Ratings Quality and Borrowing Choice

• Published date: 01 October 2019
• DOI: http://doi.org/10.1111/jofi.12820
• Author: CHRISTOPHER M. JAMES,DOMINIQUE C. BADOER,CEM DEMIROGLU
• Date: 01 October 2019

THE JOURNAL OF FINANCE •VOL. LXXIV, NO. 5 •OCTOBER 2019

Ratings Quality and Borrowing Choice

DOMINIQUE C. BADOER, CEM DEMIROGLU, and CHRISTOPHER M. JAMES∗

ABSTRACT

Past studies document that incentive conflicts may lead issuer-paid credit rating

agencies to provide optimistically biased ratings. In this paper, we present evidence

that investors question the quality of issuer-paid ratings and raise corporate bond

yields where the issuer-paid rating is more positive than benchmark investor-paid

ratings. We also find that some firms with favorable issuer-paid ratings substitute

public bonds with borrowings from informed intermediaries to mitigate the “lemons

discount” associated with poor quality ratings. Overall, our results suggest that the

quality of issuer-paid ratings has significant effects on borrowing costs and the choice

of debt.

CREDIT RATING AGENCIES ARE FINANCIAL intermediaries that facilitate ‘‘arm’s

length” debt financing by reducing information asymmetries between issuers

and investors through forward-looking opinions about the credit quality of

issuers and their specific debt obligations. By aggregating information about

credit quality from public and private sources into a single measure, rating

agencies reduce individual investors’ need to conduct detailed due diligence and

thereby broaden issuers’ investor base beyond sophisticated investors willing

and able to engage in complex credit analyses (see, for example, Faulkender

and Petersen (2006) and Sufi (2009)).

Past studies suggest, however, that reliance on compensation paid by the

issuers that they rate may lead rating agencies to provide over optimistic rat-

ings.1In this paper, we present evidence that investors treat issuer-paid rat-

ings with skepticism (question their quality), raising corporate bond yields

∗Dominique C. Badoer is with the University of Illinois at Chicago. Cem Demiroglu is with Koc¸

University.Christopher M. James is with the University of Florida. We would like to thank Philip

Bond (Editor), an anonymous Associate Editor, and two anonymous referees for useful comments

and suggestions. We would also like to thank David Brown, Denis Gromb, Tomasz Michalski,

Ramin Baghai (discussant), and seminar participants at the 2016 FIRS Conference, 10th End-

of-Year Conference of Swiss Economists Abroad, Georgia Tech, HEC Paris Workshop on Finance,

Banking, Real Economy and Trade, and the University of Missouri for helpful comments. A portion

of this paper was written while Badoer was at the University of Missouri and while James was

the Pembroke Visiting Fellow at Cambridge University.Badoer and Demiroglu have no conflicts of

interest to disclose. James has served as a paid expert retained by S&P in litigation involving the

ratings of structured products. His retention by S&P had no effect on the analysis or conclusions

of this project.

1Jiang, Stanford, and Xie (2012) find that S&P started to issue more favorable ratings for

corporate borrowers after switching from an investor-paid compensation model to an issuer-paid

model in 1974. Becker and Milbourn (2011) find that increases in Fitch market share in an industry

DOI: 10.1111/jofi.12820

2619

2620 The Journal of Finance R

when the issuer-paid rating is more favorable than the investor-paid rat-

ing, which induces some firms to substitute public bonds with intermediated

debt.

To motivate our empirical analysis, we construct a simple model of ratings

inflation. In our model, there are two rating agencies: one whose fees are paid

by issuers and that has a propensity to inflate some ratings, and another whose

fees are paid by investors and that does not inflate its ratings. The model illus-

trates how inflation in issuer-paid ratings can exacerbate information asym-

metries even when investors recognize and attempt to price it.

Our primary empirical measure of ratings quality is based on the difference

between Standard & Poor’s (S&P) and Egan-Jones Ratings Company’s (EJR)

firm-level senior debt ratings. We define a firm as overrated (underrated) if

S&P rates the issuer more (less) favorably than EJR. We compare S&P rat-

ings, which are issuer-paid, to investor-paid EJR ratings because EJR uses the

same long-term rating scale, defines its ratings in the same way as S&P, and,

like S&P, provides through-the-cycle ratings based on default probabilities.2

In robustness tests, we also measure ratings quality by comparing Moody’s

ratings to CDS-implied ratings.3

Of course, ratings disagreements between S&P and EJR may arise simply

because of differences in quantitative credit risk models or qualitative risk

assessments, and thus, S&P overrating (relative to EJR) does not necessarily

indicate ratings inflation. However, using five sets of empirical tests, we find

evidence that S&P overrating reflects, in part, ratings inflation by S&P and

not just genuine differences of opinion between the two agencies.

