Am,
“5%.:

. x? 3.. V
.7.\...JVr

. I
.. y: .lv.
n

”I

m.

. s {5. .
3......

 

«HUI.»
ail HL

 

This is to certify that the
dissertation entitled

DEBTOR-CREDITOR RELATIONSHIPS AND FINANCIAL
REPORTING' BANKS' PREFERENCE FOR BORROWERS'
ACCOUNTING CONSERVATISM

presented by

Wooseok Choi

has been accepted towards fulﬁllment
of the requirements for the

PhD. degree in Accounting and information
Systems

\ Major Prof sor’ }%ignature

’4;qu

{Date

 

 

MSU is an Affirmative Action/Equal Opportunity Institution

-—- _' -

.. ~r *'

’—

F.-~—

 

 

 

LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/DateDuopGS—pJS

DEBTOR-CREDITOR RELATIONSHIPS AND FINANCIAL REPORTING:
BANKS’ PREFERENCE FOR BORROWERS’ ACCOUNTING CONSERVATISM

By

Wooseok Choi

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Accounting and Information Systems

2004

 

BA

itpon
Obsert
consc:

htﬁds;

 

 

 

omer

mesh;
banks
3550Ci;

n0€\y

ABSTRACT

DEBTOR-CREDITOR RELATIONSHIPS AND FINANCIAL REPORTING:
BANKS’ PREFERENCE FOR BORROWERS’ ACCOUNTING CONSERVATISM

By

Wooseok Choi

This paper examines the relation between conservatism in borrowers’ financial
reporting and their dependence on bank debt financing. Using 4,114 firm-year
observations over the period 1988-2001, the analyses suggest that a ﬁrm’s accounting
conservatism is increasing in both the magnitude and length of a bank loan. This ﬁnding
holds after controlling for both the effects of other corporate governance mechanisms and
other explanations for accounting conservatism including debt covenants. These results
indicate that for bank-dependent firms, banks are an important corporate governance
mechanism in the United States. These results also indicate that the governance role of
banks is broader than the imposition of debt covenants. This paper also examines the
association between bank loan characteristics and a bank’s governance activities, ﬁnding

no evidence of an association.

Dedicated to my beloved family.

iii

 

hdpi
mdhg

monk?
thrOL

i

Axe.

men:

 

the

H Wan
sumo

Chns

ACKNOWLEDGEMENTS

I would like to thank the members of my dissertation committee for their always-
helpful advice and insight: Marilyn Johnson (Chair), Joseph Anthony, Jun—Koo Kang,
and Kathy Petroni. Without their support, I could not have finished this long journey. I
would especially like to thank Marilyn for her guidance, insight, and consideration
throughout the dissertation process and doctoral program.

I would also thank all of the other faculty members in the Department of
Accounting and Information Systems and my fellow doctoral students who have provided
me with encouragement and guidance throughout my studies at Michigan State
University.

Most importantly, I would like to thank my parents, J ong Tae Choi and Soon Hee
Hwang, and my brother and sister, In Suk Choi and Youn Jung Choi, for their love and
support. In particular, my family and I would like to thank my God and Savior Jesus

Christ for his love and sacriﬁce.

iv

LIST l

usr l

 

CHAE

CHA
DEVE

CH;

CH

CI

TABLE OF CONTENTS

LIST OF TABLES .................................................................................. vii
LIST OF FIGURES ................................................................................. viii
CHAPTER 1: INTRODUCTION .................................................................. 1
CHAPTER 2: RELATED LITERATURE AND THE MAIN HYPOTHESES
DEVELOPMENT .................................................................................... 5
2.1. Earnings management behavior around debt financing ........................... 5
2.2. Future cash flows, credit risk management, risk of overstatement, and
accounting conservatism .............................................................. 7

2.2.1. The role of borrowers’ future cash ﬂows in credit risk management...7
2.2.2. Future cash flow estimation, risk of overstatement, and accounting

conservatism .................................................................. 8
2.3. Banks’ governance activities ....................................................... 10
2.3.1. Academic evidence .......................................................... 10
2.3.2. Anecdotal evidence .......................................................... 13
2.4. Main hypotheses (Hla and Hlb) ................................................... 14
CHAPTER 3: SAMPLE AND RESEARCH DESIGN ....................................... 16
3.1. Sample and descriptive statistics ................................................... 16
3.2. Model speciﬁcation .................................................................. 19
3.2.1. Discretionary accruals ...................................................... 19
3.2.2. An estimation model, other governance mechanisms, and other
explanations for accounting conservatism ................................ 21
CHAPTER 4: EMPIRICAL RESULTS ......................................................... 27
4.1. Banks’ governance activities and conservatism in borrowers’ ﬁnancial
reporting ............................................................................... 27
4.2. Limitations ............................................................................ 28
CHAPTER 5: BANK LOAN CHARACTERISTICS ......................................... 30
5.1. Secured vs. Unsecured loan ......................................................... 30
5.2. US. banks vs. Foreign banks ...................................................... 31
5.3. Model specification and empirical results — Bank loan characteristics ........ 32
CHAPTER 6: SENSITIVITY ANALYSES .................................................... 35
6.1. A tax explanation for accounting conservatism .................................. 35
6.2. Basu (1997)’s conservatism measure .............................................. 37
6.3. Absolute values of discretionary accruals ......................................... 4O

6.4. Estimation error in accruals ........................................................... 41
6.5. Accounting conservatism before and after bank loans are made. . ..........43

CHAPTER 7: CONCLUSION ................................................................... 44

BIBLIOGRAPHY ................................................................................... 69

vi

Table

Table

Table

Table -

Table .

Table i

 

Table '
Table 1

Table 9

Table 1

 

 

 

Table 1

Table 1

LIST OF TABLES

Table 1: Sample description ....................................................................... 46
Table 2: Distribution of Firm-Year Observations by Industry ............................... 49
Table 3: Pearson correlations among independent variables ................................. 50
Table 4: Modiﬁed Jones model ................................................................... 52
Table 5: Performance-matched modified Jones model ........................................ 54
Table 6: T—tests and Wilcoxon rank-sum tests for loan characteristics

- Performance matched discretionary accrual model ................................ 56
Table 7: Test for loan characteristics ............................................................ 57
Table 8: Pearson Correlations among bank dependence and loan characteristics ......... 59
Table 9: Performance-matched modified Jones model with the measures of tax

incentives ................................................................................. 60
Table 10: Basu (1997)’s conservatism measure ............................................... 62
Table 11: Absolute values of discretionary accruals .......................................... 63
Table 12: Estimation error model — Dechow and Dichev (2002) ............................ 65

vii

LIST OF FIGURES

Figure 1: Comparisons of mean and median discretionary accruals before and after
bank loans are made ..................................................................... 67

viii

CHAPTER 1: INTRODUCTION

This paper examines the relation between bank debt financing and accounting
conservatism. Accounting research has examined the impact of a variety of different
corporate governance mechanisms including auditors, audit committees, boards of
directors, and institutional investors on a firm’s reporting decisions [e.g., DeFond and
Jiambalvo, 1991; Beasley, 1996; Becker et al., 1998; Francis et al., 1999; Bushee and
Noe, 2000; Klein 2002; Ajinkya et al., 2003]. However, prior studies have not examined
the governance role of banks, despite their signiﬁcant financing role.l

Banks are concerned that borrowers’ future cash ﬂows be of a magnitude
sufficient to repay loan principal and interest. Accrual accounting, however, requires a
certain degree of managerial discretion. Therefore, in the presence of information
asymmetries between lenders and borrowers, the use of accrual accounting to predict
future cash ﬂows creates an incentive for borrowers to inﬂuence lenders’ expectations of
future cash ﬂows by managing accruals. Such earnings management is costly to lenders
because it increases the risk that borrowers’ ﬁnancial positions are overstated. Therefore,
to reduce the risk of overstatement, banks prefer conservatism in borrowers’ ﬁnancial
reporting [e.g., FASB SFAC No. 2, 1980; Leftwich, 1983; Ahmed et al., 2002; Watts,
2003a and 2003b]. When a borrower seeks or has a significant relation with a lead bank,
that bank has a strong incentive to screen and monitor the ﬁnancial reporting behavior of

the borrower. Thus, I predict a positive relation between accounting conservatism and

 

’ An exception is the studies examining the role of debt covenants in accounting choices [e.g., Press and
Weintrop, 1990; Healy and Palepu, 1990; Begley, 1990; Beneish and Press, 1993, 1995; Sweeny, 1994;
DeFond and Jiambalvo, 1994; DeAngelo et al., 1994; Dichev and Skinner, 2002; Beatty, Ramesh, and
Weber, 2002]. These studies, however, focus primarily on debt covenants rather than the overall
governance role of debtors.

the dependence of the ﬁrm-bank relationship. I measure the dependence of the ﬁrm-bank
relationship as the magnitude and length of a bank loan. 2’3

Unlike prior research on ﬁrm-bank relationships in the United States, which
focuses on small ﬁrms [e.g., Petersen and Rajan, 1994; Berger and Udell, 1995;
Blackwell and Winter, l997]4, the empirical analyses of this paper are based on all 1988—
2001 bank loan data available from the DealScan database, which incorporates both large
and small firms. The final sample consists of 1,232 ﬁrms and 4,114 firrn-year
observations. Accounting conservatism is measured by signed discretionary accruals.
The magnitude of a bank loan is measured by the dollar value of the bank loan from the
lead bank divided by the total assets of the borrower, while the length of a bank loan is
measured by the maturity of the bank loan.

Consistent with the main hypotheses, the empirical results indicate that a firm’s
accounting conservatism is increasing in both the magnitude and length of the bank loan.
The tests include controls for the effects of other corporate governance mechanisms (e. g.,
auditors, institutional investors, 5% block holders, managerial ownership, and outside
directors), as well as other explanations for conservatism (e.g., the presence of
accounting-based debt covenants, legal liability, regulation, and taxes). This ﬁnding

indicates that for bank-dependent ﬁrms, banks are an important corporate governance

 

2 One might also argue that a ﬁrm’s bank dependence is a function of accounting conservatism, resulting in
an endogeneity problem. Banks perform two important governance activities. Banks screen loan
applicants before establishing ﬁrm—bank relationships to minimize adverse selection problems. They also
monitor borrowers after bank loans are made to minimize moral hazard problems. This paper does not
empirically discriminate between screening and monitoring because i) the focus of this paper is not on the
process of governance, but on the overall governance role of banks and ii) it is difﬁcult to obtain time-
series data for comparing accounting conservatism before and after the initiation of ﬁrm-bank relationships.
3 This paper does not address the issues of whether accounting conservatism is good or bad. Despite the
fact that FASB’s SFAC No. 2 expresses a concern that conservatism might conﬂict with some qualitative
ﬁnancial reporting characteristics, banks still have strong incentives to require ﬁrms to have conservative
ﬁnancial reporting policies for effective credit risk management.

4 Overall, these studies show that ﬁrm-bank relationships are valuable to small ﬁrms. See Chapter 2.3. for
further discussion of studies based on small ﬁrms.

mechanism in the United States.5 This paper further examines two bank loan
characteristics, collateral (secured vs. unsecured loans) and lender types (US. banks vs.
Western Europe or Asian banks), ﬁnding no evidence of an association between these
bank loan characteristics and accounting conservatism.

The results from this study shed light on two streams of research. First, this paper
demonstrates that accounting conservatism continues to play an important role in debt
contracting despite a significant decline in the use of accounting-based debt covenants
(Begley and Freedman, 2004).6 Prior studies generally argue that accounting
conservatism plays an important role in efﬁcient debt contracting especially when debt
contracts include accounting-based debt covenants. Lenders, however, value accounting
conservatism even in the absence of debt covenants because they rely on ﬁnancial
statements to estimate borrowers’ future cash ﬂows. Therefore, this paper posits that
banks will provide debtors with incentives to report conservatively even in the absence of
accounting-based debt covenants.

Second, banks are an important governance mechanism that accounting research
has not examined. Shleifer and Vishny (1997) state that despite a number of theoretical
discussions about governance by banks, there is little empirical evidence of their role.
Bank governance is closely associated with a ﬁrm’s ﬁnancial reporting because
accounting numbers play an important role in private lending agreements. The results of

this study indicate that for bank-dependent ﬁrms, banks play an important role in the US.

 

5 The corporate governance concept used in this paper is a broad one that is related to the resolution of a
general agency problem. Regarding the broad concept of corporate governance, Becht et al. (2002) state
that corporate governance is a mechanism to resolve collective action problems among dispersed investors
and to reconcile conﬂicts of interest between various corporate claimholders.

6 Begley and Freedman (2004) examine public debt contracts. Similar to their results, the results of my
study indicate that accounting—based covenants (i.e., covenants based-on ﬁnancial ratios) are not commonly
used in private debt contracts.

corporate governance system, and their contributions are incremental to other corporate
governance mechanisms. This ﬁnding is consistent with Becht et al. (2002) who
conclude their survey of the corporate governance literature by stating that there is an
emerging consensus that banks have an important role to play in corporate governance if
they are themselves well managed.

This paper proceeds as follows. Chapter 2 discusses related literature and
develops the main hypotheses. Chapter 3 describes the sample and research design.
Chapter 4 presents empirical results. Chapter 5 examines bank loan characteristics and
their association with banks’ governance activities. Chapter 6 presents sensitivity

analyses. Chapter 7 concludes the paper.

CHAPTER 2: RELATED LITERATURE AND THE MAIN HYPOTHESES

DEVELOPMENT

Although the focus of this study is on a bank’s governance role, it is important to
ﬁrst discuss why accounting conservatism is important to lenders. Therefore, Chapter 2.1.
and 2.2. discuss why ﬁrms have an incentive to manage earnings around debt ﬁnancing
and why accounting conservatism is important for banks’ credit risk management,
respectively. Then, Chapter 2.3. explains how banks provide borrowers with incentives
to maintain accounting conservatism, thereby reducing credit risk. Finally, the main

hypotheses are developed in Chapter 2.4.

2.1. Earnings management behavior around debt financing

Borrowers have various incentives to engage in earnings management behavior
before and after debt ﬁnancing.7 Before loans are made, for example, they have
incentives to manage earnings to increase borrowing capacity (e. g., higher loan amounts,
lower interest rates, and lower contracting costs). Banks typically approve loans to
borrowers based on the borrowers’ financial conditions and collateral [Fraser, Gup, and
Kolari, 2001; Mishkin, 2003; Mishkin and Eakins, 2003]. Therefore, terms of debt
contracts are likely to be affected by profits and losses (income) of borrowers, providing

strong incentives for ex-ante earnings management. Although banks’ borrower-selection

 

7 Banks (i.e., lenders) are interested in borrowers’ earnings management behavior because of the role of
accruals in predicting future cash ﬂows. Banks’ concern with the role of accruals provides borrowers with
incentives to manage earnings. The role of accruals in future cash ﬂow estimation will be discussed in
detail in Chapter 2.2.

process is well-documented, little empirical research exists about borrowers’ earnings
management behavior before loans are made.8

After loans are made, borrowers continue to have incentives to manage earnings.
One reason for ex-post earnings management is to avoid debt covenant violations. Debt
covenant violations generally result in negative consequences to borrowers, including
increases in interest rates, requests for early debt repayment, and additional restrictions
on borrowers’ activities (Beneish and Press, 1993). In addition, debt covenant violations
are associated with significant shareholder wealth losses (Beneish and Press, 1995).
Therefore, ﬁrms have incentives to reduce the probability of a covenant becoming
binding through accounting choice (Begley, 1990). Empirical findings, however, are
mixed. For ﬁrms approaching debt covenant violations, Press and Weintrop (1990),
Sweeney (1994) and DeFond and Jiambalvo (1994) ﬁnd evidence of income-increasing
accruals, while DeAngelo, DeAngelo, and Skinner (1994) find evidence of negative
accruals. Healy and Palepu (1990) ﬁnd little evidence of accounting choice in relation to
debt covenants. In addition, Dichev and Skinner (2002) ﬁnd that when debt covenants
are present in private lending agreements, they are set relatively tight, providing ﬁrms
with incentives to manage earnings.

