Do rating agencies cater evidence from rating based contracts
-
Upload
mr-hp -
Category
Economy & Finance
-
view
129 -
download
1
Transcript of Do rating agencies cater evidence from rating based contracts
Electronic copy available at: http://ssrn.com/abstract=1726943
Do Rating Agencies Cater? Evidence from Rating-Based
Contracts
Pepa Kraft ∗
New York University
Stern School of Business
August 15, 2014
Abstract
I examine whether rating agencies cater to borrowers with rating-based performance-priced loan contracts (PPrating firms). I use data from Moody’s Financial Metrics on itsquantitative adjustments for off-balance-sheet debt and qualitative adjustments for soft fac-tors. In the cross-section and for borrowers experiencing adverse economic shocks, I find thatthese adjustments are more favorable for PPrating firms than for other firms, consistent withrating agencies catering to the PPrating borrowers. I find that this catering is muted in twocircumstances when rating agencies’ reputational costs are higher than usual: (1) near theinvestment grade and prime short-term rating thresholds and (2) when Fitch Ratings alsoprovides a rating.
Keywords: Rating agency, off-balance-sheet finance, soft information, debt contracting
∗I am very grateful to the members of my dissertation committee: Ray Ball (chair), Phil Berger, Doug Diamond,Christian Leuz, and Doug Skinner, as well as Ryan Ball, Mary Barth, Utpal Bhattacharya, Alexander Bleck, Mar-shall Blume, Fabrizio Ferri, Joseph Gerakos, SP Kothari, Mathias Kronlund, Valeri Nikolaev, Maarten Petermann,Joshua Ronen, Stephen Ryan, Regina Wittenberg, Joanna Wu, Sarah Zechman, and Jerry Zimmerman. I alsothank participants at the NBER summer session on credit rating agencies, Standard & Poor’s Academic CouncilMeeting, Notre Dame Conference on Current Topics in Financial Regulation, UNC/Duke Fall Camp, Quantita-tive Management Associates research meeting as well as participants at accounting workshops at Boston College,Columbia University, London Business School, McGill University, New York University, Northwestern University,University of Chicago, University of Michigan, University of Pennsylvania, University of Rochester, University ofToronto, Stanford, and Washington University for constructive suggestions, questions, and feedback. I thank An-drew Tan and Hui Lin Tan for excellent research assistance. I am grateful for the financial support provided by theNYU Stern School of Business, the University of Chicago Booth School of Business, and the Deloitte Foundation.This paper was previously circulated under the title: The Impact of the Contractual Use of Ratings on the RatingProcess - Evidence from Rating Agency Adjustments. To contact me, email [email protected].
1
Electronic copy available at: http://ssrn.com/abstract=1726943
1 Introduction
Private loan agreements increasingly include performance-pricing provisions that raise loan inter-
est rates or trigger early payment of principal when the borrowers’ public credit ratings decrease
(Beatty and Weber 2003; Asquith et al. 2005), yielding direct and immediate adverse effects on
the borrowers’ cash flows (Nicholls 2005). Rating agencies say that they are concerned about the
potential adverse consequences of this contractual use of credit ratings for borrowers’ creditwor-
thiness (Moody’s 2001; Standard & Poor’s 2008). I test the ‘catering hypothesis’ that this concern
causes rating agencies to cater to borrowers with these loans by providing credit ratings that are
more favorable than the borrowers’ credit risk justifies.1 I further test whether reputational costs
for rating agencies limit this rating inflation. Reputational costs might lead rating agencies to treat
rating-based performance-pricing provisions in loan contracts as risk factors, due to the adverse
effects on borrowers’ cash flows when their credit ratings deteriorate.
To identify catering, I examine rating agencies’ hard and soft adjustments, which capture dif-
ferent dimensions of borrowers’ credit risk. Hard adjustments capture credit risk arising from
quantifiable factors such as off-balance-sheet debt (Moody’s 2006; Moody’s 2007; Kraft 2014).
Soft adjustments capture credit risk arising from qualitative factors such as management cred-
ibility. I infer catering when these adjustments are more favorable for borrowers with ratings-
performance pricing (PPrating firms) than for borrowers with accounting-ratio based performance
pricing (PPratio firms), all else being equal. In particular, because PPratio firms tend to be riskier
than PPrating firms, I replicate all primary analyses partitioning the sample into groups of firms
with homogeneous credit risk, and find the results are robust to this partition.
Using a sample of U.S.-domiciled, non-financial firms with information available on Moody’s
Financial Metrics and Dealscan for 2002 through 2008, I find that rating agency adjustments are
more favorable for PPrating firms than for PPratio firms, consistent with the catering hypothesis.
For example, the average adjustment for off-balance-sheet debt equals 14% of total assets for
1Borrowers may try to influence their credit ratings for reasons other than existing performance-priced loans,such as achieving better valuations or gaining access to more liquid markets. These considerations are outside thescope of this paper.
2
PPrating firms versus 21% for PPratio firms. Similarly, the average soft adjustment for PPrating
firms is only a fifth of the soft adjustment for PPratio firms. Multivariate analysis confirms that the
use of credit ratings rather than accounting ratios in performance pricing is associated with more
favorable/less unfavorable estimates of off-balance-sheet debt and soft adjustments. I further find
that PPrating firms that experience adverse economic shocks receive significantly less unfavorable
rating agency adjustments than do PPratio firms receiving such shocks, again consistent with the
catering hypothesis.
I find evidence that catering is muted in two cases where rating agencies likely bear heightened
reputational costs from catering. First, I find no evidence that rating agencies cater to PPrating
firms with ratings close to the critical investment-grade and prime short-term thresholds that act
as gateways to lower priced and more liquid debt markets. Second, I find that rating agency
adjustments are less favorable for PPrating firms with Fitch ratings, which unlike Moody’s and
Standard & Poor’s ratings are not incorporated in PPrating contracts.
This paper contributes to several literatures. First, a sizeable literature examines whether
rating agencies’ business model of collecting fees from the issuers they rate creates a conflict of
interest that leads to upwardly biased ratings in general (Partnoy 1999; Beaver et al. 2006; Mason
and Rosner 2007; Cheng and Neamtiu 2009; Becker and Milbourn 2011; Bolton et al. 2012) and
in particular for structured finance products (Mason and Rosner 2007; Benmelech and Dlugosz
2009). In the latter case, a debate exists as to whether rating inflation is due to catering or
underestimation of the credit risk of these non-traditional products (Coval et al. 2009; Ashcraft
et al. 2010; He et al. 2011; Griffin and Tang 2012). This research has not empirically investigated
the effect of debt contracts features, such as performance pricing, on catering, although Nicholls
(2005) and Manso (2013) describes the feedback loop that result from the use of credit ratings
in contracts. This study thus provides the first empirical evidence on the effect of the use of
credit ratings in debt contracts on rating agencies’ incentives in the rating process. My results
are broadly consistent with the extensively studied debt covenant hypothesis, in that borrowers
attempt to influence measures specified in debt contracts to achieve better outcomes under those
3
contracts (Watts and Zimmerman 1986; Beatty and Weber 2003; Dichev and Skinner 2002).
The study also contributes to the literature on the use of hard and soft information in con-
tracting (Stein 2002; Petersen 2004; Rajan et al. 2010). Hard information is reliable and would
evoke a consensus when presented to different parties; in contrast, soft information is generally
not verifiable for contracting purposes (Rajan and Reichelstein 2009). Consistent with Petersen
(2004)’s conjecture, I show in this paper and in Kraft (2014) that while credit ratings primarily
reflect quantitative information, they also reflect qualitative factors. These findings are broadly
complementary to Ashbaugh-Skaife et al. (2006)’s finding that borrowers with better corporate
governance receive more favorable credit ratings.
2 Hypothesis development
Ratings are benchmarks of issuers’ credit worthiness. A large proportion of private debt contracts
includes provisions that are based on issuers’ public ratings, such as rating triggers and perfor-
mance pricing. These provisions render debt contracts sensitive to rating changes. A rating trigger
is a provision in a loan agreement that initiates a specific action in the event of a rating change.
A rating downgrade might set off accelerated debt repayment or posting of collateral (SEC 2003;
Nicholls 2005).2 For a recent prominent example, the downgrade of AIG triggered some of its coun-
terparties to demand additional collateral or principal repayments.3 More generally, rating-based
performance pricing refers to rating-sensitive debt obligations whose interest payments depend
on the borrower’s public ratings. Rating-based performance pricing provisions increase contrac-
tual interest rates when borrowers’ ratings get downgraded and decrease contractual interest rates
when borrowers’ ratings get upgraded. Furthermore, parties to over-the-counter financial transac-
tions explicitly or implicitly restrict themselves to dealing with counterparties with ratings above
2Nicholls (2005) lists default and acceleration triggers in loan agreements, pricing grids, security/collateralenhancement triggers, benchmark for triggering restrictive negative covenants, calculation of borrowing base andspringing liens, and qualification of permitted assignees as rating triggers.
3See “AIG needs to address CDS portfolio to save ratings” by Reuters on February 27, 2009 and “AIG facescash crisis as stock dives 61%” by The Wall Street Journal on September 16, 2009, as well as “Downgrades andDownfall. How could a single unit of AIG cause the giant company’s near-ruin and become a fulcrum of the globalfinancial crisis?” by Washington Post staff writers Robert O’Harrow and Brady Dennis on December 31, 2009.
4
minimum levels (Moody’s 2001).
When contracts use credit ratings to enforce restrictions, changes in ratings directly and im-
mediately impact firms’ cash flows. This motivates the issuer to ask for favorable treatment by
the rating agency. Under their business model, rating agencies collect fees from the very issuers
they rate, which creates a basic conflict between providing accurate ratings and upwardly biased
ratings (Partnoy 1999; Mason and Rosner 2007; Becker and Milbourn 2011, Bolton et al. 2012).
Such a business model leads to conflicts of interest similar to the tradeoff facing other information
intermediaries which receive income from their objects of investigation, such as audit firms and
investment bank affiliated equity analysts.4 The reasoning for catering mirrors that of the debt
covenant hypothesis: Watts and Zimmerman (1986) argue that debt contracts that make covenant
thresholds a function of financial ratios give borrowers incentives to change accounting methods to
avoid costly covenant violations. Accounting ratio-based performance pricing in loan agreements
creates a continuous link between accounting ratios and interest rates, and thus performance pric-
ing creates incentives for managers to engage in income-increasing earnings management. Beatty
and Weber (2003) find that borrowers whose debt contracts allow them to make accounting changes
choose accounting methods that increase earnings. Dichev and Skinner (2002) find that borrowers’
accounting ratios are substantially more likely to be just above critical covenant thresholds rather
than below, which is consistent with the debt covenant hypothesis. Similarly, rating-based per-
formance pricing creates incentives for borrowers to implore rating agencies to cater to borrower
demands.
Rating agencies are not immune to catering by providing inflated ratings in other contexts
(Benmelech and Dlugosz 2009).5 Credit rating agencies rely on issuers for fees both at the time of
4For audit firms a large literature examines the question of auditor independence (Antle 1984; Larcker andRichardson 2004). Analysts’ economic incentives are associated with earnings adjustments, growth forecasts andrecommendations (Lin and McNichols 1998; Baik et al. 2009; Ertimur et al. 2011).
5Ashcraft et al. (2010) find that although ratings of mortgage backed securities contain useful information,ratings exhibit time-variation in their risk adjustments consistent with rating inflation in 2005-2007 and for high-risk and low-documentation loans. Coval et al. (2009) point out that ratings of CDOs are highly unreliable due tomodels that are highly sensitive to even small errors in economic projections or losses and that underestimate thecorrelation of risks across various debt securities. Griffin and Tang (2012) find evidence of upward bias in subjectiveadjustments on AAA-rated CDO tranches relative to their own model.
5
issuance and through periodic monitoring fees for as long as the issue is outstanding. In addition,
rating agencies offer related consulting services, such as pre-rating assessments (White 2002; Bolton
et al. 2012). Rating agencies may provide unduly favorable ratings, especially to issuers who
generate substantial revenues (He et al. 2011). However, it is outside the scope of this study to
disentangle the explicit mechanisms of catering. Indeed, the literature on ratings of structured
finance products has not resolved the debate whether rating inflation is due to explicit catering
for business reasons or whether credit risk is underestimated because of implicit catering due to
erroneous judgments (Coval et al. 2009; Ashcraft et al. 2010; He et al. 2011; Griffin and Tang
2012).6 In either scenario, catering – a rating process that is too favorable given the underlying
economics – would be observed.
Borrowers face changes in contractual interest rates under performance pricing that are signif-
icantly greater than the fees they pay to rating agencies. Asquith et al. (2005) find an average
increase in the contractual interest rate of 13.8 basis points for each step in the pricing grid, with an
average of 5.1 steps for interest rate increasing performance pricing loans. In contrast, borrowers
pay fees of three to four basis points of the face amount for rating agencies to rate borrowers’ cor-
porate debt.7 The potential for large interest rate movements provides borrowers with incentives
to try to influence the rating agency.