We begin by documenting that, within an S&P rating category, firms that

are overrated by S&P are riskier—based on both ex-ante firm financial

increases competition between issuer-paid rating agencies and results in the issuance of less

informative and more optimistic ratings. Comparing issuer-paid ratings with benchmark investor-

paid ratings, both Cornaggia and Cornaggia (2013) and Strobl and Xia (2012) conclude that the

issuer-paid rating model results in inflated corporate credit ratings.

2The Securities and Exchange Commission (SEC) requires all Nationally Recognized Statistical

Ratings Organizations (NRSROs) to provide public investors a general description of the procedures

and methodologies used to determine credit ratings (via Form NRSRO, reported annually). The

descriptions must include the quantitative and qualitative models and metrics used to determine

credit ratings, as well as the procedures for monitoring, reviewing, and updating credit ratings.

For the SEC’s Form NRSRO instructions, see goo.gl/Uh1sgQ. For the most recent forms for S&P

and EJR as of this writing, see goo.gl/2NQENF and goo.gl/NNtPmJ, respectively.

3Moody’s ratings are based on judgments about expected credit losses (which are determined by

loss given default [LGD] as well as the likelihood of default) and thus are not directly comparable

to EJR’s ratings. Since CDS spreads contain information about the likelihood of default and LGD,

we compare Moody’s ratings to CDS-implied ratings. We have several concerns with the CDS-

based ratings quality measure; however, and thus, we view our CDS-based results as robustness

checks. First, Moody’s ratings are ordinal ratings that are intended to be through-the-cycle ratings,

whereas CDS-based ratings are more akin to point-in-time ratings that reflect near-term forecasts

of economic conditions (Kiff, Kisser,and Schumacher (2013)). Second, CDS spreads may also reflect

liquidity conditions in the CDS market, making them a noisier measure of credit risk than ratings.

Third, the number of rated firms exceeds the number of firms with actively traded CDS contracts,

limiting the sample and raising potential endogeneity concerns.

Ratings Quality and Borrowing Choice 2621

characteristics and ex-post default rates—than both firms that are overrated

by EJR and firms that are rated the same by the two agencies. In contrast, we

find no significant difference in the default rates of firms that are overrated by

EJR (i.e., underrated by S&P) and those that are rated the same by S&P and

EJR. These results indicate that EJR ratings add to the granularity of S&P

ratings, but only when S&P is the more optimistic agency. This asymmetry is

hard to explain in the absence of incentive conflicts, and is consistent with S&P

overrating serving as an indicator of poor ratings quality.

Next, we demonstrate that the likelihood of S&P overrating is positively

related to commonly used empirical proxies for the severity of the conflicts of

interest faced by issuer-paid rating agencies. For example, we find that the

frequency of S&P overrating is significantly greater for issuers that contribute

more to S&P’s rating revenues (e.g., frequent bond issuers and issuers with

many outstanding bond issues) as well as issuers that have greater incentives

to engage in ratings shopping (e.g., high-yield [HY] issuers and issuers at the

boundary between HY and investment-grade [IG] ratings).

We also examine the time-series evolution of S&P and EJR ratings after

EJR first initiates coverage of an issuer. We find that S&P is more likely (and

faster) to converge toward EJR when it is initially the more optimistic agency

than when the two agencies agree or when S&P is the more pessimistic agency.

We do not find a similar pattern for EJR ratings. These results suggest that

S&P overrating does not arise simply from structural differences in the way

the two rating agencies assign ratings to issuers. Rather, the evidence is more

consistent with optimistic S&P ratings being less timely and thus of lower

quality than benchmark EJR ratings.

Finally,to establish a causal connection between conflicts of interest and S&P

overrating, we exploit a quasi-random shock to S&P’s incentives to engage in

ratings inflation. Fong et al. (2014) argue that exogenous reductions in a firm’s

sell-side analyst coverage due to brokerage firm mergers (or closures) give

issuer-paid rating agencies greater leeway to engage in ratings inflation with-

out being detected. Their argument is that consensus earnings forecasts are, on

average, less optimistic and more accurate when more sell-side analysts cover

an issuer, which, in turn, should make it more difficult for issuer-paid rating

agencies to issue optimistically biased ratings without losing credibility. Based

on this insight, we investigate changes in the likelihood of S&P overrating

following brokerage firm closures or mergers. Using a difference-in-differences

estimation strategy, we document economically large and statistically signif-

icant increases in the likelihood of S&P overrating following brokerage firm

mergers or closures for firms affected by these events relative to a control sam-

ple that consists of firms that have the same S&P and EJR ratings as the

affected firms but that are unaffected by such events. Taken together, these

results provide strong support for the argument that S&P overrating is a valid

measure of poor ratings quality.

After validating our ratings quality measure, we examine whether investors

recognize and price poor-quality issuer-paid ratings. We focus on bond prices