bank monitoring and accounting recognition: the case · pdf filebank monitoring and accounting...

47
Bank Monitoring and Accounting Recognition: The case of aging-report requirements Richard Frankel Olin Business School Washington University in St. Louis Campus Box 1133 One Brookings Drive St. Louis, MO 63130-4899 [email protected] Bong Hwan Kim American University Kogod School of Business 4400 Massachusetts Avenue, NW Washington, DC 20016 [email protected] Tao Ma Moore School of Business University of South Carolina 1705 College Street Columbia, SC 29208 [email protected] Xiumin Martin Olin Business School Washington University in St. Louis Campus Box 1133 One Brookings Drive St. Louis, MO 63130-4899 [email protected] First draft: December 2010 Revised: September 31 2011 We thank seminar participants at the University of Chicago.

Upload: ngodan

Post on 26-Mar-2018

218 views

Category:

Documents


2 download

TRANSCRIPT

Bank Monitoring and Accounting Recognition:

The case of aging-report requirements

Richard Frankel Olin Business School

Washington University in St. Louis Campus Box 1133

One Brookings Drive St. Louis, MO 63130-4899

[email protected]

Bong Hwan Kim American University

Kogod School of Business

4400 Massachusetts Avenue, NW

Washington, DC 20016

[email protected]

Tao Ma Moore School of Business

University of South Carolina 1705 College Street Columbia, SC 29208

[email protected]

Xiumin Martin Olin Business School

Washington University in St. Louis Campus Box 1133

One Brookings Drive St. Louis, MO 63130-4899

[email protected]

First draft: December 2010

Revised: September 31 2011

We thank seminar participants at the University of Chicago.

Bank Monitoring and Accounting Recognition:

The case of aging-report requirements

Abstract

We study changes in borrower accounting recognition surrounding initiation of loans requiring

the provision of aging schedules to the lender. Our purpose is to understand how scrutiny by

lenders of underlying transactions affects financial reporting incentives. We find that allowance

for doubtful accounts increases significantly after loan initiation controlling for current and

future write-offs, receivable turnover, and the beginning allowance balance. This increase is

more pronounced for loans with increased monitoring frequency. We also find that write-offs

are less persistent following implementation of bank monitoring, consistent with increased

timeliness. Further study of the customer base finds customer concentration declines and credit

quality of largest customers improves after initiation of borrowing base loans. Lastly, we find

borrowers increase the frequency of allowance-for-doubtful-accounts disclosure in their

quarterly financial statements after loan initiation. Our results confirm two notions. Banks add to

the oversight that already exists for public, audited companies and banks influence borrowers to

adopt more conservative accounting policies.

JEL: G14, G21, G24, G28

Key Words: Bank monitoring, borrowing-base loan, aging-report, write-off

1

I. Introduction

We study whether bank monitoring affects accounting recognition. We identify

loan contracts with covenants requiring the borrower to provide periodic accounts

receivable aging reports to the lender and measure changes in the borrower’s recognition

of allowance for doubtful accounts before and after loan initiation. We also examine

whether these changes are related to monitoring intensity measured via frequency of

aging reports. We find that borrowers report higher allowance balances after borrowing,

and the increase is more pronounced for loans when transmittal of aging reports is more

frequent. Persistence of write-offs declines after borrowing—an indication that write-

offs become more timely. Changes in accounting policy could reflect more lenient

borrower trade-credit policies. However, after an initial borrowing-base loan, sales

attributable to customers accounting for ten percent or more of sales declines and the debt

ratings of these large customers improves. Overall, our evidence suggests that

accounting policies become more conservative after initiation of aging report

requirements and this change is not explained by the change in underlying economic

factors.

Our purpose is to understand how the initiation of bank monitoring affects

financial reporting choices. Loans that require aging reports are an excellent setting for

such investigation. This venue enables us to formulate a hypothesis that directs us to

study a particular accounting policy and to use a test calibrated by our understanding of a

specific account. The majority of loans with aging report requirement in our sample are

borrowing base revolvers using accounts receivable as collateral and/or to determine the

maximum loan amount (Flannery and Wang 2011). Traditionally these loans are

2

considered to be risky (Vinter 1998, p. 371). To control the credit risk, lenders request

timely financial reports and aging schedules from the borrower to assess the quality of

borrowing base assets.

Research finds a relation between measures of borrower accounting quality and

the monitoring or screening of higher-reputation banks (Ahn and Choi 2009; Bushman

and Wittenberg-Moerman 2010). Mester et al. (2007) study receivable and inventory-

based loans made by one bank to small, management-owned firms. They find banks use

transactions accounts related to these loans to monito borrowers. In this study, we

integrate these two lines of research and test the connection between bank monitoring and

financial reporting

Prior research suggests borrowers’ accounting policies are shaped by lender

preference (e.g., Ahmed et al. 2002; Watts 2003; LaFond and Watts 2008; Khan and

Watts 2009; Frankel and Roychowdhury 2009; Bushman and Wittenberg-Moerman;

2010). However, these papers do not specify and empirically examine the institutional

means used by lenders to independently verify that the borrower is using conservative

accounting. Our study illuminates a communication channel that lenders use to obtain

information necessary to monitor whether borrower accounting choices are consonant

with lender preferences (e.g., Watts 2003). Hence, our paper contributes to this line of

research by providing empirical evidence that banks’ preferences can shape borrower

accounting choices.

By studying the initiation of loans that require an aging report be supplied to

lenders, we isolate borrowers facing bank scrutiny of a specific balance sheet item

(Banrett 1997). The focus on a specific account permits us to construct a tailored model

3

for predicted accruals (McNichols and Wilson 1988; Jackson and Liu 2010). We also

broaden the results of Mester et al. (2007) who study a set of small, private firms by

focusing on a population of audited, public companies that is typically examined by

researchers attempting to determine whether bank loans are associated with enhanced

monitoring or screening (e.g., Mikkelson and Partch 1986; James 1987; Lummer and

McConnell 1989; Best and Zhang 1993; Billet, Flannery et al. 1995).

The majority of loans in our sample are borrowing base revolvers using accounts

receivable as collateral and/or to determine the maximum loan amount (Flannery and

Wang 2011). The total accounts receivable used to compute the borrowing base is

determined by using accounts listed in the aging report which corresponds the reported

value of gross receivables on the balance sheet. Financial reporting decisions are also

likely to be of concern to the lender because loans often contain financial covenants and

because accruals provide ‘hard’ evidence to support discretionary adjustments to the

borrowing based allowed to the lender in the loan agreement. Given the information

available to the lender, a borrower wishing to maintain a lending relationship faces

increased costs when understating reserves.1 Thus, on average, we expect borrowers to

increase allowance account levels after loan initiation.

Changes in allowance account levels can also reflect changes in underlying

default risk of credit customers. If banks prefer borrowers to have a diversified portfolio

of credit customers and borrowers alter their customer portfolios accordingly, then this

1 See Bolton and Scharfstein (1990) for a model supporting the notion that borrowers have an incentive to

repay loans even when cash flows cannot be observed by the lender (or verified by a third party), because

borrower would like to continue the relationship.

4

change can lead to a reduction in allowance account levels after loan initiation.2

Alternatively, ‘unused fees’ and fixed costs associated with credit lines reduce the

incremental cost associated with additional borrowing once the line is established. This

cost structure could encourage borrowers to provide financing for customers, thereby

relaxing credit policies. The possibility of changing default risk associated with the

initiation of borrowing requires us to control for this effect. 3

Using a key word search of SEC filings between 1994 and 2006, we identify 248

firms with loan contracts requiring the borrower to periodically supply the lender with an

accounts receivable aging report.4

Our sample size is limited and contains one

observation per firm, because we include only the first aging-report-requirement loan for

each firm in our sample to capture incremental affects associated with a transition to a

regime of increased lender monitoring. We denote the year of loan initiation as year t.

Pooling data between years t-2 and t+1, we regress current allowance for doubtful

accounts on the current balance in the accounts receivable, current and future period

write-offs and other control variables. Our main variable of interest is an indicator equals

one in year t and t+1.

We find that the allowance balance increases by 0.3% of sales after loan initiation.

Because this balance averages 1.3% of sales prior to loan initiation, a 0.3% change

2 Limits are typically imposed on the amount that can be due from any customer in the computation of the

borrowing base. In addition, banks often impose a cross-aging requirement on borrowers, rendering the

entire balance owed by a given customer ineligible for inclusion in the borrowing base if one invoice is past

due. 3 Other factors can affect the observed increased in bad debt expense. If firms engaging in asset-based

borrowing tend to overstate reported allowance for doubtful accounts prior to loan initiation (Jackson and

Liu 2010), then the observed increase will be muted. On the other hand, if bank lending coupled with bank

access to inside information, reduces demand for timeliness of financial reporting (Ball et al. 2000) and

increased allowance balances provide the means for income smoothing (Jackson and Liu 2009) then we

expect increases in bad debt expense recognition. 4 An aging report lists the amounts owed to the borrower by its credit customers and sorts these customer

balances by age outstanding.

