-Statement Conservatism Wooseok Choi Abstract: This study examines the effects of a firm’s debt financing decision on the informative- ness of the income statement. This study specifically examines the association between a firm’s bank dependence and the value relevance of the income statement by investigating the income- statement conservatism of firms with bank loans. Focusing on relatively small businesses, this study finds that income-statement conservatism, measured as timely loss recognition, is increasing in a firm’s bank dependence. This study also finds that the value relevance of the income statement is increasing in a firm’s bank dependence. The findings of this paper suggest that the usefulness of the income statement varies with a firm’s bank dependence, indicating that the value relevance of the income statement is a function of a firm’s debt financing decision. The findings further suggest that bank relationships affect the value relevance of the income statement through their influence on income-statement conservatism. Keywords: bank relationships, debt financing, income statement conservatism, value relevance 1. INTRODUCTION The choice between bank and public debt is a critical financing decision for a firm. In economics and finance, therefore, banks, bank loans, and bank relationships are widely studied. In accounting, however, little is known about the effects of a firm’s debt financing decision on financial reporting. This study examines these effects by investigating the value relevance of the income statement in the context of a firm’s The author is from Korea University, Seoul, Korea. He appreciates helpful comments and suggestions from Joseph Anthony, Marilyn Johnson, Jun-Koo Kang, Tom Linsmeier, Christian Mastilak, Vicent O’Connell, Kathy Petroni, and workshop participants at Michigan State University, Korea University, and the 2003 American Accounting Association Annual Meeting. He is also grateful to Martin Walker (editor) and an anonymous referee for their constructive suggestions. He acknowledges the financial support provided by Korea University, Michigan State University, and California State University at Los Angeles. (Paper received February 2005, revised version accepted December 2006) Address for correspondence: Wooseok Choi, Korea University Business School, Anam-Dong Seongbuk-Gu, Seoul, Korea 136-701. e-mail: [email protected] C 2007 The Author Journal compilation C 2007 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA. 1

Transcript of Bank Relationships and the Value Relevance of the Income...

Page 1: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

Journal of Business Finance & Accounting, 00, 1–22, XXX/XXX 2007, 0306-686Xdoi: 10.1111/j.1468-5957.2007.02023.x

Bank Relationships and the ValueRelevance of the Income Statement:

Evidence from Income-StatementConservatism

Wooseok Choi∗

Abstract: This study examines the effects of a firm’s debt financing decision on the informative-ness of the income statement. This study specifically examines the association between a firm’sbank dependence and the value relevance of the income statement by investigating the income-statement conservatism of firms with bank loans. Focusing on relatively small businesses, this studyfinds that income-statement conservatism, measured as timely loss recognition, is increasing in afirm’s bank dependence. This study also finds that the value relevance of the income statement isincreasing in a firm’s bank dependence. The findings of this paper suggest that the usefulness ofthe income statement varies with a firm’s bank dependence, indicating that the value relevanceof the income statement is a function of a firm’s debt financing decision. The findings furthersuggest that bank relationships affect the value relevance of the income statement through theirinfluence on income-statement conservatism.

Keywords: bank relationships, debt financing, income statement conservatism, value relevance

1. INTRODUCTION

The choice between bank and public debt is a critical financing decision for a firm.In economics and finance, therefore, banks, bank loans, and bank relationships arewidely studied. In accounting, however, little is known about the effects of a firm’sdebt financing decision on financial reporting. This study examines these effects byinvestigating the value relevance of the income statement in the context of a firm’s

∗The author is from Korea University, Seoul, Korea. He appreciates helpful comments and suggestions fromJoseph Anthony, Marilyn Johnson, Jun-Koo Kang, Tom Linsmeier, Christian Mastilak, Vicent O’Connell,Kathy Petroni, and workshop participants at Michigan State University, Korea University, and the 2003American Accounting Association Annual Meeting. He is also grateful to Martin Walker (editor) and ananonymous referee for their constructive suggestions. He acknowledges the financial support provided byKorea University, Michigan State University, and California State University at Los Angeles. (Paper receivedFebruary 2005, revised version accepted December 2006)

Address for correspondence: Wooseok Choi, Korea University Business School, Anam-Dong Seongbuk-Gu,Seoul, Korea 136-701.e-mail: [email protected]

C© 2007 The AuthorJournal compilation C© 2007 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UKand 350 Main Street, Malden, MA 02148, USA. 1

Page 2: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

2 CHOI

debt financing decision. Specifically, this study examines the association between afirm’s bank dependence, income-statement conservatism, and the usefulness of theincome statement.

Prior academic and anecdotal evidence suggest that lenders such as banks preferaccounting conservatism in the presence of information asymmetries to reduce (1)the risk that borrowers’ financial positions are overstated and (2) an agency conflictbetween bondholders and shareholders (e.g., FASB, SFAC No. 2, 1980; Leftwich, 1983;Ahmed et al., 2002; Holthausen and Watts 2001; and Watts, 2003a and 2003b). However,few studies provide empirical evidence about banks’ preference for accountingconservatism. One important way to measure accounting conservatism is the timelinessof economic loss recognition (Basu, 1997; and Ball and Shivakumar, 2005). Timely lossrecognition is particularly important in debt contracts because it affects the efficiency ofdebt agreements that utilize financial statement variables (Ball and Shivakumar, 2005).Ball and Shivakumar specifically mention that ‘timely income statement incorporationof economic losses more quickly transfer decision rights from loss-making managers tolenders.’ Considering the wide use of income-statement variables in debt agreements, itis highly likely that banks prefer income-statement conservatism as measured by timelyloss recognition. Therefore, it is expected that banks prefer timelier loss recognitionand thus provide borrowers with incentives to maintain a high level of income statementconservatism.1 Accordingly, the first hypothesis specifically examines the associationbetween a firm’s bank dependence and timely loss recognition, predicting that thetimeliness of economic loss recognition increases when a firm’s bank dependenceincreases.

Prior research including Ball and Shivakumar (2005) also states that timely lossrecognition is related to the concept of value relevance, implying a positive associationbetween timely loss recognition and value relevance. In the case of debt contracts,timelier income statement recognition means that the income statement provides moreuseful information in loan agreements, implying higher value relevance of the incomestatement. As predicted in the first hypothesis, if a firm has a higher level of bank loans,the firm is likely to have a higher level of timely loss recognition. Therefore, the secondhypothesis predicts that the higher a firm’s bank dependence, the more useful theincome statement is in explaining a firm’s value because of the higher level of timelyloss recognition.

The second hypothesis is also related to the traditional role of the incomestatement. The income statement provides information about a firm’s abnormalearnings opportunities while the balance sheet provides information on liquidation orabandonment value (Burgstahler and Dichev, 1997; and Barth, Beaver and Landsman,1998). The value-relevance literature indicates that investors are more interested in afirm’s growth potential, information from the income statement, when default risk islower. Prior research shows how a significant relationship with a bank helps a firm toreduce its probability of default by: (a) increasing the ease of renegotiation (Gilson,John and Lang, 1990), (b) lowering the cost of monitoring (Diamond, 1984; andFama, 1985), and (c) providing liquidity during periods of economic stress (Kashyap,Rajan and Stein, 2002; and Saidenberg and Strahan, 1999). Collectively, the bankdependence research suggests that the uniqueness of banks and bank loans enables a

1 The incentives might include higher loan amounts, lower interest rates, lower contracting costs, andincreased renewal capacities.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 3: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

BANK RELATIONSHIPS AND THE INCOME STATEMENT 3

firm to continue its business with less worry of default risk, making income statementinformation more important. This argument is consistent with the second hypothesisthat when a firm’s bank dependence is high, the income statement will be more value-relevant than when a firm’s bank dependence is low.

The analysis in this paper is based on 3,992 firm-year observations from publiclytraded, non-regulated Mid-Cap companies (i.e., small companies with sales between 1million and 500 million dollars) during 1988–2004. This paper focuses on relativelysmall firms because small firms obtain most of their external funding from financialintermediaries, primarily commercial banks, unlike large firms that tend to relymore heavily on public capital markets. Therefore, the firm-creditor relationships areexpected to be especially important for small firms (Diamond, 1991; and Core, Wolkenand Woodburn, 1996). Bank dependence is measured as the dollar value of the bankloan from the lead bank divided by the total assets of the borrower.

