debt contract restrictiveness and timely loss … · debt contract restrictiveness and timely loss...

Debt Contract Restrictiveness and Timely Loss Recognition*

Valeri Nikolaev §

Tilburg University

[email protected]

Abstract

I examine whether firms with more covenant restrictions in their public debt contracts, exhibit recognition of economic losses in earnings in a more timely fashion. Covenants limit the ability of a manager to take actions leading to bondholder wealth expropriation. Specifically, covenants govern the transfer of control rights from shareholders to bondholders when a firm approaches financial distress. As most covenants are formulated in terms of accounting numbers and become binding when accounting performance deteriorates, I argue that covenants will be more valuable in constraining managerial opportunism when the accounting system generates timely signals of a firm’s economic health. I hypothesize that more extensive reliance on covenants in lending agreements is associated with higher demands for timely loss recognition. The results of the analysis are consistent with this hypothesis. In a sample of more than 3000 debt issues, I find that firms with the most restrictive debt contracts are twice as timely in recognizing economic losses as firms with the least restrictive debt contracts. Keywords: Timely loss recognition, debt contracting demands, restrictive covenants JEL Classification: M4, M20, G30

First version: June 2006 This version: October 2006

___________________________ * I carried out the empirical analysis in this paper while I visited the University of Chicago, GSB, which I would like to thank for its hospitality. I thank Ray Ball for encouragement and valuable discussions. I am indebted to Laurence van Lent for continuous support and guidance. Helpful comments were received from Willem Buijink, Philip Joos, S.P. Kothari, Christian Leuz, Maarten Pronk, Richard Sansing and the workshop participants at Maastricht University, Tilburg University and University of Antwerp. This study was supported by a grant (R 46-570) from the Netherlands Organization for Scientific Research (NWO). § Center for Economic Research, Tilburg University, Office B805 Warandelaan 2, 5000LE, Tilburg, the Netherlands.

Debt Contract Restrictiveness and Timely Loss Recognition

I. Introduction

I examine whether firms with more covenant restrictions in their public debt contracts

recognize economic losses in earnings in a more timely fashion. Covenants are contractual

stipulations designed to limit managerial control over a firm as it encounters financial difficulties.

Often, covenants are formulated in terms of accounting numbers and become binding when the

firm fails to meet certain thresholds of accounting performance. Covenant constraints are

designed to protect bondholders from opportunistic behavior by management, such as making

unwarranted distributions to shareholders and non-optimal investments (Jensen and Meckling,

1976; Myers, 1977; Smith and Warner, 1979). An accounting system’s failure to generate timely

signals of a firm’s economic health reduces the efficiency of using covenant restrictions in

curbing the agency costs of debt. Timely recognition of losses facilitates the early transfer of

decision rights from shareholders to bondholders as a firm approaches financial difficulties and

thus reduces the likelihood of bondholder wealth expropriation (Watts, 2003a; Ball and

Shivakumar, 2005; Ball, Robin and Sadka, 2005). Therefore, I hypothesize that more extensive

reliance on covenants in lending agreements is associated with higher demands for timely loss

recognition, i.e., accounting conservatism. I expect that the more timely a firm is in recognizing

losses, the greater is the reliance on covenants, which in turn enables the firm to reduce the

agency costs of debt.

There has been relatively little research on the role of accounting information in debt

contracts and in particular on how the microeconomic foundations of debt contracting influence

the properties of accounting information (e.g. Sloan, 2001). In particular, Guay and Verrecchia

(2006) encourage a researcher to examine the direct link between accounting-based contracts and

the degree of conservatism in public reports. This study focuses on the timeliness of economic

3

loss recognition in earnings and how it relates to firms’ debt contracting choices.1 Timely loss

recognition is the primary means of achieving conservative accounting and is often referred to as

conditional conservatism in the literature.2 Following Basu (1997), I operationalize it via a

piecewise linear regression of earnings on positive returns (as a proxy for good news) and

negative returns (as a proxy for bad news).

Accounting information is useful in contracting (Watts and Zimmerman, 1986) and the

demands from contracting parties (e.g., lenders) shape its properties. Debt contracting, in

particular, is among the most important sources of conditional conservatism (Watts, 2003a;

Holthausen and Watts, 2001; Ball, Robin and Sadka, 2005). In particular, conflicts of interests

between bondholders and shareholders are more severe when firms rely on public rather than on

private debt (or equity) because it is more difficult for outsiders to monitor and control the

manager’s actions. Accounting in public firms responds accordingly by recognizing economic

losses more promptly in earnings (Ball, Kothari and Robin, 2000; Ball and Shivakumar, 2005).

Several concerns regarding the role of accounting conservatism for debt contracting are

expressed in Guay and Verrecchia (2006). The authors argue that debt-holders’ demands for

conservative accounting information need not influence firms’ reporting policies as debt contracts

may adjust accounting numbers in accordance with their needs. The argument assumes that it is

cost-effective to substitute firms’ conservative accounting choices within GAAP with the pre-

specified adjustments to accounting information, which the parties contract upon. In addition, the

authors argue that, when investigating the link between economy-wide institutions and

accounting conservatism, it may be misleading to infer the use of accounting-based contracts

from the institutional features, as this essentially represents a joint hypothesis test. In the light of

these arguments, an empirical investigation on how the reliance on covenant restrictions

1 From a debt contracting perspective there is less demand for timeliness in gain recognition (Ball and Shivakumar 2005, Guay and Verrecchia 2006). 2 For example, see Ball and Shivakumar (2005), Ball, Robin and Sadka (2005), Ryan (2006).

4

influences the demand for conditionally conservative reporting in a cross-section of public firms

is warranted.

The efficiency-enhancing role of accounting information in debt contracts implies that

conservatism and the use of covenants complement and reinforce each other. Both covenants and

timely loss recognition can be used separately to discipline the management and signal firm’s

type to the market. However, in line with the argument in the Levine and Hughes (2005) model,

covenants are more valuable when losses are recognized in timely manner as this allows bonding

against ex post sub-optimal actions more effectively. Since covenant violation is costly,

conservative recognition of news induces early truth telling about future cash flows and thus

allows a company with lower default risk to signal its type.

I use the Mergent Fixed Investment Securities Database to retrieve information about

covenants stipulations in public debt contracts. Covenant information is available for a large

cross-section of debt issues by industrial companies. The data allow me to construct five indices

of debt contract restrictiveness. These indices measure (1) the overall restrictiveness of the

contract, (2) restrictions on new investments, (3) restrictions on the distribution of funds to

shareholders, (4) restrictions on future financing, and finally, (5) the transfer of control to

bondholders when default becomes probable. Since most of the covenants depend on accounting

information, these indices are used to proxy for the extent to which contract is linked to

accounting.

The findings suggest that firms that use more covenants in their debt contracts are

considerably more timely in their economic loss recognition. Companies with the most restrictive

debt contracts, as judged by the overall restrictiveness index, are about twice as timely in

recognizing economic losses as are the firms with the least restrictive contracts. Results are

similar when I separately examine the four indices of restrictiveness. Furthermore, the correlation

between the rank of contract restrictiveness and the estimates of timely loss recognition is as high

as 0.71 for overall restrictiveness index, 0.78 for the investment restrictions index, 0.7 for the

5

payout restrictions index, and 0.66 for financing restrictions. The results are somewhat weaker

for the transfer of control covenant index. Overall, the results are consistent with the use of

covenants increasing the demands for timelier recognition of losses.

This study contributes to the literature on the role of accounting choice and information

properties in debt contract design.3 Much of prior research focuses on firms’ incentives to make

ex post accounting choices that enable these firms to decrease the likelihood of costly covenant

violation. While some studies find evidence that accounting choices are made to avoid covenant

violations, others are inconclusive (see Fields, Lys and Vincent, 2001 for a review and

discussion). One reason for these relatively weak findings may be that debt providers anticipate

managerial incentives to make opportunistic accounting choices and thus restrict the acceptable

set of accounting practices (Watts and Zimmerman, 1990; 1986). I examine how accounting that

limits managerial choices ex post, relates to the design of debt contracts ex ante. The analysis

suggests that timely loss recognition can reduce the agency costs of debt via the use of protective

covenants. To my knowledge, this is the first paper that studies how conditionally conservative

accounting is associated with the degree of debt contract restrictiveness as measured by the

number of protective covenants.4,5

A closely related study is by Beatty, Weber and Yu (2006) who examine how

conservative modifications of accounting information specified by protective covenants are

related to conditional conservatism. They find that lenders’ demands for conservative reporting

are not fully met via the inclusion of conservative net worth covenant modifications, and the latter

are further complemented by the within-GAAP choice of timelier loss recognition.

3 Debt contract design in the context of accounting also has received attention in Begley (1994), Bharath, Sunder and Sunder (2006), Begley and Chamberlain (2005), Begley and Feltham (1999), Beatty, Ramesh and Weber (2002), Beatty and Weber (2003), Press and Weinthrop (1990), Sweeney (1994) among others. 4 Begley and Chamberlain (2005) focus on unconditional whereas I examine conditional conservatism. 5 See Ball and Shivakumar (2005a) for a discussion of conditional vs. unconditional conservatism.

6

A number of studies examine how conservatism affects the cost of debt capital and

degree of information asymmetry between bondholders and the firm (Ahmed et al, 2003, Zhang,

2005, Moerman, 2005). While these findings are important for our understanding of accounting

conservatism, the exact mechanism through which the benefits of timely loss recognition are

captured has not yet been investigated empirically. A widely held view is that a substantial part

of improvement in contract efficiency (due to a higher degree of conditional conservatism) is

realized via the use of restrictive covenants included into indentures (Watts 2003; Ball and

Shivakumar, 2005; Ball, Robin and Sadka, 2005). A direct link, however, between timely loss

recognition and debt contract design has still to be established.

The second contribution of this study is to the empirical literature on the demand for

financial reporting quality (Ball, Kothari and Robin 2001; Ball and Shivakumar 2005, Ball, Robin

and Sadka 2005). Although it has been recognized that debt contracting is causally linked with

conditional conservatism (Watts 2003a, Holthausen and Watts 2001), empirical evidence from the

field remains limited to cross-country and cross-market examinations. More work is needed to

investigate whether holding constant the properties of accounting information at macro-level,

firm’s debt contracts influence their accounting choices. The analysis in this study suggests that

firm-level debt contracts put specific demands on the properties of accounting information.

The remainder of the study is organized as follows. The next section reviews the related

literature and describes the empirical predictions. Section III outlines the research method used in

the paper. Section IV describes the data sources and variable definitions. Section V reports and

discusses the empirical findings. Finally, section VI concludes the study.

II. Hypothesis Development

I first briefly review the related literature. Subsequently, I discuss the role of debt

covenants in reducing the agency costs of debt that arise when companies approach financial

7

distress. Than, I argue that timely loss recognition improves debt contract designs that include

protective covenants. I proceed with empirical implications.

Related literature

This study relates to several lines of research. First, by investigating the link between

protective covenants and timely loss recognition, it complements the studies that examine the

efficiency gains from conservatism. Conservatism reduces the cost of debt (Ahmed et al, 2002;

Zhang, 2005). Furthermore, Ahmed et al. (2002) show that conservatism mitigates bondholder-

shareholder conflicts over dividend policy. Zhang (2005) provides evidence that conditional

conservatism causes restrictive covenants to be triggered early as the firm approaches financial

distress. Finally, Moerman (2005) demonstrates that timely incorporation of economic losses in

earnings reduces the information asymmetry component in the cost of debt in the secondary loan

market.

The second line of research investigates how demands for timely loss recognition

influence reporting practices. Ball and Shivakumar (2005, 2006b) argue that public corporations

face higher demands for timely loss recognition due to their reliance on public financing;

bondholders at arm’s length have few alternative means to become informed and therefore rely

more on publicly reported accounting information. Ball, Robin and Sadka (2005) show that

conditional conservatism is driven by the importance of public debt in the economy. Finally, in a

related study, Ball, Kothari and Robin (2000) demonstrate that international differences stemming

from the ways conflicts of interest between debt- and equity-holders are resolved, are associated

with the degree of timely loss recognition.