The incentive for earnings management to obtain a favorable cost of debt also
exists under performance pricing, a recent innovation of debt contracting. Performance
pricing links the cost of debt to certain measures of firm performance. For example, the
debt/EBITDA ratio is frequently used for determining the cost of debt under performance

pricing, providing borrowers with an incentive to manage earnings to inﬂuence EBITDA.

 

8 Beatty, Ramesh, and Weber (2002) examine the ex-ante importance of accounting changes in debt
contracts, but they do not explicitly examine ex-ante earnings management behavior.

Beatty and Weber (2003) specifically suggest that incentives to lower interest rates
through performance pricing inﬂuence borrowers’ accounting method choices. In
addition, Beatty et al. (2002) ﬁnd a complementary relation between same—variable
performance pricing and covenant provisions.

The ﬁnal reason for ex-post earnings management is to increase borrowers’
renewal capacities. It is common for ﬁrms and banks to continue debtor-creditor
relationships by renewing existing debt contracts. As in the case of new debt contracts,
the renewal decision is likely to be inﬂuenced by the ﬁnancial condition of the borrower,

providing another incentive for ex-post earnings management.

2.2. Future cash flows, credit risk management, risk of overstatement, and
accounting conservatism
2.2.1. The role of borrowers’ future cash ﬂows in credit risk management

For banks, borrowers’ ability to generate future cash ﬂows is of great concern due
to its inﬂuence on credit risk. Credit risk is the risk to a bank’s earnings and capital that a
borrower will fail to repay loan principal and interest as agreed. It is the primary cause of
bank failure and the most visible risk facing bank managers (Fraser, Gup, and Kolari,
2001). Financial institutions such as banks typically use both external rating agencies
and internal risk assessment processes to manage their credit risk. In addition, regulators
monitor credit risk.

In external, internal, and regulatory credit risk assessment systems, borrowers’
expected future cash ﬂows are considered one of the most important factors because cash

is the primary source of repayment (i.e., today’s loans are repaid with tomorrow’s cash).

For example, the credit rating approach used by Moody ’s Investor Service focuses on an
assessment of the level and predictability of a ﬁrm’s future cash generation in relation to
its commitments to repay debt holders. Moody ’s also states that the ﬁrm’s capacity to
respond favorably to uncertainty is key. Generally, the more predictable a ﬁrm’s future
cash ﬂow, the higher the rating will be. In addition, in their study of internal credit risk
rating systems at large US. banks, Tracy and Carey (1998) argue that the likely adequacy
of future cash ﬂow and thus the ability of the borrower to service its debt is critical in
internal credit risk evaluation. Finally, the Comptroller’s Handbook (Ofﬁce of the
Comptroller of the Currency, 2001) uses cash ﬂows as an important factor in evaluating
banks’ credit risk. Collectively, external, internal, and regulatory credit risk assessment
systems suggest that borrowers’ expected future cash ﬂows play a critical role in credit

risk assessment.

2.2.2. Future cash flow estimation, risk of overstatement, and accounting
conservatism

Accounting accruals affect a bank’s ability to predict borrowers’ future cash ﬂows
(and thereby credit risk) because banks generally rely on accrual-based ﬁnancial
statements for the estimation of future cash ﬂows. As discussed in Chapter 2.1., in the
presence of information asymmetries between banks and ﬁrms, firms have incentives to
manage accruals to inﬂuence lenders’ expectations of future cash ﬂows. Such accrual
manipulation, however, increases banks’ credit risk because it increases the risk that
borrowers’ future cash ﬂows are overstated. Therefore, banks have incentives to reduce

risk of overstatement, thereby reducing credit risk. Lenders prefer accounting

conservatism to reduce the risk that their assessment of a borrower’s ability to generate
future cash ﬂows is overstated due to overstatement of the borrower’s ﬁnancial position.9

Consistent with this notion, prior academic and anecdotal evidence suggests that
lenders prefer conservative accounting in the presence of information asymmetries. For
example, in his comprehensive review of the conservatism literature, Watts (2003a and
2003b) suggests that lenders such as banks are concerned with the lower bound of the
earnings and new asset distributions, and use the lower bound measures during the loan
period to monitor the borrower’s ability to 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 conservatism plays 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 higher degree of conservatism provides a greater margin
of safety for the assets that serve as loan security.

SFAC No. 2 (1980) deﬁnes conservatism as “a prudent reaction to uncertainty to
try to ensure that uncertainties and risks inherent in business situations are adequately
considered”. In the same vein, this paper deﬁnes accounting conservatism associated
with debt ﬁnancing as a borrower’s prudent ﬁnancial reporting behavior in response to a
lender’s uncertainty about the borrower’s expected future cash ﬂows and the borrower’s
ability to repay loan principal and interest as agreed. This paper, however, does not focus

on whether accounting conservatism enhances the overall quality of ﬁnancial reporting.

 

9 Relatedly, lenders might prefer accounting conservatism to reduce an agency conﬂict between
bondholders and shareholders. For example, it is likely that ﬁrms with large amounts of bank debt have
high bondholder-shareholder conﬂicts. Prior research including Ahmed et al. (2002) conﬁrms this view by
demonstrating that conservatism can be used as a mechanism for resolving the agency conﬂict.

Despite the fact that SFAC No. 2 expresses a concern that conservatism might conﬂict
with some qualitative ﬁnancial reporting characteristics, this paper posits that banks have
strong incentives to require ﬁrms to have conservative financial reporting policies for
effective credit risk management.

In summary, prior evidence on future cash ﬂows, credit risk, accrual manipulation,
and accounting conservatism suggests that banks have signiﬁcant credit risk due to the
risk of overstatement and therefore, prefer conservatism in ﬁnancial reporting for the

purpose of effective credit risk management.

2.3. Banks’ governance activities
2.3.1. Academic evidence

In Chapter 2.1., I discussed why firms have an incentive to manage earnings
around debt ﬁnancing, and in Chapter 2.2., I discussed why accounting conservatism is
important for banks’ credit risk management. The next natural question is how banks
provide incentives to borrowers to maintain accounting conservatism, and obtain lower
credit risk. Generally speaking, banks carry out their ﬁduciary responsibilities by
performing governance activities. Prior research shows that the governance activities are
costly [e.g., Townsend, 1979; Gale and Hellwig, 1985; Williamson, 1986].10 If it were
not costly for lenders to detect earnings management of borrowers, firms would have
little incentive to manage earnings.

Banks can obtain lower credit risk by choosing borrowers with a high degree of
accounting conservatism and/or by providing incentives to borrowers to maintain a high

degree of accounting conservatism. That is, banks can lower credit risk by performing

 

’0 See the costly state verification model of Townsend (1979) and subsequent studies for further discussion.

10

governance activities such as screening and/or monitoring. As discussed earlier, this
paper does not empirically discriminate between the two governance activities because
this paper focuses on the overall governance role of banks rather than the process of
governance. In addition, it is difficult to obtain time-series data for comparing
accounting conservatism before and after the initiation of ﬁrm-bank relationships.ll
Diamond (1984) and Fama (1985) focus on banks’ information generating
abilities and show why ﬁnancial intermediaries, especially banks, have advantages in
performing governance functions. The delegated monitoring theory, ﬁrst developed by
Diamond (1984), argues that compared to individual lenders and other specialized
agencies such as auditors, analysts, and rating agencies, banks have a comparative
advantage in monitoring because of low costs of delegation, scale economies in
monitoring, etc.12 Fama (1985) argues that bank loans have informational advantages
compared to market loans. Similar to internally generated funds, bank loans allow a ﬁrm
to avoid the underinvestment problem associated with information asymmetries.l3 Since
Diamond (1984) and Fama (1985), a signiﬁcant body of theoretical research has shown
that banks have advantages in monitoring borrowers over other debt ﬁnancing sources
[e.g., Berlin and Loyes, 1988; Hubrich, 1989; Diamond, 1991]. Despite a number of
theoretical discussions of governance by banks, however, there is little empirical

evidence of their role (Shleifer and Vishny, 1997).

 

it One study that distinguishes between monitoring and screening is Choi (2003). It separates monitoring
from screening by controlling for major screening devices, such as collateral and proﬁtability.

’2 For further details of this argument, see Frexis and Rochet (1997).

’3 The underinvestment problem is caused by debt overhang, the inability of a company with proﬁtable
investment opportunities to ﬁnance them because it has excessive levels of debt relative to its assets
(Milgrom and Roberts, 1992).

ll

Prior empirical research on firm-bank relationships often focuses on small
businesses because small ﬁrms obtain most of their external funding from ﬁnancial
intermediaries (i.e., commercial banks) and, in turn, ﬁrm-creditor relationships are
expected to be especially important in small ﬁrms. Petersen and Rajan (1994) show that
the primary beneﬁt of a firm-bank relationship is that the availability of ﬁnancing
increases. Berger and Udell (1995) ﬁnd results consistent with the theoretical argument
that relationship lending generates valuable information about borrower quality. Finally,
Blackwell and Winter (1997) demonstrate that ﬁrms can signiﬁcantly reduce their costs
of capital by establishing and maintaining close ﬁrm-bank relationships.

An analogy from the literature on screening and monitoring by institutional
investors is also relevant. Prior studies argue that compared to diffuse owners, owners
with concentrated equity holdings play a more active role in corporate governance [e.g.,
Shleifer and Vishny, 1986; Black, 1992; Shleifer and Vishny, 1997]. They show that
large investors have a lower marginal cost of processing information and therefore, serve
as effective monitors of management. Similarly, banks with concentrated lending are
more apt to receive sufﬁcient gains from their governance activities to cover the relevant
costs. Moreover, in 1999, Congress repealed the Glass—Stegall Act, which puts
restrictions on underwriting and dealing activities by commercial banks, allowing the
banks to compete as investment bankers, brokers, and insurers in the US. market and
play more active roles in obtaining private information.

In summary, prior research on bank loans, bank relationships, and institutional
investors suggests that through their governance activities, banks have an incentive to

reduce the risk that borrowers’ ﬁnancial condition is overstated.

12

2.3.2. Anecdotal evidence

In practice, banks perform various governance activities that provide incentives to
borrowers to maintain conservatism in ﬁnancial reporting. For example, about one-third
of large US. ﬁrms have representatives from banks on their boards (Kroszner and
Strahan, 2001). Bankers on boards could have a significant inﬂuence on the ﬁnancial
reporting of borrowers because, in the United States, the board of directors is legally
responsible for the preparation and presentation of financial reports. In addition, banks
regularly perform a loan review as part of their internal audit systems to identify potential
problems with particular loans and weakness in loan procedures (Fraser et al., 2001).

The Association of Bank Loan and Credit Officers (Robert Morris Associates: RMA)
suggests that in some cases and especially, in the case of a permanent capital loan, banks
require borrowers to provide a listing of accounts receivable and an aging to examine the
appropriateness of the values of accounts receivable. Banks also examine whether loans
are fully covered by receivables.

Another way for banks to provide incentives for borrowers to maintain accounting
conservatism might be through an amendment to a debt contract. A debt contract is
frequently amended after the contract is ﬁrst made. For instance, a debt contract of
Aviall, Inc., which was ﬁrst signed on November, 1993, was amended three times within
one year. An important change related to ﬁnancial reporting was the inclusion of
requirements for additional ﬁnancial disclosures. In addition, periodic meetings and calls
between ﬁrms’ executives and loan ofﬁcers could be other examples of banks’

governance activities. Banks frequently check whether ﬁrms have conservative liquidity

13

and leverage positions which are within industry standards (Robert Morris Associates:

RMA).

2.4. Main hypotheses (Hla and Hlb)

As suggested in Chapter 2.1. and 2.2., banks have a significant amount of credit
risk associated with uncertainty about borrowers’ ability to repay loan principal and
interest. Banks are particularly concerned with the lower bound of earnings distributions
and use the lower bound during the loan period to monitor a borrower’s ability to pay
(Watts, 2003a). This argument implies that banks have incentives to require borrowers to
maintain a high level of accounting conservatism for the purpose of effective credit risk
management. As discussed in Chapter 2.3., banks lower their credit risk by performing
governance activities, such as screening and monitoring, that provide incentives to
borrowers to use conservative financial reporting.

This paper examines the association between bank governance and conservatism
in a ﬁrm’s ﬁnancial reporting. Specifically, this paper argues that when a firm is more
dependent on a bank, the ﬁrm has a higher level of accounting conservatism. Two
aspects of the firrn-bank relationship are examined to measure a ﬁrm’s bank dependence.

The ﬁrst aspect of the ﬁrm-bank relationship is the magnitude of a bank loan.
Prior research suggests that the greater the relative magnitude of the bank loan to the ﬁrm,
the greater is the bank’s incentive to perform governance activities [e.g., Petersen and
Rajan,1994; Kang, Shivdasani, and Yamada, 2000]. In other words, the bank devotes
more effort to detecting upward manipulation of discretionary accruals when it lends

more money to the ﬁrm. Therefore, I expect that conservatism in borrowers’ ﬁnancial

l4

reporting is increasing in the magnitude of the bank loan. Accordingly, the first main

hypothesis of this paper is as follows (stated in alternative form):

Illa. Ceteris paribus, a firm’s accounting conservatism is increasing in the

magnitude of the bank loan.

The second main hypothesis examines how the length of a bank loan affects a
bank’s incentive to perform governance activities. If a ﬁrm has a longer relationship with
a bank, the ﬁrm is likely to have more chances to engage in opportunistic reporting
behavior. The bank, in turn, has a greater incentive for monitoring. For example, Rajan
and Winton (1995) argue that the existence of long-term debt with covenants increases a
bank’s incentive to monitor. Additionally, in their review of bank relationships, Ongena
and Smith (2000) state that when a bank has a longer relationship with a ﬁrm, the bank
will gain signiﬁcant monitoring benefits including advantages over obtaining private
information about its borrower. The longer relationship, in turn, allows the bank to more
effectively see through the upward manipulation of discretionary accruals. Therefore, I
expect that a bank’s incentive to perform governance activities is increasing in the length
of the bank loan. Accordingly, the second main hypothesis is as follows (stated in

alternative form):

Hlb. Ceteris paribus, a firm’s accounting conservatism is increasing in the length

of the bank loan.

15

CHAPTER 3: SAMPLE AND RESEARCH DESIGN

3.1. Sample and descriptive statistics
The initial sample comprises all bank loans for which it is possible to calculate the

bank’s percentage share of the ﬁrm’s outstanding loans from the DealScan database for
the years 1988 - 2001. Constructed by the Loan Pricing Corporation (LPC), DealScan
contains deal terms and conditions on over 72,000 loans since 1988, the majority of
which are US. loans. Deals in the database are often comprised of different “facilities”.
For example, a deal might include a term loan facility and a line of credit facility.
Similar to prior research, each facility is treated as one observation. ’4 Among all bank
loan observations with percentage share information from 1988 to 2001, 1,232 ﬁrms and
4,114 ﬁrm-year observations are selected, which meet the following requirements:

1. The ﬁrm (i.e., the borrower) is public.

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

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

4. Financial data on the firm are available from the Compustat database during the

loan period.