As hypothesized, the results indicate that the timeliness of economic loss recognitionincreases as a firm’s bank dependence increases. The results also show that the valuerelevance of the income statement increases as a firm’s bank dependence increases.The results hold even after formally controlling for a firm’s default risk. This suggeststhat bank relationships affect the value relevance of the income statement in a wayother than just default risk. Overall, the empirical results provide strong support forthe argument that the conservatism and usefulness of the income statement varies witha firm’s bank dependence.

A possible alternative explanation is the bank’s ex-ante screening of borrowers. Inthis case, bank dependence might proxy for firm-specific characteristics. The mostcommon screening device used by banks is collateral (Stiglitz and Weiss, 1981; andBester, 1985). Although recent bank competition makes bank screening more difficult(Shaffer, 1998), the sample firms’ collateral levels are further analyzed to explore thevalidity of this alternative explanation. The mean and median values of four proxies forcollateral are significantly greater for the low bank-dependent firms than for the highbank-dependent firms, which is inconsistent with ex-ante screening. Two profitabilitymeasures are also tested to examine the possibility of ex-ante screening based on theprofitability of a firm. Similar to collateral measures, the results show that the meanand median values of the profitability measures for the low bank-dependent firms aresignificantly greater than those for the high bank-dependent firms. Overall, the resultsfrom the tests for the alternative explanation show that ex-ante screening does not drivethe findings of this paper.

The primary contribution of this paper is to introduce the role of the bankrelationship into the accounting literature. Although bank relationships are widelystudied in economics and finance literature, to my knowledge there is no study on bankrelationships in the accounting literature. Given the fact that a firm’s debt financingdecision is critical and that banks and bank loans play a unique role in financingbusinesses, it is important to understand the effect of the firm-bank relationship onfinancial reporting. This study provides evidence that the conservatism and valuerelevance of the income statement is associated with a firm’s bank dependence.

This study also adds to the literature that examines accounting conservatism. Thereis theoretical evidence of a lender’s preference for accounting conservatism, but fewstudies provide empirical evidence. This paper fills this gap by providing empiricalevidence regarding the association between firm-bank relationships and income-statement conservatism.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 4: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

4 CHOI

Finally, unlike prior studies that generally focus on the importance of the balancesheet, this study finds a situation under which the income statement plays an importantrole in explaining firm value.

The paper proceeds as follows: Section 2 motivates and develops the hypotheses.Section 3 describes the sample and presents the research design and measures. Section 4presents empirical results. Section 5 examines the validity of an alternative explanationfor the results. Section 6 summarizes and concludes the paper.

2. HYPOTHESIS DEVELOPMENT

(i) Bank Loans and Accounting Conservatism

For banks, borrowers’ abilities to generate future cash flows are of great concernbecause of their abilities’ influence on credit risk. Credit risk is the risk to a bank’searnings and capital that a borrower will fail to repay the loan principal and interestas agreed. It is the primary cause of bank failure and the most visible risk facing bankmanagers (Fraser, Gup and Kolari, 2001). Therefore, banks have incentives to reducethe risk that borrowers’ financial positions are overstated, thereby reducing credit risk.With respect to financial reporting, banks prefer accounting conservatism to reducethe risk that their assessment of a borrower’s ability to generate future cash flows isoverstated due to overstatement of the borrower’s financial position.

Consistent with this notion, prior academic and anecdotal evidence suggest thatlenders prefer conservative accounting in the presence of information asymmetries.For example, Watts (2003a and 2003b) suggests that lenders such as banks areconcerned with the lower bound of the earnings and new asset distributions, and usethe lower bound measures during the loan period to monitor the borrower’s abilityto pay. He further suggests that conservatism can be used to address moral hazard,constraining management’s opportunistic behavior in financial reporting. Leftwich(1983), Ahmed et al. (2002) and Watts (2003a) argue that accounting conservatismplays an important role in efficient debt contracting. Finally, FASB’s SFAC No. 2 (1980)states that conservatism in financial reporting is desirable to bankers because a higherdegree of conservatism provides a greater margin of safety for the assets that serve asloan security.

Lenders also prefer accounting conservatism to reduce an agency conflict betweenbondholders and shareholders. For example, it is likely that firms with large amounts ofbank debt have high bondholder-shareholder conflicts and prefer more conservativefinancial reporting. Prior research including Ahmed et al. (2002) confirms this view bydemonstrating that conservatism can be used as a mechanism for resolving this type ofagency conflict.

Although prior studies on the use of accounting conservatism in debt contractingfocus on public debt, the same argument can be used in private bank debt as well.First, as Holthausen and Watts (2001) stated, SFAC No. 2 attributes the development ofconservatism to bankers and other lenders, suggesting that banks are concerned withaccounting conservatism in their lending agreements as are other lenders. Second, oneof the most important reasons why lenders prefer accounting conservatism is that itmakes debt covenants more binding. Recent evidence, including Dichev and Skinner(2002), finds that private lenders set debt covenants tightly and use them as a screeningdevice, suggesting that accounting conservatism can be importantly used in private debt

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 5: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

BANK RELATIONSHIPS AND THE INCOME STATEMENT 5

contracting. Moreover, syndicated private loans generally use more debt covenants thanpublic loans. Most bank loans are syndicated loans, further implying that accountingconservatism is an important concern to banks.

In summary, prior evidence on future cash flows, credit risk, and accountingconservatism suggests that banks have significant credit risk due to the risk ofoverstatement and, therefore, prefer conservatism in financial reporting for thepurpose of effective credit risk management.

(ii) Timely Loss Recognition and Bank Relationships- Hypothesis 1

One important way to measure accounting conservatism is the timeliness of economicloss recognition (Basu, 1997; and Ball and Shivakumar, 2005). Basu (1997) interpretsconservatism as an asymmetric verification requirement for recognizing good versusbad news in financial statements, suggesting that earnings reflect bad news more quicklythan good news. He states that debt holders demand more timely information aboutbad news that might adversely affect future cash flows because they have an asymmetricloss function. Ball and Shivakumar (2005) state that accounting income is a key factorfor evaluating financial reporting in general, suggesting that timely income-statementrecognition implies timely revision of all financial statement variables. With respectto debt agreements, they specifically suggest that timely loss recognition is related tothe efficiency of the debt agreement by affecting both ex ante loan pricing and ex postviolation of covenants based on financial statement variables. Ball and Shivakumarfurther mention that ‘timely income statement incorporation of economic losses morequickly transfer decision rights from loss-making managers to lenders.’

Therefore, as suggested by bank loans, accounting conservatism, and timelyloss recognition literature, it is highly likely that banks prefer income-statementconservatism as measured by timely loss recognition and thus provide borrowers withincentives to maintain a high level of income-statement conservatism.2 Accordingly,the first hypothesis of this paper is as follows (stated in alternative form):

H1: Ceteris paribus, the timeliness of economic loss recognition is increasing in a firm’sbank dependence.

(iii) Income-statement Conservatism, Bank Relationships, and the Value Relevanceof the Income Statement – Hypothesis 2

Prior research including Ball and Shivakumar (2005) states that timely loss recognitionis related to the concept of value relevance. It suggests that timelier recognition impliesa higher correlation of accounting numbers with market values, suggesting a positiveassociation between timely loss recognition and value relevance. In other words, timelierrecognition implies higher financial reporting quality.3 In the case of debt contracts,

2 I am grateful to the referee for suggesting the link between borrowings and income-statement conservatismfor refining the link between bank relationships and the value relevance of the income statement.3 Prior studies generally have two views about the association between accounting conservatism and valuerelevance. Consistent with the argument in this paper, prior studies, such as Watts (2003a) and Ball andShivakumar (2005), suggest a positive relation between accounting conservatism and value relevance. Onthe other hand, other studies including Lev and Zarowin (1999) suggest a negative relation. However, thispaper does not focus on issues related to whether or not accounting conservatism is good or bad. It focuseson the role of accounting conservatism in debt contracts and its relation to the usefulness of the incomestatement in loan agreements.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 6: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

6 CHOI

therefore, a timelier income statement incorporation of economic losses means that theincome statement provides more useful information in loan agreements, implying thehigher value relevance of the income statement. As predicted in the first hypothesis, ifa firm has a higher level of bank loans, the firm is likely to have a higher level of timelyloss recognition. Therefore, it can be predicted that if a firm has more bank loans,the income statement of the firm provides more useful (i.e., more value relevant)information in explaining firm value because of more timely loss recognition.