Lenders need to monitor their borrowers in order to uncover any potential catering in the rating
process. However, monitoring is not without cost. Ex ante, borrower and lender agree to contract
6There are several possible non-mutually exclusive mechanisms for catering. Under explicit catering for businessreasons, the borrower pays an inflated fee to the rating agency. The ongoing business relationship between theborrower and the rating agency results in one-way fee income paid from borrower to agency, for rating as wellas advisory business. The rating agency provides a more favorable assessment for those borrowers from which itreceives higher fees, holding reputational costs constant. Under implicit catering due to erroneous judgments, theborrower, possibly with the help of a rating advisory consultant, provides optimistic disclosures to the rating analyst,which the rating analyst processes ‘at face value’. This results in catering as well: the rating analyst, by providingtoo little effort, awards upwardly biased ratings to borrowers that provide such upwardly biased disclosures. Suchan outcome may be sustainable because certified rating agencies enjoy market power due to an SEC-granted quasi-monopoly, in which the rating analyst exerts minimal effort in information collection, and the borrower bears theonus of disclosure.
7Standard & Poor’s (2009) documents that up to 4.25 basis points are charged for corporate debt, with aminimum fee of USD 70,000. Partnoy (2006) documents fees of 3-4 basis points of the face amount for corporatedebt, which is subject to minimum fee amounts ranging from USD 30,000 to a maximum of USD 300,000. Moreis charged for complex deals (up to 10 basis points). High volume issuers receive discounts. Monitoring fees,cancellation fees, and initial confidential rating fees are in the range of USD 20,000 to USD 50,000.
6
on performance pricing because lenders can readily observe a signal of credit risk in the form of
an accounting ratio or a rating. Ratings are viewed as less manageable than borrower-generated
accounting ratios. Nonetheless, the business relationship between the borrower and the rating
agency gives rise to a conflict of interest. Lender are not part of this relationship and would incur
additional monitoring costs to assess whether ratings are biased or not.
Reputational concerns provide incentives to rating agencies to resist catering to issuer’s de-
mands (Klein and Leffler 1981; Shapiro 1983; Gorton and Winton 2003; Strausz 2005). The
economic role of rating agencies is to provide independent assessments of credit risk because the
delegation of information processing to an intermediary saves on the duplication of such monitoring
costs by dispersed bond holders (Wakeman 1984). Their primary asset is their reputation, which
is the basis for their long-term business prospects. Rating agencies are likely to take extra care in
their assessment due to reputational concerns about long-term business prospects (Klein and Lef-
fler 1981; Shapiro 1983; Gorton and Winton 2003; Strausz 2005) or due to concerns about outside
political intervention (Beaver et al. 2006). Hence I would expect less catering when reputational
costs are high.
Ratings are commonly used in performance pricing provisions in loan contracts. A traditional
loan contract is priced using a fixed interest rate or a fixed spread over a risk-free interest rate,
such as LIBOR or prime. Rating-based performance pricing explicitly links the contractual loan
interest rate to borrower’s current ratings (Asquith et al. 2005). In such rating-sensitive debt
contracts, rating changes lead to immediate changes in the contractual interest rate (for discussions
of performance pricing in debt see Beatty and Weber 2003; Asquith et al. 2005.) Borrowers enter
into such contracts with lenders for several reasons. First, these contracts help mitigate adverse
selection problems (Asquith et al. 2005.) When asymmetric information between borrower and
bank is likely to result in a misclassification of credit risk, performance pricing reduces the adverse
selection problem because the borrower and bank stipulate ex ante that the borrower’s interest rate
decreases when the borrower’s credit risk improves. This reduces re-contracting costs.8 Second,
8In a model by Hermalin and Katz 1991 renegotiation allows the contracting parties to contract over variablesthat would otherwise be non-contractible. Performance pricing reduces the need for such renegotiation.
7
rating-based performance pricing diminishes incentives to engage in claim dilution and moral
hazard problems (Asquith et al. 2005; Bhanot and Mello 2006.) Performance pricing leads to
higher contractual interest rates when the borrower’s credit risk deteriorates. Thus the threat of
ex post settling up diminishes borrowers’ incentives to engage in behavior such as claim dilution
that weakens their creditworthiness.
However, rating-based performance pricing imposes a circularity problem because credit ratings
themselves can affect the credit quality of the borrower (Manso et al. 2010; Manso 2013). A rating
downgrade leads to a higher contractual interest rate. Such interest rate step-ups exacerbate
liquidity strains at the precise moment when an issuer is least able to deal with them (Moody’s
2001). Furthermore, these step-ups can exacerbate a company’s ability to comply with its cash
flow-based covenants, such as covenants based on total debt/cash flow and senior debt/cash flow
ratios. Rating agencies say that they are concerned about the consequences of the contractual
use of ratings for borrowers’ creditworthiness (Standard & Poor’s 2008; Moody’s 2001.) If rating
agencies take into account feedback effects of rating changes, they may be slow to downgrade in
order to avoid the borrower’s inefficient liquidation. In the model in Manso (2013) rating analysts
take into account both the accuracy of the rating and the effects of the rating on the likelihood
of default arising from rating-based performance pricing because they know that rating-based
performance pricing in loans impacts the contractual interest rate, which affects the default risk
of the borrower.
3 Empirical approach
3.1 The rating process: rating agency adjustments
In this study I estimate the association between rating-based performance pricing (PPrating) and
rating agency adjustments (ADJ ). Rating agency adjustments include hard adjustments such as
estimates of off-balance-sheet debt as well as soft adjustments. Hard and soft adjustments cap-
ture quantitative and qualitative factors that impact issuers’ default risk, respectively (Moody’s
8
2006; Kraft 2014). Hard adjustments primarily comprise adjustments to reported financial state-
ments. A prominent hard adjustment is the estimation and inclusion of off-balance-sheet debt
(Kraft 2014). Higher levels of off-balance-sheet debt imply higher levels of credit risk, everything
else equal. Soft adjustments capture the impact on credit risk by factors such as management
quality, aggressive accounting, weak controls, governance risk, industry structure, and managerial
bondholder friendliness (Moody’s 2007). Soft adjustments either increase or decrease the rating
agency’s estimate of credit risk.
See Appendix A for an illustration of the rating process by Moody’s Financial Metrics for
Airgas, Inc. Moody’s rating analysts assign each industry group a rating grid that consists of
mainly quantitative factors. For Airgas, the rating grid captures assessments of its competitive
position, size, stability, profitability, leverage, and financial strength. Airgas’ adjusted financials
indicate that leverage is higher than that inferred from its reported financials. Debt-book capital
and debt/EBITDA ratios calculated using adjusted financials are substantially greater (and thus
warrant lower ratings) than those calculated using reported financials. Similarly, the cash flow
to debt ratios calculated using adjusted financials are substantially lower (and thus also warrant
lower ratings) than those calculated using reported financials. Thus, Airgas’ indicated rating on
the basis of adjusted ratios is one notch lower than the rating that the reported financials imply.
Soft adjustments lower the rating by another two notches. This illustration is typical of Moody’s
adjustments. Kraft (2014) shows that the major hard adjustment includes off-balance-sheet debt,
leading to substantially higher leverage ratios. On average, credit-risk increasing hard and soft
adjustments have an association with lower credit spreads and higher bond yields (Kraft 2014).9
9Examining rating agency adjustments allows me to investigate where, if any, the conflict of interest manifestsitself. Soft adjustments are by construction less verifiable and thus more likely to be biased than quantitativeadjustments, because ex post detection for a single firm case is difficult due to the unverifiability. For example,Rajan et al. (2010) find that as incentives for decision makers to collect value-relevant information diminish, marketparticipants rely increasingly on hard factors rather than value-relevant soft factors in the pricing of securitizedsubprime mortgages, which ultimately leads to an under-prediction of default risk in this scenario. Despite the factthat hard adjustments are less subjective than soft adjustments, even they provide discretion to rating analysts,who have to choose how big a multiplier to use to capitalize operating lease rent expense or whether to classify asecuritization as non-recourse.
9
3.2 Base model
In order to assess the association between rating-based performance pricing (PPrating) and rating
agency adjustments (ADJ ) I estimate the following model.
ADJt,i = α0 + α1PPratingt,i +∑n
βnfirmcharacteristicst,i
+∑l
θlloancharacteristicst,i (1)
Rating agency’s adjustments capture various dimensions of credit risk, but they are also subject
to bias or noise in the rating process. The bias is subject to the rating agency’s and borrower’s
incentives to provide a favorable credit risk assessment. Thus, I control for borrower characteristics
that determine credit risk and that would be reflected in the rating agency’s adjustments, such
as leverage, profitability, size, and short-term liquidity. Factors similar to those determining the
choice of debt, as well as financial reporting benefits, drive the use of off-balance-sheet finance
(Beatty et al. 1995; Mills and Newberry 2005). The proportion of debt in the capital structure
depends on the riskiness of the underlying cash flows and asset tangibility. Empirical studies on
the cross-sectional determinants of leverage find that leverage increases with fixed assets, non-
debt tax shields, growth opportunities, and firm size (Harris and Raviv 1991; Rajan and Zingales
1995). Leverage decreases with volatility, advertising expenditures, research and development
expenditures, bankruptcy probability, profitability, and product uniqueness. I expect the same
determinants to hold for off-balance-sheet debt. I focus on size, profitability, asset tangibility,
market-to-book ratio, and book-leverage. I represent the firm characteristics with size (logarithm
of revenues), leverage (total balance-sheet debt divided by total assets), interest coverage (ratio
of operating profit to interest expense), operating margin (ratio of operating profit to revenues),
return on assets (ratio of operating profit to total assets), tangibility (ratio of inventory and net
property, plant and equipment to total assets), and market-to-book ratio (market value of equity
to book value of shareholders’ equity).10 Loan characteristics include maturity and size of loan.
10Borrowers also choose off-balance-sheet debt to raise external finance because of its financial reporting treatment.
10
I estimate the model using OLS (ordered probit), where the dependent variable is the rating
agency adjustment to debt (soft or total rating agency adjustment). The empirical proxy for
rating-based contracts is the presence of a loan with performance pricing that links the contractual
interest rate to changes in the issuer’s bond rating. The variable of interest is PPrating, which
equals one if, at fiscal year-end, the borrower has at least one active loan facility outstanding that
contains a rating-based performance pricing feature. Rating agency adjustments, firm, and loan
characteristics are measured at fiscal year-end.
3.3 Contracting choice
Borrowers choose to link their interest rates to future observable events, such as changes in ratings
and accounting ratios, to mitigate adverse selection and moral hazard problems (Asquith et al.
2005). Contracts that link either payments or the posting of collateral to a deterioration of credit
risk mitigate the incentives to engage in claim dilution (Bhanot and Mello 2006; Manso et al.
2010). Ex ante, firms that choose to contract on performance pricing are more opaque than firms
raising loans without performance pricing clauses.
Conditional on contracting on performance pricing, firms and their lenders choose between
ratings and accounting ratios. Ratings are a comprehensive measure of default risk, but accounting
ratios can be timelier (Beatty and Weber 2003; Ball et al. 2008; Doyle 2008). The inclusion
of restrictions on managers’ behavior helps mitigate agency conflicts between debtholders and
managers acting on the behalf of equity holders. For example, financing covenants can be written
that restrict the issue of senior debt, the initiation of leases, or the issue of debt-like obligations
to restrict managers’ ability to dilute existing claims (Smith and Warner 1979). Contractual
adjustments incorporating off-balance-sheet debt are difficult to write and enforce. Writing an
The off-balance-sheet treatment results in financial reporting benefits, such as reporting a lower balance sheet-basedleverage ratio to comply with covenants or to appear less risky (Beatty et al. 1995; Engel et al. 1999; Mills andNewberry 2005). The rating agency’s estimate of off-balance-sheet debt is based on the borrower’s disclosures.By definition, the rating agency adjustment for off-balance-sheet debt is the amount of debt as recognized by theagency, and hence the captured amount of off-balance-sheet debt does not confer any financial reporting benefitswith respect to the rating process. Hence no controls are necessary for expected financial reporting benefits in therating process. The amount of off-balance-sheet debt may, however, confer financial reporting benefits for complyingwith debt covenants.
11
explicit contract that contains a negative covenant that prohibits the use off-balance-sheet debt
would lead to loopholes and financial engineering (Jensen and Meckling 1976; Doyle 2008). Leftwich
(1983) finds that according to best practice, lenders should consider the possibility of ‘creative
financial arrangements’ when writing debt contracts. If left unmonitored, the use of off-balance-
sheet financing allows borrowers to dilute claims of existing, on-balance-sheet debtholders, and
it leads to higher economic leverage. However, even the best-practice contracts documented in
Leftwich (1983) contain vague language that is difficult to enforce. Ratings are comprehensive
measures of credit risk that incorporate a range of factors, which renders them useful for inclusion
in contracts (Doyle 2008).