5

implies a 23% increase. This increase understates accounting recognition effects because

write-offs also rise significantly in the year of loan initiation. Thus, the income statement

effects are magnified. Bad debt expenses increase by 3.8 million dollars, on average,

after loan initiation, an increase of more than 50%. Moreover, results show a significant

decline in the serial correlation of write-off changes. This suggests that firms adopt more

timely write-off policies after initiation of aging report requirements. Consistent with the

hypothesis that bank monitoring drives this increase, our results show that the increase in

the allowance balance is higher for firms required to provide aging reports on a weekly or

monthly basis than firms with a less frequent aging report requirements. Our results are

robust to inclusion of interactions for size and analyst following and to alternative scaling

variables (assets and accounts receivable) and inclusion of variables to control for

changes in credit policies (future write-offs and accounts receivable turnover). In fact,

credit policies seem to become stricter following monitoring. After loan initiation firms

make fewer sales to their most influential customers (i.e., customers accounting for more

than ten percent of sales) and the credit quality of these customers, measured via credit

ratings, improves. Our last set of analysis finds that borrowers increase the frequency of

allowance for doubtful account disclosure in their quarterly financial statements after

loan initiation. We attribute this finding to the enhancement of the firm’s accounting

system in response to the necessity of supplying aging reports by lenders.

Our study makes several contributions. First, our study advances research that

examines accounting conservatism arising from debt contracting efficiency because our

results suggest active monitoring by lenders of a specific revenue accrual account is

associated with more conservative accounting related to that accrual by borrowers. In this

6

respect, our study complements Tan (2010) who documents increased accounting

conservatism after covenant violation and conjectures that this effect results from stepped

up bank monitoring. In addition, our study suggests that information intermediaries such

as analysts and auditors do not substitute for the unique monitoring effect of banks.

Our study also provides empirical support for the theory that banks have an

advantage in obtaining or producing private information about borrowers (LeLand and

Pyle 1977; Diamond 1984; Fama 1985). Specifically, we document effects arising when

banks obtain detailed information on borrowers’ accounts receivables that is otherwise

not observable by outside investors. Further, our results offer an explanation for the

mixed findings in the literature examining managers’ earnings management incentives to

avoid covenant violations.5 Perhaps, bank access to information can deter managers from

manipulating accounting information to avoid debt covenants. Our results demonstrate

that banks’ active monitoring can reduce such incentives of managers, highlighting the

importance of controlling for banks’ monitoring when examining debt covenant

hypothesis.

II. Background

Relation to Literature

The presence of banks suggests they perform some function intermediating

between borrowers and savers more efficiently than is available via direct exchange in

capital markets. Our research question springs from theory implying the uniqueness of

banks’ monitoring. Research argues that banks enjoy a comparative advantage in

5 The empirical results on debt covenant hypothesis are mixed. DeFond and Jiambalvo (1994) and

Sweeney (1994) find managers use discretionary accruals or income-increasing accounting changes to

avoid debt covenant constraints. Healy and Palepu (1990) and DeAngelo et al. (1994) do not find support

for the debt covenant hypothesis.

7

producing information that enables them to add value via debt-related monitoring (e.g.,

Diamond 1984). For example, Fama (1985) classifies bank debt as an insider debt

because banks have access to information from an organization’s decision process not

otherwise publicly available. Researchers have sought evidence of bank monitoring in the

reaction of borrower’s stock to loan announcements. They find significant positive

reactions. The results suggest banks have access to non-public information that allows

them to screen or to subsequently monitor borrowers (e.g., Mikkelson and Partch 1986;

James 1987; Lummer and McConnell 1989; Best and Zhang 1993; Billet et al. 1995).

Some studies (i.e., Lummer and McConnell 1989; Best and Zhang 1993) distinguish

between new loans and revisions of existing agreements and find a significant positive

reaction only for agreements that are revised favorably. This result implies that banks

gain an information advantage only after they establish a relationship with borrowers.

These studies do not provide evidence on the methods used by banks to acquire private

information.

Mester et al. (2007) fill this gap, by studying loans made by a Canadian bank to

small, management-owned firms. These borrowers maintain a checking account at the

bank and are required to provide the bank with accounts receivable and inventory

information. Mester et al. find that the transaction information available to the bank

predicts credit down grades, loan write-downs and loan reviews. Thus, the bank acts ‘as

if’ it uses transaction information to assess its loans. Related work analyzing credit lines

suggests credit line usage by borrowers reflects default risk (Jimenez et al. 2009; Sufi

2009) and predicts default (Norden and Weber 2010). The implication is that credit line

usage gives banks private information on their borrowers.

8

We link research on bank monitoring to accounting policy. We test the joint

hypothesis that bank monitoring activities produce effects and that these effects, in turn,

lead to observable alterations in borrowers’ accounting policies. The relation between

bank monitoring and accounting policy emerges from research indicating that lenders

demand conservatism.

Watts (2003) argues that debt financing spurs demand for conservative accounting.

Studies of the relation between accounting conservatism and debt financing infer debt

holders’ demand for conservatism by examining whether conservatism is correlated with

certain loan or firm characteristics. For example, Ahmed et al. (2002) document that

borrowers facing more severe debt holder-shareholder conflict are also more conservative.

Lafond and Watts (2008) and Khan and Watts (2009) find that asymmetric timeliness of

earnings is positively related to leverage. Other papers look for lender benefits associated

with conservatism. Zhang (2008) shows that accounting conservatism is associated with

both timely violations of loan covenants and lower costs of debt, suggesting that

accounting conservatism benefits both lenders and borrowers. Wittenberg-Moerman

(2008) finds that the bid-ask spread in the secondary loan market is lower for more

conservative borrowers. Nikolaev (2010) demonstrates that the intensity of covenant use

in the debt contracts is positively correlated with accounting conservatism and interprets

the positive association as evidence that debt holders demand conservatism to improve

the contracting efficiency of earnings-based covenants.6

The means by which accounting conservatism is enforced has not been explored.

A critical question is how banks assess whether accounting information provided by

6 In related research, Leftwich (1983) and Beatty et al. (2008) infer lender demand for accounting

conservatism via loan covenant computations.

9

borrowing firms is reliable. Absent covenant violations, lenders have no right to decide

accounting policies. That right resides with managers to whom it was granted by

shareholders. Researchers speculate that reputation and legal liability force borrowers to

maintain conservative accounting policies after borrowing (Beatty et al. 2008; Nikolaev

2010). Maintaining a high level of conservatism, however, is costly to borrowers because

such choices can reduce current bonuses or expedite covenant violations, enabling

lenders to exercise decision rights.7 A key component of enforcement is the ability to

verify compliance.8 In this paper, we identify a set of loan contracts with covenants

requiring borrowers to provide accounts receivable aging reports to lenders. Such

covenants indicate that banks have access to information that allows them to assess the

conservatism of borrower accounting choices with respect to accounts receivable.

Aging Reports and Banks’ Monitoring of Accounts Receivable

In addition to requiring borrowers to maintain certain financial ratios and

providing timely public financial reports, lenders can also require borrowers to grant

access to detailed financial information that is not publicly available. Loan contracts can

contain covenants granting lenders the right to inspect and review all original business

transaction documents and discuss financial matters with managers and independent

auditors. We focus on one such covenant: banks’ requirement that borrowers provide

periodic accounts receivable aging reports.

7 Evidence supporting the debt covenant hypothesis suggests that managers of the borrowing firms have

incentives to reduce accounting conservatism after borrowings to avoid costly covenant violations (Watts

and Zimmerman 1986; DeFond and Jiambalvo 1994; Sweeney 1994; Dichev and Skinner 2002; Kim 2010). 8 The ability to verify a project’s returns reduces expected deadweight liquidation costs relative to a

contract which can only use an unconditional threat of liquidation to give the borrower incentives to repay

the debt (Diamond 1996).

10

The following excerpt from Kontron Mobile Computing Inc.’s 1998 syndicated

loan contract illustrates the aging report requirement:

Borrower agrees it will: (a) Furnish to Lender in the form satisfactory to

Lender: … (iv) Within 10 days after the end of each month, an aging of accounts

receivable together with a reconciliation in a form satisfactory to Lender and an

aging of accounts payable in form acceptable to Lender, both certified as true and

accurate by an officer of the Borrower;

While this covenant requires aging reports to be provided monthly. In our sample, the

frequency of aging report provision ranges from a weekly to annual basis.

The requirement to provide aging reports is usually associated with revolving

credit lines that use accounts receivable to determine the borrowing base of the loan (i.e.,

the maximum loan amount) and/or use accounts receivable as collateral. When accounts

receivable is used as a part of the borrowing base, covenants usually require the borrower

to provide periodic borrowing base certificates to the lender documenting the

computation of the borrowing base. This computation usually begins with total accounts

receivable from which various receivables are excluded to determine “eligible accounts

receivable.” Accounts commonly excluded are

(i) receivables more than 60 (90) days past the due (invoice) date,

(ii) receivables owed by the United States or any government agency,

(iii) receivables owed by affiliates or related parties,

(iv) receivables owed by a customer with at least 50% of receivables overdue,

and

(v) receivables owned by any one customer in excess of a limit set by the

borrower.

The borrowing base is commonly 85 percent of eligible receivables, but the lender

is allowed discretion to make further adjustments based on business conditions.9 Aging

reports can help banks verify the computation on the borrowing base certificate. In

9 These modifications to GAAP-based receivables are consistent with the findings in Leftwich (1983) that

debt contracts contain clauses that make conservative adjustments to GAAP-based accounting information.

11

addition to the aging reports, banks can require additional information from borrowers

regarding accounts receivable. The following excerpt is also from Kontron Mobile

Computing Inc’s 1998 contract:

Borrower agrees to furnish to Lender, at least weekly, schedules describing

Receivables created or acquired by Borrower (including confirmatory written

assignments thereof), including copies of all invoices to account debtors and

other obligors (all herein referred to as "Customers") … Borrower shall advise

Lender promptly of any goods which are returned by Customers or otherwise

recovered involving an amount in excess of $5,000.00. Borrower shall also advise

Lender promptly of all disputes and claims by Customers involving an amount in

excess of $5,000.00 and settle or adjust them at no expense to Lender.