This prediction is also related to the traditional role of the income statement. Forexample, using an adaptation option model, Burgstahler and Dichev (1997) show thatinformation from the income statement (i.e., earnings) is more important when thefirm’s current activities are successful since the information provides a measure of theperformance of a business. Barth et al. (1998) investigate the relative valuation rolesof equity book value and net income as a function of financial health. They argue thatthe income statement increases in importance as the probability of default and theimportance of assessing liquidation value decrease. Their results imply that the incomestatement’s primary role is to provide information about a firm’s abnormal earningsopportunities. In general, the value-relevance literature indicates that investors aremore interested in a firm’s growth potential, information from the income statement,when default risk is lower.

Related to default risk, the banking literature provides several explanations as towhy bank dependence affects a firm’s default risk. Prior studies show that the bankrelationship reduces information asymmetries and the firm’s cost of financial distressby (a) increasing the ease of renegotiation (Gilson, John and Lang, 1990), (b) loweringthe cost of monitoring (Diamond, 1984; and Fama, 1985);4 and (c) providing liquidityduring periods of economic stress (Kashyap, Rajan and Stein, 2002; and Saidenbergand Strahan, 1999). Collectively, the bank dependence research suggests that a firmwith a higher level of bank loan has a lower level of default risk, making the firm’sgrowth potential, information from the income statement, more informative. Thisargument is consistent with the prediction that when bank dependence is high, theincome statement will be more value relevant than when bank dependence is low.

Therefore, as suggested by income-statement conservatism, bank dependence, andvalue relevance literature, the second hypothesis of this paper is as follows (stated inalternative form):

H2: The higher a firm’s bank dependence, the more useful the income statement isin explaining firm value.

3. SAMPLE AND RESEARCH DESIGN

(i) Sample and Descriptive Statistics

The sample comprises 3,992 firm-year observations from publicly traded Mid-Capcompanies (i.e., small companies with sales between 1 million and 500 million dollars),

4 Fama (1985) specifically argues that a bank loan could be considered a type of inside debt which issimilar to internally generated funds, allowing a firm to avoid the underinvestment problem associatedwith information asymmetries. The underinvestment problem is caused by debt overhang, the inability ofa company with profitable investment opportunities to finance them because it has excessive levels of debtrelative to its assets (Milgrom and Roberts, 1992).

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 7: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

BANK RELATIONSHIPS AND THE INCOME STATEMENT 7

which meet the following requirements (see the Appendix for a further discussion ofthe sample selection):

1. The largest lender is a bank.

2. The firm (i.e., the borrower) is a non-financial institution and non-utilitycompany.

3. The bank (i.e., the lender) is a lead lender and has the largest share of the bankloan.

4. Financial data is available from the Compustat database during the loan period.

Compared to large firms, small companies are less likely to be followed by analysts andmonitored by rating agencies or the financial press, implying that they rarely borrowin the public capital markets. Therefore, small firms obtain most of their externalfunding from financial intermediaries (i.e., commercial banks) and, in turn, firm-creditor relationships are expected to be especially important in small firms (Diamond,1991; and Core, Wolken and Woodburn, 1996). For similar reasons, bank relationshipliterature often focuses on small businesses (Petersen and Rajan, 1994; and Berger andUdell, 1995).

Due to missing values, the sample has an unbalanced panel data structure. Although3,992 firm-year observations are selected using the above criteria, the actual numberof observations used in each analysis varies, depending on the model specification(e.g., pooled OLS estimation, lag variables, Altman-Z score, etc.) and missing firm-yearobservations. If the largest lender is not a bank, then the firm is dropped.5

Information on the borrower-lender relationship is collected from the DealScandatabase for the years 1988–2004. Constructed by the Loan Pricing Corporation (LPC),DealScan contains deal terms and conditions on over 72,000 loans since 1988. Deals inthe database are often comprised of different ‘facilities’. For example, a deal mightinclude a term loan facility and a line of credit facility. Similar to prior research, eachfacility is treated as one observation (see the Appendix for further discussion of theDealScan database).

Panel A of Table 1 reports descriptive statistics for all firm-year observations, andPanel B of Table 1 reports descriptive statistics for high and low bank-dependentfirms. Firms are divided into high and low bank-dependent groups based on theirbank dependence. Consistent with bank relationship literature, bank dependence ismeasured as the dollar value of the bank loan from the lead bank divided by the totalassets of the borrower (Hoshi, Kashyap and Scharfstein, 1990; Petersen and Rajan,1994; and Kang, Shivdasani and Yamada, 2000). This measure captures the closeness ofa firm to its main lender and the concentration of a firm’s borrowing across blockholderlenders. It also measures how important a bank loan is to a firm.6

5 The focus of this study is on firm-bank relationships. Therefore, this study drops all observations when thelargest lender is not a bank. By using this criterion, this study achieves a homogenous sample which is idealfrom the perspective of addressing the core research questions outlined earlier.6 There are several reasons why the bank loan is scaled by the borrower’s total assets rather than thebank’s total assets. First, this scaler measures the extent to which borrowers are influenced by banks,which is consistent with the research questions of this paper. Second, for a lead bank, reputation is veryimportant. Therefore, a lead bank will perform monitoring activities to maintain a certain level of accountingconservatism in a borrower’s income statement, regardless of the relative size of the bank loan to its total

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 8: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

8 CHOI

Table 1Sample Description

Panel A: Descriptive Statistics for All Sample Firm-year Observations25% 75%

Median Mean Percentile Percentile N

Bank loans 20.00 43.80 7.91 55.00 3,922Maturity 15.00 31.60 6.00 34.01 3,921Market value of equity 86.94 206.96 30.50 237.38 3,833Book value of equity 50.74 74.44 19.95 111.14 3,921Net income 27.73 20.93 −14.31 10.68 3,914AltmanZ score 4.45 92.26 2.53 11.04 3,381Total Assets 111.47 155.15 46.90 243.81 3,922Bank share (%) 9.42 13.17 5.07 16.69 3,922

Panel B: Descriptive Statistics for High vs. Low Bank-dependent Firm-year ObservationsLow Bank-dependent High Bank-dependent

Median Mean Median Mean

Bank loans 10.00 29.50 18.50 36.90Maturity 37.00 45.00 36.00 42.00Market value of equity 139.35 292.64 58.29 120.78Book value of equity 76.42 95.55 35.80 53.32Net income 40.43 27.83 19.15 14.01AltmanZ score 4.66 95.48 4.27 89.06Total Assets 177.72 200.12 74.47 110.14Bank share (%) 5.07 5.00 16.69 21.34

Notes:Bank loans, market value of equity, book value of equity, net income, and total assets representdollars in millions. Maturity is in months. The high and low bank-dependent groups are classified based ona firm’s median bank share.Bank loans are total bank loans (facility amounts) in dollars in each firm-year. Maturity is a maturity ofthe bank loan. The Altman Z Score is a measure of a firm’s financial health, calculated as [1.2 (workingcapital/total assets) + 1.4 (retained earnings/total assets) + 3.3 (earnings before interest and taxes/totalassets) + 0.6 (market value of equity/book value of liabilities) + 1.0 (sales/total assets)]. Bank share is thedollar value of the bank loan from the lead bank divided by the total assets of the borrower.

The maturity is similar between high and low bank-dependent firms. The Altman-Zscores for both groups are in the safe zone.7 The median and mean values of marketvalue of equity, book value of equity, net income, and total assets for the low bank-dependent firms are larger than those values for the high bank-dependent firms.The median and mean values generally show that the bank’s screening process and,therefore, a firm’s specific characteristics do not affect the findings of this study. Thisissue will be further explored with extensive sensitivity analyses in Section 5.