In the main analysis, I estimate the regression conditional on performance pricing, so I compare
firms with rating-based performance pricing to firms with accounting-based performance pricing.
The small minority of firms whose performance pricing is based on both accounting ratios and
ratings (hybrids) is excluded. Firms that use performance pricing based on ratings are more
similar to firms that use performance pricing based on accounting ratios compared to firms that
use neither (Asquith et al. 2005). When debt contracts rely on accounting-based covenants, debt
holders are likely to provide higher incentives for timely loss recognition to the firm’s management
(Ball and Shivakumar 2006). Public bondholders will have a greater demand for timely loss
recognition than banks or other private lenders (Nikolaev 2010). This raises the concern that
differences in timely loss recognition would affect the speed with which accounting ratios reflect
the underlying economics. However, all the firms in my sample have both private debt and public
debt, as firm-years are required to have ratings available from Moody’s and loan data available
from Dealscan. Furthermore, the regressions control for leverage, and in an additional analysis the
sample is partitioned by credit risk. In the interest of external validity, I also conduct the empirical
tests for firms that have private loans without any performance pricing features. Those firms are
likely to be less opaque and thus less suitable as a control group, but their inclusion allows me to
increase the sample size.
Any observed differences in rating agency adjustments can be caused by differences in under-
12
lying firm characteristics. I employ two methods to address endogeneity concerns. First, I test
whether the hypothesized relationship between rating agency adjustments and PPrating holds for
firms matched by credit risk. Riskier firms are more likely to contract on PPratio, whereas firms
that are safer from a credit perspective are more likely to contract on PPrating. Yields spreads on
publicly traded bonds serve as a benchmark of firms’ credit risk. I partition the sample by their
issuers’ yields spreads into high, medium, and low credit risk and run equation [1] within these
partitions. The partition serves as a natural matching mechanism and ensures that firms with
similar credit risk are compared. The cost of this approach is diminished power due to smaller
sample size.
Second, I exploit variation from unexpected adverse economic shocks to investigate whether
there is a bias in rating agency adjustments. An adverse shock to companies’ cash flows, for
example, by a drop in consumer demand, decreases the value of total assets and increases a firm’s
default risk. I investigate the reaction of rating agencies to firms with adverse economic shocks by
testing whether the rating agency’s reactions differ for firms with rating-based contracts compared
with firms without such contracts. Controlling for the size of the shock, I investigate whether any
differential reaction exists for firms with rating-based contracts relative to other firms. Under the
catering hypothesis, I expect more favorable treatment for firms with rating-based contracts that
experience adverse economic shocks, all else being equal. This research design has the advantage
of allowing me to calculate a benchmark of the expected adjustment for firms that experience a
shock to their default risk. I estimate the following regression where Shockt−1 equals an extreme
change in the market value of public debt during a firm’s fiscal year:
ADJt,i = α0 + α1PPratingt,i + α2Shockt−1,i
+α3Shockt−1,i ∗ PPratingt,i +∑n
βnfirmcharacteristicst,i (2)
For each bond, I calculate daily returns during the firm’s fiscal year. Low (negative) bond
returns are a reflection of an adverse economic shock that increases credit risk. I measure the
13
size of the shock by the bottom return decile.11 Shocks are measured as changes in the market
value of public debt rather than accounting cash flows because they are less subject to accounting
discretion and more timely. An adverse economic shock increases default risk and should be
reflected equivalently in the increase in the rating agency’s risk assessment if the rating agency is
neutral for both PPrating and PPratio (or other) firms.
The net increase in default risk, rather than the gross increase in default risk, is relevant to
calculate the benchmark for the expected adjustment. The decrease in the value of public bonds
reflects the market’s anticipation of the size of the shock and how well the firm is expected to
handle it. One might think that more favorable rating agency adjustments for PPrating firms
could also imply that PPrating firms self-select into PPrating contracts because they are better at
dealing with shocks, for whatever reason. In this case, the more favorable adjustments would be
economically justified and not consistent with the catering hypothesis. However, this concern is
not valid because I measure the priced (net) increase in default risk. If the PPrating firm is better
at dealing with the shock, then the net increase in default risk will be less than the gross increase.
Hence the priced shock already includes the market’s assessment of the PPrating firm’s reaction
to the shock.
See Appendix B for a hypothetical example. The net shock is relevant to the comparison. In
the first scenario where the treatment is random, a PPrating firm is compared to PPratio firm 1.
Both should have the same change in the rating agency’s assessment of default risk. In the second
scenario, where PPrating firms choose to contract on ratings rather than accounting ratios, the
priced reaction of the PPrating firm, the net shock, is smaller than the gross shock. Because the
market anticipates that the PPrating firm is better at dealing with adverse shocks, the market
value of public debt decreases by less. Hence PPrating firm is compared to PPratio firm 2. Both
should have the same change in the rating agency’s assessment of default risk. Because the net
shock is measured by the change in the market’s assessment of default risk, any self-selection issue
is already priced, and hence a comparison of the rating agency’s assessments is meaningful.
11The deciles are less likely to be subject to data errors than the minimum daily return during the year.
14
3.4 Institutional constraints on catering
I investigate whether catering is muted when institutional constraints are present. First, I examine
whether rating agency adjustments are less favorable for issuers with ratings close to important
thresholds. Then I examine whether ratings by Fitch act as a constraining force.
A large proportion of bond investors, such as mutual funds, pension funds, and insurance
companies, use ratings by certified rating agencies to comply with rating-contingent regulation
(Coval et al. 2009; Bolton et al. 2012; Opp et al. 2013).12 To comply with such rating-contingent
regulation, investors desire high ratings. The investment grade cutoff and prime short-term ratings
are particularly important thresholds. Prime short-term credit ratings, such as P1, determine
commercial paper issuance. Furthermore, many investors cannot hold non-investment grade bonds
due to restrictions. Hence, issuers desire investment grade ratings to obtain a large investor base
and reduce the liquidity component of their cost of debt. If the regulator is myopic in the short-
term, an equilibrium with inflated ratings is feasible. Even if managers understand that investors
see through inflated ratings, they might still demand these ratings to help bond investors comply
with regulation (Bolton et al. 2012; Opp et al. 2013).
Under the catering hypothesis, I predict a stronger association between rating-based contracts
and the rating agency’s adjustment for firms close to important rating thresholds and for firms
with prime short-term credit ratings, relative to firms close to these thresholds without a rating-
based contract. I conduct a difference-in-difference analysis in order to test whether closeness to
an important rating threshold strengthens the catering incentive. On the other hand, reputational
costs as well as adverse trigger effects from rating downgrades are substantially higher at these
threshold ratings. I estimate the following regression and include an additional indicator variable
12Ratings by certified (or NRSRO) agencies are used by the SEC, federal and state legislation, and other regulatorsin the context of portfolio restrictions and capital adequacy assessments ((SEC 2003; Standard & Poor’s 2006). Forexample, money market funds can only invest in investment grade bonds. State insurance codes rely on NRSROratings to determine appropriate investments for insurers. The Federal Reserve Board and the Federal Home LoanBank System allow their members (the Federal Reserve System and federally charted savings and loans associations,respectively) to invest in investment grade securities only. The Department of Labor requires pension funds to holdcommercial paper rated above A-3. Furthermore, broker-dealers which are subject to the net capital rule useratings by certified agencies in capital adequacy tests, where the percentage reduction from stated values (securitieshaircuts) for the purpose of stock margin requirements and for net capital requirements depend on ratings.
15
that equals one if the issuer has a threshold rating (BBB-, which is the last rating above the
investment grade cutoff) or a short-term rating (P1, a prime short-term credit rating, which is
necessary to access the commercial paper market) and a term that interacts the threshold rating
and the PPrating indicator variable:
ADJt,i = α0 + α1PPratingt,i + α2thresholdt,i + α3thresholdt,i ∗ PPratingt,i
+∑n
βnfirmcharacteristicst,i +∑l
θlloancharacteristicst,i (3)
Second, I expect Moody’s and S&P to cater less if there exists a third credit rating that is
likely to constrain opportunistic behavior. Cantor and Packer (1996) finds that the probability of
obtaining a third rating is not related to uncertainty over firm default probability. A firm’s decision
to obtain a third rating is largely determined by the firm’s age and size. The importance of the
determinants of age and size in the decision to obtain a third rating underscores the importance of
spreading fixed costs. Larger firms can more easily amortize the fixed cost of a Fitch rating, while
older firms are more likely to have additional ratings because of persistence. Uncertainty does not
appear to be a major factor affecting the likelihood of obtaining a third rating. The prevalence of
additional ratings is unrelated to firms’ financial ratios such as leverage and to uncertainty about
firms’ default risk. In the same vein, Cantor and Packer (1997) finds that frequent and large debt
issuers are the most likely to obtain additional ratings. The study does not find evidence that firms
obtain additional ratings to help clear regulatory hurdles or to resolve greater ex ante uncertainty
about default risk. Xia (2014) finds S&P rating quality improves when the issuer receives a rating
by a smaller rating agency.
I estimate the following regression and include an additional indicator variable that equals
one if the issuer has a Fitch rating (FITCH and a term that interacts FITCH and the PPrating
16
indicator variable:
ADJt,i = α0 + α1PPratingt,i + α2FITCHt,i + α3FITCHt,i ∗ PPratingt,i
+∑n
βnfirmcharacteristicst,i +∑l
θlloancharacteristicst,i (4)
4 Data
I collect reported and adjusted financial statements as well as hypothetical ratings from Moody’s
Financial Metrics database for fiscal years ending during the calendar years of 2002 to 2008. For
a random subset of Financial Metrics firms, hypothetical ratings based on reported and adjusted
financial numbers are available, which allows me to compute credit analysts’ soft and total adjust-
ments. Appendix A provides an illustration of the rating process for Airgas, Inc. The final pub-
lished rating (actual rating) is a function of the reported numbers in Airgas’ financial statements,
credit analysts’ adjustments to those reported financials, such as the inclusion of off-balance-sheet
debt, and credit analysts’ qualitative adjustments.
I calculate the total adjustment (TOTAL) as the difference between actual rating and the hy-
pothetical rating implied by reported financials (indicated reported rating). The indicated reported
rating is the output of the credit analyst’s matrix of accounting ratios. These matrices include
accounting ratios such as measures of profitability and leverage and are industry-specific. The
accounting ratios in the credit risk matrix for indicated reported rating are based on the values as
reported by Airgas on the face of its financial statements, such as the reported value of debt or
total assets. As of March 31, 2008, Airgas’actual rating of BB+ is three notches below its indicated
reported rating of BBB+. Hence the numerical value of TOTAL equals three. The combined ef-
fect of the rating analyst’s hard and soft adjustments increases the analyst’s assessment of Airgas
credit risk relative to the credit risk assessment based on accounting ratios calculated from Airgas’
reported financial statements.
Conceptually, TOTAL consists of both soft and hard adjustments. I calculate the soft ad-
17
justment (SOFT ) as the difference between actual rating and the hypothetical rating implied by
adjusted financials (indicated adjusted rating). In the case of Airgas, SOFT takes a value of two,
which means that actual rating is two notches lower than indicated adjusted rating. The effect of
credit analysts’ soft adjustments increases their assessment of Airgas’ credit risk.
The hard adjustment is determined by credit analysts’ adjustments to reported financial state-
ments. Off-balance-sheet debt (OFFBS ) is the major hard adjustment (Kraft 2014). To calculate
the rating agency’s estimate of OFFBS, I calculate the difference between adjusted debt and re-
ported debt and scale the difference by total reported assets, where debt equals the sum of short-
term and long-term debt. The credit risk matrix in Appendix A shows that Airgas is awarded
an indicated reported rating of Baa1 (or BBB+ in standardized form) based on unadjusted ratios
and an indicated adjusted rating of Baa2 (BBB) based on adjusted ratios. The decrease in rating
is primarily driven by the deterioration in Airgas’ leverage ratios, namely Debt/Book Capital,
Debt/EBITDA, Retained Cash Flow/Debt, and Free Cash Flow/Debt. This illustration for Air-
gas is typical of credit rating agency adjustments to financial statements. Most hard adjustments
increase leverage as they incorporate off-balance-sheet-debt and thus lead to greater credit risk
and lower ratings (Kraft 2014).
Dealscan reports whether loan contracts have performance pricing features and whether those
are based on ratings or accounting ratios. First I link Moody’s Financial Metrics dataset to
Compustat by matching the issuers in Financial Metrics to their respective Compustat gvkey
identifiers and issuer cusips by company name. Then I employ the Dealscan-gvkey linking data
set from Chava and Roberts (2008). The merging of firm-year observations from Financial Metrics
with loans from Dealscan by gvkey creates the Financial Metrics-Dealscan sample (FMDS ). For
each firm-year in this sample, I calculate the number of active loans and determine whether any
of those include performance pricing features.