Covenants also allow the lender to access borrower accounting records and confirm the

existence of receivables. The borrower pays the cost of this investigation. These

examples provide some flavor for the nature of the information available to lenders.

Although banks do not have direct control over firm’s accounting policy, the above

excerpts indicate banks have information necessary to accurately assess the reliability of

the borrowers’ accounting choices with respect to accounts receivable.

We argue that financial reports choices are important to the lender despite the

availability of other information. First, lending agreements often contain financial

covenants.10

Second, financial statements provide information that permits third parties

(e.g., courts) to verify a state of nature has occurred and therefore can be used to justify

decisions that are potentially damaging to the borrower but that are permitted under

contract to the lender.

The reporting requirements associated with aging reports can have a direct effect

on borrowers’ accounting systems. According to BHF-Bank,

10

In a review of 30 randomly selected loan agreements from our sample, all contained accounting-based

covenants.

12

Our customers have confirmed that the borrowing base reports and our audits are

very useful from a practical point of view as they can significantly enhance the

data in their finance and accounting departments. Many of our customers believe

that our audit and borrowing base reports provide an important external analysis

of their flows of goods and cash, as the reports consistently reveal areas in which

they can optimize their companies’ business operations, both in economic and

legal terms.11

, 12

Heightened legal liability on the part of auditors and borrowers associated with accounts

receivable can also accompany the provision of aging reports. Auditors are aware that an

outside party is independently assessing the quality of accounts receivable and

documenting the age and collectability of accounts. Executives are required to certify

reports to lenders. According to a white paper by managing directors at RSM McGladrey,

Knowledge of misrepresentations of these certifications can result in civil and/or

criminal charges and should be taken very seriously by borrowers.13

In short, legal liability associated with the provision of aging reports is likely to be

another factor that leads to more conservative accounting with respect to accounts

receivable.

III. Research Design

Our goal is to draw conclusions about the effect of bank monitoring on

accounting policy for firms in our sample. We do not make inferences about the effect

that provision of aging reports would have for the general population of borrowers.

Banks likely require aging reports for firms where gathering such data is cost effective,

11

See, ‘FAQ’ on borrowing-base loans.

http://www.bhf-

bank.com/w3/imperia/md/content/internet/financialmarketscorporates/borrowing_base_faq_en.pdf?teaser=/

w3/financialmarkets_corporates/commodity_finance/borrowing_base. 12

A conversation with a senior-vice president focusing on asset-based lending at a large US bank confirms

this statement and suggests information systems with respect to receivables become more sophisticated at

lower and middle market companies and managers become more aware of problems in receivable

collection when required to provide borrowing base reports. 13

See, “Reading the fine print: What borrowers need to know about loan agreements in the new

recession.” http://mcgladrey.com/pdf/loan_agreements.pdf.

13

so one can reasonably assume that any effects related to provision of aging reports are

likely more pronounced in our sample than in the general population. We concentrate on

the effects within the selected sample precisely because it provides a more powerful

setting to observe the impact of bank monitoring of accounts receivable. Our econometric

concern with regard to factors associated with the decision to provide an aging report

therefore centers around variables jointly associated with this decision and the reported

level of bad debt expense. For example, if a manager decides to borrow from banks to

ease potential capital constraints arising from expected deterioration of customer credit in

the future, bad debt expenses will increase to reflect his expectation of lowered

collectability of accounts receivable rather than bank oversight associated with the loan.

We estimate the following model:

ALLOWit = β0 + β1 × POSTt + β2 × ARit + β3 × ALLOWit-1 + β4 × WOit

+ β5 × WOit+1 + β6 × LEVit + β7 × ARTO_INDjt

+ β8 × SD_SALE_INDjt + β9 × ALT_INDjt

+ β10 × AFjt+ β11 × ASSETjt+ ΣFIRMi +ΣYEARt + εit, (1)

where POST is a dummy variable equal to 1 if a firm-year falls in the year of or the year

after loan initiation, and 0 otherwise. The coefficient on POST (β1) is expected to be

positive reflecting the effect of bank monitoring on borrowers’ recognition of allowance

for doubtful accounts. To control for time-varying within-firm factors that simultaneously

cause firms to obtain a bank loan with an aging report requirement and drive bad debt

expense levels, we include control variables for factors that can affect firms’ bad debt

expense recognition. Following McNichols and Wilson (1988) and Jackson and Liu

(2010), we include accounts receivable (AR), prior year’s allowance for bad debt

expenses (ALLOW), contemporaneous and future write-off of accounts receivable (WO).

The variables AR, ALLOW, and WO are all scaled by contemporaneous sales. As accounts

14

receivable increases (reducing receivable turnover), we expect the balance of allowance

for doubtful accounts to increase. As the current and future write-off of accounts

receivable increase, the allowance should increase in expectation of increased credit risk.

On the other hand, the bad debt expenses will be lower if previous year’s allowance is

high. Hence, we expect the coefficient β2, β4, and β5 to be positive and the coefficient β3

to be negative.

We also include firm leverage (LEV) to control for other effects of borrowing on

reporting incentives unrelated to monitoring of receivables. For example, managers’ can

inflate earnings (decrease bad debt expenses) to avoid covenant violations (Defond and

Jiambalvo 1994) or managers can have an incentive to increase conservatism given

leverage in the absence of explicit bank monitoring. LEV is defined as total debt to assets.

To further control for factors associated with changes in the expected frequency

of credit-customer defaults, we include controls for industry factors such as industry

receivable turnover (ARTO_IND), industry standard deviation in sales (SD_SALE_IND),

and industry bankruptcy risk (ALT_IND). Analyst following (AF) and total assets (SIZE)

are also included to control for monitoring changes related to firm size or associated with

financing. In addition, we select a set of firms matched with our test firms based on

industry and receivable levels to control for market and industry-wide factors.

We also rely on the cross-section frequency of transmittal of aging report to

investigate whether results are consistent with increased bank monitoring. Specifically,

we examine whether the change in allowance of doubtful accounts after borrowing vary

with the intensity of banks’ monitoring. We partition the test sample into two subsamples

with high and low frequency of aging report, respectively, and then estimate Equation (1)

15

for these two subsamples separately. We define high frequency (HIGHFREQ) equal to 1

if banks require the borrower to provide aging reports at a monthly or weekly basis and

zero otherwise.14

The coefficient on POST is expected to be greater for the high

frequency subsample than for the low frequency subsample.

IV. Sample Selection, Descriptive Statistics, and Univariate Analysis

Sample

Test sample

We search the material contract sections of filings with the Securities and

Exchange Commission using 10K Wizard to obtain the initial sample of loans containing

covenants requiring aging reports from 1994 to 2006. 15

Our sample period begins in

1994 because 10K Wizard started providing material contracts only after 1994. We end

the sample in 2006 to provide data on post-borrowing variables. We merge this initial

loan sample with the Compustat by firm CIK number. We require each firm to have

financial information on Compustat for the two years before (t-2 and t-1) and two years

after (t and t+1) loan origination where t represents the fiscal year that a loan is originated.

This procedure results in a sample of 1,657 debt contracts with aging report covenants for

803 unique firms. To measure the effect of initiation of aging report requirements, we

only keep the first loan contract with aging report covenants within our sample period for

each firm.16

To do this, we read all 1,657 debt contracts plus 10K and 10Q notes issued

14

We treat monthly or weekly aging reports as high frequency, because most bank contracts require

quarterly financial reports. 15

Specifically, we use the key word ‘aging’ and ‘receivable’ with the condition that the two words are

separated by less than five words. We review the contracts and exclude non debt contracts such as Stock

and Asset Purchase agreements and M&A agreements. 16

Because 10K Wizard started providing loan contract data after 1994, we are less confident that loans

from 1994 and 1995 are the first instance of an aging report requirement for the firm.

16

two years before the origination year to make sure that no similar contracts existed in the

past. Furthermore, we delete contracts that are the renewals of previous contracts with

similar aging report requirements signed before 1994 because we cannot obtain these

original contracts from 10K Wizard. After this procedure, 385 debt contracts from 385

unique firms remain.

Next, we collect data on bad debt expense and write-offs of accounts receivable

from Schedule II of 10K notes. To ensure the accuracy of our data, we reconcile the

beginning balance of the allowance with the ending balance of the allowance for each

firm year. Firms missing bad debt expenses or write-offs for the two years before or the

two years after loan origination are excluded from the sample. This data restriction

eliminates 137 contracts. Our final sample consists of 248 debt contracts with 992 firm

year observations (248 unique firms) spanning from 1992 to 2008.

Our tests use annually reported values of bad debt expenses and various control

variables. Therefore, if a loan is originated between nine months before and three months

after a fiscal year t, we treat the loan originated in year t.17

Control sample

To control for industry-wide factors affecting bad debt expense recognition, we

select a set of firms that match with our borrowing firms (test sample) based on the

following procedure. We begin with the Compustat universe that excludes our test firms

and excludes firms with a borrowing-base or collateralized loan identified from Loan

Pricing Corporation over our sample period. Second, we require the matching firms to be

17

We allow a three-month buffer because firms are required to release their 10Ks within three months after

the fiscal year-end and most firms disclose in financial statement footnotes the loans originated during the

period between the fiscal-year end and the release of 10K. Hence, we assume that loans originated within 3

months after the fiscal year-end affect borrowing firms’ accounting policies for that fiscal year.