Panel A of Table 2 presents the industry composition of the sample. Bothhigh and low bank-dependent groups have similar industry distributions to theCompustat dataset. Panel B of Table 2 reports the exchange listings of the sample

assets. Finally, transaction costs of switching to another bank are associated with the relative size of the bankloan to the firms’ total assets.7 The interpretation of the Altman Z score is as follows: (i) Z > 2.99 means the firm is safe, (ii) Z < 1.80means the firm is financially distressed, and(iii) 1.80 < Z < 2.99 means the firm is in the ‘grey’ zone.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 9: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

BANK RELATIONSHIPS AND THE INCOME STATEMENT 9

Tab

le2

Dis

trib

utio

nof

Firm

-yea

rO

bser

vatio

ns

Pan

elA

:Dis

trib

utio

nof

Firm

-yea

rO

bser

vati

ons

byIn

dust

ryA

llSa

mpl

eL

owB

ank-

depe

nden

tH

igh

Ban

k-de

pend

ent

Com

pust

atYe

ar20

01

Indu

stry

Des

crip

tion

N%

N%

N%

%

Agr

icul

ture

160.

413

0.16

130.

670.

51M

inin

g21

05.

3776

3.88

134

6.88

7.00

Con

stru

ctio

n41

1.05

301.

5311

0.56

1.49

Man

ufac

turi

ng2,

085

53.3

11,

089

55.5

099

651

.10

43.3

2W

hole

sale

trad

e24

86.

3410

55.

3514

37.

344.

82R

etai

ltra

de43

111

.02

200

10.1

923

111

.85

7.65

Serv

ices

880

22.5

045

923

.39

421

21.6

022

.67

Tota

l3,

911

100.

001,

962

100.

001,

949

100.

0010

0.00

Pan

elB

:Dis

trib

utio

nof

Firm

-yea

rO

bser

vati

ons

byE

xcha

nge

Lis

ting

Stoc

kEx

chan

geH

igh

Ban

k-de

pend

ent

Low

Ban

k-de

pend

ent

Full

Sam

ple

New

York

Stoc

kE

xcha

nge

467

301

768

Am

eric

anSt

ock

Exc

hang

e20

414

534

9N

ASD

AQ

1,11

71,

265

2,38

2O

ver-

the-

coun

ter

171

249

420

Oth

ers

(reg

iona

l)3

03

Tota

l1,

962

1,96

03,

922

Not

es:

Indu

stry

mem

bers

hip

isde

term

ined

bytw

o-di

git

Stat

istic

alIn

dust

rial

Cla

ssifi

catio

n(S

IC)

code

s.O

utof

the

tota

l3,

922

firm

-yea

rob

serv

atio

ns,

11hi

ghba

nk-

depe

nden

tfir

m-y

ear

obse

rvat

ions

are

omitt

edin

the

tabl

ebe

caus

eth

eyar

ecl

assi

fied

asno

n-op

erat

ing

esta

blis

hmen

ts(S

IC99

95).

Sam

ple

firm

sar

edi

vide

din

totw

ogr

oups

,hig

han

dlo

wba

nk-d

epen

dent

,bas

edon

afir

m’s

bank

shar

e.E

xcha

nge

listin

gis

dete

rmin

edby

the

Com

pust

atex

chan

gelis

ting

code

(ZL

IST

).

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 10: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

10 CHOI

firms. There is little difference in the exchange listing between high and lowbank-dependent firms. The NASDAQ has the largest number of sample firm-yearobservations, followed by the New York Stock Exchange and the American StockExchange.

Panel A of Table 3 presents the Pearson correlations between the independentvariables used in equation (1). The largest variance inflation factor (VIF) value amongall independent variables is 5.06. The smallest tolerance (1/VIF) is 0.20. Both VIFand 1/VIF values indicate that there is no severe multicollinearity problem among theindependent variables in equation (1).

Panel B of Table 3 presents the Pearson correlations between the independentvariables used in equation (2). The largest variance inflation factor (VIF) value amongall independent variables is 7.83. The smallest tolerance (1/VIF) is 0.13. Similar toequation (1), both VIF and 1/VIF values indicate that there is no severe multicollinearityproblem among the independent variables in equation (2).

Finally, Panel C of Table 3 presents the Pearson correlations between the inde-pendent variables used in equation (3). Two out of seven VIF values are larger than10, but smaller than 20. This indicates that there might be a slight multicollinearityconcern. However, the mean VIF (5.64) is less than 6. In addition, the conditionalnumber (9.58) is less than 15.8 Both the mean VIF and the conditional number indicatethat there is a little multicollinearity problem among the independent variables inequation (3).

(ii) Model Specification – Hypothesis 1

The first hypothesis examines the relation between timely loss recognition andbank dependence. Similar to Basu (1997) and Ball and Shivakumar (2005), thefirst hypothesis is tested by estimating the coefficients in the following regressionmodel:

�NIi t = α0 + α1 D�NIi t−1 + α�2 NIi t−1 + α3(D�NIi t−1 ×� NIi t−1)

+ α4SHAREi + α5(SHAREi × D�NIi t−1) + α6(SHAREi ×� NIi t−1)

+ α7(SHAREi × D�NIi t−1 ×� NIi t−1) + εi (1)

where:i denotes firms, t denotes years for 1988–2004, �NIt is the change in earnings from yeart−1 to year t, standardized by total assets at the end of year t−1, �NIt −1 is the changein earnings from year t−2 to year t−1, standardized by total assets at the end of yeart−2, D�NI is an indicator variable that equals 1 if �NI is negative and 0 otherwise, andSHARE is the bank dependence measured as the dollar value of the bank loan fromthe lead bank divided by the total assets of the borrower.9

As Basu (1997) stated, α2 + α3 < 0 indicates timely recognition of economics losses,implying that they are recognized as transitory income decreases and hence reversed.

8 The conditional number is another method to test multicollinearity among independent variables. It equalsthe square root of the largest eigenvalue divided by the smallest eigenvalue. In general, if the conditionalnumber is larger than 15, multicollinearity is a concern. If it is greater than 30, multicollinearity is a seriousconcern.9 Standard errors in equations (1), (2), and (3) are computed based on clustering to mitigate potentialcross- and serial-correlations in the dependent variables.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 11: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

BANK RELATIONSHIPS AND THE INCOME STATEMENT 11T

able

3Pe

arso

nC

orre

latio

nsB

etw

een

Inde

pend

entV

aria

bles

(Sig

nific

ance

leve

lsin

pare

nthe

ses)

Pan

elA

:Equ

atio

n1

–In

depe

nden

tVar

iabl

esD

�N

I it−

1�

NI i

t−1

D�

NI i

t−1

×SH

AR

Ei

SHA

RE

SHA

RE

SHA

RE

D�

NI i

t−1

�

NI i

t−1

D�

NI i

t−1

�N

I it−

1�

NI i

t−1

D�

NI i

t−1

1.00

�N

I it−

1−0

.50

1.00

(0.0

0)

D�

NI i

t−1

�

NI i

t−1

−0.4

70.

661.

00(0

.00)

(0.0

0)

SHA

RE

i0.

030.

03−0

.02

1.00

(0.3

1)(0

.23)

(0.3

7)

SHA

RE

D�

NI i

t−1

0.62

−0.3

2−0

.30

0.51

1.00

(0.0

0)(0

.00)

(0.0

0)(0

.00)

SHA

RE

�N

I it−

1−0

.40

0.70

0.40

0.13

−0.4

71.

00(0

.00)

(0.0

0)(0

.00)

(0.0

0)(0

.00)

SHA

RE

D�

NI i

t−1

�

NI i

t−1

−0.3

90.

430.

61−0

.37

−0.6

90.

581.

00(0

.00)

(0.2

8)(0

.00)

(0.0

0)(0

.00)

(0.0

0)

Pan

elB

:Equ

atio

n2

–In

depe

nden

tVar

iabl

esD

Rit

1R

itR

it×

DR

itSH

AR

Ei

SHA

RE

DR

itSH

AR

Ei×

Rit

SHA

RE

DR

it×

Rit

DR

it1

1.00

Rit

−0.6

91.

00(0

.00)

Rit

×D

Rit

−0.7

50.

661.

00(0

.00)

(0.0

0)

SHA

RE

i0.

03−0

.05

−0.0

31.

00(0

.17)

(0.0

3)(0

.19)

SHA

RE

DR

it0.