Finally this study relates to the literature on design of debt contracts (e.g., Begley 1994,

Begley and Feltham 1999, Beatty et al. 2002, Beatty and Weber, 2003, Bharath et al. 2006, Press

and Weinthrop 1990, Sweeney 1994) and in particular to Beatty, Weber, and Yu (2006) who

examine conservative modifications to accounting numbers specified by debt covenants. Their

analysis suggests that firms may find it too costly to meet the demands for timely loss recognition

8

through covenant modifications, which partly addresses the concerns raised in Guay and

Verrechia (2004).

Role of bond covenants

When a firm approaches financial distress, bondholders become more vulnerable to

wealth-expropriation by managers or shareholders (Bodie and Taggert, 1978, Smith, Smithson

and Wilford, 1989, Nash, Netter, and Poulsen, 2003). Debt overhang (Myers, 1977), asset

substitution (Jensen and Meckling, 1976), and claim dilution (e.g. Nash et al., 2003) are well

known conflict-of-interest problems that are elevated in financially-troubled firms.6,7 Covenant

restrictions reduce the scope of managers and owners to take actions that harm bondholders in

sensitive areas such as investments, dividend payouts, and financing. For example, covenants

that impose bounds on the amount of debt the company can issue alleviate conflicts of interests

that involve existing bondholders. Additionally, a wide range of more specific covenants may be

used. Covenants limiting distributions of dividends to the shareholders effectively force levered

firms to invest and thus alleviate debt overhang associated with the potential unwillingness of

managers to undertake positive NPV projects. Covenants that restrict the investment policies of

firms and the lines of business they engage in reduce the likelihood of asset substitution. More

specifically, covenants can be used to constrain asset sales, mergers and acquisitions, or to

prevent firms from pursuing certain activities; all of these represent obstacles for management to

over-invest in risky projects after obtaining debt financing. Finally, covenants which place

restrictions on leases and sales-and-lease back transactions as well as negative pledge covenants

reduce claim dilution when firms issue additional debt (possibly of higher priority) and thus dilute

the value of current bondholder claims due to a higher probability of default.

6 Debt overhang, for example, is associated with project requiring a sequence of investments. After the initial investments are sunk, managers do not internalize the losses that accrue to the debt holders if late investments are not made and the projects lapses. This is more likely to happen in financial distress. More generally, debt overhang reduces incentives to invest and exert effort. 7 Asset substitution is an over-investment problem, which arises when shareholders substitute riskier assets fro the firms existing assets and expropriate value from the debt holders

9

What determines the degree of restrictiveness?

Covenants are designed to protect bondholders, covenants may become binding while the

firm is still financially healthy. Covenants can in some circumstances decrease manager’s ability

to make decisions that benefit the firm as a whole. Managers may have to forsake good

investment projects because they are not allowed to obtain additional financing. Alternatively,

covenants may prevent firms from distributing “free cash flow” to shareholders which would

otherwise reduce the agency costs (e.g., empire-building) associated with having excess cash.

Moreover, costs of technical default and renegotiation incurred by the company are significant

(Beneish and Vargus, 1993, 1995). These, altogether, make covenants a useful signaling device

through which managers may reveal their firm’s quality. The tradeoff between the costs and

benefits plays a key role in the design of debt contracts (Smith and Warner, 1976; Begley, 1994;

Nash et al., 2003).

Existing evidence is generally consistent with this tradeoff view on the use of covenant

restrictions. On the one hand, covenants are frequently when agency costs are expected to be

high. Therefore default risk, managerial entrenchment, firm governance, and size are associated

with the use of covenants (Malitz 1986, Begley, 1994, Begley and Feltham, 1999, and Chava,

Kumar and Warga, 2005). Covenants also help to reduce the cost of debt capital, presumably due

to reduced agency costs of debt (Chava, Kumar and Warga, 2005; Bradley and Roberts, 2005;

Reisel, 2005; Goyal, 2005).

On the other hand, firms forego covenant use when the costs of restricting managerial

discretion are high. Thus, firms with growth opportunities (Nash, Netter, and Poulsen, 2003,

Reisel, 2005, Chava, Kumar and Warga, 2005, and Khan and Yermack, 1998) or firms in volatile

environments (Anderson, 1999) impose fewer debt covenants.8 Breadley and Roberts (2005) find

8 A number of studies, however, suggest that high-growth, volatile firms with many investment opportunities may be perceived as more risky and thus include covenant constraints in their debt contracts (Nash, Netter and Warga, 2005).

10

that, while high-growth companies restrict the use of obtained funds, they avoid constraints

limiting their ability to raise additional funds.

Next I discuss the role of timely loss recognition is contracts that include protective

covenants. Timely loss recognition enhances the efficiency of covenant use through (1) early

transfers of decision rights to bondholders and (2) signaling a firm’s credit quality.

Distress and timely loss recognition

Timely loss recognition (conditional conservatism) represents a salient dimension of

accounting quality and is believed to play an efficiency-enhancing role in contracting.

Recognizing losses in a timely manner in earnings brings forward covenant violations in

financially distressed firms. Thresholds specified in covenants are commonly based on

accounting information immediately linked to reported earnings, book values of assets, liabilities,

and equity. Covenants can be more efficient in mitigating the agency problems associated with

financial distress if the accounting numbers incorporate adverse economic events in a more timely

fashion. This will ensure the early transfer of decision rights from managers to bondholders at the

time when bondholders face greatest expropriation risk. As timely loss recognition increases the

usefulness of accounting information in debt contracts with accounting-based covenants, this

creates a demand for timeliness in recognizing economic losses. Consistent with this idea,

economies where public debt plays relatively more important role exhibit the evidence of timelier

loss recognition (Ball, Kothari and Robin 2000, Ball, Robin and Sadka 2005, Ball and

Shivakumar 2005, 2006b)

Covenants as a signaling device and timely loss recognition

A helpful step towards understanding the concurrent use of covenants and conservative

accounting has been made in Levine and Hughes (2005), who model the use of accounting-based

debt covenants in tandem with the choice of conservative reporting. The authors demonstrate that

the use of covenants together with (conditionally) conservative reporting is an optimal contracting

11

mechanism.9 In their model, a firm seeks debt financing and signs two different contracts, a

contract with the manager and a debt contract with lenders. The compensation contract aims at

aligning managerial incentives. A firm with a lower risk of default may choose to design

incentive compensation sub-optimally to signal its type to lenders by distorting optimal operating

decisions. The introduction of a bond covenant based on earnings combined with conservative

measurement of earnings prevents having to incur these signaling costs as it forces the “lesser

type” firm into costly default early, because the latter cannot mimic the covenant threshold set by

the better type firm.10

More generally, the literature recognizes the signaling role of debt (Jensen, 1986; Harris

and Raviv, 1990; Zwiebel, 1996) and of debt covenants in particular (Garleanu and Zwiebel,

2005, Chava et al., 2005, Sridhar and Magee, 1996). Since default is costly, managers who are

privately informed about a firm’s poor future profitability or managers of firms suffering from

agency problems will separate themselves out by not including performance-based covenants into

their debt contracts; consequently, these firms will be forced to pay a higher risk premium.

However, signaling through debt contracts with accounting-based covenants is unlikely to be

successful unless the accounting information exhibits timely loss recognition. Since reported

performance is directly related to a manager’s welfare, he has incentives to introduce bias and

noise into accounting measures used in contracts (Watts and Zimmerman, 1986, 1990; Watts,

2003a). Conditional conservatism curbs managerial ability to bias accounting numbers upwards

and enables covenants to perform the signaling role better. In the absence of this property,

lenders are unlikely to rely on covenant restrictions and will look for other (costly) ways to

9 In this stylised model no distinction is made between conditional and unconditional conservatism. Nevertheless, conditional conservatism fits the spirit of the model, arguably, better as their result is driven by the downside risk facing bondholders. 10 For example, “good” firms can afford to distort incentives of the manager (e.g., pay for performance sensitivity) as the contract is publicly observed and thus represents a costly signal of firm’s “quality”.

12

mitigate agency problems that may include increasing the risk premium, reducing the maturity of

the debt or providing funds in small installments.11

Empirical predictions

Since the above suggests that the ability of covenant restrictions to detect and to prevent

agency problems depends crucially on the ability of accounting information to generate

conditionally conservative earnings numbers, I expect that inclusion of covenant restrictions into

public debt contracts is associated with higher demands for timely recognition of economic

losses. There are several potential mechanisms through which this association obtains, which I

describe in more detail below.

First, companies seeking to benefit from covenants as a contractual mechanism should

anticipate the demand and react accordingly by adopting timelier, i.e. more conditionally

conservative, financial reporting ex ante. 12 Since debt contracts require consistency of accounting

practices, adoption of more conservative policies is expected to take place in years prior to the

year when contract is signed. There are many examples of conditionally conservative accounting

policies. A company may commit ex ante to account for bad debts using the aging of receivables

method. In this method bad debt expense is based on the age of accounts receivable and as such

represents a more-timely way to account for expected losses than the percentage of sales method,

which expenses fixed percentage of credit sales irrespective of their collectibility. Increase in

collection periods, possibly due to deteriorated financial health of major customers or to relaxed

credit-granting policies, may be an early signal of default. Thus the aging of receivables method

brings recognition of adverse economic events forward and therefore is preferable from the

lender’s viewpoint. Appendix A provides a number of additional examples of timelier loss

recognition policies.

11 See Nash et al. (2003) for a detailed discussion. 12 I.e. the restrictions in the form of timely loss recognition are placed on the set of accounting practices ex ante to decrease the likelihood of ex post managerial opportunism (Watts and Zimmerman, 1986; 1990).

13

Second, firms and their lenders have long-term relations. The quality of these relations

can represent a significant value to the company. Therefore, lenders can punish the company ex

post. A firm failing to be timely in loss recognition will tarnish its reputation. This can result in

higher borrowing costs and/or more severe contract terms when renegotiations to waive the

covenants take place. Additionally, if a firm deviates from recognition of incurred losses in

timely fashion, bondholders can discipline the management by appealing the court. This implies

that borrowers will be more careful in recognizing losses that would probably be overlooked in

the absence of accounting based debt contracts.

This leads to the following hypothesis:

H1: Ceteris paribus, a greater use of covenant restrictions in public debt contracts is

associated with more timely recognition of economic losses.

III. Research Design

I construct four indices of debt contract restrictiveness to examine the hypothesized

relation between restrictions placed on managerial actions, such as financing, investments, and

distributions and the degree of asymmetric timeliness. Each index aims at capturing the degree a

debt contract is linked to accounting information and also serves as a proxy for (inverse of)

managerial flexibility in making key decisions. The following types of restrictions are

distinguished to construct individual indices:

1) Investment covenants restrict capital allocations, M&A deals and disposal of assets.

2) Distribution or payout covenants restrict payments to the shareholders or other entities.

3) Financing covenants limit managerial ability to raise funds via issuing additional debt or

common/preferred stock and sale-and-leaseback transactions.

4) Control transfer covenants facilitate transfer of control from shareholders to bondholders

when the company approaches financial distress.

14

I also construct an overall contract restrictiveness index, which is the sum of the four

indices. For more detail on the covenant restrictions that are used to construct each index, see

Appendix B.

As the costs and benefits of covenants use are in part predetermined by environmental

and innate characteristics of the firm, these characteristics are likely to reflect market forces that

influence conservative reporting other than the considerations governing the use of covenants. It

is, therefore, necessary to control for these factors. For example, prior research shows that the

book-to-market ratio correlates with the degree of asymmetric timeliness of loss recognition

(Roychowdhury and Watts, 2005); at the same time book-to-market, a proxy for growth, also

correlates with covenant use. Other characteristics, such as size, volatility, leverage, and

probability of bankruptcy are also likely to correlate with both the use of covenants and

asymmetric timeliness, which can induce a spurious relation. To mitigate the influence of these

factors, I orthogonalize the indices of contract restrictiveness by running a first stage regression.