17

in the same industry classified by two-digit SIC code as the test firm. Third, for firms that

survive the prior filters, we select the firm that has the smallest difference in accounts

receivable scaled by sales from the test firm (difference<20%). If we cannot find a

matching firm using this criterion, we relax the standard by increasing the difference to

forty percent and to sixty percent etc. If we find multiple matching firms that meet these

criteria, we select the one with the smallest difference in leverage from the test firm. In

the end, we are able to find 248 matching firms that also have bad debt expense and

write-off data available from the 10K.

Descriptive statistics

We first provide a time and industry profile of our sample borrowing firms. Panel

A of table 1 shows that the number of contracts in any given year ranges from 4 in 1994

to 42 in 2001. In particular, a total of 82 contracts (more than 32% of the entire sample)

cluster in 2000 and 2001 when business conditions are weak. This is consistent with the

observation of Rajan and Winton (1995) that collateral requirement varies inversely with

business conditions. To alleviate the concern that our results are driven by economic

conditions in a particular year, we include year fixed effects and other industry-wide

economic indicators in our empirical model. Panel B provides an industry profile of the

borrowing firms. As shown in the table, our sample represents a wide range of industries

and is similar to that shown in Flannery and Wang (2011). For example, manufacturing

industry is heavily represented in our sample (54%) compared to the Compustat Universe

firms (34.5%), but it is comparable to 51.2% reported by Flannery and Wang.

[Insert Table 1 Here]

18

Table 2 provides summary statistics on the loan characteristics for the 248 loan

contracts in our sample. As shown in panel A, the average loan amount is 52 million

dollars with a mean maturity of 2.7 years. Of the 248 loan contracts, 51% are syndicated

loans with more than one lender. The median cutoff is 90 (60) days from the invoice (due)

date of the receivables. 18

Hence, accounts receivable that are outstanding less than 90 (60)

days from the invoice (due) date of the receivables will only be considered as eligible

accounts receivable. Further, 82% of the eligible accounts receivables are used as part of

the borrowing base. Therefore, banks make several conservative adjustments to GAAP-

based accounts receivables.

Panel B presents summary statistics on the purpose of aging reports. The most

common purpose (179 contracts) is to verify eligible accounts receivables to derive the

borrowing base. In some cases, accounts receivable serve as both the collateral and the

borrowing base (113). In 66 contracts accounts receivable is used as the borrowing base

but the contract does not provide a schedule of collateral so we cannot verify whether

accounts receivable is also used as the collateral. In 40 contracts, banks require aging

reports and accounts receivables are used as collateral against firm borrowings. 29

contracts require aging reports but provide no indication that accounts receivable are used

as collateral or to set the loan amount. In general, banks appear to require aging reports to

monitor collateral and limit loan amounts to collectible collateral rather than rely on

borrower operating performance.

18

Some contracts calculate the cutoff dates based on both the invoice date and the due date of the

receivables. We also collect data on the cutoff date banks use to calculate eligible accounts receivables

when borrowing firms use accounts receivables as part of the borrowing base. We identify 179 loan

contracts with eligible accounts receivables as borrowing base. Among these 179 loan contracts, 170 (67)

contracts use invoice dates (due dates) of accounts receivables as the base to derive cutoff dates.

19

Panel C of table 2 displays the variation in the periodicity of aging report

requirements. It varies from a weekly basis reports to annual reports. 33 contracts require

borrowing firms to provide aging reports upon lender request. Available information does

not allow us to determine the frequency of such requests. The majority contracts (164 or

66% of the entire contracts) require firms to provide monthly aging reports. This

contrasts with the quarterly financial disclosures to shareholders mandated by SEC.

Hence, lenders require more frequent disclosure of information for firms in our sample

than is available to shareholders.

[Insert Table 2 Here]

Table 3 presents correlations among the dependent variable of allowance for

doubtful accounts (ALLOW), firm characteristics, and loan characteristics for our test

sample. ALLOW and firm characteristics are measured at the fiscal-year end prior to the

loan origination year. Two statistics are noteworthy. First, allowance is positively

associated with return volatility and leverage, but negatively associated with return on

assets, cash flow from operation and the presence of analyst following. These results are

consistent with the intuition that bad debt expense estimate is affected by firm

performance and risk. Second, return volatility increases the frequency of aging reports.

In contrast, higher return on assets and operating cash flow, larger assets, and the

presence of analyst following decrease the frequency of aging reports.

[Insert Table 3 Here]

Univariate analysis

In this section, we present a univariate analysis examining the effect of bank

monitoring on borrowing firms’ bad debt expense recognition. We compare the changes

20

in allowance for doubtful accounts and bad debt expenses for both the borrowing firms

and the matching firms along with changes in other firm characteristics around

origination of loans that require aging reports. We assign the borrowing firm’s

origination date to that of its matching firm. Table 4 presents summary statistics on the

characteristics for the 248 borrowing firms and their 248 matching firms in the two years

both before and in the two years after the loan origination. Allowance for uncollectible

accounts receivables (ALLOW) increases after borrowing for the test sample from 0.013

to 0.015 though this increase is statistically insignificant. For the matching sample,

ALLOW does not change after borrowing. Accounts receivable (AR) decreases in the post

period for both groups. The mean BDX increases from 0.010 (1.0% of total sales) in the

pre-period to 0.012 (1.2% of total sales) in the post-period. The difference (0.002) is

statistically significant at the 5% level. In contrast, BDX decreases from 0.014 to 0.010

for the matching firms and this decrease is statistically significant at the .01 level.

However, the allowance balance increases with bad debt expense and declines with write-

offs. Table 4 indicates that writes-offs increase by 33% (0.004/0.012). These off-setting

effects blunt the increase in the allowance balance. There are no statistically significant

Pre/Post changes in leverage (LEV) between the test sample and the control sample,

which suggests that our matching procedure seems to hold constant accounts receivable-

based lending to customers and borrowing across the two samples.

[Insert Table 4 Here]

Figure 1 illustrates the change in the allowance account balance in the four years

around loan origination for both the borrowing sample and the matching sample. The

amount for the borrowing sample starts increasing in t-1, one year before loan origination,

21

suggesting that firms expecting to borrow from banks start to adjust their accounting

policy even before the borrowing. The amount increases sharply in the loan origination

year t, and remains at a level higher than that in the pre-borrowing period. In contrast, the

allowance balance for the control sample drops in the year of loan origination and this

trend remains two years after loan origination, which could reflect overall increases in

credit quality in the customers of the industry.

[Insert Figure 1 Here]

The attenuated increase in ALLOW can be explained, in part, by the increases in

write-offs (WO) in the loan year. The increase in the write-offs can be either due to

reduced prospects for collectability, which could cause firms simultaneously to increase

bad debt expenses and borrow from banks, or changes in accounting policy. The second

explanation assumes the decision to write-off an account involves discretion by the

receivable holder and that active monitoring by banks can prod borrowers to write off

questionable accounts.19

Under this explanation, the write-offs are not a solely a function

of receivable collectability but are also subject to managers’ discretion. On the other

hand, if write-offs indicate future credit risk, a significant increase in write-offs after

borrowings is problematic for our empirical identification, because the increase can cause

a spurious correlation between a loan origination and an increase in bad debt expenses.

Inspection of the data indicates that the increase in write-offs occurs primarily in year t

and that write-off in years t+1 and t+2 resemble pre-loan levels. This evidence suggests

that the increase in write-offs is a temporary phenomenon that coincides with the year of

loan initiation. In any case, these results suggest the necessity of controlling for future

19

This is particularly possible when banks impose a cross-aging requirement on borrowers. Under these

circumstance a borrower would have an incentive to clear past due accounts out of the receivable ledger by

writing them off.

22

write-offs as well as industry performance to distinguish accounting policy changes from

credit quality changes.

A firm’s performance is likely to be negatively affected by deteriorating credit

quality of its customers. We compare borrowing firms’ operating performance to that of

control firms. As shown in the table, cash flows from operations (CFO) increase

significantly during the post-period for sample firms. Figure 2 also illustrates this point.

The mean values of CFO increases monotonically from 0.013 in t-2 to 0.032 in t+1,

suggesting firms’ operating performances improved after loan origination rather than

deteriorated. This increase contrasts with the control sample whose CFO declines; though

Table 4 indicates this decline is moderate.

[Insert Figure 2 Here]

In addition, Table 4 shows that total asset turnovers (SALES) remain the same

after borrowing. Compared to the matching firms, borrowing firms have higher revenue,

are smaller in size and are less likely to be followed by analysts. At the industry level, all

the changes in the economic indicator variables point in the same direction—that the

industries that borrowing firms belong to experience an improvement in the economic

performance. For example, the median accounts receivable turnover ratio at the industry

level (ARTO_IND) increases significantly; both the standard deviation of sales

(SALES_SD_IND) and Altman z-score (ALT_IND) at the industry level decrease

significantly.

V. Multivariate Analyses

Main regression results

23

Table 5 reports the results of testing the effect of bank monitoring on the

allowance account balance for the test sample in column (1) and for the sample

containing both the test firms and the control firms in column (2). The coefficient on

POST is positive and statistically significant in column (1), suggesting that after

borrowing, firms increase their allowance account balance significantly. In terms of

magnitude, borrowing firms experience an increase in allowance for doubtful accounts of

0.003 (0.3% of total sales). The increase is also economically significant. The average

of ALLOW in the period before borrowings is 0.013 (1.3% of total sales), and a change of

0.3% of sales in bad debt expenses represents a more than 23% increase. Given that the

average sales are $542 million, bad debt expenses increase by $1.6 million after

borrowing.