Table 1 reports that the FMDS sample contains 1,193 issuers and 6,196 issuer years. Most of
the observations are evenly split over 2002-2008. The highest industry peer group concentrations
are energy (11.3% of all firm years) and electric utilities (8.9% of all firm years). Financial services
18
are not included because they are not part of Financial Metrics. Firms from the whole distribution
of ratings are included in the sample. Around 44% (54%) of all observations have an investment
grade (speculative) rating, but most firms are concentrated in BBB-, BB- and B-ratings.
Table 2 documents that on average FMDS sample firms have total assets USD 8.8 billion,
leverage of 0.38, coverage of 7.90, operating margin of 0.10, return on assets of 0.07, and tangibility
of 0.46. Coverage ratio and operating margin are winsorized at the 1st and 99th percentile.13
OFFBS equals 17%, which implies that for the average firm a significant proportion of total assets
is financed by off-balance-sheet debt. On average, both SOFT and TOTAL reflect increases in
credit risk. TOTAL lowers the rating by almost one notch (0.96), and SOFT lowers the rating by
0.55 notches.14
Table 2 also reports firm characteristics and rating agency adjustments by type of performance
pricing. The variables PPrating and PPratio indicate whether a firm’s loan contracts incorporate
performance pricing based on ratings or accounting ratios respectively. More specifically PPrating
equals one if the firm has an active loan facility outstanding with rating-based performance pric-
ing. PPratio equals one if the firm has an active loan facility outstanding with accounting-based
performance pricing. PPratio firms are smaller and more levered than PPrating firms but have
similar profitability and tangibility as PPrating firms. Specifically, compared with the control
group of PPratio firms, firms in the PPrating subsample are bigger, with average total assets of
USD 11.2 billion (versus USD 3.5 billion), and have lower average leverage (0.30 versus 0.46) and
higher average interest rate coverage (9.5 versus 5.9), higher average operating margin (0.11 versus
0.08), and similar returns on assets (0.08 versus 0.06), as well as a similar average tangibility (0.47
versus 0.45).
The total and soft adjustments for firms with rating-based contracts decrease their ratings by
less than those for firms with accounting-ratio-based performance pricing. PPrating firms have
smaller adjustments for OFFBS, SOFT, and TOTAL than PPratio firms. On average, credit
13Firm characteristics are based on reported numbers.14Each rating is assigned a number from 1.0 for AAA to 21.0 for C. Hence a value of 1.00 of the adjustment
reflects one rating notch.
19
analysts estimate OFFBS at 14% for PPrating firms, which is less than their estimate of 21% for
PPrating firms. Likewise, TOTAL reduces the average PPrating firm’s by 0.75 notches but by
1.30 notches for the average PPratio firm. The smaller estimates of OFFBS, SOFT, and TOTAL
for PPrating firms are consistent with a more favorable treatment by the credit rating agencies.
However, they might also be a function of lower credit risk because PPrating firms are bigger
and less levered. Table 2 also reports the loan characteristics by issuer year. The average loan
has an amount of USD 2,086 million, a maturity of 63 months, and a spread of 161 basis points.
PPrating firms tend to have larger loans with shorter maturities and lower all-in-drawn spreads
than PPratio firms. However both types of loans are very likely to include accounting-based
covenants (CovAccg).
I measure an adverse economic shock by a significant decrease in the market value of the
issuer’s traded bonds. I collect bond prices from Trace and extract issue characteristics from FISD
Mergent. The sample bonds have an average offering amount of USD 436 million and an offering
yield of 5.8% (not tabulated). For each bond’s fiscal year, I calculate the bottom decile return
on an equal-weighted basis and weighted by trading volume. Table 2 documents that the 10th
percentile daily return (Shock return) for the average issue amounts to -1.2%. Shock return is
negative for more than 75% of the observations, and its minimum is -41% (not tabulated). While
some bonds are actively traded, the average bond only trades 52 days a year, and conditional on
trading, only 4.62 times per day (Bessembinder et al. 2009). I recalculate the shock variable by
using daily bond returns weighted by transaction volume (Shock return w). The distribution of
Shock return is similar to that of Shock return w.
As shown in Table 3 Panel A, of all issuer years in the FMDS sample, 78% have a performance
pricing feature (PPfeature), which is higher than the proportion reported for 1998 in Beatty et al.
(2002). The use of ratings or accounting ratios is relatively evenly distributed. Out of the 4,831
firm-years with PPfeature, around 53% exhibit PPrating and around 56% exhibit PPratio. Despite
the criticism rating agencies received during the period, I find little evidence of variation in the use
of ratings versus accounting ratios. If anything, firms and banks were more likely to incorporate
20
ratings rather than accounting ratios at the end of the sample period. A small proportion of issuers
has contracts with both PPrating and PPratio. Those hybrids are excluded in the analysis.
Table 3 Panel B reports that in most rating-based performance contracts interest rates are
allowed to step up or down (PPutroque), 12% have interest rates with step-up provisions only
(PPincrease), and 14% have interest rates with step-down provisions only (PPdecrease). The sum
of the proportions is greater than 100% because each firm year can contain several facilities with
different performance pricing schedules. Interest-rate decreasing performance pricing automatically
decreases the interest rate charged when the issuer’s credit risk improves. This feature lowers
renegotiation costs and reduces adverse selection problems (Asquith et al. 2005). Interest-rate
increasing performance pricing automatically increases the interest rate spread charged when the
issuer’s credit risk deteriorates. This feature reduces moral hazard and adverse selection problems
(Asquith et al. 2005).
In addition, Table 3 Panel B reports the potential change in interest rate spreads over Libor
at time of loan inception. MaxLessInitial is the number of basis points between the interest rate
charged on the contract at inception of the loan agreement and the maximum rate in the pricing
grid. The average difference between the maximum interest rate charged and the initial interest
rate is 44 basis points (the maximum difference amounts to 743 basis points). InitialLessMin is
the difference in basis points between the initial interest rate spread and the minimum interest
rate spread in the pricing grid. The average potential interest rate reduction is 26 basis points
(the maximum reduction is 425 basis points). These numbers for the potential interest spread
movement are significantly larger than the fees paid to rating agencies on corporate debt of three
to four basis points.
Table 4 reports the correlations between the two types of performance pricing, rating agency
adjustments, and firm characteristics. PPrating shows a significant and negative correlation with
OFFBS, whereas PPratio exhibits a significantly positive correlation with OFFBS. PPrating has
a significant negative association with SOFT and TOTAL. The reverse is true for PPratio firms.
Consistent with the univariate evidence in Table 2, PPrating firms exhibit lower rating agency
21
estimates of off-balance-sheet-debt and lower risk assessments of qualitative factors.
5 Empirical Results
Table 5 documents the estimates for the parameters from the regressions of OFFBS, SOFT and
TOTAL on PPrating and firm characteristics. Columns 1 and 2 contain the OLS estimates from
the regressions of OFFBS, and columns 3 to 6 contain the ordered probit estimates from the
regressions of SOFT and TOTAL. Standard errors are clustered by firm and include fixed effects
for utilities (electric, public, and water utilities) and energy and fixed effects for years.
Table 5 Panel A reports the estimates for the subsample of issuers that have loans with per-
formance pricing. Conditional on the presence of PPfeature, I find OFFBS and PPrating have a
significantly negative correlation in both model specifications with various control variables (mod-
els 1 and 2). Column 1 reports the estimated coefficients for a full set of control variables but
has fewer observations due to variable restrictions. Column 2 reports the coefficients for the main
control variables Leverage, Opmargin, and Tangibility. Prior research on off-balance-sheet-finance
finds that credit-constrained firms are more likely to raise off-balance-sheet debt (Beatty et al.
1995; Mills and Newberry 2005). Consistent with this claim, I find OFFBS decreases in Opmargin
and increases in Tangibility. The results suggest less profitable and more tangible-asset-intensive
firms are more willing and able to raise off-balance-sheet finance, or the rating agency makes more
conservative adjustments for these types of firms. SOFT and TOTAL have a significantly negative
association withPPrating across all model specifications. Ceteris paribus, Leverage and M2B have
negative associations with SOFT and TOTAL, whereas size, Coverage, Opmargin, and Quick have
positive associations with SOFT and TOTAL.
The results are consistent with the catering hypothesis. Unless differences in adjustments are
driven by unobservable firm characteristics, the use of rating-based performance pricing has an
association with more favorable rating agency adjustments, namely significantly lower estimates
of OFFBS, SOFT, and TOTAL.
22
Table 5 Panel B reports the results for the total FMDS sample. The results for the main effect
are very similar to the results in Panel A. I find PPrating has a significantly negative association
with OFFBS, SOFT, and TOTAL. The results for the control variables are weaker because the
control sample now includes firms that are potentially less comparable in that not all of them have
performance pricing clauses in their loans.
Table 6 reports the results for the base regressions in sample partitions of three different levels
of yield spread. YS0 denotes firms with lowest yields spread, which range from zero to 139 basis
points. YS1 denotes firms with medium yield spreads, which range from 140 to 279 basis points.
Last, YS2 denotes firms with highest yield spreads, which range from 280 to 1,654 basis points.
Columns 1–3 report the OLS estimates from the regressions of OFFBS, and columns 4–9 contain
the ordered probit estimates from the regressions of SOFT and TOTAL. The sample contains
issuers that have loans with performance pricing and the required data for yield spreads. I find
OFFBS and PPrating have no significantly negative correlation in any of the three YS partitions.
Requiring observable yields spreads decreases the sample size from 4,389 firm-years to 693 firm-
years. The coefficients for firm controls are similar to those in the base regression. The sample size
for SOFT and TOTAL regressions drops from 842 to 130 firm-year observations. PPrating has a
negative significant association with SOFT in all three sample partitions (models 4–6). Similarly,
PPrating has a negative significant association with TOTAL in all three sample partitions (models
7–9). The coefficients for firm controls are weaker than in the base regressions. The results
are consistent with the catering hypothesis. The use of rating-based performance pricing has an
association with significantly lower estimates of SOFT and TOTAL, even in small samples with
more homogeneous credit risk.
Table 7 reports the regression results of rating agency adjustments on adverse economic shocks.
Among firms that experience adverse economic shocks, I expect more favorable rating agency
adjustments under the catering hypothesis for PPrating firms: lower estimates of off-balance-sheet
debt and lower soft and total adjustments. However, such shocks only affect those borrowers
whose contractual interest rates can increase under the stipulations of the performance pricing
23
grid. Hence the shock analysis excludes performance pricing decreasing contracts and includes
performance pricing increasing (PPrating up) only. Adverse economic shocks are measured by the
magnitudes of an issuer’s 10th percentile daily bond return over the year (Shock ret). I employ
the bond level approach to maximize the probability of identifying adverse economic shocks.
Table 7 Panel A reports the regression results of rating agency adjustments on adverse eco-
nomic shocks for all issuers in the FMDS sample with available bond returns. The significantly
negative coefficient of Shock ret for OFFBS in model 1 is consistent with the interpretation that
firms’ adjustments for off-balance-sheet debt increase as they experience more adverse shocks.
The significant negative association between Shock ret and SOFT and TOTAL supports the view
that adjustments decrease credit ratings because these firms experience adverse economic shocks
(models 2–3). The coefficient of the interaction term Shock ret*PPrating up is positive but not
significantly different from zero for OFFBS (model 1). The coefficient of the interaction term
Shock ret*PPrating up is significantly positive for SOFT and TOTAL (models 2–3). More adverse
shocks have an association with incrementally more favorable adjustments for PPrating firms,
which is consistent with catering to PPrating firms. It is possible that PPrating firms are better
able to deal with adverse economic shocks than other firms, thus warranting the favorableness of
rating agency adjustments. However, the market reaction to the shock prices this possibility; for
an equally detrimental shock the market reaction for a PPrating firm would be less severe than
the market reaction for a control firm. The association between favorable adjustments and the
contractual use of ratings is observed after controlling for the size of the market reaction. Empiri-
cally, firms that experience adverse economic shocks are not more likely to contract on ratings: the
correlation between Shock ret and PPrating is not significantly different from zero, which supports
the assumption that those shocks are exogenous to the setting (not tabulated). Models 4–6 reports
the results for the analysis based on trading volume-weighted bond returns. The results are very
similar to those for simple bond returns.
Table 7 Panel B reports the probit regression results of changes in rating agency adjustments
(OFFBS δ, SOFT δ, TOTAL δ) on Shock ret. This change specification constitutes a more
24
stringent test, albeit with a loss of data points. OFFBS δ (SOFT δ, TOTAL δ) is an indicator
variable that measures whether the change in the adjustment in the fiscal year surrounding the
adverse shock increases or decreases the rating agency’s estimate of OFFBS (SOFT, TOTAL.)