Namely, I regress the indices on firm-specific characteristics to separate out unexplained variation

in covenant use, in the form of the residual from the following model:

(1), DummiesYear Amount Principal*

Maturity*I)St.Dev(IBEscore-Z*Losses ofNumber *

urns)St.Dev(Ret*Growth*tBook/Marke*Leverage*

Yield Dividend*ROA*)log(Assets* Index enessRestrictiv

12

111098

7654

3210

ξααααα

αααααααα

+++++++

+++++++=

where the dependent variable is one of the five restrictiveness indices described above.

Intuitively, the residual in (1) proxies for situations when the benefits of covenant use outweigh

their costs (or vice versa), i.e., when covenants are chosen to constrain (unobserved) agency

problems. Identifying such situations, while randomizing with respect to other factors that

influence reporting incentives, is necessary to address the research question.

The control variables in Equation (1) have been argued in prior research to affect the use

of covenants and are likely to correlate with reporting properties. Size, profitability, volatility of

15

returns, financial leverage and Altman’s bankruptcy score (Z-score) are expected to correlate with

protective covenants as they proxy for the probability of financial distress. Book-to-market ratio

and growth in assets are likely to make the use of covenants costly, as retaining flexibility is

especially valuable in these firms. Dividend yield controls for costs (i.e., negative stock price

effects) associated with the inability to pay out normal level of dividends. The variability of

accounting income is expected to be associated with higher expected costs of violating covenants.

In addition, I include two debt contract features that may be used to overcome the agency

problems (and could substitute for covenants): maturity and the size of the issue. Finally, I

include year dummies to control for possible trends in covenant use.13,14

At the second stage, I sort companies on unexplained variation of covenant use ξ and

allocate them to 10 restrictiveness portfolios. Subsequently, the degree of timely loss recognition

is assessed for all firms in each decile. Running the first stage model to control for confounding

factors avoids numerous interaction terms and allows for a parsimonious second stage model.15

Following Basu (1997), I measure timeliness of loss recognition by regressing accounting income

scaled by lagged price on annual return conditioning the relationship on the sign of economic

news (proxied by the sign of the returns). Specifically, the following model is estimated across

covenant restrictiveness deciles.

)2()0Ret(RetRet)0Ret(P/E tt1t0t101 ttt DD εββαα +<×++<+=−

13 Begley and Freeman (2004) show that the use of covenants in public debt contracts has declined. At the same time this is does not seem to be the case for all types of restrictions. The evidence in Billett, King and Mauer (2005) and Chava, Kumar, and Warga (2005), based on FISD data, suggests that while some of the covenants (e.g. dividend constraints) have become less popular, the frequency of others (e.g., certain types of investing, financing, and control transfer restrictions) have increased over time. 14 Developing a comprehensive set of variables that potentially affect the use of covenants is outside the scope of the paper; the primary reason for including these control variables is because they are potentially related to reporting quality and conservatism. 15 I do not follow an approach with introducing multiple interaction terms for the following reasons: (i) the number of interaction terms in a single stage model equals 4 variables in Basu (1997) model times 12 control variables in equation 1 which leads to over-parametrized specification (ii) there is no reason to expect linear increase in timely loss recognition across restrictiveness deciles (iii) inclusion of interactions may result in multicollinearity; (iv) the standard errors from the regression will suffer from cross-sectional dependencies.

16

where D(.) is the indicator dummy taking the value of unity when the condition inside the

parenthesis is true, Et stands for earnings at time t, Pt-1 is price (at time t-1), and Rett is annual

market-adjusted return (at time t). The degree of asymmetric timeliness is measured by

coefficient β1 (or alternatively β0+β1).

I use several measurement time horizons surrounding the debt issue. Specifically,

equation (2) is estimated over years –3 to –1, year –1, year +1 and years +1 to +3 relative to the

year of issue. The purpose of examining the years prior to the issue is twofold. First, since the

lenders scrutinize prior financial statements, this reveals whether companies anticipating debt

issues change their accounting beforehand to either create good reputation or simply commit to

more conservative accounting practices ex ante (Ball and Shivakumar, 2006b). Second, it may be

more difficult to document the asymmetric timeliness ex post as managers have incentives to

introduce additional noise into accounting numbers when trying to minimize the likelihood of

covenant violations within the accepted set.

IV. Data and Sample construction

The data are taken from the Mergent Fixed Investment Securities Database (FISD), which

is a comprehensive database of publicly traded U.S. bonds. FISD contains various bond

characteristics including the principal amount, maturity, and price, but its distinguishing feature is

that it also contains detailed information about the presence of covenants in the bond indenture.

Specifically, I use 41 indicators of various types of covenants to construct the five contract

restrictiveness indices discussed in the previous section. I sum the covenant indicators related to

investment (INVEST), financing (FINAN), distribution (DIST) decisions, and control transfers

(CONTROL) to compute the scores on each of the four separate indices. The sum of these four

indices represents the overall restrictiveness index (OVERAL). Appendix B provides more detail

on covenants used to construct the indices.

17

Balance sheet and income statement data are taken from the COMPUSTAT Industrial

Annual database and monthly return data are retrieved from CRSP. The sample construction is

summarized in Panel A of Table 1. Covenant information is available for 11,947 bond issues by

4394 industrial companies in FISD. This study excludes financial firms because they are subject

to different accounting rules and regulations. After merging FISD data with COMPUSTAT and

retaining only the first debt issue in a given year (in order to give equal weight to all companies),

the sample is reduced to 2,367 firms with 5,036 debt issues over the period of 1980-2004.16

Further requirement with regard to data availability of the variables used in the analysis reduces

the sample to 3,382 issues.

Variable Definitions

Income before extraordinary items (COMPUSTAT Item data18) scaled by the beginning

of fiscal year price (data199) times shares outstanding (data25) is used to measure the dependent

variable in Equation 2 model 1-tt /PE . Returns (Rett) are compounded over a 12 month period

starting 3 months after the beginning of the fiscal year and are adjusted subsequently by

subtracting the compounded return on a value-weighted market index.

Control Variables in Equation (1) are measured in the year prior to the debt issue and are

defined as follows:

log(Assets) = natural logarithm of total assets (data6);

ROA = return on assets defined as income before extraordinary items (data18) divided by

total assets (data6);

Dividend Yield = dividends (data21) divided by end of year market value (data199 times

data25);

Leverage = ratio of long-term debt (data9) to total assets (data6);

16 FISD data is very incomplete prior to 1980.

18

Book-to-Market = book value of equity (data60) divided by market value of equity

(data199 times data25);

Growth = growth rate in total assets (data6);

St.Dev.(Returns) = standard deviation of monthly returns in CRSP;

St.Dev.(IBEI) = standard deviation of income before extraordinary items (data18) scaled by

total assets (data6), measured over the five years preceding the issue;

Number of Losses = number of times the company had a loss over the five years preceding

a debt issue;

Z-score = Altman’s bankruptcy score;17

Maturity = Natural logarithm of years to maturity;

Principal Amount = Natural logarithm of the amount to be repaid at maturity.

Table 1, Panel B provides summary statistics for these variables. To mitigate the influence

of outliers, I winsorize 1% of observations at the tails at the population level. The mean value of

the overall restrictiveness index (OVERALL) indicates that, on average, firms include 10 out of

41 restrictions into their contracts, with a standard deviation of 4.6, which represents a substantial

cross-sectional variation. The maximum number of covenant restrictions used by a single firm in

the sample is 27, while the minimum is 0. The average company has $4.9 billion in total assets,

leverage of 33% and book-to-market of 0.45. Correlation statistics are provided in Table 2. The

restrictiveness indices INVEST, DISTR, FINAN, and CONTROL (which stand for investing,

distributions, financing and control related covenant restrictions respectively) exhibit high inter-

correlations, ranging from 0.42 to 0.73. The correlations of these indices with the overall

restrictiveness index (OVERALL) are as high as 0.93. The evidence also suggests that

correlations between debt contract restrictiveness and other control variables selected for the

analysis are substantial. For example, the correlation between the overall index and leverage or

size are 0.38 and -0.20, respectively.

17 Calculated as 3.3*data170+data12+1.4*data36+1.2*(data4-data5) divided by lagged assets data6.

19

V. Results

First, the results from estimating Equation (1) are presented and briefly discussed. I

proceed with the analysis of asymmetric timeliness over the different levels of contract

restrictiveness using return-earnings (equation 2). Finally, I present the analysis based on accrual

cash-flow relation – an alternative measure of asymmetric timeliness adopted (Ball and

Shivakumar 2005).

Covenant Equation

Table 3 presents the estimation results of the first stage equation (equation 1) for each of

the five restriction indices used as dependent variable. The explanatory power of the model is

32% for the overall restrictions index and ranges between 19% and 62%, for financing

restrictions and transfer-of-control covenants respectively, indicating that there is substantial

amount of unexplained variation in covenant use that can be exploited in the subsequent analysis.

The coefficient on size indicates that in all models covenant restrictions are used

significantly less in larger firms. ROA is positively associated with the use of covenants. While

dividend yield is not related to the overall contract restrictiveness, examination of its components

reveals that dividend paying firms try to avoid restrictions on distributions and investment

activities and the use of other control related covenants, while at the same time they constrain

financing activities more frequently. Leverage is positively related to covenant use. The

coefficient on book-to-market is significantly positive across the different types of restrictions.

Firms in a fast changing environment, as measured by the standard deviation of returns, avoid

financing restrictions, while they include covenants on distributions, investments as well as other

covenants.18 The evidence further suggests that growth in assets, the frequency of past losses and

Z-score do not appear to influence the use of covenants significantly in the presence of other

18 This parallels the evidence in Breadley and Roberts (2005), who show that high-growth firms restrict the use of funds, while avoiding restrictions on future financing. Firms with more volatile income also appear to avoid the use of covenants, in their study.

20

controls. Finally, the coefficient on maturity is significantly negative, while the principal amount

is significantly positively related to the inclusion of covenant restrictions.

The findings are broadly consistent with firms trading off the costs and benefits of

covenant restrictions. More frequent reliance on covenants in small firms (Malitz, 1986, Begley,

1994) and in firms with high leverage (Begley, 1994, Nash et al, 2003) is consistent with the

presence of a higher degree of agency problems. Similarly, dividend paying firms impose

restrictions on investment and financing activities, but at the same time seem to find it costly to

constraint their dividend policy. In line with Levine and Hughes (2005) and Chava et al. (2005),

who argue that “good” firms may signal their type via covenant inclusion, while firms with

higher default risk will find this costly, we observe that firms with stronger profitability have

more restrictive debt contracts. The findings also suggest that firms with unexercised growth

options will prefer to retain managerial flexibility and thus will avoid the use of covenants

(Begley, 1994, Nash et al., 2003, Khan and Yermack, 1998). Volatile firms value flexibility and

avoid the inclusion of financing restrictions, while at the same time these firms are perceived as

risky and thus include other types of restrictions into their lending contracts. Finally, contract

design choices, such as the size of issue, seem to substitute for covenants and thus appear to be

useful as well to control for possible conflicts of interest.

Main results

Table 4 presents the parameters of Equation (2) estimated across ten restrictiveness

deciles. Deciles are based on the residual from equation (1) and are labeled Q1 (lowest

restrictiveness) through Q10 (highest restrictiveness). Panel A of the table and Figure 1 report the

evidence based on the overall contract restrictions index, while the evidence in Panels B through

E is based on separate analyses of investment, distribution, financing, and control transfer

covenants, respectively. For brevity, β1 and R2 statistics are reported in Panel A only.

From Panel A it follows that across all four considered periods around the debt issue, also

referred to as time horizons (shown in the second column of the table), the coefficient β1 is both

21

statistically and economically higher for the tenth decile (Q10) of overall restrictiveness index as

compared to the first decile (Q1). More specifically for companies in Q1 the estimates of β1 are

0.28 for years –3 to –1 horizon, 0.30 for year –1, 0.25 for years +1, and 0.34 for years +1 to +3

horizon, while their Q10 counterparts are 0.53, 0.66, 0.53 and 0.50 respectively. The same

pattern arises when examining each of the other restrictiveness indices, as described next.