For the control variables, next period’s write-offs of accounts receivable are

positively correlated with ALLOW, suggesting that firm’s allowance balance reflects

expected credit quality of receivables. Current write-offs are negatively related to the

current allowance balance suggesting that the direct affect of write-offs on the current

allowance balance (write-offs reduce the allowance balance) exceeds any explanatory

power that current write-offs have for expected write-offs, conditional on future write-

offs. Alternatively it may suggest that the percentage credit sales method dominates the

aging method for calculating bad debt expenses in our sample, which results in a

mechanical, negative relation between write-offs and the allowance balance. The

coefficient on LEV is positive and only significant in the model that includes control

firms (β6 = 0.004 with a p-value of 0.007), providing evidence that managers tend to

increase the allowance balance as debt increases, consistent with a positive association

24

between leverage and accounting conservatism. The adjusted R-Square is 83%,

suggesting that the model explains a significant portion of the variation in allowance

balance, but this R-square also reflects the explanatory power of firm and year fixed

effects.

In column (2), the coefficient on POST continues to be positive and statistically

significant at the .05 level. In contrast, the coefficient on the interaction between POST

and CONTROL is negative and statistically significant at the .05 level and an F test of the

sum of POST and POST×CONTROL is not statistically significant. These results suggest

that test firms report an increased allowance balance after loan origination whereas this

does not occur to the matching firms. Therefore, the increase in the allowance balance for

the borrowing firms is not significantly related to industry-wide effects. In sum, we

document a significant increase in the allowance balance of firms that borrow with an

aging report requirement. We attribute this increase to bank monitoring of borrowers’

accounts receivable.

[Insert Table 5]

Bank monitoring and the persistence of write-offs

In the previous section we document that borrowers become conservative in that

they recognize more allowance for doubtful account after borrowing. In this section we

investigate whether write-offs become more conservative in the presence of bank

monitoring. Table 4 shows that write-offs increase significantly after borrowing. Thus,

isolating the allowance balance can understate the reporting implications of the

requirement to supply aging reports. We hypothesize that borrowers tend to write off

accounts receivable more fully after borrowing and this leads to a reduction in the

25

persistence of write-offs. The results of this analysis are reported in Table 6 for both the

test and the control firms. The coefficient on the lagged write-offs, WOt-1, is positive and

statistically significant suggesting that write-off is persistent during the pre-borrowing

period for both the test and the control firms. More importantly, the interaction between

POST and WOt-1 is negative and statistically significant at the .10 level for the test firms

only. For the control firms, this coefficient is negative but statistically indistinguishable

from zero. Therefore, consistent with our expectation borrowers are more conservative in

recognizing write-offs.

[Insert Table 6]

Cross-sectional analysis of bank monitoring intensity

In this section, we provide evidence on how the changes in the allowance balance

vary with banks’ monitoring intensity for our test sample. If monitoring is more intense

for loans requiring more frequent aging reports, we expect that the increase in allowance

will be greater for the high frequency subsample. Table 7 reports the results of this

analysis. The coefficient on POST is significant at the 0.05 level only in the high

frequency (HIGHFREQ=1) subsample. The increase in allowance for firms with low

monitoring frequency (quarterly or longer, including upon request) is 0.001 after loan

origination and insignificant. For firms with high frequency of aging reports (weekly or

monthly), the change in allowance is 0.004. These results are consistent with the

hypothesis that monitoring intensity affects with the observed change in the allowance

balance after the initiation of the requirement to supply aging reports to the lender.

[Insert Table 7 Here]

The real effect of bank monitoring

26

Borrowing-base loans can affect borrower operating decisions, such as the choice

of customers, as well as its financial reporting policy. In discussing the nature of

adjustments made to the accounts receivable balance when it serves as the determinant of

a borrowing base, BHF bank states that “You can influence the amount deducted by

diversifying your accounts receivable and by employing a high-quality risk management

system.”20

Borrowing base loans are often subject to a “concentration cap” that limits the

inclusion of receivables by any one customer to specific percentage of the borrowing

base.21

We therefore test if borrowers reduce sales concentration to specific customers

and improve the credit quality of their customer portfolio.

We obtain all loan contracts using accounts receivable as either borrowing base or

collateral from DealScan for the period 1992–2008.22

We then identify and keep the first

loan contract for a firm over the sample period. We then merge these firms with the

COMPUSTAT Segment File to obtain significant customers. The COMPUSTAT

Segment File contains information about sales to each customer reported by a supplier in

the footnotes under SFAS 14 and SFAS 131. To focus on significant customers we delete

customers with the percentage of sales less than 10%. We obtain borrowers’ financial

information for the six years centering on loan initiation.23

After this procedure, we

obtain 5,129 loan-years for 1,148 unique loans and use this sample to examine the change

20

Cite from FAQ document produced by BHF Bank. http://www.bhf-

bank.com/w3/imperia/md/content/internet/financialmarketscorporates/borrowing_base_faq_en.pdf?teaser=/

w3/financialmarkets_corporates/commodity_finance/borrowing_base 21

For example, the borrowing base certificate contained in the December 16, 2010 Form S-4 of Interline

Brands Inc. excludes receivables of any one customer that exceed 15% of aggregate eligible receivables

from the borrowing base. 22

The merging of our main sample of borrowers with aging report requirement with the COMPUSTAT

Segment File leads to a reduced sample of 154 unique firms. The sample size is further reduced to 60 when

we require all customers to be public firms. Therefore to increase the sample size, thus the test power, we

employ Dealscan dataset for borrowing base loans. Based on the reduced sample, we find qualitatively

similar results as reported in Table 8 though they are not statistically significant. 23

We require each borrower to have at least one observation during both the pre and the post borrowing

period.

27

in borrowers’ sales concentration. To investigate whether the credit quality of customers

changes after borrowing, we identify each customer of a borrower and merge it with the

COMPUSTAT Industry Annual File by customer’s name to obtain monthly S&P

domestic long-term issuer crediting and other financial variables. After this procedure,

we identify 1394 loan-years for 445 loans. If a customer has no S&P domestic long-term

issuer credit rating, we follow the procedure outlined in Barth et al. (2008) and develop

a credit rating that falls within 2 (AAA) – 27 (default).24,

25

Table 8 reports the results of customer concentration and credit quality analyses.

Columns (1) and (2) focus on the change in borrowers’ sales concentration. The

dependent variable of column (1) is the average of sales percentage to each customer for

a borrower’s customer portfolio and the dependent variable of column (2) is the natural

logarithm of total number of significant customers for a borrower. We expect and find

that the coefficient on POST is negative and statistically significant across both columns

suggesting that borrowers diversify account receivable by reducing sales concentration to

significant customers. In column (3), the dependent variable is the average of credit

rating for a borrower’s customer portfolio with a higher value indicating lower credit

quality. As expected, the coefficient on POST is negative and statistically significant at

the .10 level implying that after loan initiation the credit quality of a borrower’s

customers improves. In sum, we find that borrowers diversify accounts receivable and

focus on customers with lower credit risk after initiating a borrowing–base loan.

[Insert Table 8 Here]

24

The credit rating is predicted by a model including the following predictors: the natural logarithm of total

assets, ROA, leverage, a dummy variable measuring whether a firm pays dividend, a dummy variable

measuring whether a firm issues subordinated debt, and a dummy variable measuring whether a firm incurs

loss in the current period. 25

Our results are robust to the deletion of observations where customers have no credit rating.

28

Bank monitoring and disclosure frequency of allowance for doubtful accounts

As noted in section 2.2 the requirement to provide frequent aging reports to the

lender can cause the borrower to alter his internal accounting system. To the extent these

alterations reduce the borrower’s incremental cost of producing reliable quarterly

allowance estimates, we expect that borrowers will be more likely to separately disclose

these estimates in their quarterly filings. To test this prediction, we collect data on

disclosure of allowance for doubtful accounts from EDGAR 10Q for our original sample

of 248 loans. If a loan is initiated on or before 1997, we delete it because some firms do

not file with SEC through EDGAR in earlier years. If a firm discloses its allowance

account in its quarterly financial statements, we code the firm-quarter observation as 1,

and 0 otherwise. We then sum over quarters to calculate the annual frequency of

allowance for doubtful accounts disclosure (ranging between 1 and 4) and compare mean

annual disclosure frequency between the pre-borrowing and the post-borrowing periods.26

Table 9 presents the results of this analysis. Our sample for these tests is 181

borrowers after we exclude loans initiated before 1998 and borrowers that do not have

positive annual allowance balances in each year between t-2 and t+2. The mean

disclosure frequency increases monotonically from two years prior to borrowing up to

one year after borrowing.27

We see a 0.216 increase in the disclosure frequency of

allowance for doubtful accounts from the pre-period to the post- period. This increase is

about 7% of the mean disclosure frequency of the pre-borrowing period and is

statistically significant at the .03 level. In addition, the increase is 0.61 for borrowers that

26

To be included in our original sample, we require all borrowers disclosure the allowance balance in their

annual report. Thus firms must disclose at the allowance balance at least once per fiscal year. 27

Unreported results show that the median disclosure frequency remains four throughout the four-year

period surrounding a loan initiation.

29

are required to furnish aging reports more frequently to lenders (i.e., weekly or monthly).