Shock ret is more negative when the shock is worse. The statistically negative association between
Shock ret and TOTAL in model 3 implies that firms experiencing adverse shocks receive lower
assessments of credit risk. However, the interaction term Shock ret*PPrating up is statistically
positive and has almost the same absolute value of magnitude as the coefficient of Shock ret, which
suggests that PPrating up firms do not suffer lower credit risk assessments when they experience
adverse shocks. The results are qualitatively similar for SOFT δ but not statistically significant.
Models 4–6 reports the results for the analysis based on trading volume-weighted bond returns.
The results are very similar to those for simple bond returns. Overall, the direction of the coeffi-
cients is consistent with the level results in Table 7 Panel A. Rating agency adjustments capture
increases in credit risk for firms that experience adverse economic shocks. PPrating up firms’
agency adjustments however do not worsen, compared to the group of firms that does not contract
on ratings.
In a robustness test, I use changes in credit default swap (CDS) spreads as measures of credit
risk. In this alternative specification, an adverse shock is measured by an increase of the CDS
spread. Following the literature, I collect five-year CDS spreads of contracts with Modified Re-
structuring clauses, which are the most common and liquid in the US, from Markit. When the sam-
ple is restricted to firm-years with available five-year CDS spreads, sample size drops substantially.
In untabulated analysis I find that for SOFT and TOTAL the inferences remain unchanged: the
coefficients for Shock CDS and the interaction of Shock CDS and PPrating up remain significant.
Greater increases in CDS spreads are associated with greater adjustments, but the effect is miti-
gated for PPrating firms. Please note that the signs of the coefficients switch directions relative to
bond returns. Adverse shocks are increases in CDS spreads, whereas adverse shocks are measured
by lower (more negative) returns. Furthermore, I find statistically significant positive associations
between SOFT δ (TOTAL δ) and PPrating up, which implies that firms experiencing adverse
25
shocks receive assessments of greater credit risk. The interaction term Shock CDS*PPrating up
is negative but not significant. Overall, the results are largely consistent with those using bond
returns.
Next, I conduct a difference-in-difference analysis to test whether closeness to an important
rating threshold strengthens rating agency’s catering or whether reputational concerns prevail.
The rating thresholds I consider are the BBB- ratings as well as the short-term rating P1. Table
8 reports the results from the multivariate analysis for the FMDS sample.15 PPrating has a
negative association with OFFBS. However, the coefficient of the interaction term BBB-*PPrating
is significantly positive (model 1). Similarly, the interaction term P1*PPrating is significantly
positive (model 2). This implies that firms with rating-based contracts receive lower estimates
of off-balance-sheet debt than firms without such contracts. However, those PPrating firms that
are close to the BBB- cutoff or enjoy a P1-rating experience greater adjustments to their off-
balance-sheet debt than their counterparts that are not close to these rating thresholds. As in the
base regressions, I find that PPrating has negative associations with SOFT and TOTAL (models
3–6). Both interaction terms BBB-*PPrating and P1*PPrating are positive, and the latter is
statistically significant. This suggests that closeness to rating threshold leads to lower credit risk
assessments: those PPrating firms that have P1-ratings experience statistically greater soft and
total adjustments. The coefficients of the controls for firm characteristics have similar signs and
levels of significance as those in the base regression. The multivariate evidence is not consistent
with increased catering for firms near important short-term rating thresholds. In contrast, the
use of ratings in contracts for firms close to rating thresholds has an association with a more
unfavorable assessment of credit risk. This can be explained with higher reputational costs for the
credit rating agency at these threshold ratings.
Next, I conduct a difference-in-difference analysis to test whether the existence of Fitch ratings
for a given issuer weakens rating agency’s catering. Table 9 reports the results. Columns 1, 3, and
15There is not sufficient data to include P1 rating interactions if the sample includes firms with performancepricing only. The untabulated results with respect to the BBB- interaction term are similar to the results for fullFMDS sample.
26
5 report the results for the sample of firms with PPfeature embedded in their loans. Columns 2, 4,
and 6 report results for the FMDS sample. Both FITCH and PPrating have negative associations
with OFFBS. The coefficient of the interaction term FITCH*PPrating is significantly positive
(model 1). This implies that while firms with rating-based contracts and firms with FITCH
ratings receive lower estimates of off-balance-sheet debt, those PPrating firms that have a FITCH
rating experience a greater adjustments to their off-balance-sheet debt than their counterparts
without FITCH ratings. The coefficients for the FMDS sample have the same signs but lower
statistical significance (model 2).
Similarly, the coefficients of the interaction term FITCH*PPrating are significantly positive in
both the subsample of PPfeature firms and the FMDS sample (models 3–4). Last, the interaction
term FITCH*PPrating is significantly positive in the FMDS sample (model 6). This suggests that
the existence of a FITCH rating acts as a constraint on catering as it is associated with assessments
of greater credit risk: thosePPrating firms that have FITCH ratings experience statistically greater
off-balance-sheet debt, soft, and total adjustments. This can be explained with higher reputational
costs for the credit rating agency when another credit rating agency provides ratings.16
6 Conclusion
This study examines whether rating agencies cater to issuers with rating-based contracts. Rating-
based contracts link cash payouts to changes in ratings and thus make issuers more sensitive to
their public debt ratings. I examine the relation between rating-based debt contracts and rating
agency adjustments: hard adjustments in the form of the agency’s estimate of off-balance-sheet
debt, as well as soft adjustments for qualitative factors. I find evidence that rating agencies provide
more favorable adjustments to issuers with rating-based contracts relative to issuers with similar
contracts based on accounting ratios and other issuers with private loan agreements. Furthermore,
the documented negative association between rating-based debt contracts and rating agency ad-
16However, these results can also be explained if competition among credit rating agencies leads to lower qualityratings (Becker and Milbourn 2011). The debate on the impact of competition on rating quality is still open.
27
justments continues to hold for subsamples with more homogeneous credit risk. Last, I examine
rating agency adjustments in response to unexpected adverse economic shocks to firms, and I
find a differential reaction by the rating agency, which is consistent with catering to firms with
rating-based contracts. Firms with rating-based contracts receive more favorable rating agency
adjustments after experiencing adverse shocks to credit risk than firms without such contracts.
The evidence from the difference-in-difference analysis for rating thresholds shows that impor-
tant rating thresholds, such as the investment grade rating and prime short-term ratings that allow
firms access to more liquid markets, do not result in catering for firms with rating-based contracts.
In contrast, the adjustments for firms are more unfavorable than for other firms near important
rating thresholds. The reputational costs for the rating agency are likely to be more substantial at
these important rating thresholds. Similarly, the existence of a third credit rating mutes catering
incentives.
A lot of unanswered questions remain. Performance pricing is prevalent among firms that
issue private debt. However, rating triggers that link the posting of collateral or trigger early
repayment result in an even greater sensitivity of firms’ cash flows to changes in ratings. So rating-
based performance pricing might not be the most powerful setting to study catering arising from
rating-based contracts; however, the performance pricing data are available for a large sample.
Furthermore, this study examines only one aspect that could give rise to catering. A higher
sensitivity to rating changes could also result from the dependence on the public markets to issue
debt in order to raise external financing, or the existence of a financial subsidiary that relies more
heavily on ratings for its business than a firm not active in financial services.
28
References
Antle, R. (1984). Auditor independence. Journal of Accounting Research 2 (1), 1–20.
Ashbaugh-Skaife, H., D. W. Collins, and R. LaFond (2006). The effects of corporate governanceon firms’ credit ratings. Journal of Accounting and Economics 42 (1-2), 203–243.
Ashcraft, A., P. Goldsmith-Pinkham, and J. Vickery (2010). MBS ratings and the mortgage creditboom. Working Paper, Federal Reserve Bank of New York.
Asquith, P., A. Beatty, and J. Weber (2005). Performance pricing in bank debt contracts. Journalof Accounting & Economics 40 (1-3), 101–128.
Baik, B., D. B. Farber, and K. Petroni (2009). Analysts’ incentives and street earnings. Journalof Accounting Research 47 (1), 45–69.
Ball, R., R. M. Bushman, and F. P. Vasvari (2008). The debt-contracting value of accountinginformation and loan syndicate structure. Journal of Accounting Research 46 (2), 247–287.
Ball, R. and L. Shivakumar (2006). The role of accruals in asymmetrically timely gain and lossrecognition. Journal of Accounting Research 44 (2), 207–242.
Beatty, A., P. G. Berger, and J. Magliolo (1995). Motives for forming research & developmentfinancing organizations. Journal of Accounting & Economics 19, 411–442.
Beatty, A., I. D. Dichev, and J. Weber (2002). The role and characteristics of accounting-basedperformance pricing in private debt contracts. Working Paper.
Beatty, A. and J. Weber (2003). The effects of debt contracting on voluntary accounting methodchanges. The Accounting Review 78 (1), 119–142.
Beaver, W. H., C. Shakespeare, and M. T. Soliman (2006). Differential properties in the ratings ofcertified versus non-certified bond-rating agencies. Journal of Accounting & Economics 42 (3),303–334.
Becker, B. and T. Milbourn (2011). How did increased competition affect credit ratings? Journalof Financial Economics 101 (3), 493–514.
Benmelech, E. and J. Dlugosz (2009). The alchemy of CDO credit ratings. Journal of MonetaryEconomics 56 (5), 617–634.
Bessembinder, H., K. M. Kahle, W. F. Maxwell, and D. Xu (2009). Measuring abnormal bondperformance. Review of Financial Studies 22 (10), 4219–4258.
Bhanot, K. and A. S. Mello (2006). Should corporate debt include a rating trigger? Journal ofFinancial Economics 79 (1), 69–98.
Bolton, P., X. Freixas, and J. Shapiro (2012). The credit ratings game. The Journal of Fi-nance 67 (1), 85–111.
29
Cantor, R. and F. Packer (1996). Multiple ratings and credit standards: Differences in opinion inthe credit rating industry. Federal Reserve Bank of New York Staff Reports 12.
Cantor, R. and F. Packer (1997). Differences of opinion and selection bias in the credit ratingindustry. Journal of Banking and Finance 21 (10), 1395–1417.
Chava, S. and M. R. Roberts (2008). How does financing impact investment? The role of debtcovenants. Journal of Finance 63 (5), 2085–2121.
Cheng, M. and M. Neamtiu (2009). An empirical analysis of changes in credit rating properties:Timeliness, accuracy and volatility. Journal of Accounting and Economics 47 (1-2), 108–130.
Coval, J. D., J. W. Jurek, and E. Stafford (2009). The economics of structured finance. Journalof Economic Perspectives 23 (1), 3–25.
Dichev, I. D. and D. J. Skinner (2002). Large-sample evidence on the debt covenant hypothesis.Journal of Accounting Research 40 (4), 1091–1123.
Doyle, J. (2008). Credit risk measurement and pricing in performance-based debt contracts. Un-published dissertation.
Engel, E., M. Erickson, and E. Maydew (1999). Debt-equity-hybrid securities. Journal of Account-ing Research 37 (2), 249–274.
Ertimur, Y., V. Muslu, and F. Zhang (2011). Why are recommendations optimistic? Evidencefrom analysts’ coverage initiations. Review of Accounting Studies 16 (4), 679–718.
Gorton, G. and A. Winton (2003). Financial intermediation. In G. Constantinides, M. Harris, andR. Stulz (Eds.), Handbook of the Economics of Finance (1 ed.), Volume 1, Part 1, Chapter 08,pp. 431–552. Elsevier.
Griffin, J. M. and D. Y. Tang (2012). Did subjectivity play a role in CDO credit ratings? TheJournal of Finance 67 (4), 1293–1328.
Harris, M. and A. Raviv (1991). The theory of capital structure. Journal of Finance 46 (1),297–355.
He, J., J. Qian, and P. E. Strahan (2011). Credit ratings and the evolution of the mortgage-backedsecurities market. American Economic Review 101 (3), 131–35.
Hermalin, B. E. and M. L. Katz (1991). Moral hazard and verifiability: The effects of renegotiationin agency. Econometrica 59 (6), 1735 – 1753.
Jensen, M. C. and W. H. Meckling (1976). Theory of the firm: Managerial behavior, agency costsand ownership structure. Journal of Financial Economics 3 (4), 305–360.
Klein, B. and K. B. Leffler (1981). The role of market forces in assuring contractual performance.Journal of Political Economy 89 (4), 615–41.
30
Kraft, P. (2014). Rating agency adjustments to GAAP financial statements and their effect onratings and bond yields. The Accounting Review , forthcoming.
Larcker, D. F. and S. A. Richardson (2004). Fees paid to audit firms, accrual choices, and corporategovernance. Journal of Accounting Research 42 (3), 625–658.
Leftwich, R. (1983). Accounting information in private markets: Evidence from private lendingagreements. The Accounting Review 58 (1), 23–42.