The analysis of the investment restrictions (Panel B) indicates that the differences

calculated by subtracting the estimate of β1 for companies in Q10 from β1 estimated for those in

Q1 are 0.12, 0.29, 0.27 and 0.22 for years –3 to –1, year –1, years +1, and years +1 to +3

horizons, respectively. Similarly, the differences based on the index of financing constraints

(Panel D) are 0.26, 0.35, 0.58, and 0.39, respectively for years –3 to –1, year –1, years +1, and

years +1 to +3 horizons. All these differences are statistically significant at conventional levels.

The evidence based on the distribution restrictions index (Panel C) suggests that differences in

the Q10 and Q1 estimates of β1 are positive but small in magnitude and statistically insignificant.

This result is driven by the very large coefficients in Q1, which drop considerable in magnitude in

the second decile of distribution restrictions Q2. The results suggest an increasing pattern in β1 as

we move towards the higher distribution restrictiveness deciles. Finally, the analysis of transfer

of control covenants index (Panel E) reveals that over years –3 to –1 and year –1 horizons, Q10-

Q1 differences in β1 are positive and statistically significant (0.22 and 0.30 respectively). These

differences become statistically indistinguishable from zero for years +1 and years +1 to +3,

respectively, but they still remain positive.

Overall, the evidence in Table 4 supports the hypothesis that firms which rely more on

covenants exhibit increased levels of timely loss recognition. In addition, an examination of the

coefficients of determination (R2) reveals patterns which are generally increasing. This too is

further consistent with the hypothesis that the use of covenants is positively related to timeliness

of negative economic news in accounting income.

22

The focus thus far has been on the highest and the lowest deciles of debt contract

restrictiveness. Note that patterns in β1 estimates not always appear to be linear. The patterns in

the coefficients for the lower part of the restrictions’ distribution are rather flat and in some cases

are even somewhat downward sloping, which are subsequently followed by substantial increases

as we move towards the highest contract restrictiveness decile.19 Such increase is consistent with

the notion of complementarity, e.g., as the timely loss recognition increases the marginal gain in

contracting efficiency due to covenants also goes up and hence both mechanisms reinforce each

other. On the other hand, a slightly downward sloping relation in the lowest restriction portfolios

is consistent with covenants being a substitute for the asymmetric timeliness at the other

extreme.20 Figure 1 documents these patterns and also reveals that the estimated parameters from

Equation 2 are quite volatile.21 To assess formally whether there is a significant trend in the

asymmetric timeliness estimates across the restrictiveness deciles Q1-Q10, I perform additional

correlation and regression analyses as discussed next.

I use decile rank (d) ranging from 1 for the lowest restrictiveness decile (Q1) to 10 for the

highest decile of contract restrictiveness (Q10), and based on ten observations (for each

combination of time horizon and type of restriction index in Table 4) and I calculate both Pearson

and Spearman correlations between d and β1, d and β0+β1, as well as between d and R2, and I also

run the following regression

υφφβ +⋅+= ddr 10

ˆ

19 A related issue is whether there is a certain threshold beyond which further increases in conditional conservatism become prohibitively costly as they would provoke too frequent covenant violations. However, the findings suggest that even if such threshold exists, in practice it is not binding. 20 This substitution, however, is not expected to be due to debt contracting, as argued in the previous sections, but may be due to the usefulness of conservatism in curbing agency problems unrelated to debt contracts (e.g., compensation contracts), which in turn may diminish the overall level of the agency problems and thus lead to a lesser use of covenants. 21 This need not be surprising as the number of observations in the analysis is rather limited, while the estimates in the piecewise linear regression of returns on earnings are substantially influenced by the extremes of the distribution.

23

where d ∈ {1,…,10} is the restrictiveness decile rank and drβ ∈{β0, β1, β0+β1, R

2} measures the

degree of timely loss recognition. Coefficient 1φ indicates an average increase in the timeliness

of loss recognition that occurs moving to the next contract restrictiveness decile.

Table 5 presents the results of this analysis. Panel A reports correlations based on the

overall restrictiveness index. The results reveal that decile rank (d) strongly correlates with β1

(and β0+β1) estimates, with Pearson correlation coefficients 0.60, 0.38, 0.47, and 0.54 (0.63, 0.34,

0.72, and 0.71) for years –3 to –1, year –1, years +1, and years +1 to +3 horizons, respectively.

The evidence based on Spearman correlations yields very similar results. As for the regression

analysis, the estimates of 1φ for the overall restrictiveness index are similar across all time

horizons and imply that as contract restrictiveness increases by 10% (one decile), on average the

magnitude of β1 (β0+β1) coefficient increases by approximately 0.02 (0.015 to 0.03 depending on

the horizon). The magnitude of this effect is economically substantial and, while based on only

ten observations, the estimates in Panel A are mostly statistically significant.

I repeat the analysis for separate components that comprise the overall restrictiveness

index and find similar results. For brevity, Panel B reports Pearson correlations only. The

evidence on Investing restrictions indicates that the asymmetric timeliness estimates correlate

positively with decile rank d. The correlation coefficients depend on the time horizon (years –3

to –1, year –1, years +1 and years +1 to +3) and are as high as 0.51, 0.51, 0.57, and 0.77 for β1

and 0.58, 0.36, 0.75, and 0.78 for the β0+β1. Analysis of restriction on distributions reveals

correlation coefficients of 0.47, 0.58, 0.62, and 0.61 for β1, and 0.50, 0.52, 0.73, and 0.68 for

β0+β1, respectively, for years –3 to –1, year –1, years +1 and years +1 to +3. The evidence based

on financing restrictiveness yields similar findings: depending on the horizon, the correlation

coefficients are 0.46, 0.50, 0.41, and 0.48 for β1 and 0.50, 0.35, 0.59, and 0.67 for β0+β1. The

magnitudes of 1φ coefficient (not tabulated) are similar across the indices of investing,

24

distribution and financing restrictiveness and range from 0.015 to 0.038 across the 3 panels.

Finally, as in Table 4, the results based on the index of control transfer covenants indicate a

positive correlation between decile rank d and asymmetric timeliness estimates only for years –3

to –1 and year –1.

The combined evidence in Tables 4 and 5 suggests that we cannot reject the positive

association between covenants use and asymmetric timeliness. This evidence is consistent with

firms limiting their discretion ex ante, i.e. in the years leading up to a debt issue.22

Accrual-Based Measure of Asymmetric Timeliness

Following Ball and Shivakumar (2005), I also use an alternative measure of asymmetric

timeliness based on the relation between cash flows and accruals. Accruals are used to

incorporate revisions of the present value of future cash flows into the income statement.

Therefore, in addition to reducing the noise in cash flows from operations (Dechow, 1994),

accruals also help to recognize economic gains and losses in a timely manner. And since the

demand for economic loss recognition is higher, in general, accruals will be used to recognize

economic losses in a timely manner, while economic gains are more likely to be accounted for on

a cash basis, i.e., when realized. As a result, a positive but asymmetric correlation arises between

cash flows and accruals. The following piecewise-linear regression is estimated to measure the

degree of asymmetric timeliness of loss recognition, 1β :

)3(,)0CFO(CFOCFO)0CFO(ACC 1010 tttttt DD εββαα +<×++<+=

Following Ball and Shivakumar (2005b) and Collins and Hribar (2002), I use cash flow

statement information available in COMPUSTAT for the period of 1987-2004. CFOt is year t

cash flow from operations (COMPUSTAT item data308). Accruals (ACCt) are measured by

subtracting CFOt from cash flow statement earnings (data123). Both variables are scaled by

22 It may happen that this effect is in part due to prior debt issues, as contacts usually do not differ substantially within the same firm. A direct examination of this is difficult due to truncation and limited coverage of FISD before the 90s, while some maturities exceed 50 years.

25

lagged total assets. 0)(CFOt <D is a dummy variable taking the value of unity when CFOt is

negative and zero otherwise. The coefficient 1β is expected to be positive as accruals are used to

recognize economic losses in a more timely fashion (Ball and Shivakumar, 2005a, 2005b). 23

Equation (3) is estimated over ten restrictiveness deciles (Q1-Q10) constructed based on

the five indices of contract restrictiveness. The results are reported in Table 6 and also depicted

in Figure 2. The tenor of the results is the same as of those reported in Table 4. Analysis of the

overall restrictiveness index reveals that in the highest restrictiveness decile Q10 the magnitudes

of 1β coefficient are 1.00, 1.34, 0.77, and 1.09 for years –3 to –1, year –1, year +1 and years +1 to

+3 respectively, while their lowest decile (Q1) counterparts are 0.45, 0.34, 0.42, and 0.67. The

differences between these estimates are both statistically and economically significant.

The differences in asymmetric timeliness estimates between Q10 and Q1 are especially

pronounced for investing restrictions, which are significant at the 1% level and attain values of up

to 1.40. The evidence based on the indices of distribution and financing restrictions, as well as on

the index of control transfer covenants indicates that in years prior to a debt issue there is a

substantial and statistically significant increase in asymmetric timeliness from decile Q1 to Q10.24

In the post issue periods (year +1 and year +1 to year +3), 1β increases in, on average, smaller

increments moving from Q1 to Q10.

Finally, Table 7 presents correlations of decile ranks with the estimated parameters.

Correlations between the decile rank (d) and 1β estimate based on the overall restrictions (Panel

A) are 0.81 for years –3 to –1 and year –1 and are statistically significant at 1%. However, the

correlation coefficients decline to 0.46 and 0.16 for years +1 and +1 to +3 respectively and are no

longer statistically significant (based on 10 observations). This pattern is similar when different

23 Coefficient 0β is predicted to be negative; it indicates the noise reducing role of accruals (Dechow et al.,

1998) 24 One exception is payout restriction in year –1, where the difference between Q10 and Q1 is negative. However, examination of Q9 and Q8 strongly indicates an increasing pattern, so the result is likely to be driven by noise in the estimation.

26

types of restrictions are examined separately (Panel B). Overall, the correlation coefficients

computed for the horizons prior to the debt issue (years –3 to –1 and year –1) are substantial, at

about 0.50 on average. In years following the issue, the trends in the estimates of β1 or β0+β1

become weaker, consistent with management introducing noise into accounting formation.

VI. Discussion and Limitations

I argue that the use of restrictive covenants in public debt contracts is associated with

higher demands for the timely recognition of economic losses in accounting income. The purpose

of covenants is to transfer control rights from shareholders to bondholders when a firm

approaches financial distress, thus limiting the ability of a manager to take actions leading to

bondholder wealth expropriation. As most covenants either explicitly or implicitly depend on

accounting information and become binding when accounting performance deteriorates, I argue

that covenants will be more effective in constraining opportunistic behavior by the manager (and

in preventing the associated agency problems) when firm’s accounting system is designed in a

way that it produces timely signals of the firm’s economic health. Therefore, when contracts rely

on covenants as a contractual mechanism to protect bondholders, higher demands with respect to

timely loss recognition arise.

The analysis shows that companies with the most restrictive contracts are about two times

as timely in recognizing economic losses into earnings as companies with lower degree of

restrictions. The results are robust to measures of asymmetric timeliness derived from return-

earnings and accrual-cash-flows relations. Overall, I conclude that the demands of contracting

parties, at the firm level, are predictably associated the degree of timeliness in recognition of

economic losses.

The study is subject to several potential limitations. First, no attempt is made to

determine causality in the relation between timely loss recognition and debt contract

restrictiveness. It is not established in the study whether the timely loss recognition affects debt

27

contract design or whether it is the contractual features, such as covenants that induce companies

to record timely losses. Nevertheless in under both scenarios the lenders are concerned with

timely loss recognition and thus reporting properties relate to contract design. This study

documents that the choice of the degree of contract restrictiveness and the choice of the degree of

asymmetric timeliness go together, i.e., are complementary.

Second, I argue that a company that relies on covenants will adopt more timely loss

recognition ex ante. Such adoption can take place prior to the year in which the debt contract is

closed.25 Alternatively, debt contracts themselves may require accounting information to satisfy

certain conditions and may specify adjustments to accounting information (Core and Verrecchia

2006). Leftwich (1983) finds that debt contracts adjust accounting information to be more

conservative. Still, the information in this case is usually backed out from GAAP numbers.