Due to missing values, the sample has an unbalanced panel data structure. Although
4,114 ﬁrm-year observations are selected by using the above criteria, the actual number
of observations used in each analysis varies with the model speciﬁcation (e.g., the

inclusion of lead and lag variables, the availability of corporate governance variables, the

 

"’ One problem with the DealScan database is that coverage improves during the late 19805, so if a ﬁrm has
a signiﬁcant relationship with a bank before the late 1980s, it is difﬁcult to ﬁnd an exact date for the
initiation of the ﬁrm-bank relationship. In addition, DealScan provides information about loans at the date
of origination, but little information about subsequent events such as rollovers (Strahan, 1999). For further
discussion of the DealScan database, see Strahan (1999), and Dichev and Skinner (2002).

16

use of a random effects model, and the availability of data for ratio calculations, etc.).15
The sample is further restricted to those ﬁrms that have a minimum of ﬁve years of data
on all necessary variables because of the nature of some of tests. Most loans in the ﬁnal
sample are long-term senior debts.

Panel A of Table 1 reports descriptive statistics for the sample. Consistent with
the bank relationship literature [e.g., Hoshi, Kashyap, and Scharfstein, 1990; Petersen and
Rajan,1994; Kang, Shivdasani, and Yamada, 2000], the ﬁrst measure of bank dependence,
the magnitude of a bank loan is the dollar value of the bank loan from the lead bank
divided by the total assets of the borrower.16 This measure captures the closeness of a
ﬁrm to its main lender and the concentration of a firm’s borrowing across blockholder
lenders. It also measures how important a bank loan is to a ﬁrm. There are several
reasons why the bank loan is scaled by borrowers’ total assets rather than banks’ total
assets. First, this scaling measures the extent to which borrowers are inﬂuenced by
banks’ governance activities, which is consistent with the research questions of this paper.
Second, for a lead bank, a reputation is very important. Therefore, a lead bank will
perform governance activities regardless of the relative size of the bank loan to its total
assets. Finally, transaction costs of switching to another bank are associated with the
relative size of the bank loan to the ﬁrrns’ total assets. The second measure of bank
dependence, the length of a bank loan, is the maturity of the bank loan. The second

measure captures the length of a ﬁrm’s borrowing relation with a lead lender.

 

’5 To mitigate the potential adverse effect of data errors and outliers, the analyses are also performed with
the data winsorized at the 1% level. The results are qualitatively similar to those reported in the tables.

’6 To mitigate the potential adverse effect of data errors in the DealScan database, the ﬁrst bank
dependence measure is set to one if it exceeds one. The values of less than 15 observations are changed by
the imposition of this restriction.

l7

The median (mean) value of total bank loans per ﬁrm-year is $20.4 million
($104.37 million), while the median (mean) total bank loans from the lead lender is $15
million ($31.60 million). The median (mean) magnitude of the bank loan is 0.06 (0.10),
indicating that the median (mean) value of the bank loan from the lead lender is
approximately 6% (10%) of total assets. The median (mean) length of the bank loan is
60 (67) months. The median (mean) ﬁnancial leverage, interest coverage, ROA, and the
Altman-Z score are 0.56 (0.56), 5.75 (16.36), 0.05 (0.03), and 2.94 (3.76) respectively.17
Finally, the median (mean) market-to-book ratio is 1.78 (2.67), implying that on average,
a ﬁrm’s market value is about twice as high as its book value.

Panel B of Table 1 reports descriptive statistics for low versus high bank
dependent ﬁrms. Consistent with prior research, high bank dependent ﬁrms are relatively
small ﬁrms compared to low bank dependent ﬁrms and have smaller free cash ﬂows than
low bank dependent ﬁrms. Other descriptive statistics are similar between low and high
bank dependent ﬁrms. Table 2 shows that the industry composition of the sample is
similar to that of ﬁrms on Compustat in 2001. Additionally, the industry compositions of
the low and high bank dependent ﬁrm sub—samples are similar to each other. Finally,
Table 3 presents the Pearson correlations among the independent variables used in the

estimation models. In general, the correlations are low in absolute magnitude.’8

 

’7 The interpretation of the Altman-Z score is as follows: i) Z > 2.99 means the ﬁrm is safe, ii) Z < 1.80
means the ﬁrm is ﬁnancially distressed, and iii) 1.80 < Z < 2.99 means the ﬁrm is in the ‘grey’ zone.

'8 The variance inﬂation factor (VIF) is measured to examine possible multicollinearity problems among
independent variables. See Chapter 4 for further discussion.

18

3.2. Model specification
3.2.1. Discretionary accruals

I use discretionary accruals to measure accounting conservatism. To examine the
relationship between accounting conservatism and bank dependence, I then regress
signed discretionary accruals on the bank-dependence measures and control variables. A
primary reason why discretionary accrual models are chosen is their relation to an
important governance activity performed by banks. As stated in Chapter 2.3.2., the
Association of Bank Loan and Credit Ofﬁcers (Robert Morris Associates: RMA) suggests
that it is not uncommon for banks to determine the appropriateness of the values of
accounts receivable by performing an aging of specific accounts receivable. With respect
to the level of discretionary accruals, this examination is particularly important because
the modiﬁed Jones model assumes that the change in credit sales is the result of
managerial discretion.

An alternative approach to measuring conservatism is the earnings-price
regression developed by Basu (1997). Discretionary accruals are used in this paper
because the focus is on conservatism in discretionary ﬁnancial reporting behavior around
bank debt ﬁnancing. In contrast, Basu (1997) emphasizes the impact of conservatism on
the asymmetric incorporation of good and bad news in stock price.”

The ﬁrst model is the cross-sectional version of the modiﬁed Jones model [Jones

(1991) and Dechow, Sloan, and Sweeney (1995)], which separates total accruals into

discretionary and non-discretionary accruals. 20 Unlike the Jones (1991) model, which

 

’9 Basu (l997)’s conservatism measure is examined in Chapter 6 as a sensitivity analysis.

2° The cross-sectional version means that the modiﬁed Jones model is estimated for each two-digit SIC
group for each year. Subramanyam (1996) shows that the parameter estimates are better speciﬁed for the
cross-sectional modiﬁed Jones model than their time series counterparts.

l9

assumes that none of the change in revenues is attributable to managerial discretion, the
modiﬁed Jones model assumes that the change in credit sales results from managerial

discretion as follows:

TAit/AssetitJ = ﬂdI/AssetiIJ) + ﬂﬂARevi, - AARit)/Asseti,_1
+ ﬂzPPEit/AssetitJ + git (1)

where,

TAit = ﬁrm i’s total accruals in year t, calculated as change in current assets
(ACAir) - Change in current liabilities (ACLio — change in cash (ACashit) +
change in debt in current liabilities (AST_DEBTi0 - depreciation and
amortization expense (ADeprit),

AssetiH = firm i’s total assest at the beginning of year t,

ARevit = ﬁrm i’s change in revenue,

AARit = ﬁrm i’s change in accounts receivable,

PPEit = ﬁrm i’s gross property, plant, and equipment,

i, t = ﬁrm and year index, respectively.

The above deﬁnitions follow those of Kothari et al. (2002), and Francis et al. (2002).
Accounting conservatism is measured as the value of the residuals from equation (1).
The lower the value of the residuals, the higher the accounting conservatism.

The second measure of discretionary accruals is performance-matched
discretionary accruals, i.e., the value of [the discretionary accruals (residuals) from

equation (1) - median discretionary accruals for the ﬁrm’s ROA deciles]. In a recent

20

study, Kothari et al. (2002) ﬁnd that inferences about earnings management using a
performance-matched discretionary accrual measure are likely to be more reliable than
those from traditional models of discretionary accruals, such as the Jones or modiﬁed
Jones models. Therefore, I also measure the degree of earnings management by using the
performance-matched modiﬁed Jones model. Francis et al. (2002) adapt the Kothari et a1.
(2002) treatment firm procedure to a portfolio adjustment procedure that is better suited
for a broad sample study with no well-deﬁned treatment groups. Francis et al. (2002) use
a time-series version of the modiﬁed Jones model. In contrast, this paper uses the cross-
sectional version of the modiﬁed Jones model for each two-di git SIC group in each year
to obtain residuals, and then standardizes the residuals by the median values of the
residuals from the performance-matching procedure. The sample is divided into deciles

based on return on assets (ROA).

3.2.2. An estimation model, other governance mechanisms, and other explanations
for accounting conservatism
After measuring discretionary accruals, I then test the main hypotheses by

estimating the following random effects regression model:

DA_Jones” or DA_Perfl-t = a0 + a,BD_MAG,-, + azBD_LEN,-t + agINSTITUTIONi,
+ m5%BLOCK,-, + a5INSIDERi, + adoaooit + a7BIGi, +q,FIN,-,
+ agLEVERAGEi, + armCOVERAGEl-t + auALTMANit+ ale/Bi,
+ 013WSETSI'; + a14R0Ait + a15FREECF it + git (2)

where,

DA_Jones = the value of discretionary accruals from the modiﬁed Jones Model,

21

DA_Perf = the value of discretionary accruals from the performance matched
discretionary accrual model,

BD_MAG = the magnitude of a bank loan measured as (the dollar value of the bank

loan from the lead lender / total assets),

BD_LEN = the length of a bank loan measured as a maturity of the bank loan,

INSTITUTION = proportion of institutional ownership,

5%BLOCK = proportion of 5% block holder ownership,

INSIDER = proportion of managerial ownership,

OBOD = proportion of outside directors on a ﬁrm’s board,

BIG = an indicator variable that equals 1 if a ﬁrm’s auditor is a Big 5 accounting

ﬁrm, and 0 if a ﬁrm’s auditor is a non-Big 5 accounting ﬁrm,

FIN = an indicator variable that equals 1 if a deal includes covenants based on

ﬁnancial ratios, and 0 if a deal dose not include ﬁnancial ratios,

LEVERAGE = ﬁnancial leverage measured as (total liabilities / total assets),

COVERAGE = the interest coverage ratio measured as (operating income / interest

expenses),

ALTMAN = 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/total assets) + 0.6 (market value of equity/book value of
liabilities) + 1.0 (sales/total assets) ],

M/B = the market-to-book ratio measured as (market value of equity / book value of

equity),2’

LASSETS = the natural log of the book value of assets,

ROA = the return on assets measured as (income before extraordinary items /

beginning total assets),

FREECF = free cash ﬂow measured as [operating income before depreciation —

interest expenses — (income taxes — deferred tax & invest tax credit) —
dividends],

i, t = ﬁrm and year index, respectively.

 

2’ I also use the book-to~market ratio instead of the market-to-book ratio. The results are qualitatively
similar.

22

The random effects model allows me to control for the unobserved ﬁrm effects, which is
appropriate when analyzing an unbalanced panel dataset. The Hausman (1978) test
indicates that the random effects model is appropriate when the control variable, BIG, is
omitted.”23 Therefore, the random effects regression is estimated without BIG. For a
comparison, equation (2) is also estimated using pooled OLS with controlling for the
clustering effects of years.“ The pooled OLS is estimated including BIG.

The estimation model includes important control variables that proxy for the
presence of other corporate governance mechanisms (i.e., INSTITUTION, 5%BLOCK,
INSIDER, OBOD, and BIG). Prior researchers have hypothesized that ﬁnancial
reporting quality is associated with monitoring incentives arising from institutional
ownership and block holdings [e.g., Bushee and Noe, 2000; Klein, 2002; Ajinkya et al.,
2003], managerial ownership [e.g., Warfield et al., 1995; Klein, 2002], and board
composition [e.g., Beasley, 1996; Klein, 2002; Ajinkya et al., 2003]. Few studies,
however, directly examine the association between signed discretionary accruals and
these other corporate governance mechanisms. Prior research on institutional ownership
and block holdings generally suggest that there is a positive association between ﬁnancial
reporting quality and institutional ownership and/or block holdings, implying that
sophisticated institutional owners are less likely to be fooled by management of accruals.
The ﬁndings from prior research on managerial ownership are mixed. For example,

Warﬁeld et al. (1995) suggest a negative relation between managerial ownership and

 

22 The Hausman (1978) test examines the appropriateness of the ﬁxed and random effects models, based on
the difference between the ﬁxed effect and random effect estimates. The ﬁxed effects model cannot be
used to estimate equation (2) because the ﬁxed effects estimation automatically drops all time-invariant
variables.

23 About 94% of sample ﬁrm-year observations have a Big 5 auditor, indicating that there is little variation
in BIG.

2’ One problem associated with using pooled OLS is that I must ignore all within-ﬁrm effects. In the case
of a panel dataset, ignoring within-ﬁrm effects might lead to incorrect interpretation of the coefﬁcients.

23

earnings management while Healy (1985) suggests a positive relation. Klein (2002)
states that the relation is unclear. Finally, prior research on board composition generally
suggests that there is a positive relation between board independence and ﬁnancial
reporting quality.

In addition, prior research has examined the relation between audit quality as
indicated by Big versus non-Big 5 or 6 auditors and earnings management [e.g., DeFond
and Jiambalvo, 1991; Becker et al, 1998; Francis et al, 1999]. Prior research generally
shows that clients of Big 6 auditors are less likely to engage in earnings management than
clients of Big 6 auditors. The control variable, BIG is included to control for the effect of
audit quality. The corporate governance variables except BIG are collected from the
Compact D/SEC database.25 Similar to prior research, the mid-year of each loan period is
used to measure governance structure for each ﬁrm because a ﬁrm’s governance structure
is generally sticky (Core et al., 1999). The identity of the auditor is obtained from the
Compustat database.

Prior studies including Watts (2003a and 2003b) generally suggest four
explanations for accounting conservatism: contracting, litigation, regulation, and taxes.
This study argues that banks provide general contracting incentives for conservatism in
borrowers’ ﬁnancial reporting beyond the direct contracting incentives in debt covenants.
Thus, to control for the direct incentives of debt covenants, an indicator variable for the

presence of accounting-based debt covenants, FIN, is included in the estimation

 

25 The Compact D/SEC database provides access to ﬁnancial and other information contained within annual
reports, proxy statements, and 10-K/20-F ﬁlings for over 12,000 public US companies. The database is
used with permission of the Compact D/SEC.

24

model?”27 There is an additional contracting explanation for accounting conservatism,
compensation contracts. I do not control for the effect of this explanation because it is
difﬁcult to obtain compensation information for each of the sample ﬁrms. In addition,
compensation contracts are likely to be related to various governance variables, such as
managerial ownership, institutional ownership, and block holder ownership, for which I
include explicit controls. Prior research also suggests that litigation risk and regulation
provide incentives to use conservative ﬁnancial reporting. Industry membership is
typically used to proxy for the ﬁrm’s litigation risk and regulatory environment (Francis
et al., 1994). As discussed earlier, the second model of discretionary accruals, the
performance matched model, is estimated for each two-di git SIC group, mitigating the
effects of the litigation and regulation explanations. Finally, prior research argues that
ﬁrms have incentives to use conservative ﬁnancial reporting to reduce taxes. This tax
explanation will be examined in Chapter 6.1. as a sensitivity analysis by using a
conelation between book income and taxable income and marginal tax rate. It is tested in
the sensitivity analyses chapter because of the reduction in sample size that results when
the tax measure is included.