62−0

.43

−0.4

70.

561.

00(0

.00)

(0.0

0)(0

.00)

(0.0

0)

SHA

RE

Rit

−0.5

10.

710.

500.

12−0

.51

1.00

(0.0

0)(0

.00)

(0.0

0)(0

.00)

(0.0

0)

SHA

RE

DR

it×

Rit

−0.4

80.

430.

66−0

.44

−0.7

90.

551.

00(0

.00)

(0.0

0)(0

.00)

(0.0

0)(0

.00)

(0.0

0)

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 12: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

12 CHOI

Tab

le3

(Con

tinu

ed)

Pan

elC

:Equ

atio

n3

–In

depe

nden

tVar

iabl

esB

VE

itB

VE

×D

NE

GB

itN

I itN

DN

EG

I itB

VE

×SH

AR

Eit

NI×

SHA

RE

itA

LTZ

it

BV

Eit

1.00

BV

DN

EG

Bit

0.30

1.00

(0.0

0)N

I it0.

27−0

.13

1.00

(0.0

0)(0

.00)

NI×

DN

EG

I it0.

10−0

.07

0.94

1.00

(0.0

0)(0

.00)

(0.0

0)B

VE

×SH

AR

Eit

0.62

0.19

0.18

0.09

1.00

(0.0

0)(0

.00)

(0.0

0)(0

.00)

NI×

SHA

RE

it0.

25−0

.17

0.66

0.57

0.29

1.00

(0.2

8)(0

.00)

(0.0

0)(0

.00)

(0.0

0)A

LTZ

it0.

070.

030.

02−0

.00

0.03

0.02

1.00

(0.0

0)(0

.05)

(0.3

6)(0

.97)

(0.1

3)(0

.12)

Not

es:

Pane

lA:

ide

note

sfir

ms,

tde

note

sye

ars

for

1988

–200

4,�

NI t

−1is

chan

gein

earn

ings

from

year

t−2

toye

art−

1,st

anda

rdiz

edby

tota

lass

ets

aten

dof

year

t−2,

D�

NI

isan

indi

cato

rva

riab

leth

ateq

uals

1if

�N

Iis

nega

tive

and

0ot

herw

ise,

and

SHA

RE

isth

edo

llar

valu

eof

the

bank

loan

from

the

lead

bank

divi

ded

byth

eto

tala

sset

sof

the

borr

ower

.Pa

nelB

:i

deno

tes

firm

s,t

deno

tes

year

sfo

r19

88–2

004,

Ris

the

buy-

and-

hold

retu

rns

from

nine

mon

ths

befo

refis

caly

ear-

end

toth

ree

mon

ths

afte

rfis

caly

ear-

end,

DR

isan

indi

cato

rva

riab

leth

ateq

uals

1if

Ris

nega

tive,

and

0ot

herw

ise,

and

SHA

RE

isba

nkde

pend

ence

mea

sure

das

the

dolla

rva

lue

ofth

eba

nklo

anfr

omth

ele

adba

nkdi

vide

dby

the

tota

lass

ets

ofth

ebo

rrow

er.

Pane

lC:

iden

otes

com

pani

es,t

deno

tes

year

sfo

r19

88–2

004,

BV

Eis

book

valu

eof

equi

ty,N

Iis

neti

ncom

e,D

NE

GB

isan

indi

cato

rva

riab

leth

ateq

uals

1if

BV

Eis

nega

tive

and

0ot

herw

ise,

DN

EG

Iis

anin

dica

tor

vari

able

that

equa

ls1

ifN

Iis

nega

tive

and

0ot

herw

ise,

SHA

RE

isth

edo

llar

valu

eof

the

bank

loan

from

the

lead

bank

divi

ded

byth

eto

tala

sset

soft

hebo

rrow

er,a

ndA

LTZ

isth

eA

ltman

Zsc

ore

mea

sure

das

[1.2

(wor

king

capi

tal/

tota

lass

ets)

+1.

4(r

etai

ned

earn

ings

/tot

alas

sets

)+

3.3

(ear

ning

sbe

fore

inte

rest

and

taxe

s/to

tala

sset

s)+

0.6

(mar

ketv

alue

ofeq

uity

/boo

kva

lue

oflia

bilit

ies)

+1.

0(s

ales

/tot

alas

sets

)].

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 13: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

BANK RELATIONSHIPS AND THE INCOME STATEMENT 13

α3 < 0 implies that economic losses are recognized in a more timely manner thangains.10 The first hypothesis predicts that the coefficient on SHAREi × D�NIit −1 ×�NIit −1, α7, is negative, meaning that high bank-dependent firms are more likely torecognize economic loses in a timely manner than the low bank-dependent firms.Similar to Ball and Shivakumar (2005), there is no prediction on a difference in gainrecognition between high and low bank dependent firms, offering no prediction onα6.

Equation (1) measures accounting conservatism as the transitory component ofchanges in net income. It has a potential limitation because the transitory componentof changes in net income can arise either from noise that reverses over time or fromconservatism. Therefore, to mitigate this limitation and check the robustness of theresults from equation (1), the first hypothesis is re-examined by using an alternativemeasure of conservatism.11

The alternative measure is Basu’s (1997) reverse regression of earnings on stockreturns. The estimation model is as follows:

Xit/Pit−1 = α0 + α1DRi t + α2 Rit + α3(Rit × DRi t ) + α4SHAREi + α5(SHAREi × DRi t )

+ α6(SHAREi × Rit ) + α7(SHAREi × DRi t × Rit ) + εi t (2)

where:i denotes firms, t denotes years for 1988–2004, X is the earnings per share measuredas income before extraordinary items/(number of shares outstanding × stock splitadjustment factor), P is the price per share at the beginning of the fiscal year dividedby the stock split adjustment factor, R is the buy-and-hold returns from nine monthsbefore the fiscal year-end to three months after the fiscal year-end, DR is an indicatorvariable that equals 1 if R is negative, and 0 otherwise, and SHARE is bank dependencemeasured as the dollar value of the bank loan from the lead bank divided by the totalassets of the borrower.

Based on the sign of the returns, Basu divides the sample into a ‘good news’ group(i.e., firm-year observations with positive returns) and a ‘bad news’ group (i.e., firm-year observations with negative returns). He predicts that under conservatism, thecoefficient on R∗DR, α3, will be significantly positive, suggesting that the earnings-return relation is stronger during bad news periods. He also predicts that α0 and α2will be positive and significant. This paper further predicts that α7 will be positive andsignificant, indicating that the extent to which earnings reflects bad news more quicklythan good news is increasing in a firm’s bank dependence.

(iii) Model Specification – Hypothesis 2

The second hypothesis examines the association between bank dependence and theinformativeness of the income statement. It is tested by using a revised version of Barth

10 As for economic gains, timely recognition implies α2 < 0 while untimely recognition implies α2 >

0. Untimely recognition implies that economic gains are recognized as persistent positive components ofaccounting income that are unlikely to be reversed.11 I am grateful to the referee for indicating the potential limitation and suggesting the alternative measure.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 14: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

14 CHOI

et al. (1998). Specifically, the following regression model is tested:

MVEi t = α0 + α1BVEi t + α2(BVE × DNEGB)i t + α3NIi t + α4(NI × DNEGI)i t

+ α5(BVE × SHARE)i t + α6(NI × SHARE)i t + α7ALT Zi t

+ α8(BVE × ALT Z)i t + α9(NI × ALT Z)i t + εi t (3)

where:i denotes companies, t denotes years for 1988–2004, MVE is the market value of equity,BVE is the book value of equity, NI is the net income, DNEGB is an indicator variable thatequals 1 if BVE is negative and 0 otherwise, DNEGI is an indicator variable that equals1 if NI is negative and 0 otherwise, SHARE is bank dependence measured as the dollarvalue of the bank loan from the lead bank divided by the total assets of the borrower,and ALT Z is the Altman Z score measured as [1.2 (working capital/total assets) +1.4 (retained earnings/total assets) + 3.3 (earnings before interest and taxes/totalassets) + 0.6 (market value of equity/book value of liabilities) + 1.0 (sales/totalassets)].