Since it is likely to be very costly to specify a comprehensive set of accounting adjustments in a

contract (Holthausen and Leftwich, 1983), self-imposed constraints in the form of conditional

conservatism should prove useful. This is consistent with the evidence in Beatty, Weber and Yu

(2006) who find that the demands for conservatism are not entirely met via conservative

adjustments to accounting information.

The empirical evidence presented above suggests that companies adjust their accounting

in the years preceding the debt issue. This is consistent with Ball and Shivakumar (2006b) who

argue that initial public offerings experience much of scrutiny by investors and thus adopt

timelier reporting. What’s more, in the years following the debt issue, I find weaker evidence of

timely loss recognition when I rely on accrual based regressions. A possible explanation for this

is that firms exert discretion over the accounting as well as real activities to reduce the likelihood

of covenant violations. This can introduce additional noise in accruals and more importantly into

the cash flows which results in errors-in-variables problem. The latter makes it more difficult to

25 Observing income numbers from past years before the issue, can give creditors some idea of how timely the firm’s accounting system is.

28

find significant changes in asymmetric timeliness when moving from one level of covenant

restriction to the next.

Finally, the restrictiveness indices I use in the analysis count the number of covenants

included in debt contracts, which need not imply more restrictive contracts per se. The contracts

may provide a slack in meeting covenant thresholds. Sridhar and Magee (1997) show that

accounting information quality and tightness of covenants are substitutes. However the analysis

there applies to the tightness of covenants and not to their inclusion, whereas, consistent with the

arguments in Levine and Hughes, 2005 and others, I argue that covenant restrictions and timely

loss recognition are complementary. If the restrictiveness indices correlate positively with the

tightness of covenants, this may bias my results. However, the bias should generally work

against finding the hypothesized relation.

29

References

Ahmed, A., Billings, B., Morton, R., Harris, M., 2003, The role of accounting conservatism in

mitigating bondholder-shareholder conflicts over dividend policy and in reducing debt

costs, The Accounting Review 77, 867-890.

Altman, E., 1968, Financial Ratios, discriminant analysis and the prediction of corporate

bankruptcy, Journal of Finance 23, 589-609.

Ball, R., Kothari, S., Robin, A., 2000, The effect of international institutional factors on

properties of accounting earnings, Journal of Accounting & Economics 29, 1-51.

Ball, R., Shivakumar, L., 2005, Earnings quality in UK private firms: Comparative loss

recognition timeliness, Journal of Accounting & Economics 39, 83-128.

Ball, R., Shivakumar, L., 2006a, The role of accruals in asymmetrically timely gain and loss

recognition, Journal of Accounting Research 44, 207-242.

Ball, R., Shivakumar, L., 2006b, Earnings quality at initial public offerings, working paper,

University of Chicago.

Ball, R., Robin, A., Sadka, G., 2005, Is accounting conservatism due to debt or share

markets? A test of “Contracting” versus “Value Relevance” theories in accounting,

working paper, University of Chicago.

Basu, S., 1997, The conservatism principle and asymmetric timeliness of earnings, Journal of

Accounting & Economics 24, 3-37.

Bharath, S., Sunder, J., Sunder, S., 2006, Accounting quality and debt contracting, working

paper, University of Michigan.

Beatty, A., Ramesh, K., Weber, J., 2002, The importance of accounting changes in debt

contracts: The cost of flexibility in covenant calculations, Journal of Accounting &

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.

30

Beatty, A., Weber, J., 2002, Performance pricing in debt contracts, working paper,

Massachusetts Institute of Technology.

Beatty, A., Weber, J., 2006, Conservatism and debt, working paper, Ohio State University.

Begley, J., 1994, Restrictive covenants included in public debt agreements: An empirical

investigation, working Paper, University of British Columbia.

Begley, J., Chamberlain, S., 2005, The use of debt covenants in public debt: The role of

accounting quality and reputation, working paper, University of British Columbia.

Begley, J., Feltham, G., 1999, An empirical examination of the relation between debt

contracts and management incentives, Journal of Accounting & Economics 27, 229-

259.

Begley, J., Freeman, R., 2004, The changing role of accounting numbers in public lending

agreements, Accounting Horizons 18, 81-96.

Billet, M., King, T., Mauer, D., 2005, Growth opportunities and the choice of leverage, debt

maturity, and covenants, working paper, University of Iowa.

Bradley, M. Roberts, M., 2004, The structure and pricing of corporate debt covenants,

working paper, Duke University.

Bodie, Z., Taggert, R., 1978, Future Investment and the value of the call provision on a bond,

Journal of Finance 33, 1187–1200.

Bushman, R., Piotroski, J., 2006, Financial reporting incentives for conservative accounting:

the influence of legal and political institutions. Journal of Accounting & Economics 42,

107-148

Chava, S., Kumar, P., Warga, A., 2004, Agency costs and the pricing of bond covenants,

working Paper, University of Huston.

DeAngelo, H., DeAngelo, L., Skinner, D., 1994, Accounting choices in troubled companies,

Journal of Accounting & Economics 17, 113-143.

DeAngelo, H., DeAngely, L., Wruck, K., 2002, Asset liquidity, debt covenants, and

managerial discretion in financial distress: the collapse of L.A. gear, Journal of

Financial Economics 64, 3-34.

31

Dechow, P., 1994, Accounting earnings and cash flows as measures of firm performance: The

role of accounting accruals, Journal of Accounting & Economics 18, 3-42.

DeFond, M., Jiambalvo, J., 1994, Debt covenant violation and manipulation of accruals,


Dichev, I., Skinner, D., 2002, Large-sample evidence on the debt covenant hypothesis,


Garleanu, N., Zwiebel, J., 2005, Design and renegotiation of debt covenants, working paper,

University of Pennsylvania.

Goyal, V., 2005, Market discipline of bank risk: Evidence from subordinated debt contracts,

Journal of Financial Intermediation, 14, 318-350.

Guay, W., Verrecchia, R., 2006, Discussion of an economic framework for conservative

accounting and Bushman and Piortoski (2006), Journal of Accounting & Economics 42,

149-165.

Harris, M., and Raviv, A., 1990, Capital structure and the informational role of debt, Journal

of Finance 45, 321-349.

Healy, P., Palepu, K., 1990, Effectiveness of accounting-based dividend covenants, Journal of


Holthausen, R., Leftwich, R., 1983, The economic consequences of accounting choice:

implications of costly contracting and monitoring, Journal of Accounting & Economics

39, 295-327.

Holthausen, R., Watts, R., 2001, The relevance of the value-relevance literature for financial

accounting standard setting, Journal of Accounting & Economics 31, 3-75.

Fields, T., Lys, T., Vincent, L., 2001, Empirical research on accounting choice, Journal of


Garleanu, N., Zwiebel, J., 2005. Design and renegotiation of debt covenants. University of

Pennsylvania and Stanford University working paper.

Hribar, P., Collins, D., 2002, Errors in estimating accruals: Implications for empirical

research, Journal of Accounting Research 40, 105-134.

32

Jensen, M., Meckling, W., 1976, Theory of the firm: Managerial behavior, agency costs and

ownership structure, Journal of Financial Economics 3, 305-360.

Jensen, M., 1986, Agency costs of free cash flow, corporate finance, and the market for

takeovers, American Economic Review 76, 323-329.

Kahan, M. Yermack, D., 1998, Investment opportunities and the design of debt securities,

Journal of Law, Economics, and Organization 14, 136-151.

Leftwich, R., 1983, Accounting information in private markets: Evidence from private

lending agreements, The Accounting Review 58, 23-42.

Levine, C., Hughes, J., 2005, Management compensation and earnings-based covenants as

signaling devices in credit markets, Journal of Corporate Finance 11, 832-850.

Malitz, I., 1986, On financial contracting: The determinants of bond covenants, Financial

Management 15, 18-25.

Myers, S., 1977, Determinants of corporate borrowing, Journal of Financial Economics 5,

147-175.

Nash, R., Netter, J., Poulsen, A., 2003, Determinants of contractual relations between

shareholders and bondholders; investments opportunities and restrictive covenants,

Journal of Corporate Finance 9, 201-232.

Press, E., Weinthrop, J., 1990, Accounting-based covenants in public and private debt

agreements, Journal of Accounting & Economics 12, 65-95.

Reisel, N., 2004. On the value of restrictive covenants: An empirical investigation of public

bond issues. Working paper, Rutgers University.

Ryan, S., 2006, Identifying conditional conservatism, European Accounting Review,

forthcoming.

Roychowdhury, S., Watts, R., 2006, Asymmetric timeliness of earnings as a measure of

conservatism, working paper, Massachusetts Institute of Technology and the University

of Rochester.

Sloan, R., 2001, Financial accounting and corporate governance: A discussion, Journal of


33

Smith, C., Smithson, C., Wilford, D., 1989, Managing financial risk, Journal of Applied

Corporate Finance 2, 27-48.

Smith, C., Warner, J., 1979, On financial contracting: An analysis of bond covenants, Journal

of Financial Economics 7, 117-161.

Sridhar, S., Magee, R., 1997, Financial contracts, opportunism, and disclosure management,

Review of Accounting Studies 1, 225–258.

Sweeney, A., 1994, Debt-covenant violations and managers' accounting responses, Journal of


Watts, R., 2003a, Conservatism in accounting part 1: Explanations and implications,

Accounting Horizons 17, 207-221.

Watts, R., 2003b, Conservatism in accounting Part II: Evidence and research opportunities,

Accounting Horizons 4, 287-301.

Watts, R., Zimmerman, J., 1986, Positive Accounting Theory, Englewood Cliffs, N.J.:

Prentice-Hall.

Watts, R., Zimmerman, J., 1990, Positive accounting theory: a ten year perspective, The

Accounting Review 65, 131-156.

Wittenberg-Moerman, R., 2005, The role of information asymmetry and financial reporting

quality in debt contracting: Evidence from the secondary loan market, working paper,

University of Chicago.

Zhang, J., 2005, Efficiency gains from accounting conservatism: Benefits to lenders and

borrowers, working paper, Massachusetts Institute of Technology.

Zwiebel, J., 1996, Dynamic capital structure and managerial entrenchment, American

Economic Review 86, 1198-1215.

34

Appendix A: Examples of timely accounting policies:

First example concerns accounting treatments of maintenance and betterments of

long-lived assets. Firms often have a choice either to capitalize large maintenance expenses

necessitated by deterioration in asset’s condition (which would result in higher income) or to

expense them. Since the discretion over this choice may be used to avoid recognition of

economic losses, a commitment to treat the costs associated with betterments whenever, e.g.,

the betterment does not involve a purchase of new equipment, as maintenance expenses,

rather than capitalize them, would be a way to become more timely in loss recognition.

More generally, firms may become more timely in loss recognition by choosing

accounting policies that slow down the recognition of gains while speeding up the recognition

of losses. As Basu (1997) points out, firms may revise the estimates of asset’s useful life

upwards, which would have a positive impact on income over multiple years. Since these

revisions are subjective, creditors would prefer the company to commit against following

such a practice.

Commitment accelerated depreciation in the industries where fixed assets loose most

of their economic value in the beginning of their useful life, particularly when asset value

depends on whether the firm subsequently exercises its growth options (i.e. whenever useful

life may be much longer then effective economic life of an asset) is another example of timely

recognition of economic losses. As the amount of firm’s revenue the asset contributes

decreases quickly (after the firm begins exploitation) stock price is generally going to reflect

this and therefore the firm will be more timely by shifting future depreciation expenses

forward, i.e., by adopting accelerated depreciation.

Definitions of materiality thresholds adopted by a firm is another area that can have a

lot of influence on timely loss recognition. Materiality implies that transactions that

immaterial amounts/transactions need not be accounted for by using principles of financial

accounting. This means that managers have a fair amount of discretion in accounting for

immaterial transactions and implies that investors will often miss important information

because companies deem it ‘immaterial’. Abusing materiality may help a company to avoid

violation of covenants. A commitment to decrease materiality thresholds with respect to

accounting for losses and adverse economic events (e.g. treat virtually any loss as material

unless it becomes prohibitively costly to do so) would be another way to commit to timelier

loss recognition.