In addition, other control variables that might affect a firm’s financial reporting
are included in the estimation model. Financial leverage, interest coverage, and the
Altman Z-Score are included to control for a firm’s ﬁnancial health. The market-to-book
ratio and the log of the book value of assets are included to control for a ﬁrm’s growth

opportunities and size, respectively. Finally, RCA and free cash ﬂow are included to

 

2‘ Results when observations with ﬁnancial ratio-based covenants are dropped are qualitatively the same as
results using the indicator variable.

27 In addition, Begley and Feltham (1999) show that the CEO ownership fraction is related to the use of
debt covenants restricting dividends and borrowing. Managerial ownership is explicitly controlled in the
estimation model of this paper.

25

control for a firm’s performance and the ability to finance operations from internal
savings, respectively.

To mitigate the possible effects of heteroskedasticity, the main hypotheses are
tested with the White (1980) heteroskedasticity robust standard errors. The ﬁrst main
hypothesis (Hla) predicts that the coefﬁcient on BD_MAG, or], will be negative and
signiﬁcant, indicating that a ﬁrm’s accounting conservatism is increasing in the
magnitude of the bank loan. The second main hypothesis (Hlb) predicts that the
coefﬁcient on BD_LEN, (12, will be negative and signiﬁcant, indicating that a ﬁrm’s

accounting conservatism is increasing in the length of the bank loan.

26

CHAPTER 4: EMPIRICAL RESULTS

4.1. Banks’ governance activities and conservatism in borrowers’ financial
reporting

The empirical tests of the main hypotheses examine the association between
banks’ governance activities, as indicated by the magnitude and length of the bank loan,
and accounting conservatism, as measured by discretionary accruals. The two
discretionary accrual models used to measure a ﬁrm’s accounting conservatism are the
modiﬁed Jones model and the performance-matched modified Jones model. Table 4 and
5 report results from equation (2) based on the modiﬁed Jones model and the
performance-matched modiﬁed Jones model, respectively. Consistent with the ﬁrst main
hypothesis (Hla) that a firm’s accounting conservatism is increasing in the magnitude of
the bank loan, the coefficient on BD_MAG in Table 4, -O. 1747, is negative and
statistically signiﬁcant (p-value < 0.01). The coefﬁcient on BD_MAG in Table 5, -
0.1506, is also negative and statistically significant (p-value < 0.01). Consistent with the
second main hypothesis (Hlb) that a ﬁrm’s accounting conservatism is increasing in the
length of the bank loan, the coefﬁcient on BD_LEN in Table 4, -0.0005 is negative and
statistically signiﬁcant (p-value = 0.029). The coefficient on BD_LEN in Table 5, -
0.0005, is also negative and statistically signiﬁcant (p-value = 0.012).28 The results from
a pooled OLS that controls for the clustering effects of years are similar to those from the

random effects estimations. Irrespective of whether discretionary accruals are measured

 

2’ The variance inﬂation factor (VIF) is measured to check for multicollinearity among independent
variables. The largest VIF, 2.78, is less than 10, suggesting that the variance of an estimator is not
signiﬁcantly inﬂated by collinearity between independent variables. In addition, 1N IF measure shows that
over 60% of the variances of the most independent variables are independent of all the other independent
variables. Both VIF measures indicate that multicollinearity is not a serious problem in this estimation.

27

using the modiﬁed Jones or performance-matched modiﬁed Jones model, the results
indicate that for bank-dependent ﬁrms, banks are an important governance mechanism.
The coefﬁcients on the other corporate governance control variables are generally
not signiﬁcant, attesting to the importance of bank governance for this sample of ﬁrms.
FIN is not signiﬁcant, implying that accounting-based debt covenants are not a
signiﬁcant determinant of accounting conservatism for this sample.29 The signiﬁcant
results for the main variables from the performance-matched modiﬁed Jones model imply
that the litigation and regulation environment do not signiﬁcantly affect the main results.
Among other control variables, ALTMAN and RCA are statistically signiﬁcant in the
modiﬁed Jones model while LASSETS and RCA are statistically signiﬁcant in the
performance-matched modiﬁed Jones model. The negatively signiﬁcant coefﬁcient of
LASSETS indicates that larger ﬁrms have more conservative accounting. Because high
bank dependent ﬁrms are relatively small, the negative sign on LASSETS implies that the
main results (BD_MAG and BD_LEN) are not attributable to ﬁrm size. In summary, a
ﬁrm’s accounting conservatism is increasing in both the magnitude and length of the

bank loan, supporting the main hypotheses, Hla and Hlb.

4.2. Limitations

The interpretation of the results is subject to two caveats. First, the ﬁndings of
this paper show that there is an association between bank governance and accounting
conservatism. However, it is difﬁcult to know if banks make ﬁrms conservative or ﬁrms

choose to be conservative. Although the results demonstrate that banks play an important

 

2’ As mentioned earlier, results when observations with ﬁnancial ratio-based covenants are omitted are
unchanged.

28

governance role, the exact governance mechanism and its direction is left to future
studies.

Second, the ﬁrst measure of bank dependence, the magnitude of the bank loan, is
scaled by the total assets of a borrower. This measure is consistent with the research
questions of this paper and is widely accepted in prior studies. Alternatively, one might
argue that scaling by the total assets of the lender is a better measure of a bank’s
incentive to perform governance activities. DealScan does not provide sufficient
information to obtain the value of a lead bank’s total assets. Therefore, as a sensitivity
test, I use as a proxy for a lead bank’s total assets, the sum of all bank loans per each lead
bank within the sample, and scale the bank loan by this proxy. I do not ﬁnd a signiﬁcant
coefﬁcient on the magnitude of the bank loan, although the length variable is still
signiﬁcant. Possible explanations would be 1) as discussed above, the alternative
measure is not appropriate for addressing the research questions of this paper, 2) the
proxy is noisy and not representative because it is calculated within the sample and by
using limited loan information, and/or 3) the current measure (scaled by the total assts of
a borrower) is not an appropriate proxy for bank dependence and is confounding my

analysis.

29

CHAPTER 5: BANK LOAN CHARACTERISTICS

In this chapter, I will discuss two bank loan characteristics, lender types and
collateral, and explain how these loan characteristics might be expected to affect a bank’s
governance activities and therefore, a firm’s accounting conservatism. Irnportantly, one
should note that the purpose of the tests of the loan characteristics is not to examine a
bank’s governance activities in general, but, rather, to examine the relative intensity of a
bank’s governance activities conditional on various types of bank loans. The former
question, a bank’s overall corporate governance role per se, is examined by testing the

two main hypotheses.

5.1. Secured vs. Unsecured loan

In their study of the role of accounting conservatism in mitigating bondholder-
shareholder conﬂicts over dividend policy, Ahmed et al. (2002) mention that, besides
accounting conservatism, there are other mechanisms for resolving the agency conﬂict,
such as secured debt, leaving to future research the examination of such other
mechanisms. In response to their concern, the ﬁrst loan characteristic examined in this
paper is whether a loan is secured.

A secured loan is a loan that is collateralized by a borrower’s tangible assets,
while an unsecured loan is a loan that is backed only by a borrower’s promise to pay. If a
loan is secured, a lender can liquidate the collateral in the event that the borrower defaults
on the loan. Therefore, the existence of collateral signiﬁcantly reduces the lender’s credit

risk. The reduced credit risk, in turn, reduces incentives for the lender to perform

30

governance activities. In an extreme case when a loan is fully secured (i.e., fully
collateralized), the collateral serves as a substitute monitoring mechanism and the lender
has little incentive to perform governance activities.

On the other hand, although a secured loan reduces a lender’s incentive to
perform governance activities, it also requires a lender to monitor the collateral per se (as
opposed to overall credit risk). For example, Rajan and Winton (1995) show that using
collateral may give lenders an incentive to monitor. If this is the case, the intensity of
bank governance is not reduced by using a secured loan. There is still the same incentive
for a bank to perform governance activities because of the transaction costs of seizing and
selling the collateral. In addition, due to macro shocks, the value of collateral is likely to
be lowest at the time of default, providing a bank with another incentive to perform
governance activities. Accordingly, the prediction of the association between secured

loans and banks’ governance (in turn, accounting conservatism) is non-directional.

5.2. US. banks vs. Foreign banks

The United States can be categorized as a capital market-centered system, while
Japan and Continental Europe can be categorized as bank-centered systems. The term,
“capital market centered”, in this paper does not mean that ﬁrms obtain more funds from
the stock market than from banks. 30 It is a relative concept and therefore, implies that
banks in bank-centered countries play more active roles in disciplining a borrower’s

management than banks in capital-market centered countries. Prior research ﬁnds

 

3° According to Mishkin and Eakins (2003), banks are the most important source of external funds used to
ﬁnance US. corporations, accounting for 40.2% of the external ﬁnancing of American business in the
1970-1996 period. For the same period, equity accounts for 9.2% of the external ﬁnancing, while bonds
and nonbank loans accounts for 35.5% and 15.1%, respectively. These ﬁgures are, however, based on all
US. ﬁrms, while the current paper focuses only on public ﬁrms.

31

evidence that banks in bank-centered countries traditionally perform very active roles in a
ﬁrm’s corporate governance system [e.g., Kaplan, 1994; Kaplan and Minton, 1994; Kang
and Shivdasani, 1995; Elsas and Krahnen, 1998].

Today more than 250 foreign banks do business in the United States under the
federal regulation of foreign banking activities codified in the International Banking Act
of 1978 (Fraser, Gup, and Kolari, 2001). In light of their active roles in firm governance
in their home countries, it is likely that foreign banks from bank-centered countries
perform more intensive monitoring and screening of US. borrowers than US. banks do.
Additionally, foreign banks may perceive information asymmetries to be greater when
loans are made to companies domiciled in other countries and may, therefore, increase
screening and monitoring activities.

Alternatively, US. banks may have advantages over foreign banks in monitoring
and screening borrowers because they have fewer information asymmetries. US. banks
have less uncertainty in their operations than foreign banks because they have more
information about domestic borrowers and the domestic economy. Therefore, as in the
case of the ﬁrst loan characteristic, the prediction of the association between lender types

and banks’ governance is non-directional.

5.3. Model specification and empirical results - Bank loan characteristics

Similar to the test of the main hypotheses, the random effects model without BIG
is used as the primary test of the association between accounting conservatism and loan
characteristics. Pooled OLS that include BIG is also estimated. I do not include the two

loan characteristic variables in tests of the main hypotheses because the inclusion of these

32

variables reduces the sample size by one-third, reducing the power of the test.31 The test
of the signiﬁcance of the loan characteristic hypotheses is performed by estimating the

following regression model:

DA_Perfi, = or; + a,BD_MAG,-, + azBD_LEN,-t + a3INSTIT UT IONi,
+ a45%BL0CK,-, + a5INSIDERl-t + 030800;, + ayBIGi, +a8FINi,
+ agLEVERAGEi, + amCOVERAGEi, + aruALTMANit-t- aIZM/Bi, + a13LASSETSi,
+ aMROAl-t + a15FREECFi,+ a16SECUREi, + chSBANKit + 8;, (3)

where,

DA_Perf = the value of discretionary accruals from the Performance matched
discretionary accrual model,

SECURE = an indicator variable that equals 1 if a loan is a collateralized loan, and

0 if a loan is a non—collateralized loan,

USBANK = an indicator variable that equals 1 if a lender is a US. bank, and 0 if a
lender is an Westem European or Asian/Paciﬁc bank,”2

i, t = ﬁrm and year index, respectively,

other variables are deﬁned in equation (2).

In equation (3), the variables of interest are SECURE and USBANK. Since the two bank
loan characteristics could have both increasing and/or decreasing effects on the intensity

of banks’ governance activities, I do not make a directional prediction.

 

3’ One might argue that if the loan characteristics are not included in equations (2), there might be a
correlated-omitted-variables problem, making the interpretation difﬁcult. There is, however, no problem
interpreting the results from equations (2) even though the loan characteristics are not included in the
models because the random effects model implicitly considers the unobserved effects in the calculation of
the coefﬁcients. For further discussion of the panel data analyses, see Wooldridge (2002).

32 One problem with this variable is that the DealScan database categorizes foreign banks into Western
Europe, Asian, etc. without specifying names of their home countries. Therefore, it is possible that there is
measurement error in my USB AN K variable. For example, Western Europe includes not only the bank-
centered Continental countries, but also the capital market-centered United Kingdom.

33

 

Table 6 presents the results from two univariate tests. Both the t-test and
Wicoxon rank-sum test show that mean and median values of discretionary accruals are
lower for ﬁrms with unsecured loans and for ﬁrms with Western Europe or Asian bank
loans, but the differences are generally insigniﬁcant.33

The estimations of equation (3) are reported in Table 7. In the random effects
estimation, the coefﬁcients on BD_MAG, -O.1825 (p-value = 0.014), and BD_LEN, -
0.0007 (p-value = 0.026), are still negative and statistically signiﬁcant. However, the
coefﬁcients on SECURE and USBAN K are not statistically signiﬁcant. Similar to the
random effects estimation, the pooled OLS estimation shows that the coefficients on
SECURE and USBANK are not significant while the coefﬁcients on BD_MAG,
BD_LEN, and FIN are negative and signiﬁcant. The signiﬁcant coefﬁcient on FIN
implies that when within-ﬁrm effects are ignored in the panel dataset, accounting
conservatism is higher for a ﬁrm that has a deal with covenants based on ﬁnancial ratios.
Finally, Table 8 presents the Pearson correlations between bank dependence measures
and loan characteristics. These correlations are low in absolute magnitude.

In summary, the results from the multivariate tests show that bank loan
characteristics are not signiﬁcantly associated with a bank’s governance role and in turn,
a ﬁrm’s accounting conservatism. These results imply that the magnitude and length of
the bank loan play a more important role in a bank’s governance activities than bank loan

characteristics.

 

’3 The Spearman rank correlations among the loan characteristic variables are less than 0.10.

34

 

CHAPTER 6: SENSITIVITY ANALYSES

6.1. A tax explanation for accounting conservatism

The link between reported income and taxable income provides a firm with an
incentive to use conservative ﬁnancial reporting to reduce taxes as long as the ﬁrm is
proﬁtable, has taxable income, and interest rates are positive (Watts, 2003a). Therefore,
it is possible that tax incentives affect the ﬁndings of this paper. To rule out this
possibility, I reexamine the main hypotheses, controlling for a ﬁrm’s tax incentive.

Watts (2003b) states that when the links between tax accounting and ﬁnancial
reporting become closer, there are incentives for conservative financial reporting.
Therefore, I use a correlation between pre-tax book income and taxable income as the
ﬁrst measure of a ﬁrm’s tax incentive. I expect that the greater the correlation between
pre-tax book income and taxable income, the greater is the ﬁrm’s incentive to use
conservative ﬁnancial reporting. Therefore, I predict a positive relation between the
correlation and accounting conservatism if tax incentives exist for the sample ﬁrms.
Since discretionary accruals are inversely related to accounting conservatism, this implies
a negative relation between the correlation and discretionary accruals. Following Hanlon

et a1. (2003), pre-book income and taxable income are measured as follows:

Pre-tax book income = Pre-tax book income - Minority interest,
Taxable income = (Current federal and foreign income taxes) / the top U. S. statutory

tax rate — the change in net operating loss carry forwards.34

 

3’ The top US. statutory tax rates used are 34% for 1988-1993 and 35% for 1994-2001. For further
discussion of book income and taxable income, see Hanlon et al. (2003).

35

 

The correlation between pre-tax book income and taxable income is measured for the
sample period (i.e., from 1988 to 2001) per each ﬁrm. The main hypotheses are
reexamined by including the correlation as an additional control.