Lin, H. and M. F. McNichols (1998). Underwriting relationships, analysts’ earnings forecasts andinvestment recommendations. Journal of Accounting and Economics 25 (1), 101–127.
Manso, G. (2013). Feedback effects of credit ratings. Journal of Financial Economics 109 (2), 535– 548.
Manso, G., B. Strulovici, and A. Tchistyi (2010). Performance-sensitive debt. Review of FinancialStudies 23 (5), 1819–1854.
Mason, J. R. and J. Rosner (2007). Where did the risk go? How misapplied bond ratings causemortgage backed securities and collateralized debt obligation market disruptions. Working Pa-per, Drexel University.
Mills, L. F. and K. J. Newberry (2005). Firms’ off-balance sheet and hybrid debt financing:Evidence from their book-tax reporting differences. Journal of Accounting Research 43 (2), 251–282.
Moody’s (2001). The unintended consequences of rating triggers.
Moody’s (2006). Moody’s approach to global standard adjustments in the analysis of financialstatements for non-financial corporations - part I.
Moody’s (2007). Financial reporting and credit ratings (Presentation by Greg Jonas), CAREconference, Napa California.
Nicholls, C. C. (2005). Public and private uses of credit ratings. In Credit Rating Agencies: Needfor Reform in Canada? Capital Markets Institute (cmi).
Nikolaev, V. (2010). Debt covenants and accounting conservatism. Journal of Accounting Re-search 48 (1), 51–89.
Opp, C. C., M. M. Opp, and M. Harris (2013). Rating agencies in the face of regulation. Journalof Financial Economics 108 (1), 46 – 61.
Partnoy, F. (1999). The Siskel and Ebert of financial markets?: Two thumbs down for the creditrating agencies. Washington University Law Quarterly 77 (3), 619–712.
Partnoy, F. (2006). How and why rating agencies are not like other gatekeepers. In Y. Fuchitaand R. E. Litan (Eds.), San Diego Legal Studies Research Paper Series No. 07-46. BrookingsInstitution Press and the Nomura Institute of Capital Market Research.
31
Petersen, M. A. (2004). Information: Hard and soft. Working Paper, Northwestern University.
Rajan, M. V. and S. Reichelstein (2009). Objective versus subjective indicators of mangerialperformance. The Accounting Review 84 (1), 209–237.
Rajan, R. G. and L. Zingales (1995). What do we know about capital structure? Some evidencefrom international data. Journal of Finance 50 (5), 1421–1460.
Rajan, U., A. Seru, and V. Vig (2010). The failure of models that predict failure: Distance,incentives and default. Working Paper.
SEC (2003). Report on the role and function of credit rating agencies in the operation of thesecurities markets.
Shapiro, C. (1983). Premiums for high quality products as returns to reputations. The QuarterlyJournal of Economics 98 (4), 659–680.
Smith, C. W. and J. B. Warner (1979). On financial contracting: an analysis of bond covenants.Journal of Financial Economics 7, 117–161.
Standard & Poor’s (2006). Corporate ratings criteria.
Standard & Poor’s (2008). Evaluating liquidity triggers in insurance enterprises.
Standard & Poor’s (2009). Standard & poor’s rating services U.S. ratings fee disclosure.
Stein, J. C. (2002). Information production and capital allocation: Decentralized versus hierarchi-cal firms. Journal of Finance 57 (5), 1891–1921.
Strausz, R. (2005). Honest certification and the threat of capture. Journal of Industrial Organi-zation 23 (1-2), 45–62.
Wakeman, L. M. (1984). The real function of bond rating agencies. In M. C. Jensen and C. W.Smith (Eds.), The Modern Theory of Corporate Finance. New York, N.Y.: North-Holland Pub-lishing Company.
Watts, R. and J. Zimmerman (1986). Positive Accounting Theory. Englewood Cliffs, NJ: PrenticeHall.
White, L. (2002). The credit rating industry: An industrial organization analysis. In R. Levich,G. Majnoni, and C. Reinhart (Eds.), Ratings, Rating Agencies and the Global Financial System.Kluwer.
Xia, H. (2014). Can investor-paid credit rating agencies improve the information quality of issuer-paid rating agencies? Journal of Financial Economics 111 (2), 450 – 468.
32
Appendix AIllustration of rating process for Airgas, Inc.
FY Ending March 31, 2008
Rating
(Moody's)
Rating
(Standardized) Numerical value
Indicated Rating (based on reported financials) Baa1 BBB+ 8
Indicated Rating (based on adjusted financials) Baa2 BBB 9
Actual Rating (based on adjusted financials and soft factors) Ba1 BB+ 11
SOFT (soft adjustment) 2
TOTAL (total adjustment) 3
Weight
Reported
financials
Adjusted
financials
BUSINESS PROFILE
Business Position Assessment 8.3% A A
SIZE & STABILITY
Revenues (USD Billion) 8.3% Baa Baa
Number of Divisions of Equal Size 8.3% Baa Baa
Stability of EBITDA 8.3% Caa Caa
COST POSITION
EBITDA Margin (3‐yr Average) 8.3% A A
EBIT/Average Assets (3‐yr Average) 8.3% A A
Contingencies as % of Cash from Operations (3‐yr Average) 8.3% Aa Aa
MANAGEMENT QUALITY
Debt / Book Capital 8.3% Baa Ba
Debt / EBITDA (3‐yr Average) 8.3% Baa Ba
FINANCIAL STRENGTH
EBITDA / Interest Expense (3‐yr Average) 8.3% Baa Baa
Retained Cash Flow / Debt (3‐yr Average) 8.3% A Baa
Free Cash Flow / Debt (3‐yr Average) 8.3% Baa Ba
WEIGHTED AVERAGE Baa1 Baa2
Source: Moody's Financial Metrics
Indicated Rating
33
Appendix BShock analysis (hypothetical example)
Firm type: PPrating firm PPratio firm 1 PPratio firm 2
Shock, gross ‐100 ‐100 ‐70
If treatment is random
Shock, net ‐100 ‐100 ‐70
Rating agency adjustment X X Y < X
If PPrating firm is better at dealing with shock (self‐selection)
Shock, net ‐70 ‐100 ‐70
Rating agency adjustment Y X Y
34
Table 1Sample distribution
No. observations FMDS Industries (top 25) % of sample
N (issuers) 1,193 Energy 11.3%
N (issuer years) 6,196 Electric Utilities 8.9%
Manufacturing 7.8%
Year N Retail 7.3%
2002 836 Services 6.7%
2003 892 Media 5.6%
2004 943 Consumer Products 5.2%
2005 976 Chemicals 4.5%
2006 917 Healthcare 4.2%
2007 875 Technology 4.0%
2008 757 Metals, Mining & Steel 2.7%
Total 6,196 Automotive 2.7%
Telecommunications 2.4%
Rating N Gaming / Lodging 2.4%
AAA 36 Aerospace / Defense 2.3%
AA 107 Pharmaceuticals 2.3%
A 854 Homebuilding 1.7%
BBB 1,723 Forest Products 1.6%
BB 1,758 Restaurants 1.6%
B 1,507 Packaging 1.4%
CCC 194 Wholesale Distribution 1.2%
CC 13 Wholesale Power 1.2%
C 4 Apparel 1.2%
Total 6,196 Rail Roads & Trucking 1.1%
Agriculture 1.1%
The table reports the sample distribution for the FMDS sample by year, rating and
industry. Rating is the long‐term Moody's issuer rating on filing date. Industries are
classified according to Moody's classification scheme.
35
Table 2Sample summary statistics for subsamples
Sample means FMDS PPrating=1 PPratio=1
Firm characteristics (millions USD)
Total assets 8,789 11,236 3,505
Revenues 7,611 9,553 3,370
Ratios
Leverage 0.38 0.30 0.46
Coverage 7.90 9.50 5.90
Opmargin 0.10 0.11 0.08
ROA 0.07 0.08 0.06
Tangibility 0.46 0.47 0.45
Rating agency adjustment for off‐balance sheet debt (% of total assets)
OFFBS 17.0% 14.0% 21.0%
Implied rating agency adjustments (notches)
SOFT 0.55 0.23 1.10
TOTAL 0.96 0.75 1.30
Loan characteristics
Loan amount (millions USD) 2,086 2,644 1,905
Loan maturity (months) 63 60 67
Allindrawn spread (bps) 161.0 100.0 226.0
CovAccg 0.81 0.95 0.98
Adverse return (Shock_ret)
Shock_ret ‐1.2% ‐1.1% ‐1.4%
Shock_ret_w ‐1.2% ‐1.1% ‐1.4%
The table reports the statistics for the FMDS sample. PPrating equals 1 if issuer year has a facility that includes a performance
pricing clause based on a rating. PPratio equals 1 if issuer year has a facility that includes a performance pricing clause based
on an accounting ratio. Leverage is the ratio of total debt to total assets. Coverage is the ratio of operating profit to interest
expense, winsorized at 1%. Opmargin is the ratio of operating profit to revenues, winsorized at 1%. ROA is the ratio of
operating profit to total assets. Tangibility is the ratio of inventory and net PPE to total assets. OFFBS equals adjusted total
debt less reported total debt, divided by total assets. SOFT equals the difference between the actual rating and the implied
adjusted rating. TOTAL equals the difference between the actual rating and the implied reported rating. CovAccg equals 1 if
loan has an accounting‐based covenant. Shock_ret equals the tenth percentile of daily bond return by issuer‐year bond.
Shock_ret_w equals the tenth percentile of daily bond return by issuer‐year bond (return based on price weighted by trading
volume).
36
Table 3Performance pricing by year and type
Panel A
Year Total obs N Pct of total N Pct of PPfeat N Pct of PPfeat N Pct of PPfeat
2002 836 603 72% 321 53% 340 56% 58 10%
2003 892 653 73% 315 48% 385 59% 47 7%
2004 943 733 78% 367 50% 424 58% 58 8%
2005 976 781 80% 398 51% 451 58% 68 9%
2006 917 749 82% 401 54% 422 56% 74 10%
2007 875 708 81% 395 56% 388 55% 75 11%
2008 757 604 80% 355 59% 311 51% 62 10%
Total 6,196 4,831 78% 2,552 53% 2,721 56% 442 9%
Panel B
Performance pricing ‐ by direction (proportion) Sensitivity to interest rate (bps over Libor)
PPrating PPratio mean min max
PP_increase 12% 19% MaxLessInitial 44 0 743
PP_decrease 14% 56% InitiallessMin 26 0 425
PP_utroque 90% 66%
This table reports the statistics for the FMDS sample. PPfeature equals 1 if issuer year has a facility that includes a performance pricing clause.
PPrating equals 1 if issuer year has a facility that includes a performance pricing clause based on a rating. PPratio equals 1 if issuer year has a
facility that includes a performance pricing clause based on an accounting ratio (including user conditions). Hybrid equals 1 if issuer year has
facilities with both performance pricing based on ratings and accounting ratios. PP_increase equals 1 if issuer year has performance pricing
clause with initial interest rate equal to minimum interest rate in grid. PP_decrease equals 1 if issuer year has performance pricing clause with
initial interest rate equal to maximum interest rate in grid. PP_utroque equals 1 if issuer year has performance pricing clause with initial interest
rate between maximum and minium interest rate in grid. MaxLessInitial equals the number of basis points between the rate charged on the
contract at the inception of the loan agreement and the maximum rate in the performance pricing grid. InitiallessMin equals the number of
basis points between the rate charged on the contract at the inception of the loan agreement and the minimum rate in the performance pricing
grid.
Hybrid=1PPfeature=1 PPrating=1 PPratio=1
37
Table 4Correlation matrix
(1) (2) (3) (4) (5) (6) (7) (8) (9)
PPrating (1) 1.0000
PPratio (2) ‐0.4672* 1.0000
OFFBS (3) ‐0.0279* 0.0315* 1.0000
SOFT (4) ‐0.1557* 0.2610* 0.0601* 1.0000
TOTAL (5) ‐0.0897* 0.1662* 0.1936* 0.8673* 1.0000
Ln(Revenues) (6) 0.4168* ‐0.4504* 0.0972* ‐0.0739* 0.0113 1.0000
Leverage (7) ‐0.2511* 0.3313* ‐0.0083 ‐0.0136 ‐0.1635* ‐0.3959* 1.0000
Coverage (8) 0.2936* ‐0.3271* ‐0.0971* 0.0531* 0.1669* 0.3977* ‐0.6669* 1.0000
Opmargin (9) 0.1319* ‐0.1580* ‐0.3550* 0.0135 0.0254* ‐0.0124 ‐0.1664* 0.5840* 1.0000
Tangibility (10) ‐0.0844* 0.0922* ‐0.1633* ‐0.1265* ‐0.1296* ‐0.0902* 0.1116* ‐0.0434* 0.0213*
This table reports the Spearman rank correlation coefficients for the FMDS sample. PPrating equals 1 if issuer year has a facility that
includes a performance pricing clause based on a rating. PPratio equals 1 if issuer year has a facility that includes a performance
pricing clause based on an accounting ratio. OFFBS equals adjusted total debt less reported total debt, divided by total assets. SOFT
(soft adjustment) equals the difference between the actual rating and the implied rating from adjusted financials. TOTAL (total
adjustment) equals the difference between the actual rating and the implied rating from reported financials. Leverage is the ratio of
total debt to total assets. Coverage is the ratio of operating profit to interest expense, winsorized at 1%. Opmargin is the ratio of
operating profit to revenues, winsorized at 1%. Tangibility is the ratio of inventory and net PPE to total assets. The * denotes
significance at the 5% level.