In a similar vein, in the context of accounting for pensions, SFAS 87, which is

designed to smooth volatility of income due to fluctuations of the value of the fund, leaves

space for improvements in timely loss recognition. Realized returns on a pension fund’s

assets always differ from the expected return used to calculate the pension expense. This,

together with a number of other factors, such as changes in actuarial assumptions with respect

35

to life expectancy, turnover, salary growth, results in the need for deferrals, i.e. the presence

of unrecognized gains or losses. Unless deferred gains or losses exceed (a substantial)

materiality threshold26 firms are not required to start recognizing (amortizing) the gain/loss.

A firm may adopt a policy to expense a loss (or start its amortization) whenever the deferred

gain evolves into a deferred loss (i.e. has been absorbed). While this may be too extreme,

nothing prevents a firm from coming up with a policy to recognize deferred losses resulting

from pension accounting sooner then deferred gains, for example, using faster depreciation

rates for losses that for gains.

Additionally managers often have a choice over parameters used to calculate pension

expense. Comprix and Muller (2006) provide evidence consistent with managers

manipulating expected rates of returns to increase their compensation. In similar fashion

assumptions about expected returns can be used to avoid covenant violations. A firm can

become more timely in recognition of economic losses by ex ante specifying the rules for

determination of the expected return on fund’s assets. For instance some firms use moving

average over actual realizations, which increase firm’s expense when the fund is performing

poorly. Such policy may be considered as more timely, in a cross-section of firms.

Accounting for marketable securities also leaves room for timeliness of losses

recognition. Firms have discretion with respect to (re)classifying marketable securities as

trading or available-for-sale. Unrealized losses for the latter do not enter the income

statement as presumably the company has no intentions of selling them. These need not be

true intentions, however, and may be just a way to increase reported income and thus avoid

covenant violations. Adoption of a policy not to reclassify available-for-sale securities as

trading (i.e., recognize all losses for marketable securities in the income statement) whenever

this results in increased profit due to the recognition of unrealized gain; or alternatively, a

commitment not to classify trading securities as available-for-sale in case this leads to

recognition of an unrealized loss would represent a way to be more timely in loss recognition.

Additionally, companies have discretion in whether to treat unrealized losses in value of

available-for-sale securities as permanent or temporary. The former necessitates expensing,

which creditors would generally prefer. A commitment to always treat unrealized losses as

permanent would be another way to adopt timelier loss recognition.

Contingent liabilities represent yet another way of having more timely loss

recognition. Consider a situation when a court jury finds a company liable for punitive

damages of $7 million (with subsequent settlement for $5 million) and the company files an

immediate appeal, while recognizing zero amount of liability on the balance sheet or income

statement (instead the firm makes only a footnote disclosure). The lenders are likely to prefer 26 10 percent of the greater of the projected benefit obligation or the market-related value of plan assets.

36

the alternative treatment of recognizing an immediate loss of 7 million and wait for a

subsequent decision of the court. It is often the case that while exact the amount of the

liability cannot be determined, the maximum amount is well-defined (e.g. fines for

environmental pollution) in which case adopting a policy to recognize the maximum amount

would clearly represent a more timely accounting for losses.

Finally, accounting for changes in replacement costs of obsolete inventory often

requires estimating its net realizable value, which in turn leads to (subjective) assessments of

future demands, market conditions, etc. An optimistic assessment will prevent a timely

writedowns of inventory (to the replacement costs). Having a policy, for example, to assume

that future demands for obsolete inventory are zero would avoid problems associated with

discretionary assessments of future demands and would recognize economic losses in a more

timely fashion. Last but not least, the choice of LIFO over FIFO to value inventory is more

timely in the context of increasing inventory costs.

37

Appendix B: Index Construction

Investment Restrictions Covenant Index includes indicators for the following covenants:

1. A consolidation or merger of the issuer with another entity is restricted

(consolidation_merger).

2. Restricts issuer’s investment policy to prevent risky investments (investments).

3. Restrictions on the ability of an issuer to sell assets or restrictions on the issuer’s use

of the proceeds from the sale of assets (sale_assets).

4. Restricts subsidiaries’ investments (su_investments_unrestricted_subs).

5. Issuer is restricted in certain business dealings with its subsidiaries

(transaction_affiliates)

6. Issuer must use proceeds from sale of subsidiaries’ assets to reduce debt

(su_sale_xfer_assets_unrestricted).

Distributions Restrictions Covenant Index includes indicators for the following

covenants:

1. Payments made to shareholders or other entities may be limited to a certain

percentage of net income or some other ratio (dividends_related_payments).

2. Restricts issuer’s freedom to make payments (other than dividend related payments)

to shareholders and others (restricted_payments).

3. Limits the subsidiaries’ payment of dividends to a certain percentage of net income or

some other ratio (su_dividends_related_payments).

Financing Restrictions Covenant Index includes indicators for the following covenants:

1. Restricts issuer from issuing additional funded debt (funded_debt).

2. To issue additional debt the issuer must have achieved or maintained certain

profitability levels (net_earnings_test_issuance).

3. Restricts user from incurring additional debt with limits on absolute dollar amount of

debt outstanding or percentage total capital (indebtedness).

4. Restricts issuer to the type or amount of property used in a sale leaseback transaction

and may restrict its use of the proceeds of the sale (sales_leaseback).

5. Restricts issuer to the amount of senior debt it may issue in the future

(senior_debt_issuance).

6. Restricts issuer from issuing additional common stock (stock_issuance_issuer).

7. The issuer cannot issue secured debt unless it secures the current issue on pari passu

basis (negative_pledge_covenant).

38

8. Restricts the issuer from transferring, selling, or disposing of it’s own common stock

or the common stock of a subsidiary (stock_transfer_sale_disp).

9. Restricts issuance of junior or subordinated debt (subordinated_debt_issuance).

10. Restricts the total indebtedness of the subsidiaries (su_indebtedness).

11. Indicates subsidiaries are restricted from borrowing, except from the parent

(su_borrowing_restricted).

12. Restricts issuer’s subsidiaries from issuing additional funded debt (su_funded_debt).

13. Restricts issuer from issuing additional common stock in restricted subsidiaries

(su_stock_issuance).

14. Restricts subsidiaries’ ability to issue preferred stock (su_preferred_stock_issuance).

15. Subsidiary is restricted from issuing guarantees for the payment of interest and/or

principal of certain debt obligations (su_subsidiary_guarantee).

16. Restricts subsidiaries from selling then leasing back assets that provide security for

the debt holder (su_sales_leaseback).

17. If issuer’s net worth (as defined) falls below minimum level, certain bond provisions

are triggered (declining_net_worth).

18. Issuer must maintain a minimum specified net worth (maintenance_net_worth).

19. Issuer is required to have a ratio of earnings available for fixed charges, of at least a

minimum specified level (fixed_charge_coverage).

20. Restricts total indebtedness of the issuer (leverage_test).

21. Limits subsidiaries’ leverage (su_leverage_test).

22. In the case of default, the bondholders have the legal right to sell mortgaged property

to satisfy their unpaid obligations.(liens).

23. Restricts subsidiaries from acquiring liens on their property (su_liens).

24. Subsidiaries are required to maintain a minimum ratio of net income to fixed charges

(su_fixed_charge_coverage).

Control Covenant Index includes indicators for the following covenants:

1. A bondholder protective covenant that will activate an event of default in their issue,

if an event of default has occurred under any other debt of the company

(cross_default).

2. A bondholder protective covenant that allows the holder to accelerate their debt, if

any other debt of the organization has been accelerated due to an event of default

(cross_acceleration).

3. Upon a change of control in the issuer, bondholders have the option of selling the

issue back to the issuer (change_control_put_provisions).

39

4. The issue’s change of control provisions are triggered if an investor controls more

than this percentage of the issuer’s stock (voting_power_percentage).

5. The issue’s change of control provisions are triggered if the issuer’s employee

retirement plan controls more than this percentage of the issuer’s stock

(voting_power_percentage_erp).

6. A decline in the credit rating of the issuer (or issue) triggers a bondholder put

provision (rating_decline_trigger_put).

7. Property acquired after the sale of current debt issues will be included in the current

issuer’s mortgage (after_acquired_property_clause).

8. Indicates if restricted subsidiaries may be reclassified as an unrestricted subsidiaries

(su_subsidiary_redesignation).

40

T a b l e 1 Sample and Descriptive Statistics

Note: Both Compustat and FISD variables are winsorised at 1% the population level

Panel A: Sample construction # Obs. Issues by Industrial Companies with available Covenant data before merging to Compustat 11947

Number of Firms before merging to Compustat 4394

Number of Issues that Merge to Compustat 7310

Number of Issuers that Merge to Compustat 2367

Retaining only one Issue per year results in following # of issues 5036

Requiring non-missing control variables 3382

Panel B: Descriptive Statistics

Variable # Obs. Mean St. Dev. Min Max

OVERAL 4999 8.370 4.565 0 26

INVEST 4999 2.236 0.941 0 6

DISTR 4999 0.736 0.944 0 3

FINAN 4999 3.026 2.502 0 15

CONTROL 4999 2.372 1.108 0 5

Log of Assets (lassets) 4734 7.331 1.670 -0.805 10.41

Assets 4734 4895 7784 0.447 33207

ROA 4564 -0.003 0.274 -2.896 0.488

Dividend yield (divyld) 4386 0.011 0.017 0 0.091

Leverage (lev) 4726 0.333 0.224 0 1.001

Book-to-Market (btm) 4124 0.455 0.659 -4.17 4.584

Growth in Assets 4281 1.436 1.103 0.297 10.80

Std.Dev. of Returns (stdret) 4067 0.029 0.015 0.007 0.143 Number of losses over the last 5 years (nloss) 4281 0.941 1.369 0 5

Altman’s Z-score 4176 1.672 2.421 -46.28 9.08

Std. Dev. of IBEI (stdibei) 4018 0.078 0.239 0 3.698

Log of Maturity (lmatur) 4987 2.274 0.589 0 4

Principal 4999 6.851 0.438 3.219 6.908

41

T a b l e 2 Correlation Statistics

Table reports Pearson correlation statistics computed for the variables used in equation 1. See section 3 of the paper for more detail on variables definitions

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17)

(1) OVERAL 1.00

(2) INVEST 0.77 1.00

(3) DISTR 0.85 0.66 1.00

(4) FINAN 0.93 0.59 0.73 1.00

(5) CONTROL 0.65 0.42 0.43 0.43 1.00

(6) Log(Assets) -0.20 -0.19 -0.37 -0.08 -0.14 1.00

(7) ROA -0.07 -0.07 -0.08 -0.01 -0.11 0.18 1.00

(8) Dividend Yield -0.20 -0.22 -0.24 -0.06 -0.33 0.33 0.13 1.00

(9) Leverage 0.38 0.31 0.43 0.31 0.24 -0.16 -0.12 -0.13 1.00

(10) Book-to-Market 0.01 -0.03 0.03 0.04 -0.06 0.05 0.07 0.13 -0.19 1.00

(11) Growth in Assets 0.05 0.03 0.07 -0.01 0.11 -0.18 -0.30 -0.18 0.07 -0.04 1.00

(12) St.Dev. (Returns) 0.17 0.18 0.23 0.03 0.28 -0.38 -0.38 -0.37 0.14 -0.12 0.31 1.00

(13) Number of losses in 5 preceding years

0.17 0.19 0.21 0.07 0.20 -0.23 -0.28 -0.30 0.32 -0.15 0.14 0.42 1.00

(14) Z-score -0.09 -0.09 -0.10 -0.04 -0.12 0.07 0.57 0.10 -0.26 0.10 -0.19 -0.30 -0.42 1.00

(15) St.Dev.(IBEI) 0.00 0.03 0.02 -0.05 0.07 -0.21 -0.28 -0.14 0.05 -0.08 0.65 0.27 0.29 -0.28 1.00

(16) Log(Maturity) -0.20 -0.13 -0.15 -0.15 -0.23 0.02 0.11 0.13 -0.08 0.02 -0.06 -0.24 -0.11 0.10 -0.03 1.00

(17) Log(Principal) 0.06 0.04 0.06 0.07 0.01 -0.06 0.00 -0.02 0.00 -0.01 0.01 -0.01 -0.01 0.04 0.01 -0.05 1.00