In addition to the ﬁrst measure, I use the marginal tax rate as the second measure
of a ﬁrm’s tax incentive. In their comprehensive survey of empirical tax research in
accounting, Shackelford and Shevlin (2001) state that one needs a marginal tax rate
estimate to control for possible tax effects. Shevlin (1990) and Graham (1996a) develop
a simulated marginal tax rate measure and Graham (1996b) shows that the simulated tax
rate is the best proxy for the true marginal tax rate. Watts (2003b) states that when taxes
increase, incentives to report conservative ﬁnancial reporting increases. Therefore, I
predict that the higher the marginal tax rate, the greater is the firm’s incentive to use
conservative ﬁnancial reporting. Similar to the ﬁrst measure, I predict a negative relation
between the marginal tax rate and discretionary accruals. The main hypotheses are
reexamined by using the simulated corporate marginal tax rate after interest expenses as
an additional control.35 The interaction term between the marginal tax rate and the book-
tax correlation is also included because the partial effect of the correlation on
conservatism (holding all other variables ﬁxed) might depend on the marginal tax rate
and vice versa.36 Accordingly, the estimation model with the measures of a ﬁrm’s tax

incentive is as follows:

 

’5 The simulated tax rates are available on http://www.duke.edu/~j graham. The results are qualitatively the
same when the simulated corporate marginal tax rate before interest expenses is used. For further
discussion of the simulated tax rate, see Graham (1996b).

’6 I do not have a directional prediction for the interaction term because it could have positive, negative,
and no effect on the dependent variable. For example, even though the book-tax correlation is high, a ﬁrm
might have no or little incentive to use conservative ﬁnancial reporting if it has the low marginal tax rate.
On the other hand, a ﬁrm might have more incentive to use conservative ﬁnancial accounting if it has both
a high correlation and high marginal tax rate.

36

 

DA_Perf“ = or; + a,BD_MAG,-, + azBD_LENi, + agINSTITUTIONl-t
+ m5%BLOCK,-, + a5INSIDERi, + a6080Di, + ayBIGi, +otgFINl-t
+ agLEVERAGEl-t + a10COVERAGEit+ a1 [ALTMANit + aIZM/Bit
+ a13LASSETSi, + aMROAi, + a15FREECFit+ a16C0RRl-t + anMTRl-t

+ 013C ORR“ *MTRit + 8," (4)
where,

DA_Perf = the value of discretionary accruals from the Performance matched
discretionary accrual model,
CORR = the correlation between pre-tax book income and taxable income
measured for 1988 — 2001,
MTR = the corporate marginal tax rate after interest expenses,

i, t = ﬁrm and year index, respectively,

 

other variables are deﬁned in equation (2). I

As reported in Table 9, the main results are unchanged after controlling for the
effect of the tax incentive. Consistent with Hla and Hlb, the coefﬁcients on BD_MAG, —
0.1487 and BD_LEN, -0.0006, are negative and statistically signiﬁcant (p-value = 0.054
and 0.015, respectively). Although the signs are negative as predicted, the coefﬁcients on
CORR and MTR, are not statistically signiﬁcant, implying that the tax incentive is not a

determinant of accounting conservatism for this sample of ﬁrms.

6.2. Basu (1997)’s conservatism measure

Basu (1997) interprets conservatism as an asymmetric veriﬁcation requirement
for recognizing good versus bad news in ﬁnancial statements, suggesting that earnings
reﬂects bad news more quickly than good news. Regarding debtholders, he states that
debtholders demand more timely information about bad news that might adversely affect

future cash ﬂows because they have an asymmetric loss function. Therefore, in this

37

chapter, I examine the association between accounting conservatism based on the Basu
(1997)’s conservatism measure and the magnitude and length of the bank loan. The
association is measured by using a modiﬁed version of the Basu (1997)’s model as

follows:
Xi/Pit-1 = a0 + G’IDRit + azRir + agRit’I‘DRit + aaBD_MAG*Rit*DRit
+ a5BD_LEN*Rit*DRi, + 81', (5)

where,

Xit = the earnings per share measured as income before extraordinary items /

 

(number of shares outstanding*stock split adjustment factor),
Pip] = the price per share at the beginning of the ﬁscal year divided by stock split
adjustment factor,
Rit = the buy-and-hold returns from nine months before ﬁscal year-end to three
months after ﬁscal year-end,
DRit 2 an indicator variable that equal 1 if Rit < 0, and 0 otherwise,

BD_MAG = the magnitude of a bank loan measured as (the dollar value of the bank
loan from the lead lender / total assets),
BD_LEN = the length of a bank loan measured as a maturity of the bank loan,

i, t = ﬁrm and year index, respectively.

Based on the sign of returns, Basu divides the sample into ‘good news’ group (i.e., firm-
year observations with positive returns) and ‘bad news’ group (i.e., ﬁrm-year
observations with negative returns). He predicts that under conservatism, the coefﬁcient

on R*DR, a3, will be signiﬁcantly positive, suggesting that the earnings-retum relation is

stronger during the bad news period. He also predicts that (to and or; will be positive and

38

signiﬁcant. This paper further predicts that our and (15 will be positive and signiﬁcant,
indicating that the extent to which earnings reﬂects bad news more quickly than good
news is increasing in both the magnitude and length of the bank loan.

Table 10 presents results from equation (5). Consistent with the Basu’s prediction,
a3, 0.1570, is signiﬁcantly positive (p-value < 0.01), implying that earnings is more
sensitive to negative returns than to positive returns. Both (10 and or; are signiﬁcantly
positive consistent with Basu’s predictions. As predicted, the coefﬁcient on
BD_MAG*R*DR, 0.2044, is positive and signiﬁcant (p-value = 0.084), implying that the
extent to which earnings reﬂects bad news more quickly than good news is increasing in
the magnitude of the bank loan. The coefﬁcient on BD_LEN*R*DR, 0.008, however, is
positive, but insigniﬁcant (p-value = 0.123).

Basu (l997)’s conservatism measure might not be appropriate for examining the
research questions of this paper because of the following reasons. First, this paper
focuses on accounting conservatism related to a ﬁrm’ discretionary ﬁnancial reporting
behavior. In other words, the focus of this paper is on conservatism in a ﬁrm’s ﬁnancial
reporting per se, rather than on how fast earnings reﬂect speciﬁc news. Second, Basu’s
measure relies on stock price movements to distinguish between good and bad news
(Givoly and Hayn, 2000). This paper, however, focuses on debt ﬁnancing, which is not

directly related to the stock market.37

 

’7 In their study of the joint effect of investment and conservative accounting on the accounting rate of
return, Penman and Zhang (2002) measure the effect of conservative accounting on the balance sheet.
Their measure of conservatism focuses on how GAAP accounting for inventories, R&D expenditures, and
advertising expenditures differentially impacts ﬁrms’ ﬁnancial statements. As a result, it is not relevant to
this paper, which focuses on conservatism related to a ﬁrm’s discretionary ﬁnancial reporting behavior.

39

 

6.3. Absolute values of discretionary accruals

Francis et al. (2002) deﬁne the quality of accounting accruals as the absolute
value of discretionary accruals. The absolute value might be of interest to banks because
the quality of accruals is related to the predictability of future cash ﬂows. Speciﬁcally,
accruals can be used to estimate borrowers’ future cash ﬂow generating ability, and the
absolute value of discretionary accruals can be one measure of the quality of accruals.
The higher the absolute value of the discretionary accruals, the lower the quality of
accruals. Therefore, I reexamine the main hypotheses using the absolute value of

discretionary accruals as a dependent variable. Table 11 reports results using the absolute

 

value of discretionary accruals. Results are consistent with predictions that the quality of ,
accruals (the absolute value of discretionary accruals) is increasing (decreasing) in both
magnitude and length of the bank loan. In the modiﬁed Jones model, the coefﬁcients on
BD_STR and BD_LEN, -0.1196 and 00007, are negative and statistically signiﬁcant (p-
value = 0.03 and <0.01, respectively). Similarly, in the performance-matched modiﬁed
Jones model, the coefﬁcients on BD_STR and BD_LEN, -0.0867 and -0.0006, are
negative and statistically significant (p-value = 0.09 and <0.01, respectively).

Given the signiﬁcance of the absolute value of discretionary accruals results, one
might argue that the quality of accruals is a primary driver of the main results of this
paper rather than accounting conservatism. To address this concern, I ﬁrst divide the
sample into two groups based on the signs of discretionary accruals, and then examine
the main hypotheses using the absolute value of discretionary accruals for positive and
negative discretionary accrual groups. For the positive discretionary accrual group, the

coefﬁcients on BD_STR and BD_LEN, -0.2016 and 00009 are negative and statistically

40

signiﬁcant (p-value = 0.026 and <0.01, respectively). In contrast, neither coefﬁcient is
signiﬁcant for the negative discretionary accrual group. These results imply that banks
are concerned about income-increasing discretionary accruals and perform governance
activities to reduce the level of positive discretionary accruals, but do not monitor the
level of negative discretionary accruals. This finding is consistent with the main
prediction that banks prefer accounting conservatism, demonstrating that accounting

conservatism is a primary driver of the main ﬁndings of this paper.

6.4. Estimation error in accruals

Banks might be concerned with estimation error in accruals because it is likely to
be related to the predictability of ﬁrms’ future cash ﬂows. Dechow and Dichev (2002)
measure the quality of accounting accruals as the magnitude of estimation error in
accruals. They argue that the quality of accruals and earnings is decreasing in the
magnitude of estimation error in accruals. One problem with this measure in relation to
the current study is that it is not appropriate to apply their data requirements (i.e., the
availability of at least eight years of data) to my sample. First, I lose too many
observations, and second, most sample loans are a ﬁve-year length, so the estimation
error is measured using a longer period than the length of the loan. As part of my
sensitivity analyses, I reexamine the main hypotheses using the estimation error model
and the less restrictive requirement that there be at least ﬁve years of data.

Following Dechow and Dichev (2002) and Francis et al. (2002), the estimation

error is calculated as the standard error of the residuals from the following regression:

TCAit = ﬂo ‘I' ﬂICFOit-I + ﬂzCFOit 'I' ﬂ3CFOit+1 +5” (6)

41

 

m.

where,

TCAit = firm i’s total current accruals in year t, calculated as ACAit — ACLit —
ACashit + AST_DEBTit,

CFOiH = ﬁrm i’s cash ﬂow from operations in year t-l, calculated as ﬁrm i’s net

income before extraordinary items in year t-l — ﬁrm i’s total accruals in year

t-l,
CFOit = ﬁrm i’s cash ﬂow from operations in year t , calculated as ﬁrm i’s net
income before extraordinary items in year t — ﬁrm i’s total accruals in year t,
CFOit+1 = ﬁrm i’s cash ﬂow from operations in year t+l , calculated as ﬁrm i’s net

income before extraordinary items in year t+1 — ﬁrm i’s total accruals in
year t+l,

i, t = ﬁrm and year index, respectively,

other variables are deﬁned in equation (1),

all variables are deﬂated by total assets in t-l.

Equation (6) is estimated annually for each two-di git SIC group with at least 20
observations.

The results using the estimation error model (Dechow and Dichev, 2002) with a
less restrictive data requirement (i.e., at least ﬁve years of data) are similar to the
modiﬁed Jones and the performance matched model for BD_MAG.38 As reported in
Table 12, the coefficient on BD_MAG, -0.0312, is statistically signiﬁcant (p-value =
0.037), indicating that accrual estimation errors (the quality of accruals) are decreasing

(increasing) in the magnitude of the bank loan. As for BD_LEN, however, the results

 

’8 McNichols also measures the quality of accounting accruals using a modiﬁed version of the Dechow and
Dichev (2002) model. The results using the McNichols (2002) model are qualitatively the same as those
using the Dechow and Dichev (2002) model.

42

using the estimation error models are no longer significant at conventional levels

although the signs are still negative.

6.5. Accounting conservatism before and after bank loans are made

In this chapter, I examine accounting conservatism as measured by discretionary
accruals before and after bank loans are made. The purpose of this analysis is to examine
changes in accounting conservatism before and after bank loans are made between high
and low bank-dependent firms. This analysis, however, is descriptive because other
relevant variables are not controlled. Moreover, the loan dates do not necessarily
represent the initiation of the ﬁrm-bank relationship.

Panel A of Figure 1 presents the comparison of mean discretionary accruals
between high and low bank-dependent ﬁrms. Consistent with the prediction that high
bank-dependent ﬁrms have low discretionary accruals, the mean discretionary accruals
for high bank-dependent ﬁrms is lower than those for low bank-dependent firms before
and after bank loans are made. After bank loans are made, the difference becomes larger.
Panel B of Figure B presents the comparison of median discretionary accruals between
high and low bank-dependent firms. Before bank loans are made, high bank-dependent
ﬁrms have higher median discretionary accruals than low bank-dependent ﬁrms. After
bank loans are made, however, high bank-dependent ﬁrms have signiﬁcantly lower
median discretionary accruals than low bank-dependent ﬁrms. This ﬁgure might indicate
that bank monitoring plays an important role in governing borrowers while their
screening does not, consistent with the notion that recent bank competition makes banks’

ex-ante screening more difﬁcult.

43

CHAPTER 7: CONCLUSION

This study provides evidence on the association between a firm’s accounting
conservatism and the ﬁrm’s bank dependence. Prior accounting research has examined
the speciﬁc role of debt covenants, but has not examined the more general governance
role of banks, despite their signiﬁcant role in the US. economy. Using 4,114 ﬁrm-year
observations over the period 1988-2001, this study finds a positive relationship between
conservatism in borrowers’ ﬁnancial reporting and both the magnitude and length of the
bank loan. In contrast, other corporate governance mechanisms, such as auditors,
institutional investors, 5% block holders, managerial ownership, and outside directors are
generally unassociated with the accounting conservatism of the sample ﬁrms. In addition,
my ﬁndings cannot be explained by other explanations for accounting conservatism such
as accounting-based debt covenants, litigation, regulation, and taxes.

These results are important for at least four reasons. First, the results indicate that
for bank-dependent ﬁrms, banks play an important role in monitoring financial reporting
quality, and their contributions are incremental to those of other corporate governance
mechanisms. This ﬁnding helps us to better understand the relation between US.
corporate governance mechanisms and ﬁnancial reporting quality. Second, the results
imply that the incentives creditors provide for conservatism in borrowers’ financial
reporting extend beyond the incentives in debt covenants. The results further imply that
banks induce borrowers to have higher accounting conservatism by performing
governance activities such as screening and/or monitoring. Third, this paper ﬁnds that

the magnitude and length of the bank loan play a more important role in banks’

governance activities than do loan characteristics. Speciﬁcally, there is little evidence of
a relation between borrowers’ accounting conservatism and loan characteristics such as
collateral and lender type. Finally, this paper ﬁnds that the length of the bank loan is
positively related to the intensity of banks’ governance activities. This result is relevant
to ongoing debates about the impact of the length of monitoring relationships on the
quality of the monitoring.

The results suggest several additional issues for future research. Of particular
interest is to distinguish between screening and monitoring, the two major governance
activities discussed in the current paper. This paper does not empirically discriminate
between the two governance activities. Instead, it focuses on the overall governance role
of banks. The two activities are, however, different in that screening is performed before
bank loans are made to avoid adverse selection problems, while monitoring is performed
after bank loans are made to prevent moral hazard problems. Therefore, an examination
of the differential roles of the two governance activities would further our understanding
of a bank’s governance role. In addition, an examination of other explanations for
accounting conservatism such as compensation contracts in relation to debt ﬁnancing
would further our understanding of the association between a firm’s ﬁnancial reporting
behavior and debtor-creditor relationships. Finally, an examination of the effects of ﬁrm-
bank relationships on other important accounting issues (e. g., valuation) would provide a

more complete understanding of the accounting consequences of private debt ﬁnancing.