The Altman Z score is included to formally control for the probability of default.The control for the probability of default is important for the following reason: Thispaper argues that bank relationships affect the value relevance of the income statementbecause of the impact of these relationships on income-statement conservatism. Thepaper further argues that the impact of bank relationships on the value relevanceof the income statement is a function of the traditional importance of the incomestatement as in the context of predicting default risk. Therefore, to directly examineonly the effect of income-statement conservatism on value relevance, this paperformally controls for the probability of default using the Altman Z score. Thenegative net income indicator is included since prior research suggests that lossfirms have different pricing multiples (Hayn, 1995; Barth et al., 1998; and Collinset al., 1999). The second hypothesis predicts that the coefficient on NI∗SHARE,α6, is positive since the importance of NI increases as a firm’s bank dependenceincreases.

4. EMPIRICAL RESULTS

The results of the tests for the first hypothesis are reported in Table 4 and Table 5.The tests focus on the association between income-statement conservatism and bankdependence. Table 4 reports results from equation (1). Specifically, it reports thefindings from the test of differences in timely loss recognition conditional on a firm’sbank dependence. The findings are consistent with this study’s main prediction.Consistent with the first hypothesis that the timeliness of economic loss recognitionincreases with a firm’s bank dependence, the coefficient on SHAREi × D�NIit −1 ×�NIit −1, −1.6246, is negative and statistically significant at less than 1% significancelevel (p-value < 0.01). As stated before, the current paper makes no predictions about adifference of gain recognition between high and low bank-dependent firms. Therefore,no prediction is provided for the coefficient on SHAREi × �NIit −1. Similar to Basu(1997) and Ball and Shivakumar (2005), the F -test for α2 + α3 shows that the sum ofthe coefficients is negative and statistically significant (F-statistic = 10.22 and p-value< 0.01). In addition, the F -test for α3 + α7 shows that the sum of the coefficientsis negative and statistically significant (F-statistic = 8.72 and p-value < 0.01). Both

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 15: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

BANK RELATIONSHIPS AND THE INCOME STATEMENT 15

Table 4Income Statement Conservatism and Bank Dependence

Regression of equation (1):�NIi t = α0 + α1 D�NIi t−1 + α�

2 NIi t−1 + α3(D�NIi t−1 ×� NIi t−1) + α4SHAREi + α5(SHAREi × D�NIi t−1)

+ α6(SHAREi ×� NIi t−1)+ α7(SHAREi × D�NIi t−1 ×� NIi t−1) + εi

Independent Variables Coefficient t-statistics

D�NIit −1 −0.0056 −0.56�NIit −1 −0.1435 −2.11∗∗

D�NIit −1 × �NIit −1 −0.0857 −0.90SHAREi −0.0106 −0.27SHAREi × D�NIit −1 −0.0380 −0.54SHAREi × �NIit −1 0.1762 0.53SHAREi × D�NIit −1 × �NIit −1 −1.6246 −2.66∗∗∗

R2 0.0870

N 1,558

Notes:∗, ∗∗ and ∗∗∗ indicate significance at the 1%, 5% and 10% levels, respectively, for two-tailed tests.The regression is winsorized at 1% on each side for �NIt and �NIt −1.i is a company index, and t is a year index. �NIt is the change in earnings from year t−1 to year t,standardized by total assets at end of year t−1. Other variables are defined in Table 3.

results indicate that for bank-dependent firms, economic losses are recognized in atimelier manner than gains as transitory income increases. Unlike Basu (1997) andBall and Shivakumar (2005), the coefficient on �NIit −1 is negative and statisticallysignificant, implying that for bank-dependent firms, economic gains are also recognizedin a timely fashion although the degree of timeliness is less than that for economiclosses.12

Table 5 reports results from equation (2) using an alternative measure of conser-vatism. The results are also consistent with the first hypothesis. The coefficient onSHAREi × DRit × Rit , 0.4868, is positive and statistically significant (p-value = 0.095),implying that the extent to which earnings reflects bad news more quickly than goodnews increases as a firm’s bank dependence increases. Similar to Basu (1997), thecoefficients on the intercept, R, and R∗DR variables are all significantly positive. Theseresults confirm the first hypothesis that the timeliness of economic loss recognition isincreasing in a firm’s bank dependence.13

The results from the test of the second hypothesis are shown in Table 6. Thecoefficient on NI × SHARE, 6.3487, is positive and statistically significant (p-value =0.019). Consistent with the second hypothesis, this result indicates that the value

12 One caution for this interpretation is that this result might be due to the limitation of the estimationmodel, equation (1), discussed in Section 3.13 In addition, Ball and Shivakumar’s (2006) non-linear regression of accruals on cash flows is examinedalthough it is not reported. The results are also consistent with the first hypothesis.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 16: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

16 CHOI

Table 5Income Statement Conservatism and Bank Dependence – Alternative Measure of

Conservatism

Regression of equation (2):Xit/Pit−1 = α0 + α1DRi t + α2 Rit + α3(Rit × DRi t ) + α4SHAREi + α5(SHAREi × DRi t )

+ α6(SHAREi × Rit ) + α7(SHAREi × DRi t × Rit ) + εi t

Independent Variables Coefficient t-statistics

Intercept 0.0436 4.63∗∗∗

DRit 0.0092 0.48Rit 0.0229 2.02∗∗

Rit × DRit 0.2131 4.79∗∗∗

SHAREi −0.0390 −0.48SHAREi × DRit 0.0023 0.02SHAREi × Rit −0.1181 −1.13SHAREi × DRit × Rit 0.4868 1.67∗

R2 0.1090N 2,290

Notes:∗∗∗, ∗∗ and ∗ indicate significance at the 1%, 5% and 10% levels, respectively, for two-tailed tests.The regression is winsorized at 1% on each side for Xit /Pit −1 and Rit .i is a company index, and t is a year index. X is the earnings per share measured as income beforeextraordinary items/(number of shares outstanding ∗ stock split adjustment factor) and P is the priceper share at the beginning of the fiscal year divided by a stock split adjustment factor. Other variables aredefined in Table 3.

relevance of NI increases as a firm’s bank dependence increases.14,15 The coefficientis positive and significant after controlling for default risk (i.e., Altman Z score) inthe estimation model, implying that bank relationships affect the value relevanceof the income statement in ways which extend beyond the impact of default risk.These findings are consistent with the intuition underlying the first hypothesis that theimpact of bank relationships on income-statement conservatism may be of sufficientimportance to simultaneously impact the value relevance of the income statement.Finally, similar to prior studies, the coefficients on BVE and NI are both positive andsignificant, indicating that the balance sheet and income statement generally playimportant roles in explaining firm value.

In summary, the results from the test of the second hypothesis show that the highera firm’s bank dependence, the greater the weight placed on the income statement in

14 Although there is no prediction for the coefficient on BVE × SHARE, the negative and significantcoefficient on BVE × SHARE might indicate that the importance of BVE decreases as a firm’s bankdependence increases. In addition, consistent with prior studies, the coefficient for negative net incomeobservations (NI

∗DNEG) is negative and statistically significant, reflecting the transitory nature of losses.

Barth et al. (1998) specifically state that the coefficient for negative net income observations is negativebecause ‘the limited liability feature of common shares suggests each incremental dollar of loss has adiminishing relation with share prices.’15 The coefficients on BVE and NI are similar to prior studies in terms of magnitude and significance. Thecoefficient on the book value of equity in prior studies ranges from 0.16 to 0.92 and that on earnings rangesfrom 3.44 – 9.31.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 17: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

BANK RELATIONSHIPS AND THE INCOME STATEMENT 17

Table 6Bank Dependence and the Value Relevance of the Income Statement

Regression of equation (3):MVEi t = α0 + α1BVEi t + α2(BVE × DNEGB)i t + α3NIi t + α4(NI × DNEGI)i t

+ α5(BVE × SHARE)i t + α6(NI × SHARE)i t + α7ALT Zi t

+ α8(BVE × ALT Z)i t + α9(NI × ALT Z)i t + εi t

Independent Variables Coefficient t-statistics

BVEit 1.4657 3.22∗∗∗

BVE × DNEGBit 0.8506 1.87∗

NIit 12.0431 7.12∗∗∗

NI × DNEGIit −13.7649 −8.17∗∗∗

BVE × SHAREit −8.2941 −3.04∗∗∗

NI × SHAREit 6.3487 2.34∗∗∗

ALT Zit 0.0305 1.02BVE × ALT Zit 0.0003 0.68NI × ALT Zit −0.0003 −0.86

R2 0.3288N 3,381

Notes:∗∗∗, ∗∗ and ∗ indicate significance at the 1%, 5% and 10% levels, respectively, for two-tailed tests.i is a company index, and t is a year index. MVE is market value of equity. Other variables are defined inTable 3.

explaining firm value. Overall, these findings suggest that the usefulness of the incomestatement varies with a firm’s bank dependence.