42

T a b l e 3 Determinants of Covenant Restrictiveness

Covenant data is taken from Fixed Income Securities Database (which contains more than 40 covenant indicators). 5 different contract restrictiveness indexes are constructed: Overall covenant index, investment restrictions covenant index, distributions restrictions covenant index, financing restriction covenant index and control covenant index. Each index is constructed by summing up corresponding covenant indicators for a particular contract. The following multiple regression is estimated:

(1), DummiesYear

Amount Principal*Maturity*score-Z*Losses ofNumber *

urns)St.Dev(Ret*Growth*tBook/Marke*Leverage*

Yield Dividend*ROA*)log(Assets* Index enessRestrictiv

111098

7654

3210

ξαααα

αααααααα

++++++

+++++++=

Dependent Variable

Independent Variable Statistic

Overall Restrictions

Investment Restrictions

Distribution Restrictions

Financing Restrictions

Control Covenants

Log of Assets Estimate -0.350*** -0.061*** -0.133*** -0.054* -0.101*** t-value (-7.34) (-5.95) (-13.38) (-1.88) (-11.05) ROA Estimate 1.062** 0.171* 0.260*** 0.566** 0.065 t-value (2.40) (1.79) (2.81) (2.11) (0.76) Dividend Yield Estimate 1.186 -3.272*** -2.163** 10.894*** -4.274*** t-value (0.26) (-3.27) (-2.23) (3.87) (-4.81) Leverage Estimate 6.444*** 0.934*** 1.512*** 3.282*** 0.716*** t-value (19.17) (12.88) (21.48) (16.08) (11.13) Book-to-Market Estimate 0.730*** 0.093*** 0.187*** 0.347*** 0.104*** t-value (7.02) (4.14) (8.59) (5.49) (5.20) Growth in Assets Estimate -0.012 -0.029 0.019 -0.025 0.022 t-value (-0.14) (-1.54) (1.07) (-0.47) (1.32) St. Dev. Of Returns Estimate 10.121* 2.679* 5.396*** -0.521** 2.567* t-value (1.77) (2.18) (4.52) (-0.15) (2.35) # of Losses in 5 prior years Estimate 0.058 0.022* 0.018 -0.003 0.020* t-value (0.06) (0.02) (0.02) (0.00) (0.02) Olson’s Z-score Estimate 0.030 -0.004 -0.002 0.017 0.019** t-value (0.74) (-0.44) (-0.21) (0.69) (2.39) St. Dev. Of IBEI Estimate -0.944*** -0.035 -0.268*** -0.432** -0.210*** t-value (-2.87) (-0.49) (-3.89) (-2.16) (-3.32) Log of Maturity Estimate -0.490*** -0.027 -0.082*** -0.345*** -0.036* t-value (-4.71) (-1.21) (-3.76) (-5.46) (-1.81) Log of Principal Amount Estimate 0.607*** 0.078** 0.056* 0.360*** 0.114*** t-value (3.99) (2.37) (1.76) (3.89) (3.90) Adjusted R-squared 31.6% 19.9% 27.6% 19.2% 62.1% Number of Observations 3382 3382 3382 3382 3382

43

T a b l e 4 Asymmetric Timeliness of Earnings: Stock Returns Evidence

Estimates from piecewise-linear regression of earnings on returns conditional on the sign of “economic news” are estimated across 10 deciles of contract restrictiveness indices Q1-Q10. 5 types of covenant restriction indices are considered: Overall restrictions, Investment restrictions, Distribution restrictions, Financing restrictions and Other Control covenants. All of them are measured by taking a residual from model (1) reported in table 2. Initially each index is constructed by summing up covenant indicators present in a particular contract. Earnings are measured by income before extraordinary items (data18) scaled by year -1 market value (Compustat items data199*data25); returns are compounded over 12 months starting 3 months after the fiscal year end are adjusted by subtracting return on value weighted market index compounded over the same period. 0.5% of scaled earnings and returns are excluded from both tails to mitigate the influence of outliers. ***, **, * indicate the significance at 1, 5 and 10 % levels respectively. The model is:

)2(,)0Ret(*Ret*Ret*)0Ret(*P/E tt1t0t101 ttt DD εββαα +<++<+=−

Years relative to issue Coef. Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10 Q10-Q1

Panel A: Overall Restrictions

-3 to -1

�

0 -0.044 -0.041 -0.026 -0.004 -0.013 -0.041 -0.022 -0.031 -0.014 -0.054 -0.010

�

1 0.279 0.217 0.197 0.252 0.194 0.242 0.236 0.221 0.358 0.525 0.245*** R2 0.06 0.06 0.06 0.09 0.05 0.05 0.06 0.04 0.07 0.10

-1

�

0 -0.016 -0.004 -0.017 -0.050 0.002 -0.056 -0.046 -0.040 0.025 -0.109 -0.093***

�

1 0.295 0.215 0.257 0.295 0.033 0.156 0.281 0.196 0.254 0.661 0.366*** R2 0.06 0.08 0.09 0.10 0.01 0.05 0.06 0.02 0.06 0.13

+1

�

0 -0.095 -0.040 -0.183 -0.108 0.042 -0.070 0.019 -0.023 -0.059 -0.005 0.090

�

1 0.248 0.296 0.398 0.322 0.091 0.299 0.219 0.257 0.602 0.525 0.277** R2 0.04 0.12 0.23 0.11 0.03 0.14 0.10 0.06 0.22 0.20

+1 to +3

�

0 -0.095 -0.076 -0.130 -0.091 -0.037 -0.020 -0.015 -0.020 -0.105 -0.048 0.048

�

1 0.339 0.364 0.318 0.390 0.246 0.232 0.260 0.520 0.621 0.501 0.162*

R2 0.06 0.11 0.09 0.11 0.07 0.08 0.08 0.12 0.20 0.11

Panel B: Investment Restrictions

-3 to -1

�

1 0.319 0.277 0.196 0.144 0.184 0.242 0.254 0.374 0.297 0.444 0.124*

-1

�

1 0.275 0.279 0.175 0.116 0.044 0.261 0.218 0.322 0.276 0.567 0.292*** +1

�

1 0.374 0.252 0.136 0.490 0.131 0.311 0.362 0.536 0.381 0.645 0.271*

+1 to +3

�

1 0.356 0.300 0.263 0.358 0.284 0.265 0.405 0.553 0.519 0.580 0.224**

Panel C: Distributions Restrictions -3 to -1

�

1 0.368 0.171 0.141 0.142 0.108 0.169 0.388 0.367 0.395 0.320 -0.049

-1

�

1 0.316 0.128 0.113 0.144 0.017 0.100 0.410 0.295 0.407 0.444 0.128

+1

�

1 0.374 0.090 0.331 0.187 0.358 0.427 0.280 0.487 0.448 0.439 0.065 +1 to +3

�

1 0.430 0.239 0.260 0.228 0.306 0.292 0.545 0.558 0.520 0.429 -0.001

44

table 3 continued


Panel D: Financing Restrictions

-3 to -1

�

1 0.299 0.204 0.208 0.273 0.259 0.248 0.141 0.241 0.305 0.556 0.257***

-1

�

1 0.299 0.237 0.154 0.226 0.218 0.120 0.295 0.262 0.251 0.644 0.345*** +1

�

1 0.224 0.153 0.666 0.182 0.221 0.138 0.274 0.239 0.405 0.805 0.581***

+1 to +3

�

1 0.332 0.236 0.564 0.284 0.224 0.239 0.425 0.299 0.487 0.719 0.387***

Panel E: Control Covenants

-3 to -1

�

1 0.230 0.356 0.310 0.300 0.269 0.333 0.216 0.133 0.314 0.450 0.219*** -1

�

1 0.188 0.265 0.384 0.283 0.169 0.285 0.275 0.114 0.416 0.483 0.295***

+1

�

1 0.306 0.438 0.379 0.596 0.230 0.565 0.417 0.185 0.379 0.345 0.039

+1 to +3

�

1 0.466 0.388 0.460 0.465 0.346 0.527 0.407 0.183 0.311 0.439 -0.027

45

T a b l e 5 Trends in Asymmetric Timeliness Estimate from Basu (1997) model estimated across

10 Debt Contract Restrictiveness Deciles. This table analyzes trends in coefficients and R-squareds (reported in Table 3) from the piecewise-linear regression of earnings on returns (allowing for differential effect of positive vs. negative returns on earnings) estimated around the year of debt issue across 10 deciles of each of 5 contract restrictiveness proxies. Namely the table reports Pearson and Spearman correlations between abnormal contract restrictiveness decile ranks and their corresponding estimated coefficients (from Table 3). In addition the table reports the slope from the regression of estimated regression coefficients from Table 3 on their decile ranks. I.e. the following model is estimated


ˆ

where d is decile rank ∈ {1,..,10} and βdr∈{β0, β0, β0+β1, R

2}r is the decile estimate for type of statistic restrictiveness measure r. Significance levels for Pearson and Spearman correlation coefficients are based on 10 observations (i.e. 8 degrees of freedom); for coefficient φ 1 the significance is reported based on Normal distribution as φ 1 is a linear combination of asymptotically normal estimates of βd

r. Refer to table 3 for additional details about the sample.

Panel A: Overall Restrictiveness Horizon Statistic ρPearson

P-value ρSpearman P-value φ1 t-stat P-value

-3 to t-1 E0 -0.009 (0.981) -0.006 (0.987) 0.000 -0.02 (0.981) E1 0.604 (0.064) 0.418 (0.229) 0.020 2.14 (0.032) E0+E1 0.639 (0.047) 0.527 (0.117) 0.020 2.35 (0.019) R2 0.265 (0.458) 0.139 (0.701) 0.002 0.78 (0.436) -1 E0 -0.391 (0.264) -0.297 (0.405) -0.005 -1.20 (0.229) E1 0.382 (0.276) -0.006 (0.987) 0.020 1.17 (0.242) E0+E1 0.340 (0.337) 0.200 (0.580) 0.015 1.02 (0.307) R2 0.032 (0.931) 0.018 (0.960) 0.000 0.09 (0.929) +1 E0 0.485 (0.156) 0.503 (0.138) 0.011 1.57 (0.117) E1 0.471 (0.169) 0.358 (0.310) 0.023 1.51 (0.131) E0+E1 0.723 (0.018) 0.624 (0.054) 0.034 2.96 (0.003) R2 0.359 (0.308) 0.261 (0.467) 0.009 1.09 (0.276) +3 to +1 E0 0.458 (0.183) 0.394 (0.260) 0.006 1.46 (0.145) E1 0.544 (0.104) 0.406 (0.244) 0.023 1.83 (0.067) E0+E1 0.711 (0.021) 0.600 (0.067) 0.030 2.86 (0.004) R2 0.531 (0.114) 0.539 (0.108) 0.007 1.77 (0.076) Panel B: Correlations for Separate Restriction Types � Investment Restrict. Distribution Restrict. Financing Restrict. Control Transfer Horizon Statistic ρPearson

P-value ρPearson P-value ρPearson

P-value ρPearson P-value

-3 to -1 E � � 0.006 (0.99) -0.014 (0.97) -0.066 (0.86) 0.637 (0.05) E � � 0.513 (0.13) 0.47 (0.17) 0.461 (0.18) 0.13 (0.72) E � �E � � 0.582 (0.08) 0.496 (0.15) 0.497 (0.14) 0.406 (0.24)

-1 E � � -0.715 (0.02) -0.464 (0.18) -0.566 (0.09) 0.08 (0.83) E � � 0.509 (0.13) 0.581 (0.08) 0.495 (0.15) 0.391 (0.26) E � �E � � 0.362 (0.30) 0.52 (0.12) 0.351 (0.32) 0.43 (0.22)

+1 E � � 0.242 (0.50) 0.408 (0.24) 0.198 (0.58) 0.46 (0.18) E � � 0.571 (0.09) 0.617 (0.06) 0.406 (0.24) -0.175 (0.63) E � �E � � 0.757 (0.01) 0.734 (0.02) 0.59 (0.07) 0.208 (0.56)