In contrast, this increase is only 0.10 for borrowers required to provide less frequent

aging reports (i.e., quarterly or longer). The increase is statistically significant only for

the high frequency group. Therefore, we find evidence consistent with our expectation in

that borrowers increase the disclosure frequency of allowance for doubtful accounts after

loan initiation. We attribute this increase to the enhanced accounting system resulting

from frequent requirement of aging reports by lenders.28

[Insert Table 9 Here]

VI. Robustness Check

Additional control variables

As shown in Table 3 firm size and the presence of analyst following are highly

correlated with multi-lender loans and the frequency of aging reports. Therefore their

interactions with POST can be a correlated omitted variable in our cross-sectional tests.

To address this concern we rerun regressions in Table 6 and include these two interaction

terms. Our results are robust to this procedure suggesting that other governance

mechanisms do not substitute the role of bank monitoring of borrowers’ accounts

receivable.

Scaling variables by accounts receivable or assets

All the results presented are based on scaling bad debt expense and other

independent variables by sales. To investigate whether our results are driven by the

28

An alternative explanation for this increase can be that the amount of allowance for doubtful accounts

increases after borrowing as seen from Table 4 and Table 5 and this increase can push firms beyond the

materiality threshold leading to disclosure. To rule out this explanation, we compare the mean allowance of

firm-years disclosing this account once to firm-years disclosing this account four times. Though we see a

higher mean of allowance account balance for firm-years disclosing four times annually, the difference

between these two groups is not statistically significant. We find similar results when we compare firm-

years with decreasing disclosure frequency to firm-years with increasing disclosure frequency.

30

choice of scaling factor, we replace sales with accounts receivable (and, alternatively,

book value of assets) as the scaling variable. The variable AR is removed from the

equation when receivables serves as the scaling variable, because including AR can cause

a mechanical negative association between the dependent variable and AR. Unreported

results show that our findings are not sensitive to the choice of scaling variable.

VII. Conclusion

We study changes in accounting recognition related to accounts receivable

surrounding the initiation of loans requiring the provision of aging schedules to the lender.

We find that the allowance for doubtful accounts balance increases significantly after

loan initiation controlling for write-offs, receivable turnover, and firm and year fixed

effects. This increase is more pronounced for loans characterized by increased

monitoring intensity. In addition, we find that write-offs are less persistent implying

increased write-off timeliness. Borrowing is also associated with real affects. More

specifically, the initiation of borrowing base loan is associated with reduced sales to

larger customers and improved credit quality of larger customers. Lastly, we demonstrate

that borrowing with an aging reports requirement is associated with an increase in the

frequency of disclosing allowance for doubtful accounts in the quarterly financial

statements by borrowers.

Our results provide direct confirmation of two widely held beliefs in banking and

accounting research. The first is that banks monitor firms. Such monitoring is thought to

be an important reason for the existence of banks. With some notable exceptions, prior

research provides indirect evidence of this monitoring by examining stock-price reactions

to bank loan announcements and by studying accruals before and after loans. The second

31

is that banks demand conservative accounting and that these demands affect firm

accounting policies. Our results suggest that banks’ influence is unique in that it is not

overwhelmed by other monitoring mechanisms in place.

32

REFERENCES

Ahmed, A. B., Billings, R., Morton, and M. Stanford. 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.

Ahn, S., and W. Choi. 2009. The Role of Bank Monitoring in Corporate Governance:

Evidence from Borrowers’ Earnings Management Behavior. Journal of Banking and

Finance 33: 425 – 434.

Ball, R., S. P. Kothari, and A. Robin. 2000. The Effect of International Institutional

Factors on Properties of Accounting Earnings. Journal of Accounting and

Economics 29: 1-51.

Barnett, W. 1997. What’s In A Name? A Brief Overview of Asset-Based Lending. The

Secured Lender 53: 80-82.

Barth, M., L., Hodder, and S. Stubben. 2008. Fair Value Accounting for Liabilities and

Own Credit Risk. The Accounting Review 83: 629-664.

Beatty, A., J. Weber, and J. Yu. 2008. Conservatism and Debt. Journal of Accounting

and Economics 45: 154-174.

Best, R., and H. Zhang. 1993. Alternative Information Sources and The Information

Content of Bank Loans. Journal of Finance 4: 1507-1523.

Billet, M. T., Flannery, M. J., and Garfinkel, J. A. 1995. The effect of lender identity on a

borrowing firm’s equity return. Journal of Finance 50: 699–718.

Bolton, P., and D. Scharfstein. 1990. A Theory of Predation Based on Agency Problem in

Financial Contracting. The American Economic Review 80: 93-106.

Bushman, R., and R. Wittenberg-Moerman. 2010. The Role of Bank Reputation in

“Certifying” Future Performance Implications of Borrowers’ Accounting Numbers.

Working paper, University of North Carolina and University of Chicago.

DeAngelo, H., L. DeAngelo, and D. J. Skinner. 1994. Accounting Choice in Troubled

Companies. Journal of Accounting and Economics 17: 113–43.

DeFond, M., and J. Jiambalvo. 1994. Debt Covenant Violation and Manipulation of

Accruals. Journal of Accounting and Economics 17: 145-176.

Diamond, D. W. 1984. 1984. Financial Intermediation and Delegated Monitoring. Review

of Economic Studies 51: 393-414.

33

Diamond, D.W. 1996. Financial intermediation as delegated monitoring: a simple example,

debt maturity structure and liquidity risk. Federal Reserve Bank of Richmond Economic

Quarterly 82/3: 51–66.

Dichev, I., and D. Skinner. 2002. Large-Sample Evidence on The Debt Covenant

Hypothesis. Journal of Accounting Research 40: 1091–1123.

Fama, E. 1985. What’s Different About Banks? Journal of Monetary Economics 15: 29-

39.

Flannery, M., and X. Wang. 2011. Borrowing Base Revolvers: Liquidity for Risky Firms.

Working Paper, University of Florida.

Frankel, R., and Roychowdhury, S. 2009. Are all special items equally special? The

predictive role of conservatism. Working paper, Washington University in St. Louis

and Boston College.

Healy, P. M., and K. G. Palepu. 1990. Effectiveness of Accounting-Based Dividend

Covenants. Journal of Accounting and Economics 12: 97–133.

Jackson, S. B., and X. Liu. 2010. The Allowance for Uncollectible Accounts,

Conservatism, and Earnings Management. Journal of Accounting Research 48: 565-

601.

James, C. 1987. Some Evidence of The Uniqueness of Bank Loans. Journal of financial

economics 19: 217-235.

Jimenez, G., J. Lopez, and J. Saurina. 2009. Empirical Analysis of Corporate Credit

Lines. Review of Financial Studies 22: 2059-5098.

Kahn, M., and R. Watts. 2009. Estimation And Empirical Properties of A Firm-Year

Measure of Accounting Conservatism. Journal of Accounting and Economics 48:

132-150.

Kim, B. H. 2010. Ex-Post Change in Conservatism and Debt-Covenant Slack. Working

paper, American University.

LaFond, R., and R. Watts. 2008. The Information Role of Conservatism. The Accounting

Review 83: 447-478.

Leftwich, R. 1983. Accounting Information in Private Markets: Evidence From Private

Lending Agreements. Accounting Review 58: 23–42.

Leland, H., and D. Pyle. 1977. Informational Asymmetries, Financial Structure, and

Financial Intermediation. Journal of Finance, 32: 371-415.

34

Lummer, S., and J. McConnell. 1989. Further Evidence on The Bank Lending Process and

The Capital Market Response to Bank Loan Agreements. Journal of Financial

Economics 25: 99-122.

McNichols, M., and P. Wilson. 1988. Evidence of Earnings Management from the

Provision for Bad Debts. Journal of Accounting Research 26: 1–31.

Mester, L. J., L. L. Nakamura, and M. Renaut. 2007. Transactions Accounts And Loan

Monitoring. Review of Financial Studies, 20, 529-556.

Mikkelson, W., and M. Partch. 1986. Valuation Effects of Securities Offerings and The

Issuance Process. Journal of Financial Economics 15: 31-60.

Nikolaev, V. 2010. Debt Covenants and Accounting Conservatism. Journal of

Accounting Research 48: 137-175.

Norden, L., and M., Weber. 2010. Credit Line Usage, Checking Account Activity, And

Default Risk of Bank Borrowers. Review of Financial Studies, 23: 3665-3699.

Rajan, R. and Winton, A. 1995. Covenants and collateral as incentives to monitor,

Journal of Finance 50: 1113-1146.

Sweeney, A. 1994. Debt-Covenant Violations and Managers’ Accounting Responses.

Journal of Accounting and Economics 17: 281–308.

Sufi, A. 2009. Bank Line of Credit in Corporate Finance: An Empirical Analysis. Review

of Financial Studies 22: 1057-1088.

Tan, L. 2010. Creditor Control, State of Nature Verification, and Financial Reporting

Conservatism. Working paper, Northwestern University.

Vinter, G. D. 1998. Project finance: A legal guide. Sweet & Maxwell, London.

Watts, R. 2003. Conservatism in Accounting, Part I: Explanations and Implications.

Accounting Horizons 17: 207–221.

Watts, R., and J. Zimmerman. 1986. Positive Accounting Theory. Prentice-Hall,

Englewood Cliffs, NJ.

Wittenberg-Moerman, R. 2008. The Role of Information Asymmetry and Financial

Reporting Quality in Debt Contracting: Evidence Form The Secondary Loan Market.

Journal of Accounting and Economics 46: 240-260.

Zhang, J. 2008. The Contracting Benefits of Accounting Conservatism to Lenders and

Borrowers. Journal of Accounting and Economics 45: 27-54.