45

Table 1
Sample description

Panel A: Descriptive statistics for all sample ﬁrm-year observations

 

 

25% 75%

Median Mean Percentile Percentile N
Bank loans - Total 20.40 104.37 6.50 75.00 4,114
Bank loans - Lead lender 15.00 31.60 6.00 34.01 4,114
Bank dependence - Magnitude 0.06 0.10 0.02 0.13 4,114
Bank dependence - Length 60 67 47 84 3,400
Short-term debt/Total Asset 0.02 0.05 0.00 0.05 4,108
Long-term debt/Total Asset 0.22 0.25 0.09 0.36 4,108
Common equity/Total Asset 0.44 0.43 0.30 0.59 4,105
Total Assets 250.90 1,287.91 79.01 848.85 4,108
Financial leverage 0.56 0.56 0.41 0.69 4,114
Interest coverage 5.75 16.36 2.95 11.28 4,019
Altman-Z score 2.94 3.76 1.93 4.35 4,031
Market-to-book ratio 1.78 2.67 1.05 2.93 4,063
ROA 0.05 0.03 0.01 0.08 4,108
Free cash ﬂow 23.53 148.09 5.94 86.73 4,017
Institutional investors 0.40 0.42 0.21 0.62 3,931
5% Block holders 0.37 0.40 0.17 0.61 3,931
Managerial ownership 0.12 0.20 0.02 0.30 3,923
Outside directors 0.67 0.66 0.57 0.78 3,958
Big-5 auditor 1 0.94 1 1 4,144
FIN 0 0.03 0 0 4,1 13
SECURE 1 0.80 1 1 2,612
USBANK 1 0.89 1 1 4,060

 

46

Table l - continued
Sample description

Panel B: Descriptive statistics for high vs. low bank dependent ﬁrm-year observations

 

 

 

Low bank-dependent High bank-dependent
(N = 2,057) (N = 2,057)

Median Mean Median Mean

Bank loans - Total 25.00 132.55 20.00 76.20
Bank loans - Lead lender 15.00 28.81 17.00 34.38
Bank dependence — Magnitude 0.02 0.03 0.13 0.17
Bank dependence — Length 60 71 60 63
Short-term debt/Total Asset 0.02 0.04 0.02 0.05
Long-term debt/Total Asset 0.23 0.26 0.20 0.24
Common equity/Total Asset 0.41 0.41 0.47 0.45
Total Assets 628.08 2,292.03 109.54 283.79
Financial leverage 0.58 0.58 0.53 0.55
Interest coverage 5.97 15.42 5.47 17.30
Altman-Z score 2.86 3.99 3.01 3.56
Market-to-book ratio 1.95 2.82 1.63 2.52
ROA 0.05 0.03 0.04 0.03
Free cash ﬂow 60.90 263.13 10.05 32.76
Institutional investors 0.51 0.50 0.31 0.34
5% Block holders 0.35 0.40 0.38 0.41
Managerial ownership 0.08 0.17 0.16 0.23
Outside directors 0.70 0.68 0.67 0.64
Big-5 auditor 1 0.96 1 0.92
FIN 0 0.04 0 0.02
SECURE 1 0.77 1 0.82
USBANK 1 0.84 1 0.94

 

47

 

‘ll

Table 1 - continued
Sample description

 

Bank loans (total and lead lenders), short- and long-term debt, common equity, market value of equity,
and total assets represent dollars in millions. Bank dependence - Length is in months.

Bank loans (total) are total bank loans in dollars in each ﬁrm-year. Bank loans (lead lender) are bank
loans in dollars from a lead lender in each ﬁrm-year. Bank dependence - Magnitude is the magnitude of a
bank loan measured as (the dollar value of the bank loan from the lead lender / total assets). Bank
dependence — Length is the length of a bank loan measured as a maturity of the bank loan. Financial
leverage is measured as (total liabilities / total assets). The interest coverage ratio is measured as
(operating income before depreciation / interest expenses). The Altman Z-Score is a measure of a ﬁrm’s
ﬁnancial health, calculated as [ 1.2 (working capital/total assets) + 1.4 (retained earnings/total assets) +
3.3 (earnings before interest and taxes/total assets) + 0.6 (market value of equity/book value of liabilities)
+ 1.0 (sales/total assets) ]. The market-to-book ratio is a measure of a ﬁrm’s growth opportunity,
calculated as (market value of equity / book value of equity). ROA is the return on assets measured as
(income before extraordinary items / beginning total assets). Free cash ﬂow is measured as [operating
income before depreciation — interest expenses - (income taxes — deferred tax & invest tax credit) -
dividends]. Institutional investors is proportion of institutional ownership. 5% Block holders is
proportion of 5% block holder ownership. Managerial ownership is proportion of managerial ownership.
Outside directors is proportion of outside directors on a ﬁrm’s board. Big-5 auditor is an indicator
variable that equals 1 if a ﬁrm’s auditor is a Big 5 accounting ﬁrm, and 0 if a ﬁrm’s auditor is a non-Big 5
accounting ﬁrm. FIN is an indicator variable that equals 1 if a deal includes ﬁnancial ratios, and 0 if a
deal dose not include ﬁnancial ratios. SECURE is an indicator variable that equals 1 if a loan is a
collateralized loan, and 0 if a loan is a non-collateralized loan. USBANK is an indicator variable that
equals 1 if a lender is a US. bank, and 0 if a lender is an Western European or Asian/Paciﬁc bank.

48

.ooeoeeonov x53 co woman Jeep—Homepages 32 can 5.: .3:on 95 9:. woe—>6 03 were 295m .oﬁawame—o 8: 2a meoaatomno
0 .3033030 Boxer—w :— v .89 mo 50 .380 A29 eeaaocamﬂu 353v:— Roam—“Sm “$6.95. .3 35.58% m. 95.53808 .3965

 

 

 

 

49

code code Vmod code Vmo.m code we .v :30...
50.3 mm .2 9cm 3.3 cum 9.: S b 803.com
was 3.3 mmm omd >3 8.9 mg numb 250%
$6 3.0 o3 and h: cad 5mm 03.5 232055
mmﬁv mmdm mm #4 Sdm N3; No.3 EQN wetBoEEa—Z
92 wwd 3 wwd f wwd om coco—5300
9:. 5mm 0: mod mmd mmﬁ Sm wee:
Ed cod o 33 a 85 a 23383
s as 2 as 2 as 2 83:33 bases
Sew Sex aeeuzemmﬁ Bennemen maﬁa =<
Neaaamﬁob «55 ~35 «SS 23

 

 

bases 2 83328.5 a8>-§_m .8 5233“:—
N 2.3

@900 80.00
00.0 00.0-
80.8
00.0

00.—

00.0

00.0

 

Hmmm< mSZ HA< >00 >m-H

ABE—.0530 5 0.0.5— 8535550 833.3.» 05050005 wee—:0 meets—ounce sewn—«om
m 030,—.

$0.00
no.0-
€0.00
5.0
80.8
0N.0

00.0

80.8
0N0
$0.00
00.0
80.8
3.0-
80.8
S .0-

00.0

80.00
00.0
80.8
00.0
ANNE
.000-
80.8
00.0-
80.00
00.0

00.0

80.8
0N0
868
5.0
20.8
~00-
8.8.00
5.0
A2 .00
No.0
80.8
#00

00.0

80.8
0nd
€0.00
m0.0
80 .8
no.0-
A0000
00.0-
80.8
0_ .0
Awm.0v
N00
80.8
_ _.0

004

80.8
0N0-
$0.00
00.0-
2 0.00
no.0
3900
_0.0
80.8
v0.0-
808
m0.0
$0.00
8.0-
80.8
2 .0-

00.—

 

a 00m QOmO QmZH

80.8
w0.0-
8N8
N00-
$0.00
v0.0-
8000
3.0-
@000
No.0
$0.00
8.0
2.000
~00
80.00
00.0-
$0.00
0m .0

00.0

8
an

80.8
mm .0
30.00
.000
80.8
00.0
80.00
m0.0
80.8
00.0-
80.8
00 .0
80.00
00 .0
80.8
3.0
80.8
N00-
808
m0.0

00.0

HmZH

80.8
2.0
0&8
00.0
80.00
8.0-
80.00
3.0-
80.00
8.0
$28
8.0-
08.8
8.0-
80.00
8.0
80.8
8.0-
$0.00
3.0-
80.8
8.0

00.0

25
lam

80.8
wvd-
$0.00
00.0-
80.8
00.0-
80.8
00.0-
$00.00
3.0-
80.8
no.0-
80.8
E .0-
80.8
N0 .0-
80.8
2 .0
20.00
00.0
80.8
mm.0-
80.8
00.0-
00.0
00.2
lam

Hmmm<

HA<

>00

>m§

2E

00m

QOmO

QmZH

4m exam

HmZH

ZmAIQm

O<EIQm

50

.1

30000030

I 000000 x3 000000 0% 0000 0000000 I 00000 0000000 I 00000900 00000000 I 00000000000 000000 0000000 w000000£ 00 000000000 300 0000 000 3 000000
.3030 300 w00000w00 \ 3003 000000000000 00000 0000000 00 000000000 30000 00 000000 05 3 <00

.3030 00 030> 0.000 0.0 00 w0_ 000000 05 3 Hmmm<

000000 00 030> 3000 1000.00 00 030.» 000—0000 00 000000000 0000 0.03-00-000300 0.0 3 03

A 3030 30000—000 0.0 + 00000000 .00 30> 300033000 00 30> 000.0000 0.0 + 30000 3050000

000 00000000 000000 30000000 m.m + 30000 3000000000 0003000 .0.— + 30000 30500000 30.0030 «.0 H 00 000000000 Boom-N 00002 0.0 3 0.3
00000900 00000000 \ 0000000 w00000000 00 00000008 0000 0w000>00 00000000 000 3 >00

.3030 300 \ 00000000 300 00 000000000 0w000>0_ 300000 3 >04

00000 300000 002000 000 0000 .000 0 .0 0 000 00000 300000 0000—000 _000 0 00 0 300.00 00.0 003000, 00000000 00 3 25

.0000 w00000000 m w0m-000 0 3 003000 0.0000 0 0 0 000 .0000 0.0000080 m 30 0 3 0000000 3.000 0 .0 0 30000 0000 0300003 000000000 00 3 03
00000 0.0000 0 00 000000000 00300 00 000000000 3 0000

000000030 300w00000 .00 000000000 3 QmZH

000000030 000.00 3003 00m 00 000000000 3 gm 00m

000000030 000030000 .00 000000000 3 .570

.002 3000 05 00 5002000 0 00 000000000 000— 3000 0 .00 5900. 0.0 3 ZmAIQm

.3030 300 \ 00003 000— 0.0 000 000. 0.000 05 00 0:_0> 00:00 0.00 00 000000000 000— 3000 0 00 00033000 05 3 0<2I00

20.00 80.00 Gm .00 2000 0.00.00 A0000 $0.00 A0000 80.00 A0000 A0000 A0000 80.00 A0000
00.0 mm .0 00.0 m0.0- 00.0- 00 .0 00.0 00.0 m0 .0 m 0 .0- m— .0- 2.10 00 .0 w 0 .0- ”Human
80.00 30.00 A0000 30.00 A0000 $00.00 80.00 80.00 $0.00 @000 A0000 $0.00 00.0
00. 0 w0.0 _0.0- NN.0 0N.0 00 .0- 8.0 00.0- 00.0- 00.0 N00- 3 .0 00.0 00.0- <00
“0d 2m: 03
<00 Hmmm< 002 0.400 >00 >0..— ZE 000 0000 DmZH 00m 0070 IQm I00

 

 

 

51

Table 4

Modiﬁed Jones model

Results of equation (2):

 

 

 

Random effects estimation Pooled OLS
Independent
Variables Coefﬁcient z-statistics Coefﬁcient t-statistics
BD_MAG -0. 1747 -2.93 * * * -0.0890 -2.83 ***
BD_LEN -0.0005 -2. 19* * -0.0002 -2.11* *

INSTITUTION 00273 -0.79 —0.0036 -0.23

5%BLOCK -0.0288 -1.09 -0.0015 -0.09

INSIDER 00014 -0.04 0.0010 0.04

OBOD -0.0324 -0.80 -0.0160 -0.78

BIG - - -0.0020 -0.10

FIN 0.0078 0.20 -0.0091 -0.69

LEVERAGE 0.0027 0.05 -0.0120 -0.68

COVERAGE -0.0001 -1.19 -0.0001 -1.40

ALTMAN 0.0033 2.1 1** 0.0020 1.19

M/B -0.0001 -0.40 -0.0000 -0.03

LASSETS -0.0100 -1.61 -0.0071 - l .64

ROA 0.0934 3.85*** 0.1676 1.60

FREECF 0.0000 1.06 0.000 1.54

R2 0.0223 0.0165
N 2,931 2,930

 

52

 

Table 4 - continued
Modiﬁed Jones model

 

***, ***, and * indicate signiﬁcance at the 1 percent, 5 percent, and 10 percent levels, respectively, for
two-tailed tests.

A dependent variable is the value of discretionary accruals from the modiﬁed Jones Model.
BD_MAG is the magnitude of a bank loan measured as (the dollar value of the bank loan from the lead
lender / total assets).

BD_LEN is the length of a bank loan measured as a maturity of the bank loan.

INSTITUTION is proportion of institutional ownership.

5%BLOCK is proportion of 5% block holder ownership.

INSIDER is proportion of managerial ownership.

OBOD is proportion of outside directors on a ﬁrm’s board.

BIG is an indicator variable that equals 1 if a ﬁrm’s auditor is a Big 5 accounting ﬁrm, and 0 if a ﬁrm’s
auditor is a non-Big 5 accounting ﬁrm.

FIN is an indicator variable that equals 1 if a deal includes ﬁnancial ratios, and 0 if a deal dose not
include ﬁnancial ratios.

LEVERAGE is ﬁnancial leverage measured as (total liabilities / total assets).

COVERAGE is the interest coverage ratio measured as (operating income / interest expenses).
ALTMAN 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/total assets) + 0.6 (market value of
equity/book value of liabilities) + 1.0 (sales/total assets) ].

M/B is the market-to-book ratio measured as (market value of equity / book value of equity).
LASSETS is the natural log of the book value of assets.

ROA is the return on assets measured as (income before extraordinary items / beginning total assets).
FREECF is free cash ﬂow measured as [operating income before depreciation — interest expenses —
(income taxes — deferred tax & invest tax credit) - dividends].