5. AN ALTERNATIVE EXPLANATION – EX-ANTE SCREENING

A possible alternative explanation for the results reported in Section 4 is the bank’sex-ante screening of borrowers. In this case, bank dependence might proxy for the firmcharacteristics used by banks to screen loan applicants. The most common screeningdevice used by banks is collateral (Stiglitz and Weiss, 1981; and Bester, 1985). Commonlyused collateral includes real property, equipment, inventories, accounts receivable,and securities and savings accounts. Since a collateral’s value varies by items (e.g.,90 percent of the value of high-graded municipal bonds is counted as collateral butonly 75 percent of the value of major common stock is included (Freixas and Rochet,1997)), four proxies for a firm’s collateral value are used: (i) receivables, (ii) inventories,(iii) property, plant, and equipment, and (iv) book value of equity. Then, whetherthe collateral value is different between high bank-dependent firms and low bank-dependent firms is examined. If the alternative explanation is valid, the collateralvalues of the high bank-dependent firm should be greater than those of the low bank-dependent firm, implying that banks ex-ante screen borrowers based on the borrowers’collateral values.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 18: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

18 CHOI

Table 7Mean and Median Tests for the Alternative Explanation – Ex-ante Screening

Panel A: Test for the Difference of Collateral Between the Low and High Bank-dependent FirmsMean Test a Median Test b

Variables Sample Mean p-value Median p-value σ

Receivables Low 36.58 <0.01 23.62 <0.01 38.43High 21.71 11.97 27.66

Inventories Low 32.30 <0.01 13.92 <0.01 44.63High 20.52 7.85 33.80

Property, plant, & Low 113.56 <0.01 63.08 <0.01 136.47equipment High 65.95 32.94 92.74

Book value of equity Low 95.55 <0.01 76.42 <0.01 91.56High 53.32 35.80 59.30

Panel B: Test for the Difference of Profitability Between the Low and High Bank-dependent FirmsMean Test a Median Test b

Variables Sample Mean p-value Median p-value σ

Net income/Total Low −0.01 <0.01 0.04 0.04 0.25assets High −0.04 0.03 0.35

Retained earnings/Total assets Low −0.02 <0.01 0.14 <0.01 0.74

High −0.17 0.10 1.30

Notes:a The mean test refers to the standard t-test for differences in means. It assumes unequal variances.b The median test refers to Wilcoxon two-sample tests.Receivables, inventories, property, plant, & equipment, and book value of equity represent dollars inmillions. Sample firms are divided into two groups, high and low bank-dependent, based on a firm’s bankloan. Low represents low bank-dependent firms. High represents high bank-dependent firms.

Panel A of Table 7 presents the mean and median tests for the low and high bank-dependent samples.16 The mean and median values of all four proxies for the low bank-dependent firms are significantly greater than those for the high bank-dependent firms,which is inconsistent with the alternative explanation. Therefore, the results indicatethat the screening process based on collateral values does not explain the findings ofthis paper.

The profitability measures between high bank-dependent firms and low bank-dependent firms are further examined because it is also possible that banks ex-antescreen borrowers based on the borrowers’ profitability. Two measures are used: netincome divided by total assets and retained earnings divided by total assets. Similarto collateral measures, the result shown in Panel B of Table 7 shows that the mean

16 The high and low bank-dependent groups are classified based on a firm’s median bank share. The resultsare unaffected if the classification is based on quartiles. The mean and median tests refer to the standardt-test for differences in means and Wilcoxon two-sample tests, respectively.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 19: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

BANK RELATIONSHIPS AND THE INCOME STATEMENT 19

and median values of the profitability measures for the low bank-dependent firms aresignificantly greater than those for the high bank-dependent firms, which is inconsistentwith the alternative explanation.

In addition, recent bank competition makes bank screening more difficult. Priorstudies, including Shaffer (1998), show that there is a negative relationship betweenbank competition and the degree of screening. Therefore, in today’s competitivebanking environment, it is less likely that the ex-ante screening process influences thenature of a firm’s business.

Overall, the tests show that firm-specific characteristics such as collateral andprofitability are not the alternative explanation for the findings of this paper.Therefore, bank dependence is not a proxy for other firm attributes that influencethe informativeness of the income statement.

6. SUMMARY AND CONCLUSION

This paper examines the effects of a firm’s debt financing decision on the informa-tiveness of the income statement by investigating the association between a firm’sbank dependence and income statement conservatism. Prior academic and anecdotalevidence suggest that banks prefer accounting conservatism to secure their loans andreduce credit risk. Banks particularly value borrowers’ timely loss recognition for thepurpose of efficient debt contracting. As a consequence, banks provide borrowerswith incentives to maintain a high level of income-statement conservatism measuredas timely loss recognition. Therefore, this study predicts that the more a firm is bank-dependent, the higher the timeliness of economic loss recognition is in the firm’sincome statement. Prior research further indicates that there is a positive associationbetween timely loss recognition and the value relevance of financial statements. Inother words, timelier loss recognition in the income statement implies higher valuerelevance of the income statement. Therefore, combining this argument with the firstprediction, this study also predicts that the value relevance of the income statement isincreasing in a firm’s bank dependence. The second prediction is also related to therole of the income statement in value relevance literature. According to the literature,information from the income statement is more value relevant when a firm’s defaultrisk is lower. The uniqueness of the bank loan, such as the ease of renegotiation, lowercost of monitoring, and liquidity provision, enables a firm to continue its businesswith less worry about the risk of default. Therefore, when a firm’s bank dependenceis higher, the firm’s default risk would be lower, making the income statement morevalue relevant. This argument is consistent with the second prediction.

Focusing on relatively small businesses, the analyses report consistent results withthe two predictions. The results show that the timeliness of economic loss recognitionis increasing in a firm’s bank dependence. In addition, the results show that the highera firm’s bank dependence, the more useful the income statement is in explaining firmvalue. This result indicates that the value relevance of the income statement increasesas bank dependence increases. Overall, the findings of this paper suggest that theusefulness of the income statement varies with a firm’s bank dependence. The findingssuggest that the value relevance of the income statement is a function of a firm’s debtfinancing decision. This paper specifically argues that the effect of a bank relationshipon the value relevance is mainly due to the effect of the bank relationship on incomestatement conservatism.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 20: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

20 CHOI

This study contributes to the accounting literature by introducing the role ofthe bank relationship in financial reporting. Given the importance of a firm’s debtfinancing decision and the uniqueness of banks and bank loans, it is important tounderstand the effect of the bank relationship on financial reporting. By investigatingthe effect of bank dependence on the usefulness of the income statement, this studysheds light on the role of a firm’s financing decision in financial reporting.

It also adds to the literature examining accounting conservatism by providingempirical evidence of a lender’s preference for accounting conservatism. In addition,compared to prior studies that generally focus on the importance of the balance sheet,this study focuses on a situation where the income statement plays a major role invaluing a firm.

There are several ways to expand this study. First, this paper generally focuseson the positive side of bank relationships. There is, however, a dark side to bankrelationships. For example, banks might earn quasi-rents from borrowers, makingthe diversification of financing sources more beneficial to certain firms. Therefore,in addition to the positive side, exploring the dark side of bank relationships mightalso be important to better understand the effect of bank relationships on financialreporting. Second, to investigate the association in an international setting will provideus with another insight. Unlike the US, which is a capital market-centered economy,there are many bank-centered economies such as Japan and Germany. By comparingthe two systems (bank-centered and capital market-centered), we might be able to gainadditional significant insights into the effect of the financing methods on financialreporting in relation to international accounting standards. Finally, a more extensiveexamination of accounting conservatism using large samples, different motivations,and different measures could further our understanding of the association betweenbank relationships and accounting conservatism.