38
Table 5Panel A: Regression analysis
Model 1 2 3 4 5 6
Regression type OLS OLS Oprobit Oprobit Oprobit Oprobit
Dependent variable OFFBS OFFBS SOFT SOFT TOTAL TOTAL
PPrating ‐0.084** ‐0.082** ‐1.096** ‐1.103** ‐0.915** ‐0.924**
[‐3.790] [‐3.748] [‐7.016] [‐7.543] [‐5.976] [‐6.421]
Ln(revenues) 0.021** 0.036** 0.159* 0.168** 0.155* 0.176**
[2.618] [3.995] [2.434] [2.920] [2.330] [3.061]
Leverage ‐0.053 0.106 ‐0.559+ ‐0.777** ‐1.188** ‐1.230**
[‐0.792] [1.157] [‐1.707] [‐3.069] [‐3.367] [‐4.592]
Coverage 0.000 0.008* 0.008*
[0.745] [2.160] [2.032]
Opmargin ‐0.186** ‐0.185** 0.787 0.937* 0.644 0.824+
[‐4.430] [‐4.951] [1.557] [2.001] [1.233] [1.729]
Tangibility 0.129** 0.184** ‐0.309 ‐0.457+ ‐0.235 ‐0.283
[3.202] [4.134] [‐1.104] [‐1.832] [‐0.819] [‐1.128]
Quick ‐0.057** 0.331** 0.253**
[‐4.773] [3.888] [2.900]
M2B 0.000 ‐0.001** ‐0.001**
[‐0.426] [‐7.092] [‐5.468]
Loan amount ‐0.000** ‐0.000** 0.000 0.000 0.000 0.000
[‐5.088] [‐5.093] [1.295] [0.554] [0.871] [0.158]
Loan maturity ‐0.001* ‐0.001* 0.001 0.001 0.001 0.001
[‐2.476] [‐2.496] [0.376] [0.385] [0.372] [0.280]
0.008 ‐0.351*
Constant [0.063] [‐2.360]
Observations 3,787 4,389 747 842 743 838
(Pseudo) R‐squared 0.120 0.120 0.068 0.052 0.055 0.045
The estimates are for the OLS (columns 1‐2) and ordered probit (columns 3‐6) parameters using the FMDS
sample, conditional on having PPfeature and excluding hybrids. OFFBS equals adjusted total debt less reported
total debt, divided by total assets. SOFT equals the difference between the actual rating and the implied rating
from adjusted financials. TOTAL equals the difference between the actual rating and the implied rating from
reported financials. PPrating equals 1 if issuer year has a facility that includes a performance pricing clause based
on a rating. Leverage is the ratio of total debt to total assets. Coverage is the ratio of operating profit to interest
expense, winsorized at 1%. Opmargin is the ratio of operating profit to revenues, winsorized at 1%. Tangibility is
the ratio of inventory and net PPE to total assets. Quick is the ratio of cash, marketable securities, and accounts
receivable to current liabilities, winsorized at 1%. M2B is the ratio of market value of equity to book value of
equity. Loan amount is the amount in millions USD. Loan maturity is the maturity of the loan in months. Industry
fixed effects for utilities and energy and year fixed effects are included. Robust t ‐ and z‐statistics in brackets.
Standard errors clustered by firm. The + indicates significance at 10%; the * significance at 5%; the **
significance at 1%.
39
Table 5 (continued)Panel B: Regression analysis
Model 1 2 3 4 5 6
Regression type OLS OLS Oprobit Oprobit Oprobit Oprobit
Dependent variable OFFBS OFFBS SOFT SOFT TOTAL TOTAL
PPrating ‐0.035** ‐0.030* ‐0.436** ‐0.480** ‐0.359** ‐0.400**
[‐2.709] [‐2.182] [‐3.817] [‐4.230] [‐3.200] [‐3.589]
Ln(revenues) 0.006 0.025* ‐0.061 ‐0.052 ‐0.026 ‐0.012
[0.932] [2.147] [‐1.221] [‐1.110] [‐0.520] [‐0.252]
Leverage ‐0.014 0.194 ‐0.218 ‐0.401+ ‐0.780** ‐0.842**
[‐0.265] [1.416] [‐0.861] [‐1.864] [‐2.680] [‐3.613]
Coverage 0.001 0.001 0.001
[1.034] [0.387] [0.308]
Opmargin ‐0.231** ‐0.210** 0.788+ 0.773+ 0.673 0.714+
[‐5.905] [‐5.556] [1.684] [1.859] [1.397] [1.685]
Tangibility 0.154** 0.193** ‐0.303 ‐0.417+ ‐0.197 ‐0.237
[4.115] [4.683] [‐1.228] [‐1.893] [‐0.779] [‐1.067]
Quick ‐0.046** 0.256** 0.214**
[‐4.141] [3.338] [2.880]
M2B 0.000 ‐0.001** ‐0.001**
[0.432] [‐6.058] [‐5.213]
Loan amount ‐0.000** ‐0.000** 0.000* 0.000+ 0.000 0.000
[‐5.134] [‐4.061] [2.167] [1.654] [1.164] [0.617]
Loan maturity 0.000 0.000 0.003 0.003 0.003 0.003
[‐0.557] [‐0.648] [1.100] [1.354] [1.307] [1.417]
0.141 ‐0.277
Constant [1.324] [‐1.300]
Observations 4,884 5,721 928 1,035 922 1,029
(Pseudo) R‐squared 0.110 0.120 0.035 0.027 0.031 0.025
The estimates are for the OLS (columns 1‐2) and ordered probit (columns 3‐6) parameters using the FMDS
sample, excluding hybrids. OFFBS equals adjusted total debt less reported total debt, divided by total assets.
SOFT equals the difference between the actual rating and the implied rating from adjusted financials. TOTAL
equals the difference between the actual rating and the implied rating from reported financials. PPrating equals
1 if issuer year has a facility that includes a performance pricing clause based on a rating. Leverage is the ratio of
total debt to total assets. Coverage is the ratio of operating profit to interest expense, winsorized at 1%.
Opmargin is the ratio of operating profit to revenues, winsorized at 1%. Tangibility is the ratio of inventory and
net PPE to total assets. Quick is the ratio of cash, marketable securities, and accounts receivable to current
liabilities, winsorized at 1%. M2B is the ratio of market value of equity to book value of equity. Loan amount is
the amount in millions USD. Loan maturity is the maturity of the loan in months. Industry fixed effects for utilities
and energy and year fixed effects are included. Robust t‐ and z‐statistics in brackets. Standard errors clustered by
firm. The + indicates significance at 10%; the * significance at 5%; the ** significance at 1%.
40
Table 6Sample partition by credit risk
Model 1 2 3 4 5 6 7 8 9
Dependent variable OFFBS OFFBS OFFBS SOFT SOFT SOFT TOTAL TOTAL TOTAL
Regression type OLS OLS OLS oprobit oprobit oprobit oprobit oprobit oprobit
PARTITION BY YS YS0 YS1 YS2 YS0 YS1 YS2 YS0 YS1 YS2
PPrating 0.038 0.017 ‐0.045 ‐13.833** ‐0.904 ‐1.367** ‐11.058** ‐0.868+ ‐1.546**
[1.285] [0.615] [‐1.399] [‐9.523] [‐1.571] [‐2.851] [‐3.995] [‐1.705] [‐3.113]
Ln(revenues) 0.013 0.002 0.004 0.626 ‐0.080 0.052 ‐0.050 ‐0.129 0.121
[0.846] [0.177] [0.314] [0.745] [‐0.394] [0.194] [‐0.075] [‐0.580] [0.478]
Leverage ‐0.240* ‐0.128 ‐0.145 1.927 ‐2.144 3.607* ‐10.131+ ‐3.799** 3.718*
[‐2.591] [‐1.487] [‐1.378] [0.454] [‐1.440] [1.990] [‐1.804] [‐2.644] [2.096]
Coverage 0.000 ‐0.001 0.000 0.058 0.008 0.080+ ‐0.032 0.003 0.113**
[‐0.259] [‐1.347] [0.195] [0.532] [1.635] [1.705] [‐0.271] [0.598] [2.869]
Opmargin ‐0.195+ ‐0.224* ‐0.061 2.319 ‐2.368 1.743 ‐4.378 ‐3.076 0.821
[‐1.883] [‐2.343] [‐0.725] [0.240] [‐1.189] [0.984] [‐0.477] [‐1.477] [0.469]
Tangibility 0.116 0.046 0.167* ‐2.776 ‐0.634 ‐2.163* ‐0.531 ‐1.671+ ‐1.373
[1.601] [0.751] [1.983] [‐0.887] [‐0.598] [‐2.466] [‐0.208] [‐1.662] [‐1.624]
Loan amount 0.000 0.000 ‐0.000* 0.000 0.000+ 0.000 0.000 0.000* ‐0.000+
[‐1.213] [‐0.219] [‐2.097] [‐0.343] [1.737] [‐1.209] [0.029] [2.087] [‐1.736]
Loan maturity 0.000 0.000 0.001 0.002 0.007 0.035** 0.022 0.012 0.035**
[1.192] [‐1.383] [0.972] [0.140] [0.705] [3.496] [1.439] [1.259] [2.869]
Constant 0.003 0.174 0.337
[0.012] [1.206] [1.152]
Observations 202 242 249 24 59 47 24 59 47
R‐squared 0.250 0.140 0.230 0.364 0.124 0.182 0.322 0.156 0.197
The estimates are for the OLS (columns 1‐3) and ordered probit (columns 4‐9) parameters using the FMDS sample, conditional on having PPfeature
and excluding hybrids. The sample is partioned by YS. YS equals the difference between offering yield and yield on a comparable treasury security
(in basis points). OFFBS equals adjusted total debt less reported total debt, divided by total assets. SOFT equals the difference between the actual
rating and the implied rating from adjusted financials. TOTAL equals the difference between the actual rating and the implied rating from reported
financials. PPrating equals 1 if issuer year has a facility that includes a performance pricing clause based on a rating. Leverage is the ratio of total
debt to total assets. Coverage is the ratio of operating profit to interest expense, winsorized at 1%. Opmargin is the ratio of operating profit to
revenues, winsorized at 1%. Tangibility is the ratio of inventory and net PPE to total assets. Loan amount is the amount in millions USD. Loan
maturity is the maturity of the loan in months. Industry fixed effects for utilities and energy and year fixed effects are included. Robust t‐ and z‐
statistics in brackets. Standard errors clustered by firm. The + indicates significance at 10%; the * significance at 5%; the ** significance at 1%.
41
Table 7Panel A: Regression analysis with adverse shocks
Model 1 2 3 4 5 6
Regression type OLS Oprobit Oprobit OLS Oprobit Oprobit
Dependent variable OFFBS SOFT TOTAL OFFBS SOFT TOTAL
PPrating_up ‐0.003 0.162 0.256 ‐0.003 0.157 0.244
[0.21] [1.01] [1.52] [0.17] [1.00] [1.46]
Shock_ret ‐0.924+ ‐21.471** ‐26.711** ‐1.003* ‐20.038** ‐25.432**
[1.92] [3.62] [4.37] [2.05] [3.36] [4.09]
Shock_ret*PPrating_up 0.817 21.209** 27.830** 0.867 20.774** 26.757**
[1.52] [2.64] [3.35] [1.61] [2.62] [3.22]
Ln(revenues) ‐0.003 ‐0.053 ‐0.043 ‐0.003 ‐0.055 ‐0.045
[0.45] [0.92] [0.72] [0.45] [0.96] [0.75]
Leverage 0.008 ‐0.382 ‐0.881+ 0.008 ‐0.397 ‐0.903+
[0.14] [1.01] [1.89] [0.14] [1.05] [1.95]
Coverage ‐0.001** ‐0.001 ‐0.004 ‐0.001** ‐0.001 ‐0.004
[3.44] [0.26] [1.17] [3.44] [0.27] [1.19]
Opmargin ‐0.110* 0.162 ‐0.234 ‐0.109* 0.148 ‐0.249
[2.27] [0.26] [0.41] [2.26] [0.24] [0.44]
Tangibility 0.097** ‐0.877** ‐0.756+ 0.097** ‐0.872** ‐0.750+
[2.71] [2.78] [1.95] [2.71] [2.76] [1.93]
Constant 0.118
[1.08] [1.08]
Observations 5,459 1,301 1,294 5,459 1,301 1,294
(Pseudo) R‐squared 0.110 0.033 0.030 0.110 0.023 0.029
Bond returns (Shock_ret)
Bond returns, weighted by
volume (Shock_ret_w)
The estimates are for the OLS (columns 1, 4) and ordered probit (columns 2‐3, 5‐6) model
parameters using the FMDS‐Trace sample, excluding hybrids. OFFBS equals adjusted total debt less
reported total debt, divided by total assets. SOFT (soft adjustment) equals the difference between
the actual rating and the implied rating from adjusted financials. TOTAL (total adjustment) equals
the difference between the actual rating and the implied rating from reported financials.