35

FIGURE 1

This figure shows the time trend of mean and median of allowance for doubtful accounts scaled by total

sales around the year of loan origination for borrowers with loans that require aging reports (test sample) in

Panel A and for firms that do not face aging reports requirements (control sample) in Panel B. Year 0

indicates the year when a loan is originated.

Panel A: Test sample (N = 248)

0.004

0.006

0.008

0.01

0.012

0.014

0.016

-2 -1 0 1 2

Test sample

mean allowance median allowance

Panel B: Control sample (N = 248)

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

-2 -1 0 1 2

Control sample

mean allowance median allowance

36

FIGURE 2

This figure shows the time trend of cash from operation scaled by total assets around the year of loan

origination (year = 0) for borrowers with loans that require aging reports (test sample) in Panel A and for

firms that do not have provision of aging report requirements (control sample).

Panel A: Test sample (N = 248)

0.01

0.02

0.03

0.04

0.05

0.06

-2 -1 0 1 2

Test sample

mean cash flow/assets median cash flow/assets

Panel B: Control sample (N = 248)

0.001

0.011

0.021

0.031

0.041

0.051

0.061

0.071

-2 -1 0 1 2

Control sample

mean cash flow/assets median cash flow/assets

37

TABLE 1

Time and Industry Profile of Sample

This table describes the yearly distribution of our sample of borrowers with loans that require aging reports

in Panel A and its industry profile in Panel B.

Panel A: Time profile of sample

year Frequency Percentage Year Frequency Pecentage

1994 4 1.60% 2001 42 16.80%

1995 12 4.80% 2002 19 7.60%

1996 23 9.20% 2003 7 2.80%

1997 23 9.20% 2004 8 3.20%

1998 31 12.40% 2005 8 3.20%

1999 27 10.80% 2006 6 2.40%

2000 40 16.00%

Panel B: Industry profile of sample

Industry FrequencyFrequency

PersentageFrequency

Frequency

Percentage

Agriculture, Forestry, &Fishing 0 0.0% 57 0.3%

Mining 6 2.4% 1,183 6.7%

Construction 2 0.8% 162 0.9%

Manufacturing 134 54.0% 6,073 34.5%

Transportation & Public Utilities 5 2.0% 1,610 9.1%

Wholesale Trade 16 6.5% 539 3.1%

Retail Trade 8 3.2% 900 5.1%

Finance, Insurance, &Real Estate 2 0.8% 3,801 21.6%

Services 75 30.2% 3,068 17.4%

Nonclassifiable Establishments 0 0.0% 217 1.2%

Total 248 100.0% 17,608 100.0%

Sample Firms Compustat Universe Firms

38

TABLE 2

Characteristics of Loan Contracts with Aging Report Requirements

This table presents loan characteristics in Panel A, the purpose of aging reports in Panel B, and the

frequency of aging reports in Panel C. LOAN_AMOUNT is the size of a loan in millions of dollars;

MATURITY is the maturity of a loan in years; MULTILENDERS is a indicator variable equal to one for a

syndicated loan with multiple lenders and zero otherwise; CUTOFF_INVOICE is the maximum number of

days following the invoice date allowable for a customer receivable to be included as eligible accounts

receivable in the computation of the borrowing base; CUTOFF_DUE is the maximum number of days

following the due date of the invoice for a customer accounts receivable to be included in the computation

of the borrowing base; PCT_BASE is the percentage of eligible accounts receivable used as the borrowing

base.

Lower Upper

N Mean Quartile Median Quartile Std Dev

LOAN AMOUNT 242 52.712 18.000 18.000 50.000 102.262

MATURITY 242 2.790 2.000 3.000 3.500 1.434

MULTILENDERS 248 0.508 0.000 1.000 1.000 0.501

CUTOFF_INVOICE 170 104.612 90.000 90.000 120.000 29.000

CUTOFF_DUE 67 69.701 60.000 60.000 90.000 18.152

PCT_BASE 179 81.778 80.000 80.000 85.000 6.555

Panel A. Loan characteristics

Purpose Frequency

Borrowing Base Only 66

Collateral Only 40

Borrowing Base and

Collateral113

Other 29

Total 248

Panel B: Purpose of aging report

Panel C. Frequency of aging reports

Frequency Number Percentage

Weekly 4 1.6%

Monthly 164 66.1%

Quarterly 39 15.7%

Semi-Annually 2 0.8%

Annually 6 2.4%

By Request 33 13.3%

Total 248 100%

39

TABLE 3 Correlation between Variables

This table reports Pearson correlation below the diagonal and Spearman correlation above the diagonal for the test sample. A firm is included in the test sample

when a loan contract when required aging reports can be identified. Firm characteristics are measured at the fiscal-year end immediately prior to the loan

origination year. ALLOW is the allowance for uncollectible accounts receivable; LEV is leverage, defined as total debt (long-term and short-term) divided by

assets. CFO is cash flow from operation scaled by assets; ASSET is natural logarithm of book value of assets; AF is an indicator variable equal to 1 if the firm

has positive analyst following and zero otherwise. MULTILENDER is an indicator variable equal one if a bank loan has multiple lenders and zero otherwise;

HIGHFREQ is an indicator variable equal one if a bank loan requires borrowers to furbish aging reports on a weekly or monthly basis and zero otherwise.

RETVOL is the variance of monthly returns. Correlations with significance 5% (two-tailed) are in bold.

Variable ALLOW RETVOL ROA CFO LN(ASSETS) LEVERAGE AF MULTI HIGHFREQ

ALLOW 0.201 -0.220 -0.095 -0.002 0.034 -0.060 -0.004 -0.048

RETVOL 0.107 -0.318 -0.192 -0.174 -0.044 -0.118 -0.084 0.070

ROA -0.229 -0.354 0.401 0.063 -0.172 0.187 0.009 -0.064

CFO -0.176 -0.254 0.470 0.171 -0.060 0.145 0.114 -0.109

LN(ASSETS) 0.011 -0.185 0.152 0.218 0.297 0.357 0.585 -0.188

LEV 0.124 0.048 -0.021 -0.043 0.272 -0.054 0.252 -0.028

AF -0.130 -0.141 0.173 0.148 0.345 -0.080 0.141 -0.060

HIGHFREQ -0.034 0.075 -0.070 -0.095 -0.197 -0.027 -0.060 -0.108

40

TABLE 4

Firm Characteristics before and after the Loan Initiation

This table reports mean statistics for firm characteristics in the two years before and two years (after and including) the year of loan initiation for both the test

sample and the control sample. ALLOW is the allowance for uncollectible accounts receivable; AR is the gross accounts receivable; WO is the write-offs of

uncollectible accounts receivable; BDX is bad debt expenses. ALLOW, AR, BDX, and WO are scaled by contemporaneous sales. LEV is leverage, defined as

total debt (long-term and short-term) divided by assets. SALES is total sales scaled by assets; CFO is cash flow from operation scaled by assets; ASSET is

natural logarithm of book value of assets; ‘No. Ana Follow’ is the number of analysts following the borrower measured at the fiscal year end before loan

origination. ARTO_IND is industry median accounts receivable turnover ratio, defined as sales divided by average gross accounts receivable; SALE_SD_IND is

industry median standard deviation of sales using quarterly data for all firms in the same industry with available data in Compustat; ALT_IND is industry median

Altman (1968) z-score computed using all firms in the industry with available data in Compustat. Industry classification is based on two-digit SIC codes. ***, **,

and * indicate the statistical significance for the difference of the mean values at the level of 1%, 5%, and 10%, respectively.

Pre Post Pre Post (2) - (1) (4) - (3) (1) - (3) (2) - (4)

(1) (2) (3) (4) (5) (6) (7) (8)

ALLOW 0.013 0.015 0.016 0.016 0.002 0.000 -0.003 0.001

AR 0.191 0.179 0.185 0.169 -0.012** -0.016*** 0.006 0.010**

BDX 0.010 0.012 0.014 0.010 0.002** -0.004*** -0.004 0.002

WO 0.012 0.016 0.013 0.013 0.004*** 0.000 -0.001 0.003

LEV 0.252 0.267 0.250 0.244 0.015 -0.006 0.002 0.023

SALES 1.576 1.543 1.270 1.237 -0.033 -0.033 0.306*** 0.306***

ROA -0.027 -0.065 -0.032 -0.035 -0.038*** -0.003 0.005 -0.030*

CFO 0.013 0.032 0.043 0.041 0.019** -0.002 -0.029** -0.009

ASSET 4.559 4.842 5.355 5.509 0.283*** 0.154*** -0.796*** 0.667***

No. Ana Follow 2.237 2.332 4.528 4.463 0.095 -0.065 -2.291*** -2.131***

ARTO_IND 1.974 2.014 1.974 2.014 0.040*** 0.040***

SALES_SD_IND 0.033 0.031 0.033 0.031 -0.002*** -0.002***

ALT_IND 3.248 2.821 3.248 2.821 -0.427*** -0.427***

Test sample Control sample Mean Diff.

41

TABLE 5

Bank Monitoring and Allowance for Doubtful Account

This table reports the results of regressing allowance for doubtful account on various independent variables.