APPENDIX

Sample Selection

The sample firm-year observations are selected from the DealScan database by usingthe following sample selection procedure:

1. Facilities that meet the following requirements:(13,737 firm-year observations)

(1) The borrower (the firm) is public.(2) The lender is a lead lender.(3) The lender is a bank.(3) Bank share information is available from the DealScan database.(4) Facility-end-date is no later than December 31, 2004.

2. Among 13,737 observations from 1, facilities that meet the following require-ments:

(7,890 firm-year observations)(1) Financial data for the borrower is available from Compustat during the loanperiod.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 21: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

BANK RELATIONSHIPS AND THE INCOME STATEMENT 21

3. Among 7,890 observations from 2, facilities that meet the following requirements:(3,992 firm-year observations)

(1) The borrower (the firm) is a Mid-Cap company.(2) The borrower (the firm) is a non-financial institution and non-utilitycompany.

The DealScan database lists each credit facility as a separate record field. Therefore,it is possible that a single borrower may have a credit facility that is syndicated amongmultiple banks or multiple facilities among multiple banks (Dahiya et al., 2003). If aborrower has multiple bank-lead lenders, only one lead lender that has the largest shareis selected as a lead lender for the borrower. The largest lead lender for a facility is thesame for a loan period because a bank share is not changed during the loan period.Each facility during the loan period is treated as one observation.

Note that the largest lender is selected based only on the information from theDealScan database. Therefore, it does not mean that the largest lender has the largestshare for a borrower’s total liability. It means that the largest bank-lender has the largestbank loan share for a specific borrower during a loan period.

REFERENCES

Ahmed, A., B. Billings, R. Morton and M. Stanford-Harris (2002), ‘The Role of AccountingConservatism in Mitigating Bondholder-shareholder Conflicts Over Dividend Policy andin Reducing Debt Costs’, The Accounting Review, Vol. 77, pp. 867–90.

Altman, E. (1968), ‘Financial Ratios, Discriminant Analysis and the Prediction of CorporateBankruptcy’, Journal of Finance, Vol. 23, pp. 589–609.

Ball, R. and L. Shivakumar (2005), ‘Earnings Quality in UK Private Firms: Comparative LossRecognition Timeliness’, Journal of Accounting and Economics, Vol. 39, pp. 83–128.

——— ——— (2006), ‘The Role of Accruals in Asymmetrically Timely Gain and LossRecognition’, Journal of Accounting Research, Vol. 44, pp. 207–42.

Barth, M., W. Beaver and W. Landsman (1998), ‘Relative Valuation Roles of Equity Book Valueand Net Income as a Function of Financial Health’, Journal of Accounting and Economics,Vol. 25, pp. 1–34.

Basu, S. (1997), ‘The Conservatism Principle and the Asymmetric Timeliness of Earnings’,Journal of Accounting and Economics, Vol. 24, pp. 3–37.

Berger, A. and G. Udell (1995), ‘Relationship Lending and Lines of Credit in Small FirmFinance’, Journal of Business, Vol. 68, pp. 351–81.

Bester, H. (1985), ‘Screening vs. Rationing in Credit Markets with Imperfect Information’,American Economic Review, Vol. 75, pp. 850–55.

Burgstahler, D. and I. Dichev (1997), ‘Earnings, Adaptation and Equity Value’, The AccountingReview, Vol. 72, pp. 187–215.

Cole, R., J. Wolken and D. Woodburn (1996), ‘Bank and Nonbank Competition for SmallBusiness Credit: Evidence from the 1987 and 1993 National Survey of Small BusinessFinances’, Federal Reserve Bulletin, Vol. 82, pp. 983–95.

Collins, D., M. Pincus and H. Xie (1999), ‘Equity Valuation and Negative Earnings: The Roleof Book Value of Equity’, The Accounting Review, Vol. 74, pp. 29–61.

Dahiya, S., A. Saunder and A. Srinivasan (2003), ‘Financial Distress and Bank LendingRelationships’, Journal of Finance, Vol. 58, pp. 375–399.

Diamond, D. (1984), ‘Financial Intermediation and Delegated Monitoring’, Review of FinancialStudies, Vol. 51, pp. 393–414.

——— (1991), ‘Monitoring and Reputation: The Choice Between Bank Loans and DirectlyPlaced Debt’, Journal of Political Economy, Vol. 99, pp. 689–721.

Dichev, I. and D. Skinner. (2002), ‘Large-sample Evidence on the Debt Covenant Hypothesis’,Journal of Accounting Research, Vol. 40, pp. 1091–123.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007

Page 22: Bank Relationships and the Value Relevance of the Income ...homepages.rpi.edu/home/17/wuq2/yesterday/public...the income statement varies with a firm’s bank dependence, indicating

jbfa˙2023 JBFA2006.cls (1994/07/13 v1.2u Standard LaTeX document class) 5-3-2007 :747

22 CHOI

Fama, E. (1985), ‘What’s Different About Banks?’, Journal of Monetary Economics, Vol. 15,pp. 29–36.

Fraser, D., B. Gup and J. Kolari (2001), Commercial Banking: The Management of Risk (South-Western College Publishing, a division of Thomson Learning, Cincinnati, Ohio).

Freixas, X. and J. Rochet (1997), Microeconomics of Banking (MIT Press, Cambridge, MA).Gilson, S., K. John and L. Lang (1990), ‘Troubled Debt Restructuring: An Empirical Study

of Private Reorganization of Firms in Default’, Journal of Financial Economics, Vol. 27,pp. 315–53.

Hayn, C. (1995), ‘The Information Content of Losses’, Journal of Accounting and Economics,Vol. 20, pp. 125–53.

Holthausen, R. and R. Watts (2001), ‘The Relevance of the Value-relevance Literature forFinancial Accounting Standard Setting’, Journal of Accounting and Economics, Vol. 31,pp. 3–75.

Hoshi, T., A. Kashyap and D. Scharfstein (1990), ‘The Role of Banks in Reducing the Costs ofFinancial Distress in Japan’, Journal of Financial Economics, Vol. 27. pp. 67–88.

Kang, J., A. Shivdasani and T. Yamada (2000), ‘The Effect of Bank Relationships on InvestmentDecisions: An Investigation of Japanese Takeover Bids’, Journal of Finance, Vol. 55,pp. 2197–218.

Kashyap, A., R. Rajan and J. Stein (2002), ‘Banks as Liquidity Providers: An Explanation forthe Coexistence of Lending and Deposit-taking’, Journal of Finance, Vol. 57, pp. 33–73.

Leftwich, R. (1983), ‘Accounting Information in Private Markets: Evidence from PrivateLending Agreements’, The Accounting Review, Vol. 58, pp. 23–42.

Lev, B. and P. Zarowin (1999), ‘The Boundaries of Financial Reporting and How to ExtendThem’, Journal of Accounting Research, Vol. 37, pp. 353–85.

Milgrom, P. and J. Roberts (1992), Economics, Organization & Management (Prentice-Hall,Englewood Cliffs, NJ).

Petersen, M. and R. Rajan (1994), ‘The Benefits of Lending Relationships: Evidence fromSmall Business Data’, Journal of Finance, Vol. 49, pp. 3–37.

Saidenberg, M. and P. Strahan (1999), ‘Are Banks Still Important for Financing LargeBusinesses?’, Current Issues in Economics and Finance (August, Federal Reserve Bank ofNew York).

Shaffer, S. (1998), ‘The Winner’s Curse in Banking’, Journal of Financial Intermediation, Vol. 7,pp. 359–92.

Stiglitz, J. and A. Weiss (1981), ‘Credit Rationing in Markets with Imperfect Information’,American Economic Review, Vol. 71, pp. 393–410.

Watts, R. (2003a), ‘Conservatism in Accounting Part I: Explanation and Implications’,Accounting Horizon, Vol. 17, pp. 207–21.

——— (2003b), ‘Conservatism in Accounting Part II: Evidence and Research Opportunities’,Accounting Horizon, Vol. 17, pp. 287–301.

C© 2007 The AuthorJournal compilation C© Blackwell Publishing Ltd. 2007