+3 to +1 E � � 0.154 (0.67) 0.339 (0.34) 0.203 (0.58) 0.613 (0.06) E � � 0.774 (0.01) 0.614 (0.06) 0.484 (0.16) -0.403 (0.25) E � �E � � 0.782 (0.01) 0.677 (0.03) 0.667 (0.04) 0.003 (0.99)

46

T a b l e 6

Asymmetric Timeliness and Smoothing Effect of Earnings: Accruals Evidence Estimates from piecewise linear regression of accruals on cash-flows conditional on “cash-flow news” are estimated across 10 deciles of contract restrictiveness indices Q1-Q10. 5 types of covenant restriction indices are considered: Overall restrictions, Investment restrictions, Distribution restrictions, Financing restrictions and Other Control covenants. All of them are measured by taking a residual from model (1) reported in table 2. Initially each index is constructed by summing up a certain type of covenant indicators present in a particular contract. Accruals are measured by subtracting cash flow from operations (data308) minus cash flow statement earnings (data123) both scaled by lagged total assets. 0.5% of scaled accruals and cash flows are excluded to mitigate the influence of outliers. ***, **, * indicate the significance at 1, 5 and 10 % levels respectively. The model is:

)3(,)0CFO(*Assets/CFO*Assets/CFO*)0CFO(*Assets/Accruals 1110101 ttttttttt DD εββαα +<++<+= −−−


Panel A: Overall Restrictions

-3 to -1

�

0 -0.285 -0.431 -0.288 -0.504 -0.527 -0.432 -0.527 -0.486 -0.588 -0.657 -0.372***

�

1 0.451 0.316 0.586 0.808 0.864 1.071 0.816 0.847 0.923 1.004 0.552*** R2 0.05 0.09 0.11 0.22 0.15 0.19 0.18 0.14 0.23 0.14

-1

�

0 -0.288 -0.365 -0.412 -0.556 -0.506 -0.46 -0.532 -0.609 -0.638 -0.866 -0.578***

�

1 0.343 0.35 0.977 1.216 0.735 1.264 1.737 1.167 1.777 1.346 1.002*** R2 0.05 0.03 0.28 0.29 0.07 0.25 0.13 0.28 0.49 0.26

+1

�

0 -0.452 -0.307 -0.376 -0.394 -0.797 -0.417 -0.362 -0.473 -0.522 -0.45 0.002

�

1 0.425 0.493 0.2 -0.13 0.528 0.291 0.982 0.491 0.466 0.772 0.348 R2 0.06 0.02 0.17 0.09 0.29 0.26 0.09 0.07 0.12 0.09

+1 to +3

�

0 -0.484 -0.342 -0.424 -0.418 -0.587 -0.446 -0.303 -0.386 -0.510 -0.481 0.004

�

1 0.674 0.575 0.260 0.071 1.123 0.266 0.358 0.407 0.363 1.087 0.413**

R2 0.08 0.05 0.18 0.08 0.32 0.22 0.08 0.06 0.14 0.12

Panel B: Investment Restrictions

-3 to -1

�

1 0.508 0.537 0.654 0.798 0.941 0.067 0.663 0.515 0.826 1.303 0.795***

-1

�

1 0.419 0.253 0.823 1.229 1.172 -1.018 1.124 0.804 1.083 1.817 1.398*** +1

�

1 0.315 -0.18 0.573 0.348 0.125 0.569 0.335 -0.291 0.546 0.974 0.660***

+1 to +3

�

1 0.624 -0.17 0.797 0.495 0.765 0.392 0.428 -0.188 0.477 1.111 0.487***

Panel C: Distributions Restrictions -3 to -1

�

1 0.558 0.516 0.562 0.761 0.736 0.326 1.299 0.7 0.829 1.153 0.596***

-1

�

1 0.664 0.545 1.171 0.993 1.276 0.083 0.799 1.379 1.578 0.002 -0.662***

+1

�

1 0.382 0.57 0.583 0.2 0.23 -1.826 0.631 0.458 1.371 0.493 0.112 +1 to +3

�

1 0.527 0.781 0.431 0.842 0.291 -1.269 0.687 0.557 0.891 0.57 0.043

47

table 5 continued


Panel D: Financing Restrictions

-3 to -1

�

1 0.435 0.59 0.523 0.518 0.935 0.897 0.524 1.365 0.869 0.906 0.471***

-1

�

1 0.583 0.715 0.494 0.639 1.03 1.254 1.495 1.754 1.381 1.05 0.467** +1

�

1 0.482 0.316 0.606 0.386 0.579 0.155 -0.802 0.114 0.799 0.437 -0.045

+1 to +3

�

1 0.745 0.536 0.193 0.951 0.503 0.209 -0.086 0.864 0.509 0.867 0.121

Panel E: Control Covenants

-3 to -1

�

1 0.364 0.543 0.816 0.612 1.02 0.688 0.811 1.023 0.801 0.949 0.585*** -1

�

1 0.403 0.497 0.931 0.741 0.609 1.551 1.206 1.238 1.188 1.469 1.067***

+1

�

1 0.708 0.39 0.417 0.422 0.473 0.468 0.175 0.364 0.405 0.542 -0.166

+1 to +3

�

1 0.494 0.431 0.381 0.857 0.685 0.386 0.836 0.088 0.478 0.453 -0.041

48

T a b l e 7 Trends in Asymmetric Timeliness Estimate from Accrual-Return Model Estimated across

10 Debt Contract Restrictiveness Deciles. This table analyzes trends in coefficients and R-squareds (reported in Table 5) from the piecewise-linear regression of Accruals on Cash flow (allowing for differential effect of positive vs. negative cash flows on accrual component of earnings) estimated around the year of debt issue across 10 deciles of each of 5 contract restrictiveness proxies. Namely the table reports Pearson and Spearman correlations between abnormal contract restrictiveness decile ranks and their corresponding estimated coefficients (from Table 5). In addition the table reports the slope from the regression of estimated regression coefficients from Table 5 on their decile ranks. I.e. the following model is estimated


ˆ

where d is decile rank ∈ {1,..,10} and βdr∈{β0, β0, β0+β1, R

2}r is the decile estimate for type of statistic restrictiveness measure r. Significance levels for Pearson and Spearman correlation coefficients are based on 10 observations (i.e. 8 degrees of freedom); for coefficient φ 1 the significance is reported based on Normal distribution as φ 1 is a linear combination of asymptotically normal estimates of βd

r. Refer to table 3 for additional details about the sample.

Panel A: Overall Contract Restrictions Horizon Statistic ρPearson

P-value ρSpearman P-value φ1 t-stat P-value

t-3 to t-1 E0 -0.836 (0.003) -0.855 (0.002) -0.033 -4.30 (0.000) E1 0.814 (0.004) 0.806 (0.005) 0.065 3.97 (0.000) E0+E1 0.525 (0.119) 0.673 (0.033) 0.032 1.74 (0.081) R2 0.594 (0.070) 0.539 (0.108) 0.011 2.09 (0.037)

t-1 E0 -0.904 (0.000) -0.915 (0.000) -0.048 -5.97 (0.000) E1 0.809 (0.005) 0.830 (0.003) 0.134 3.89 (0.000) E0+E1 0.628 (0.052) 0.503 (0.138) 0.085 2.28 (0.022) R2 0.620 (0.056) 0.552 (0.098) 0.029 2.24 (0.025)

t+1 E0 -0.204 (0.571) -0.370 (0.293) -0.009 -0.59 (0.555) E1 0.455 (0.186) 0.442 (0.200) 0.045 1.45 (0.148) E0+E1 0.344 (0.330) 0.333 (0.347) 0.036 1.04 (0.300) R2 0.096 (0.792) 0.297 (0.405) 0.003 0.27 (0.785)

t+3 to t+1 E0 -0.104 (0.775) -0.103 (0.777) -0.003 -0.30 (0.768) E1 0.155 (0.668) 0.103 (0.777) 0.018 0.44 (0.657) E0+E1 0.146 (0.687) 0.103 (0.777) 0.015 0.42 (0.676) R2 0.042 (0.908) 0.139 (0.701) 0.001 0.12 (0.905) Panel B: Correlations for Separate Restriction Types � Investment Restrict. Distribution Restrict. Financing Restrict. Control Transfer Horizon Statistic ρPearson

P-value ρPearson P-value ρPearson

P-value ρPearson P-value

-3 to -1 E � � -0.668 (0.035) -0.817 (0.004) -0.854 (0.002) -0.418 (0.230) E � � 0.408 (0.242) 0.589 (0.073) 0.659 (0.038) 0.718 (0.019) E � �E � � 0.225 (0.533) 0.356 (0.312) 0.429 (0.216) 0.632 (0.050)

-1 E � � -0.668 (0.035) -0.663 (0.037) -0.770 (0.009) -0.741 (0.014) E � � 0.378 (0.281) 0.019 (0.959) 0.769 (0.009) 0.826 (0.003) E � �E � � 0.262 (0.464) -0.140 (0.700) 0.495 (0.146) 0.728 (0.017)

+1 E � � -0.136 (0.709) -0.055 (0.880) 0.141 (0.697) 0.007 (0.984) E � � 0.347 (0.325) 0.117 (0.747) -0.145 (0.690) -0.325 (0.360) E � �E � � 0.358 (0.310) 0.098 (0.788) -0.115 (0.751) -0.226 (0.531)

+3 to +1 E � � 0.088 (0.810) 0.112 (0.758) 0.147 (0.686) -0.437 (0.206) E � � 0.155 (0.670) -0.007 (0.985) 0.046 (0.900) -0.149 (0.682) E � �E � � 0.189 (0.601) 0.013 (0.971) 0.101 (0.780) -0.286 (0.423)

49

Figure 1: Timely Loss Recognition and Overall Contract Restrictiveness: Stock Returns Evidence

The figure displays β1 coefficients from the piecewise linear regression estimated over 10 deciles Q1-Q10 based on the orthogonalized overall contract restrictiveness index. The evidence is based on table 4, which is to be consulted for additional details.

)2(,)0Ret(*Ret*Ret*)0Ret(*P/E tt1t0t101 ttt DD εββαα +<++<+=− Figure 1a

Period: t-3 to t-1

0.000

0.100

0.200

0.300

0.400

0.500

0.600

1 2 3 4 5 6 7 8 9 10

decile

Figure 1b Period: t-1

0.000

0.100

0.200

0.300

0.400

0.500

0.600

0.700

1 2 3 4 5 6 7 8 9 10

deci le

Figure 1c

Period: t+1 to t+3

0.000

0.100

0.200

0.300

0.400

0.500

0.600

0.700

1 2 3 4 5 6 7 8 9 10

decile

Figure 1d Period: t+1

0.000

0.100

0.200

0.300

0.400

0.500

0.600

0.700

1 2 3 4 5 6 7 8 9 10

deci le

50

Figure 2: Timely Loss Recognition and Overall Contract Restrictiveness: Case of Accruals

The figure displays β1 coefficients from the accrual-cashflow piecewise linear regression estimated over 10 deciles (Q1-Q10) based on the orthogonalized overall contract restrictiveness index. The evidence is based on the coefficients in Table 6, which is to be consulted for additional details.

)2(,)0CFO(*Assets/CFO*Assets/CFO*)0CFO(*Assets/Accruals 1110101 ttttttttt DD εββαα +<++<+= −−−

Figure 2a

Period: t-3 to t-1

0.000

0.200

0.400

0.600

0.800

1.000

1.200

1 2 3 4 5 6 7 8 9 10

decile

Figure 2b Period: t-1

0.000

0.200

0.400

0.600

0.800

1.000

1.200

1.400

1.600

1.800

2.000

1 2 3 4 5 6 7 8 9 10

deci le

Figure 2c

Period: t+1 to t+3

0.000

0.200

0.400

0.600

0.800

1.000

1.200

1 2 3 4 5 6 7 8 9 10

decile

Figure 2d Period: t+1

-0.200

0.000

0.200

0.400

0.600

0.800

1.000

1.200

1 2 3 4 5 6 7 8 9 10

deci le

debt contract restrictiveness and timely loss … · debt contract restrictiveness and timely loss...

Documents