Each sample firm is included in the regression four times. Two years prior to loan initiation and two years

after. The year of the loan is considered the first year after loan initiation. ALLOW is the allowance for

uncollectible accounts receivable; POST is an indicator variable equal one if the fiscal year is in the

initiation year or after and zero otherwise; AR is gross accounts receivable; WO is the write-offs of

uncollectible accounts receivable; ALLOW, AR, and WO are scaled by contemporaneous sales. LEV is the

leverage, defined as total debt divided by assets. ARTO_IND is the industry median accounts receivable

turnover ratio, defined as sales divided by average gross accounts receivable; SALE_SD_IND is the

industry median standard deviation of sales using quarterly data for all firms in the industry with available

data in Compustat; ALT_IND is the industry median Altman (1968) z-score computed using all firms in the

industry with available data in Compustat. AF is an indicator variable equal to 1 if the firm has positive

analyst following and zero otherwise. ASSET is natural logarithm of book value of assets; Industry

classification is based on the two-digit SIC codes. Standard errors are clustered at the firm level. ***, **,

and * indicate the statistical significance at the level of 1%, 5%, and 10%, respectively.

Dependent Variable = ALLOW t/SALESt Predicted

Sign Coeff p value Coeff p value

Intercept ? 0.037 0.305 0.063 0.013

POST + 0.003 0.004 0.003 0.029

POST×CONTROL -0.004 0.028

ARt + 0.019 0.221 0.022 0.242

WOt ? -0.115 0.019 -0.125 0.023

WOt+1 + 0.197 0.000 0.192 0.000

ARt × CONTROL -0.019 0.508

WOt× CONTROL 0.345 0.000

WOt+1× CONTROL -0.050 0.518

LEVt ? 0.006 0.539 0.004 0.007

ARTO_INDt - -0.013 0.286 -0.012 0.008

SALE_SD_INDt + 0.002 0.989 -0.053 0.098

ALT_INDt - -0.001 0.367 -0.001 0.001

AFt ? -0.001 0.128 -0.001 0.000

ASSETt ? -0.002 0.516 -0.006 0.002

Firm Dummy

Year Dummy

N

Adj-R2

0.83

(1)

Yes

Yes

992

(2)

Yes

Yes

1984

0.77

42

TABLE 6

The Change in the Persistence of Write-offs and Bank Monitoring

This table reports regression results examining the effect of banks’ monitoring incentive on the persistence

of write-offs. WO is the write-offs of uncollectible accounts receivable; POST is an indicator variable equal

one if the fiscal year is in the initiation year or after and zero otherwise; AR is gross accounts receivable;

ALLOW, AR, and WO are scaled by contemporaneous sales. LEV is the leverage, defined as total debt

divided by assets. ARTO_IND is the industry median accounts receivable turnover ratio, defined as sales

divided by average gross accounts receivable; SALE_SD_IND is the industry median standard deviation of

sales using quarterly data for all firms in the industry with available data in Compustat; ALT_IND is the

industry median Altman (1968) z-score computed using all firms in the industry with available data in

Compustat. AF is an indicator variable equal to 1 if the firm has positive analyst following and zero

otherwise. ASSET is natural logarithm of book value of assets measured at the fiscal year end before loan

origination; ROA is net income before extraordinary item scaled by total assets. Industry classification is

based on the two-digit SIC codes. Standard errors are clustered at the firm level. ***, **, and * indicate the

statistical significance at the level of 1%, 5%, and 10%, respectively.

Dependent Variable = WOt Predicted

Sign Coeff p value Coeff p value

Intercept ? -0.010 0.426 0.001 0.934

WOt-1 + 1.112 < 0.0001 0.823 < 0.0001

POST + 0.000 0.949 -0.001 0.405

POST * WOt-1 - -0.206 0.067 -0.025 0.793

ARt + 0.029 0.338 0.019 0.374

LEVt ? 0.006 0.430 0.006 0.202

ARTO_INDt - 0.001 0.803 0.004 0.086

SALE_SD_INDt + 0.022 0.687 -0.116 0.086

ALT_INDt - 0.000 0.893 0.000 0.942

AFt ? 0.000 0.671 0.000 0.632

ASSETt ? 0.001 0.272 -0.001 0.306

ROAt - -0.030 0.013 -0.004 0.948

Year Dummy

N

Adj-R2

Test Firms Control Firms

0.78

Yes

744

Yes

744

0.57

43

TABLE 7

Bank Monitoring Intensity and Allowance for Doubtful Accounts

This table reports regression results examining the effect of banks’ monitoring incentive on borrowing

firms’ bad debt expenses. MULTILENDER is an indicator variable equal one if a bank loan has multiple

lenders and zero otherwise; HIGHFREQ is an indicator variable equal one if a bank loan requires

borrowers to furbish aging reports on a weekly or monthly basis and zero otherwise. ALLOW is the

allowance for uncollectible accounts receivable; POST is an indicator variable equal one if the fiscal year is

in the initiation year or after and zero otherwise; AR is gross accounts receivable; WO is the write-offs of

uncollectible accounts receivable; ALLOW, AR, and WO are scaled by contemporaneous sales. LEV is the

leverage, defined as total debt divided by assets. ARTO_IND is the industry median accounts receivable

turnover ratio, defined as sales divided by average gross accounts receivable; SALE_SD_IND is the

industry median standard deviation of sales using quarterly data for all firms in the industry with available

data in Compustat; ALT_IND is the industry median Altman (1968) z-score computed using all firms in the

industry with available data in Compustat. AF is an indicator variable equal to 1 if the firm has positive

analyst following and zero otherwise. ASSET is natural logarithm of book value of assets; Industry

classification is based on the two-digit SIC codes. Standard errors are clustered at the firm level. ***, **,

and * indicate the statistical significance at the level of 1%, 5%, and 10%, respectively.

Dependent Variable = ALLOW/SALESt Predicted

Sign Coeff p value Coeff p value

Intercept ? 0.052 0.296 0.041 0.344

POST + 0.001 0.701 0.004 0.013

ARt + 0.001 0.984 0.024 0.195

WOt + -0.165 0.077 -0.098 0.146

WOt+1 + 0.191 0.000 0.205 0.002

LEVt ? -0.003 0.773 0.014 0.200

ARTO_INDt - -0.016 0.404 -0.016 0.351

SALE_SD_INDt + 0.362 0.465 -0.059 0.741

ALT_INDt - -0.002 0.303 -0.001 0.518

AFt ? -0.001 0.066 0.000 0.743

ASSETt ? -0.001 0.772 0.000 0.869

Firm Dummy

Year Dummy

N

Adj-R2

HIGHFREQ = 0 HIGHFREQ = 1

(i) (ii)

Yes Yes

Yes Yes

320 672

0.78 0.76

44

TABLE 8

The Change in Customer Choice after Borrowing

This table reports regression results of examining the change in sales concentration and customer credit risk after initiation of a borrowing-base loan.

AVGSALEPCT is the average percentage of sales to a customer with the minimum of 10%; LNUMCSTMER is the natural logarithm of the number of

customers with a percentage of sales exceeding10%; CSTMRATING is the weighted average monthly customers' S&P domestic issuer credit rating, weighted by

the percentage of a firm's sales to that customer. If a customer has no credit rating, then the credit rating is predicted by a model using natural logarithm of total

assets, ROA, leverage, a dummy variable measuring whether a firm pays dividend, a dummy variable measuring whether a firm issues subordinated debt, and a

dummy variable measuring whether a firm incurs loss in the current period; POST is an indicator variable equal one if the fiscal year is in the loan initiation year

or the following year and zero otherwise; ASSET is natural logarithm of suppliers’ book value of assets; ROA is suppliers’ net income over suppliers’ total assets;

CFO is suppliers’ cash flow from operation over is total assets; CREDITRATING is suppliers’ monthly average of the S&P domestic long-term issuer credit

rating. If a supplier has no credit rating, then the credit rating is replaced by the predicted credit rating using the same procedure described above; Credit rating

ranges between 2 (S&P rating =AAA) and 27 (S&P rating = D). MKTSHARE is the supplier’s sales over the total sales of all firms in the same three digit SIC

industry. Industry classification of industry dummy is based on the two-digit SIC codes. Standard errors are clustered at the firm level. ***, **, and * indicate the

statistical significance at the level of 1%, 5%, and 10%, respectively.

Coeff p value Coeff p value Coeff p value

Intercept 0.504 0.000 1.322 0.000 3.042 0.003

POST -0.012 0.025 -0.020 0.083 -0.162 0.037

ASSETt -0.007 0.012 0.003 0.664 -0.225 0.000

ROAt 0.024 0.062 0.051 0.041 0.256 0.159

CFOt -0.056 0.007 -0.029 0.492 -0.375 0.207

CREDITRATINGt 0.001 0.548 0.003 0.172 -0.006 0.522

MKTSHAREt 0.029 0.494 0.241 0.171 0.144 0.779

Industry Dummy

Year Dummy

N

Adj-R2

Dependent Variable = AVGSALEPCTt Dependent Variable = LNUMCSTMERt Dependent Variable = CSTMRATINGt

(3)

Yes

(1)

Yes

Yes

1394

0.13

(2)

Yes

Yes

5129

0.1

Yes

0.08

5129

45

Table 9

The Change in Disclosure Frequency of Allowance after Borrowing

This table reports univariate analysis of the change in the disclosure frequency of allowance for doubtful

accounts in the two years prior to borrowing (-2) through the two years after borrowing (+2). Disclosure is

collected from EDGAR-10Q and the sample contains 181 loans. If a borrower discloses the balance of

allowance, then the disclosure frequency is coded as one, zero otherwise. Quarterly disclosure of

allowance is cumulated to arrive at the annual frequency. The mean annual frequency is presented in each

cell and the difference in mean is also reported.

-2 -1 0 1 2 Mean diff p-value

Frequency 2.839 2.966 3.116 3.121 3.044 0.216 0.0353

Pre-Period Post-period Difference (Post - Pre)