53

 

Table 5
Performance-matched modiﬁed Jones model

Results of equation (2):

 

 

 

Random eﬂects estimation Pooled OLS
Independent
Variables Coefficient z-statistics Coefficient t-statistics
BD_MAG -0. 1506 -2.70*** -0.0796 -2.74***
BD_LEN -0.0005 -2.53*** -0.0002 -2.58**
INSTITUTION -0.0250 -0.75 -0.0013 -0.08
5%BLOCK -0.0236 -0.95 0.0010 0.07
INSIDER -0.0130 -0.40 -0.0073 -0.35
OBOD -0.0276 -0.72 -0.0148 -0.72
BIG - - -0.0042 -0.25
FIN 0.0022 0.06 -0.0131 -1.26
LEVERAGE 0.0136 0.47 -0.0069 -0.37
COVERAGE -0.0001 -1.22 -0.0001 - l .39
ALTMAN 0.0022 1.46 0.0015 0.83
M/B -0.0001 -0.37 -0.0000 -0.04
LASSETS -0.0110 -1.83* -0.0079 - l .97*
ROA 0.0545 2.3 l ** 0.0651 0.77
FREECF 0.0000 1.00 0.0000 1.72
R2 0.0214 0.0080
N 3,127 3,126

 

54

Table 5 - continued
Performance-matched modiﬁed Jones model

 

***, **. and * indicate signiﬁcance at the 1 percent, 5 percent, and 10 percent levels, respectively, for
two-tailed tests.

A dependent variable is the value of discretionary accruals from the performance matched
discretionary accrual model.

BD_MAG is the magnitude of a bank loan measured as (the dollar value of the bank loan from the
lead lender / total assets).

BD_LEN is the length of a bank loan measured as a maturity of the bank loan.

INSTITUTION is proportion of institutional ownership.

5%BLOCK is proportion of 5% block holder ownership.

INSIDER is proportion of managerial ownership.

OBOD is proportion of outside directors on a ﬁrm’s board.

BIG is an indicator variable that equals 1 if a ﬁrm’s auditor is a Big 5 accounting ﬁrm, and 0 if a
ﬁrm’s auditor is a non-Big 5 accounting ﬁrm.

FIN is an indicator variable that equals 1 if a deal includes ﬁnancial ratios, and 0 if a deal dose not
include ﬁnancial ratios.

LEVERAGE is ﬁnancial leverage measured as (total liabilities / total assets).

COVERAGE is the interest coverage ratio measured as (operating income / interest expenses).
ALTMAN 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/total assets) + 0.6 (market value of
equity/book value of liabilities) + 1.0 (sales/total assets) ].

M/B is the market-to-book ratio measured as (market value of equity / book value of equity).
LASSETS is the natural log of the book value of assets.

ROA is the return on assets measured as (income before extraordinary items / beginning total assets).
FREECF is free cash ﬂow measured as [operating income before depreciation - interest expenses —
(income taxes - deferred tax & invest tax credit) — dividends].

55

Table 6
T-tests and Wilcoxon rank-sum tests for loan characteristics
- Performance matched discretionary accrual model

 

 

 

Wilcoxon
T- test 3 tests b
Loan
characteristics Categories Mean p-value Median p-value o
Collateral Secured 0.0230 0.20 -0.0027 0.09 0.2618
Unsecured 0.0094 -0.0046 0.2046
Lender types US. banks 0.0135 0.09 0.0573 0.23 0.1333
Western
Europe/
Asian banks 0.0012 0.0512 0.2311

 

Mean and median are the mean and median values of the values of discretionary accruals from the
Performance matched modiﬁed Jones model, respectively. The results are qualitatively the same
when the values from other accrual models are used.

' The t-test is the standard t-test for differences in means.
" The Wilcoxon rank-sum test (also known as a Mann-Whitney U test) is the nonparametric test for

differences in medians.

56

Table 7
Test for loan characteristics

Results of equation (3):

 

 

Random effects estimation Pooled OLS

Independent

Variables Cmfﬁgjgm z-ﬂaﬂ'ﬂigs g :chfigigm t-smn‘m'cs
BD_MAG -0.1825 -2.46*** -0.0832 -2.16**
BD_LEN -0.0007 -2.26** -0.0004 -1.96*
INSTITUTION -0.0327 -0.68 -0.0009 -0.03
5%BLOCK -0.0692 -1.93 -0.0300 -1. 12
INSIDER -0.0082 -0. 18 0.0021 0.07
OBOD -0.0638 -l.l6 -0.0415 -1.26
BIG - - 0.0016 0.08
FIN -0.0061 -0.13 -0.0238 -2.75** *
LEVERAGE 0.0262 0.69 0.0005 0.02
COVERAGE -0.0001 -0.95 -0.0001 -0.83
ALTMAN 0.0017 0.92 0.0007 0.39
MB -0.0001 -0.37 0.0000 0.07
LASSETS -0.0088 -0.97 -0.0080 41.16
ROA 0.0448 1.66* 0.0554 0.51
FREECF 0.0000 0.01 -0.0000 -0.02
SECURE 0.0062 0.26 -0.0046 -0.33
USBAN K 0.0067 0.18 -0.0003 -0.02
R2 0.0307 0.0097
N 1,968 1 ,967

 

57

Table 7 - continued
Test for loan characteristics

 

***, **, and * indicate signiﬁcance at the 1 percent, 5 percent, and 10 percent levels, respectively, for
two-tailed tests.

A dependent variable is the value of discretionary accruals from the performance matched discretionary
accrual model.

BD_MAG is the magnitude of a bank loan measured as (the dollar value of the bank loan from the lead
lender / total assets).

BD_LEN is the length of a bank loan measured as a maturity of the bank loan.

INSTITUTION is proportion of institutional ownership.

5%BLOCK is proportion of 5% block holder ownership.

INSIDER is proportion of managerial ownership.

OBOD is proportion of outside directors on a ﬁrm’s board.

BIG is an indicator variable that equals 1 if a ﬁrm’s auditor is a Big 5 accounting ﬁrm, and 0 if a ﬁrm’s
auditor is a non-Big 5 accounting ﬁrm.

FIN is an indicator variable that equals 1 if a deal includes ﬁnancial ratios, and 0 if a deal dose not
include ﬁnancial ratios.

LEVERAGE is ﬁnancial leverage measured as (total liabilities / total assets).

COVERAGE is the interest coverage ratio measured as (operating income / interest expenses).
ALTMAN 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/total assets) + 0.6 (market value of
equity/book value of liabilities) + 1.0 (sales/total assets) ].

W8 is the market-to-book ratio measured as (market value of equity / book value of equity).
LASSETS is the natural log of the book value of assets.

ROA is the return on assets measured as (income before extraordinary items / beginning total assets).
FREECF is free cash ﬂow measured as [operating income before depreciation - interest expenses -
(income taxes — deferred tax & invest tax credit) - dividends].

SECURE is an indicator variable that equals 1 if a loan is a collateralized loan, and 0 if a loan is a non-
collateralized loan.

USBAN K is an indicator variable that equals 1 if a lender is a US. bank, and 0 if a lender is an Western
European or Asian/Paciﬁc bank.

58

Table 8
Pearson Correlations among bank dependence and loan characteristics
(Signiﬁcance levels in parentheses)

 

 

BD_MAG BD_LEN SECURE USBANK
BD_MAG 1.00
BD_LEN -0.06 1 .00
(0.00)
SECURE 0.10 0.07 1.00
(0.00) (0.00)
USBANK 0.12 -0.03 -0.05 1.00
(0.00) (0.06) (0.01)

 

BD_MAG is the magnitude of a bank loan measured as (the dollar value of the bank loan from the
lead lender / total assets).

BD_LEN is the length of a bank loan measured as a maturity of the bank loan.

SECURE is an indicator variable that equals 1 if a loan is a collateralized loan, and 0 if a loan is a
non-collateralized loan.

USBAN K is an indicator variable that equals 1 if a lender is a US. bank, and 0 if a lender is an
Western European or Asian/Paciﬁc bank.

59

Table 9

Performance-matched modiﬁed Jones model with the measures of tax incentives

Results of equation (4):

 

 

 

Random effects estimation Pooled OLS
Independent ﬂ
Variables Coefficient z-statistics Coefﬁcient t-statistics F]
BD_MAG -0.1487 -1.93** -0.0772 ~2.47** .-
BD_LEN -0.0006 -2.43*** -0.0003 - l .73*
INSTITUTION -0.03 l 2 -0.71 0.0144 0.61 Lj
5 %BLOCK -0.0162 -0.50 0.0091 0.45
INSIDER -0.0726 -1.70* -0.0506 -2.41**
OBOD -0.0325 -0.65 -0.0319 -1.28
BIG - - -0.0144 -0.62
FIN 0.0118 0.25 0.0156 1.11
LEVERAGE -0.0090 -0.24 -0.0133 -0.66
COVERAGE -0.0001 -1.89** -0.0001 -2.36**
ALTMAN 0.0022 1.18 0.0027 1.29
M/B -0.0002 -0.84 -0.0001 -0.70
LASSETS -0.0142 -1.73* -0.0104 -2.32**
ROA 0.0292 1.16 0.0145 -0.09

 

 

60

Table 9 - continued
Performance-matched modiﬁed Jones model with the measures of tax incentives

 

 

 

Random eﬂects estimation Pooled OLS
Independent
Variables Coefﬁcient z-statistics Coefficient t-statistics
FREECF 0.0000 1.35 0.0000 1.90*
CORR -0.0277 -1.06 -0.0314 -1.69
MTR -0.0728 -0.91 -0.1196 - l .20
CORR*MTR 0.1349 1.40 0.1618 1.73
R2 0.0290 0.0152
N 1,828 1,827

 

***, **, and * indicate signiﬁcance at the 1 percent, 5 percent, and 10 percent levels, respectively,
for two-tailed tests.

A dependent variable is the value of discretionary accruals from the performance matched
discretionary accrual model.

BD_MAG is the magnitude of a bank loan measured as (the dollar value of the bank loan from the
lead lender / total assets).

BD_LEN is the length of a bank loan measured as a maturity of the bank loan.

INSTITUTION is proportion of institutional ownership.

5%BLOCK is proportion of 5% block holder ownership.

INSIDER is proportion of managerial ownership.

OBOD is proportion of outside directors on a ﬁrm’s board.

BIG is an indicator variable that equals 1 if a ﬁrm’s auditor is a Big 5 accounting ﬁrm, and 0 if a
ﬁrm’s auditor is a non-Big 5 accounting ﬁrm.

FIN is an indicator variable that equals 1 if a deal includes ﬁnancial ratios, and 0 if a deal dose not
include ﬁnancial ratios.

LEVERAGE is ﬁnancial leverage measured as (total liabilities / total assets).

COVERAGE is the interest coverage ratio measured as (operating income / interest expenses).
ALTMAN 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/total assets) + 0.6 (market value of
equity/book value of liabilities) + 1.0 (sales/total assets) ].

MB is the market-to-book ratio measured as (market value of equity / book value of equity).
LASSETS is the natural log of the book value of assets.

ROA is the return on assets measured as (income before extraordinary items / beginning total
assets).

FREECF is free cash ﬂow measured as [operating income before depreciation - interest expenses —
(income taxes — deferred tax & invest tax credit) — dividends].

CORR is the correlation between pre-tax book income and taxable income measured for 1988 -
2001.

MTR is the corporate marginal tax rate after interest expenses.

61

 

 

Table 10
Basu (1997)’s conservatism measure

Results of equation (5):

 

 

 

Basu (I 997) Modified Basu (I 997)
Independent
Variables Coefﬁcient t-statistics Coefﬁcient t-statistics
Intercept 0.0446 7.56*** 0.0485 8.32***
DR -0.0106 -0.92 -0.0020 -0.17
R 0.0160 3.32*** 0.0148 3.18***
R*DR 0.1879 6.94*** 0.1570 3.65***
BD_MAG*R*DR 0.2044 1.73*
BD_LEN*R*DR 0.0008 1.54
Adjusted R2 0.0458 0.0616
N 3,827 3,147

 

A dependent variable is the earnings per share. measured as [income before extraordinary items /
(number of shares outstanding*stock split adjustment factor)], divided by [the price per share at the
beginning of the ﬁscal year / stock split adjustment factor].

R is the buy-and-hold returns from nine months before ﬁscal year-end to three months after ﬁscal
year-end.

DR is an indicator variable that equal 1 if R < 0, and 0 otherwise.

BD_MAG is the magnitude of a bank loan measured as (the dollar value of the bank loan from the
lead lender / total assets).

BD_LEN is the length of a bank loan measured as a maturity of the bank loan.

62

Table 11

Absolute values of discretionary accruals

 

 

 

 

Modified Jones Performance matched
Independent
Variables Coefﬁcient t-statistics Coefﬁcient t-statistics
BD_MAG -0.1196 -2.22** -0.0867 -1.71*
BD_LEN -0.0007 -3.1 l*** -0.0006 -3.21***
INSTITUTION -0.05 86 -1.79* -0.0473 -1.50
5%BLOCK -0.0185 -0.74 -0.0219 -0.92
INSIDER -0.0062 -0. 19 -0.0077 -0.25
OBOD 0.0045 0.12 0.0045 0.12
FIN -0.0004 -0.01 0.0036 0.10
LEVERAGE 0.0305 1.16 0.0199 0.80
COVERAGE -0.0000 -0.21 0.0000 0.01
ALTMAN 0.0019 1.40 0.0019 1.46
MB -0.0001 -0.31 -0.0001 -0.45
LASSETS -0.0178 -3.12*** -0.0169 -3.04***
ROA -0.0149 -0.75 0.0031 0.16
FREECF 0.0000 1.43 0.0000 1.42
R2 0.0461 0.0419
N 2,930 3,127

 

63

Table 11 - continued
Absolute values of discretionary accruals

 

***, **, and * indicate signiﬁcance at the 1 percent, 5 percent, and 10 percent levels, respectively, for
two-tailed tests.

A dependent variable is the absolute value of discretionary accruals from the modiﬁed Jones or

performance matched discretionary accrual model.

BD_MAG is the magnitude of a bank loan measured as (the dollar value of the bank loan from the

lead lender / total assets).

BD_LEN is the length of a bank loan measured as a maturity of the bank loan. . ‘
INSTITUTION is proportion of institutional ownership. F1
5%BLOCK is proportion of 5% block holder ownership.

INSIDER is proportion of managerial ownership.

OBOD is proportion of outside directors on a ﬁrm’s board.

FIN is an indicator variable that equals 1 if a deal includes ﬁnancial ratios, and 0 if a deal dose not
include ﬁnancial ratios.

LEVERAGE is ﬁnancial leverage measured as (total liabilities / total assets). ,
COVERAGE is the interest coverage ratio measured as (operating income / interest expenses). j
ALTMAN is the Altman Z-Score measured as [ 1.2 (working capital/total assets) + 1.4 (retained i1."
earnings/total assets) + 3.3 (earnings before interest and taxes/total assets) + 0.6 (market value of
equity/book value of liabilities) + 1.0 (sales/total assets) ].

M/B is the market-to-book ratio measured as (market value of equity / book value of equity).
LASSETS is the natural log of the book value of assets.

ROA is the return on assets measured as (income before extraordinary items / beginning total assets).
FREECF is free cash ﬂow measured as [operating income before depreciation — interest expenses —
(income taxes — deferred tax & invest tax credit) — dividends].

i is a company index, and t is a year index.

 

 

3! vi .1 1E, .11.“ #1.!qu I .
rt -

Table 12

Estimation error model — Dechow and Dichev (2002)

Results of equation (6):

 

 

Independent
Variables Coefﬁcient z-statistics
BD_MAG -0.0312 -2.08**
BD_LEN -0.0001 -1.23
INSTITUTION 0.0136 1.66*
5%BLOCK 0.0010 0.17
INSIDER 0.0077 0.96
OBOD 0.0175 1.78*
FIN 0.0026 0.31
LEVERAGE -0.0037 -0.50
COVERAGE 0.0000 1.18
ALTMAN 0.0006 165*
WB 0.0000 0.20
LASSETS -0.0031 -2.01**
ROA 0.0029 0.55
FREECF 0.0000 2.04**
R2 0.0613
N 1,547

 

65

 

Table 12 - continued
Estimation error model — Dechow and Dichev (2002)

 

***, **, and * indicate signiﬁcance at the 1 percent, 5 percent, and 10 percent levels, respectively, for
two-tailed tests.