PP_rating_up equals 1 if issuer year is classified as PP_increase or PP_utroque. Shock_ret equals
the tenth percentile of daily bond return by issuer‐year bond. Shock_ret_w equals the tenth
percentile of daily bond return by issuer‐year bond (return based on price weighted by trading
volume). The interaction terms Shock_ret*PPrating_up (Shock_ret_w*PPrating_up) measure the
diff‐in‐diff. Leverage is the ratio of total debt to total assets. Coverage is the ratio of operating
profit to interest expense, winsorized at 1%. Opmargin is the ratio of operating profit to revenues,
winsorized at 1%. Tangibility is the ratio of inventory and net PPE to total assets. Robust t‐ and z‐
statistics in brackets. Industry fixed effects for utilities and energy and year fixed effects are
included. Standard errors clustered by firm. The + indicates significance at 10%; the * significance
at 5%; the ** significance at 1%.
42
Table 7Panel B: Regression analysis with adverse shocks for changes in adjustments
Model 1 2 3 4 5 6
Regression type Probit Probit Probit Probit Probit Probit
Dependent variable OFFBS_δ SOFT_δ TOTAL_δ OFFBS_δ SOFT_δ TOTAL_δ
PPrating_up ‐0.073 ‐0.125 ‐0.027 ‐0.073 ‐0.099 ‐0.01
[0.75] [0.43] [0.10] [0.75] [0.34] [0.04]
Shock_ret 2.736 ‐12.123 ‐28.776** 2.756 ‐17.363 ‐31.432**
[1.47] [0.92] [2.72] [1.49] [1.31] [2.92]
Shock_ret*PPrating_up ‐2.28 6.459 23.854+ ‐2.338 9.628 26.395+
[0.69] [0.40] [1.75] [0.72] [0.60] [1.94]
Ln(revenues) 0.023 ‐0.025 ‐0.087 0.023 ‐0.021 ‐0.086
[0.61] [0.26] [1.00] [0.61] [0.22] [0.99]
Leverage ‐0.442+ ‐1.607* ‐1.367* ‐0.443+ ‐1.681* ‐1.450*
[1.84] [2.30] [2.38] [1.85] [2.38] [2.49]
Coverage 0.004 ‐0.007 ‐0.006 0.004 ‐0.008 ‐0.007
[1.40] [1.25] [1.01] [1.40] [1.32] [1.06]
Opmargin 0.497 ‐0.093 ‐1.293 0.498 ‐0.066 ‐1.287
[1.26] [0.10] [1.46] [1.26] [0.07] [1.45]
Tangibility 0.689** ‐0.36 ‐0.664 0.689** ‐0.362 ‐0.655
[2.64] [0.63] [1.21] [2.65] [0.63] [1.20]
Constant 0.338 0.891 1.748 0.337 0.797 1.737
[0.53] [0.53] [1.17] [0.53] [0.47] [1.17]
Observations 5,287 697 683 5,287 697 683
Pseudo R‐squared 0.068 0.070 0.142 0.068 0.049 0.144
The estimates are for the probit model parameters using the FMDS‐Trace sample (excluding hybrids).
OFFBS_delta equals 1 if OFFBS at fiscal‐year‐end is greater than OFFBS at prior fiscal‐year end and 0
if it is smaller. SOFT_delta equals 1 if SOFT at fiscal year‐end is greater than SOFT at prior fiscal year‐
end and 0 if it is smaller. TOTAL_delta equals 1 if TOTAL at fiscal year‐end is greater than TOTAL at
prior fiscal year‐end and 0 if it is smaller. PPrating_up equals 1 if issuer‐year is classified as
PP_increase or PP_utroque. Shock_ret equals the tenth percentile of daily bond return by issuer‐year
bond. Shock_ret_w equals the tenth percentile of daily bond return by issuer‐year bond (return
based on price weighted by trading volume). The interaction term Shock*PPrating_up measures the
diff‐in‐diff. Leverage is the ratio of total debt to total assets. Coverage is the ratio of operating profit
to interest expense, winsorized at 1%. Opmargin is the ratio of operating profit to revenues,
winsorized at 1%. Tangibility is the ratio of inventory and net PPE to total assets. Industry fixed
effects for utilities and energy and year fixed effects are included. Robust z‐statistics in brackets.
Standard errors clustered by firm. The + indicates significance at 10%; the * significance at 5%; the **
significance at 1%.
Bond returns (Shock_ret)
Bond returns, weighted by
volume (Shock_ret_w)
43
Table 8Regression analysis for rating thresholds
Model 1 2 3 4 5 6
Regression type OLS OLS Oprobit Oprobit Oprobit Oprobit
Dependent variable OFFBS OFFBS SOFT SOFT TOTAL TOTAL
PPrating ‐0.041** ‐0.037+ ‐0.459** ‐0.640** ‐0.390** ‐0.472*
[‐3.032] [‐1.934] [‐3.816] [‐2.861] [‐3.273] [‐2.269]
BBB‐ ‐0.042* ‐0.422 ‐0.319
[‐2.034] [‐1.511] [‐1.458]
BBB‐*PPrating 0.057* 0.428 0.394
[2.265] [1.282] [1.347]
P1 ‐0.088** ‐1.336** ‐1.163**
[‐3.450] [‐4.480] [‐4.392]
P1*PPrating 0.064* 0.850** 0.636*
[2.426] [2.692] [2.189]
Ln(revenues) 0.006 0.020** ‐0.059 ‐0.041 ‐0.025 0.000
[1.016] [2.619] [‐1.159] [‐0.523] [‐0.484] [0.006]
Leverage ‐0.017 ‐0.008 ‐0.253 ‐0.151 ‐0.814** ‐0.794+
[‐0.321] [‐0.147] [‐0.989] [‐0.413] [‐2.751] [‐1.676]
Coverage 0.001 0.001 0.001 0.005 0.001 0.003
[1.052] [1.136] [0.421] [1.221] [0.367] [0.727]
Opmargin ‐0.230** ‐0.195** 0.776+ 1.014+ 0.673 0.894
[‐5.907] [‐3.553] [1.670] [1.778] [1.408] [1.612]
Tangibility 0.153** 0.148** ‐0.308 ‐0.701+ ‐0.201 ‐0.406
[4.105] [3.055] [‐1.243] [‐1.782] [‐0.794] [‐1.015]
Quick ‐0.046** ‐0.033* 0.262** 0.345** 0.217** 0.160
[‐4.143] [‐2.189] [3.434] [2.637] [2.921] [1.158]
M2B 0.000 0.000 ‐0.001** 0.003 ‐0.001** 0.002
[0.428] [0.115] [‐6.135] [1.027] [‐5.266] [0.555]
Loan amount ‐0.000** ‐0.000** 0.000* 0.000* 0.000 0.000
[‐5.138] [‐4.140] [2.115] [2.474] [1.136] [0.991]
Loan maturity 0.000 0.000 0.003 0.007+ 0.003 0.007+
[‐0.554] [‐0.683] [1.065] [1.818] [1.285] [1.946]
Constant 0.138 ‐0.090
[1.302] [‐0.776]
Observations 4,884 2,043 928 398 922 394
(Pseudo) R‐squared 0.110 0.130 0.036 0.072 0.031 0.045
The estimates are for the OLS (columns 1‐2) and ordered probit (columns 3‐6) model parameters using
the FMDS sample, excluding hybrids. OFFBS equals adjusted total debt less reported total debt, divided
by total assets. SOFT equals the difference between the actual rating and the implied rating from
adjusted financials. TOTAL equals the difference between the actual rating and the implied rating from
reported financials. PPrating equals 1 if issuer year has a facility that includes a performance pricing
clause based on a rating. BBB‐ equals 1 if rating = BBB‐, and 0 otherwise. P1 equals 1 if rating = P1, and 0
otherwise. The interaction term THRESHOLD*PPrating measures the diff‐in‐diff. Leverage is the ratio of
total debt to total assets. Coverage is the ratio of operating profit to interest expense, winsorized at 1%.
Opmargin is the ratio of operating profit to revenues, winsorized at 1%. Tangibility is the ratio of
inventory and net PPE to total assets. Quick is the ratio of cash, marketable securities, and accounts
receivable to current liabilities, winsorized at 1%. M2B is the ratio of market value of equity to book
value of equity. Loan amount is the amount in millions USD. Loan maturity is the maturity of the loan in
months. Industry fixed effects for utilities and energy and year fixed effects are included. Robust t‐ and z‐
statistics in brackets. Standard errors clustered by firm. The + indicates significance at 10%; the *
significance at 5%; the ** significance at 1%.
44
Table 9Interaction with FITCH
Model 1 2 3 4 5 6
Dependent variable OFFBS OFFBS SOFT SOFT TOTAL TOTAL
Regression type OLS OLS oprobit oprobit oprobit oprobit
Sample PPfeature FMDS PPfeature FMDS PPfeature FMDS
PPrating ‐0.083** ‐0.032+ ‐1.181** ‐0.648** ‐0.979** ‐0.526**
[‐3.469] [‐1.734] [‐7.492] [‐4.917] [‐6.345] [‐4.033]
FITCH ‐0.116** ‐0.067** ‐0.219 ‐0.317+ ‐0.175 ‐0.229
[‐3.061] [‐3.001] [‐1.089] [‐1.830] [‐0.869] [‐1.473]
FITCH*PPrating 0.068+ 0.027 0.392+ 0.593** 0.280 0.444*
[1.819] [1.144] [1.678] [2.835] [1.190] [2.254]
Ln(revenues) 0.041** 0.029** 0.137* ‐0.051 0.149* ‐0.013
[4.579] [2.715] [2.371] [‐1.043] [2.556] [‐0.274]
Leverage 0.121 0.207 ‐0.596* ‐0.368+ ‐1.091** ‐0.819**
[1.282] [1.473] [‐2.288] [‐1.693] [‐4.003] [‐3.486]
Coverage 0.001 0.001 0.011** 0.003 0.010** 0.003
[1.116] [1.329] [3.088] [0.913] [2.847] [0.868]
Opmargin ‐0.207** ‐0.236** 0.683 0.726+ 0.589 0.658
[‐4.963] [‐5.991] [1.441] [1.657] [1.223] [1.476]
Tangibility 0.192** 0.197** ‐0.506* ‐0.409+ ‐0.333 ‐0.230
[4.297] [4.776] [‐1.990] [‐1.831] [‐1.312] [‐1.023]
Loan amount ‐0.000** ‐0.000** 0.000 0.000+ 0.000 0.000
[‐4.919] [‐4.103] [0.763] [1.733] [0.380] [0.650]
Loan maturity ‐0.001* 0.000 0.002 0.003 0.001 0.004
[‐2.215] [‐0.526] [0.530] [1.518] [0.370] [1.529]
Constant ‐0.437** ‐0.345+
[‐2.943] [‐1.690]
Observations 4,389 5,721 842 1,035 838 1,029
R‐squared 0.130 0.130 0.059 0.031 0.049 0.028
The estimates are for the OLS (columns 1‐2) and ordered probit (columns 3‐6) model parameters. Columns
1,3,5 are based on the FMDS sample, conditional on PPfeature. Columns 2,4,6 are based on FDMS sample.
OFFBS equals adjusted total debt less reported total debt, divided by total assets. SOFT equals the difference
between the actual rating and the implied rating from adjusted financials. TOTAL equals the difference
between the actual rating and the implied rating from reported financials. PPrating equals 1 if issuer year has
a facility that includes a performance pricing clause based on a rating. FITCH equals 1 if issuer‐year has Fitch
rating, and 0 otherwise. The interaction term FITCH*PPrating measures the diff‐in‐diff. Leverage is the ratio of
total debt to total assets. Coverage is the ratio of operating profit to interest expense, winsorized at 1%.
Opmargin is the ratio of operating profit to revenues, winsorized at 1%. Tangibility is the ratio of inventory
and net PPE to total assets. Loan amount is the amount in millions USD. Loan maturity is the maturity of the
loan in months. Industry fixed effects for utilities and energy and year fixed effects are included. Robust t‐ and
z‐statistics in brackets. Standard errors clustered by firm. The + indicates significance at 10%; the *
significance at 5%; the ** significance at 1%.
45