A dependent variable is the estimation error calculated as equation (6).

BD_MAG is the magnitude of a bank loan measured as (the dollar value of the bank loan from the lead
lender / total assets).

BD_LEN is the length of a bank loan measured as a maturity of the bank loan.

INSTITUTION is proportion of institutional ownership.

5%BLOCK is proportion of 5% block holder ownership.

INSIDER is proportion of managerial ownership.

OBOD is proportion of outside directors on a ﬁrm’s board.

FIN is an indicator variable that equals 1 if a deal includes ﬁnancial ratios, and 0 if a deal dose not
include ﬁnancial ratios.

LEVERAGE is ﬁnancial leverage measured as (total liabilities / total assets).

COVERAGE is the interest coverage ratio measured as (operating income / interest expenses).
ALTMAN 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/total assets) 4» 0.6 (market value of
equity/book value of liabilities) + 1.0(sales/tota1 assets) ].

M/B is the market-to-book ratio measured as (market value of equity / book value of equity).
LASSETS is the natural log of the book value of assets.

ROA is the return on assets measured as (income before extraordinary items / beginning total assets).
FREECF is free cash ﬂow measured as [operating income before depreciation — interest expenses -
(income taxes - deferred tax & invest tax credit) — dividends].

i is a company index, and t is a year index.

66

 

Panel A: Mean discretionary accruals from Year -2 to Year +2

0.0700

0.0600

0.0500

0.0400

0.0300

0.0200

0.0100

0.0000

-0.0100

-0.0200

Figure 1
Comparisons of mean and median discretionary accruals
before and after bank loans are made

Mean discretionary accruals

 

 

 

 

 

 

 

‘ 7
\ I
\
s
\

 

 

I
I
\ I
\ I
\ I
\ I
\ I
\ I
T

 

 

 

 

Year -2 V Year 0 Year +1 Year +2
Year -1
""""" v ' ' ' ' ‘ ' ' Low bank-dependent ﬁrms

High bank-dependent ﬁrms

 

Year 0 represents the year when bank loans are made.

The sample ﬁrms are divided into four groups based on their bank dependence. The low bank-dependent
ﬁrms are the ﬁrms in the ﬁrst quartile and the high bank-dependent ﬁrms are the ﬁrms in the fourth quartile.

The number of observations for low and high bank-dependent ﬁrms is 1,028 each.

67

 

Panel B

0.0150

0.0100

0.0050

0.0000

-0.0050

-0.0100

-0.0150

......... ‘.------.

.' Median discretionary accruals from Year -2 to Year +2

Median discretionary accruals

 

 

 

 

 

 

 

 

‘\
I’ ‘s
‘0
l
ear +1 Year +2

 

 

 

Low bank-dependent ﬁrms

High bank-dependent firms

Year 0 represents the year when bank loans are made.

The sample ﬁrms are divided into four groups based on their bank dependence. The low bank-dependent
ﬁrms are the ﬁrms in the ﬁrst quartile and the high bank-dependent ﬁrms are the ﬁrms in the fourth quartile.

The number of observations for low and high bank-dependent ﬁrms is 1,028 each.

68

 

{Wm .
L! 12;: .

BIBLIOGRAPHY

Ahmed, A., Billings, B., Morton, R., Stanford-Harris, M., 2002. The role of accounting
conservatism in mitigating bondholder-shareholder conﬂicts over dividend policy and in
reducing debt costs. The Accounting Review 77, 867-890.

Ajninkya, B., Bhojraj, S., Sengupta, P., 2003. The governance effect of institutional
investors and outsider directors on the properties of management earnings forecasts. T
Working paper, University of Florida. '-

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

 

Beasley, M., 1996. An empirical analysis of the relation between the board of director i a
composition and financial statement fraud. The Accounting Review 71, 443-465. L:

Beatty, A., Ramesh, K., Weber, J ., 2002. The importance of accounting changes in debt
contracts: The cost of ﬂexibility in covenants calculations. Journal of Accounting and
Economics 33, 205-227.

Beatty, A., Weber, J ., 2003. The effects of debt contracting on voluntary accounting
method changes. The Accounting Review 78, 119-142.

Beaver, W., 1998. Financial reporting: An accounting revolution. Prentice Hall, New
Jersey.

Becht, M., Bolton, P., Roell, A, 2002. Corporate Governance and Control. National
Bureau of Economic Research (NBER) Working Paper 9371, Cambridge, MA.

Becker, C., Defond, M., J iambalvo, J ., Subramanyam, K., 1998. The effect of audit
quality on earnings management. Contemporary Accounting Research 15, 1-24.

Begley, J ., 1990. Debt covenants and accounting choice. Journal of Accounting and
Economics 12, 125-139.

Begley, J ., Feltham, G., 1999. An empirical examination pf the relation between debt
contracts and management incentives. Journal of Accounting and Economics 27, 229-

259.

Begley, J ., Freedman, R., 2004. The changing role of accounting numbers in public
lending agreements. Accounting Horizons 18, 81-96.

69

 

Berlin, M, Loyes, J ., 1988. Bond covenants and delegated monitoring. Journal of
Finance 43, 397-412.

Berger, A., Udell, G., 1995. Relationship lending and lines of credit in small ﬁrm ﬁnance.
Journal of Business 68, 351-381.

Black, B., 1992. Institutional investors and corporate governance: The case for
institutional voice. Journal of Applied Corporate Finance 5, 19-32.

Blackwell, D., Winters, D., 1997. Banking relationships and the effects of monitoring on
loan pricing. The Journal of Financial Research 20, 275-289.

Bushee, B., Noe, C., 2000. Corporate disclosure practices, institutional investors, and
stock return volatility. Journal of Accounting Research 38, 171-202.

Choi, W., 2003. Debt ﬁnancing and the value relevance of the balance sheet and income
statement: The role of bank dependence. Working paper, Michigan State University.

 

Core, J., Holthausen, R., Larcker, D., 1999. Corporate governance, chief executive E
ofﬁcer compensation, and ﬁrm performance. Journal of Financial Economics 51, 371-406.

DeAngelo, H., DeAngelo, L., Skinner, D., 1994. Accounting choice in troubled
companies. Journal of Accounting and Economics 17, 113-143.

Dechow, P., Dichev, I., 2002. The quality of accruals and earnings: The role of accrual
estimation errors. The Accounting Review 77, 35-59.

Dechow, P., Skinner, D., 2000. Earnings management: Reconciling the views of
accounting academics, practitioners, and regulators. Accounting Horizons 14, 235-250.

Dechow, P., Sloan, R., Sweeney, A., 1995. Detecting earnings management. The
Accounting Review 70, 193-225.

DeFond, M., Jiambalvo, J. 1991. Incidence and circumstances of accounting errors. The
Accounting Review 66, 643-655.

DeFond, M., Jiambalvo, J. 1994. Debt covenant effects and the manipulation of accruals.
Journal of Accounting and Economics 17, 145-176.

Diamond, D., 1984. Financial interrnediation and delegated monitoring. Review of
Financial Studies 51, 393-414.

Diamond, D., 1991. Monitoring and reputation: The choice between bank loans and
directly placed debt. Journal of Political Economy 99, 689-721.

70

Dichev, I., Skinner, D., 2002. Large-sample evidence on the debt covenants hypothesis.
Journal of Accounting Research 40, 1091-1123.

Elsas, R., Krahnen, J ., 1998. Is relationship lending special? Evidence from credit-ﬁle
data in Germany. Journal of Banking and Finance 22, 1283-1316.

Fama, E., 1985. What’s different about banks? Journal of Monetary Economics 15, 29-
36.

Francis, J ., LaFond, R., Olsson, P., Schipper, K., 2002. The market pricing of earnings
quality. Working paper, Duke University.

J

 

Francis, J ., Philbrick, D., Schipper, K., 1994. Shareholder litigation and corporate i: q;
disclosures. Journal of Accounting Research 32, 1994.
Francis, J., Maydew, E., Sparks, H., 1999. The role of Big 6 auditors in the credible

reporting of accruals. Auditing: A Journal of Practice & Theory 18, 17-34. l
Francis, J ., LaFond, R., Olsson, P., Schipper, K., 2003. Accounting anomalies and U

information uncertainty. Working paper, Duke University.

Frankel, R., McNichols, M., Wilson, G., 1995. Discretionary disclosure and external
ﬁnancing. The Accounting Review 70, 135-150.

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

Freixas, X., Rochet, J., 1997. Microeconomics of Banking. MIT Press, Cambridge, MA.

Gale, D., Hellwi g, M., 1985. Incentive compatible debt contracts: The one-period
problem. Review of Economic Studies 52, 647-663.

Givoly, D., Hayn, C., 2000. The changing time-series properties of earnings, cash ﬂows
and accruals: Has ﬁnancial reporting more conservative? Journal of Accounting and

Economics 29, 287-320.

Graham, J ., 1996a. Debt and the marginal tax rate. Journal of Financial Economics 41,
41-74.

Graham, J ., 1996b. Proxies for the marginal tax rate. Journal of Financial Economics 42,
187-221.

Guay, W., Kothari, 8., Watts, L., 1996. A market-based evaluation of discretionary
accrual models. Journal of Accounting Research 34, 83-105.

71

Hanlon, M., Kelley, S., Shevlin, T., 2003. Evidence on the possible information loss of
conforming book income and taxable income. Working paper, University of Washington.

Haubrich, J ., 1989. Financial intermediation, delegated monitoring, and long-term
relationships. Journal of Banking and Finance 13, 9-20.

Hausman, J ., 1978. Speciﬁcation tests in econometrics. Econometrica 46, 1251-1271.

Healy, P., Palepu, K., 1990. The effectiveness of accounting-based dividend covenants.
Journal of Accounting and Economics 12, 97-123.

Healy, P., Wahlen, M., 1999. A review of the earnings management literature and its I'm-:1.
implications for standard setting. Accounting Horizons 13, 365-383. -“~

Hoshi, T., Kashyap, A., Scharfstein, D., 1990. The role of banks in reducing the costs of
ﬁnancial distress in Japan. Journal of Financial Economics 27, 67-88.

 

Jones, J., 1991. Earnings management during important relief investigations. Journal of S J
Accounting Research 29, 193-228. E. _ :.

Kang, J ., Shivdasani, A., 1995. Firm performance, corporate governance, and top
executive turnover in Japan. Journal of Financial Economics 38, 29—58.

Kang, J ., Shivdasani, A., Yamada, T., 2000. The effect of bank relationships on
investment decisions: An investigation of Japanese takeover bids. Journal of Finance 55,
2197-2218.

Kaplan, S., 1994 Top executive rewards and ﬁrm performance: A comparison of Japan
and the United States. Journal of Political Economy 102, 510—546.

Kaplan, S., Minton, B., 1994. Appointments of outsiders to Japanese boards:
Determinants and implications for managers. Journal of Financial Economics 36, 225—
258.

Klein, A., 2002. Audit committee, board of director characteristics, and earnings
management. Journal of Accounting and Economics 33, 375-400.

Kothari, S., 2001. Capital markets research in accounting. Journal of Accounting and
Economics 31, 105-231.

Kothari, S., Leone, A., Wasley, C., 2002. Performance matched discretionary accrual
measures. Working paper, Massachusetts Institute of Technology.

Kroszner, S., Strahan, P., 2001. Bankers on boards of directors: Monitoring, conﬂicts of
interest and lender liability. Journal of Financial Economics 62, 415—452.

72

Lang, M., Lundholm, R., 2000. Voluntary disclosure and equity offering: Reducing
information asymmetry or hyping the stock. Contemporary Accounting Research 17,
623-662.

Leftwich, R., 1983. Accounting information in private markets: Evidence from private
lending agreements. The Accounting Review 58, 23-42.

McNichols, M., 2002. Discussion of the quality of accruals and earnings: The role of
accrual estimation errors. The Accounting Review 77 , 61-69.

Milgrom, P., Roberts, J ., 1992. Economics, Organization & Management. Prentice-Hall,
Englewood Cliffs, NJ. :—

Mishikin, F ., 2002. The economics of money, banking, and ﬁnancial markets. Addison ;
Wesley, Boston. '

 

Mishkin, F., Eakins, S., 2003. Financial markets and Institutions. Addison Wesley, ‘
Boston. '

Mugasha, A., 1997. The law of multi-bank financing. McGill-Queen’s University Press,
Montreal.

Ongena, S., Smith, D., 2000. Bank relationships: A review. The performance of
Financial Institutions, Cambridge University Press, Cambridge.

Petersen, M., Rajan, R., 1994. The beneﬁts of lending relationships: Evidence from
small business data. Journal of Finance 49, 3-37.

Press, E., Weintrop, J ., 1990. Accounting-based constraints in public and private debt
agreements: Their association with leverage and impact on accounting choice. Journal of

Accounting and Economics 12, 65-95.

Rajan, R., Winton, A., 1995. Covenants and collateral as incentives to monitor. Journal
of Finance 50, 1113-1135.

Shackelford, D., Shevlin, T., 2001. Empirical tax research in accounting. Journal of
Accounting and Economics 31, 321-387.

Shevlin, T., 1990. Estimating corporate marginal tax rates with asymmetric tax treatment
of gains and losses. Journal of the American Taxation Association 11, 51-67.

Shleifer, A., Vishny, R., 1986. Large shareholders and corporate control. Journal of
Political Economy 94, 461-488.

Shleifer, A., Vishny, R., 1997. A survey of corporate governance. Journal of ﬁnance 52,
737-783.

73

Strahan, P., 1999. Borrower risk and the price and nonprice terms of bank loans.
Working Paper, Federal Reserve Bank of New York.

Subramanyam, K., 1996. The pricing of discretionary accruals. Journal of Accounting
and Economics 22, 249-281.

Sweeney, A. 1994. Debt covenant violations and managers' accounting responses.
Journal of Accounting and Economics 17, 281-308.

Tech, 8., Welch, I, Wong, T., 1998. Earnings management and the post-issue
performance of seasoned equity offerings. Journal of Financial Economics 50, 63-99.

The Ofﬁce of the Comptroller of the Currency, 2001. Rating credit risk. Comptroller’s r21
Handbook, Department of Treasury. '

Townsend, R., 1979. Optimal contracts and competitive markets with costly state
veriﬁcation. Journal of Economic Theory 21, 265-293. [J-

 

Treacy, W., Carey, M., 1998. Credit risk rating at large US. banks. Federal Reserve
Bulletin (November), 897-921.

Warﬁeld, T., Wild, J ., Wild, K., 1995. Managerial ownership, accounting choices, and
informativeness of earnings. Journal of Accounting and Economics 20, 61-92.

Watts, R., 20033. Conservatism in accounting part I: Explanation and implications.
Accounting Horizon 17, 207-221.

Watts, R., 2003b. Conservatism in accounting part H: Evidence and research
opportunities. Accounting Horizon 17, 287-301.

White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a
direct test for heteroskedasticity. Econometrica 48, 817-838.

Williamson, 8., 1986. Costly monitoring, ﬁnancial intermediation, and equilibrium credit
rationing. Journal of Monetary Economics 18, 159-179.

Wooldridge, J ., 2002. Econometric analysis of cross section and panel data. The MIT
Press, Cambridge, Massachusetts.

74