Вы находитесь на странице: 1из 34

The Society for Financial Studies

The Real Effects of Debt Certification: Evidence from the Introduction of Bank Loan Ratings
Author(s): Amir Sufi
Source: The Review of Financial Studies, Vol. 22, No. 4 (Apr., 2009), pp. 1659-1691
Published by: Oxford University Press. Sponsor: The Society for Financial Studies.
Stable URL: http://www.jstor.org/stable/30225706 .
Accessed: 30/09/2013 13:05
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Oxford University Press and The Society for Financial Studies are collaborating with JSTOR to digitize,
preserve and extend access to The Review of Financial Studies.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effects of Debt Certification:

Evidencefrom the Introductionof Bank
Loan Ratings
Amir Sufi
Schoolof Business,Universityof Chicago
I examine the introductionof syndicatedbank loan ratings by Moody's and Standard&
Poor's in 1995 to evaluate whether third-partyrating agencies affect firm financial and
investmentpolicy. The introductionof bank loan ratingsleads to an increase in the use of
debt by firmsthatobtaina rating,andalso increasesin firms'assetgrowth,cash acquisitions,
andinvestmentin workingcapital.Consistentwith a causaleffect of the ratings,the increase
in debt usage and investment is concentratedin the set of borrowerswho are of lower
credit quality and do not have an issuer credit rating before 1995. A loan-level analysis
demonstratesthat previously unratedborrowerswho obtain a loan rating gain increased
access to the capital of less-informedinvestors.The results suggest that third-partydebt
certificationhas real effects on firminvestmentpolicy. (JEL G31, G32, G34, G21)

Can third-partyratingagencies affectfirmfinancialandinvestmentpolicy? The

increasingworldwide presence of such agencies suggests that they provide a
valuable service for borrowersand creditors.Moody's and Standard& Poor's
(S&P) are the most visible agencies in the US, but organizationssuch as Dun
& Bradstreetand Equifax also provide credit scores for private businesses.
Djankov,McLiesh, and Shleifer (2007) show thatratingagencies are common
worldwide;theyreportthatprivatecreditbureausexist in 71 of the 129 countries
in their study.
From a theoreticalperspective, it is unclear whether third-partydebt certification services should have real effects on the economy. The model by
Holmstrom and Tirole (1997), for example, argues that an intermediarythat
exerts unobservableeffort in certifying a borrowermust have the incentive to
exert such effort. Holding capital in the borrower'sproject provides such an
incentive. Indeed, the fundamentallink between holding a financial stake in
the borrowerandcertifyingthe borrowerdefinesthe modem commercialbank,

I thankDouglasDiamond,MichaelFaulkender,
at the University
of Virginia(McIntire),
the University
of Illinois,the University
the FRBNY/Wharton/RFS
conferenceon theCorporate
ChicagoGSB,the NBERSummerInstitute,
of FinancialIntermediaries,
& Poor's,andtheAmerican
to A. Sufi,Graduate
Schoolof Business,
of Chicago,5807SouthWoodlawn
O The Author2007. Publishedby OxfordUniversityPresson behalfof TheSocietyfor FinancialStudies.
All rights reserved.For Permissions,please e-mail:journals.permissions@oxfordjournals.org.
doi: 10.1093/rfs/hhm061
Advance Access publicationDecember 11, 2007

This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReviewof Financial Studies/ v 22 n 4 2009


0. -3 ........

- - - --.-




........................ .............. ... .. .... .............. ... ..... ......... ..... ... ......... .... .. ...



















- -- -Asset weightedfraction

Fractionof firmsthatever obtaina loan ratingby year.The sample includes all publicly listed firmsthatexist in
1994 through 1996.

and helps explainwhy bankstypically hold partof the loans they originate(see
Sufi, 2007). Unlike a bank, a third-partyratingagency does not hold the debt
of the borrowersthey certify, and relies on reputationalone when providing
certificationservices. The key empiricalquestion I seek to answer is the following: can a third-partyratingagency improvethe allocation of credit in the
economy withoutholding the debt they certify? Despite the increasingimportance of third-partyratingagencies bothin the US and worldwide,thereis little
empiricalresearchthat addressesthis question.
I attemptto bridge this gap in the existing researchby analyzing the introduction of syndicatedbank loan ratings by Moody's and S&P in 1995. S&P
states on theirWeb site thatloan ratingshelp borrowers".. . as ratings(1) help
to expanda loan's initial investorbase and secondary-marketliquidity;and (2)
facilitatepurchaseby institutionalinvestors,CDOs, specializedloan funds,and
otherbuyersthatneed ratingsor do not have large internalcredit staffs"(S&P,
2005). Loan ratingsprovideinformationto potentialsyndicatepartnersduring
the syndication process, and they provide informationto secondary market
participantsafterthe loan is originated.
I focus on the introductionof loan ratingsto assess the impactof third-party
debt certificationon financialandrealoutcomesof borrowers.The introduction
of loan ratingsoffers a promisingenvironmentfor two reasons.First,the scope
of the programis large. As Figure 1 demonstrates,by 2004, almost 30% of
public firms that existed in 1995 had obtaineda loan ratingand over 70% of
asset-weighted public firms had obtained a loan rating. Second, loan ratings
were unavailableto all firmsbefore 1995, andimmediatelybecame availableto
all firmsthereafter.Giventhatfirmswere restrictedfrom obtainingloan ratings


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification

before 1995, the innovationoffers promise in terms of identifying the causal

effect of ratingson firm outcomes.
The empiricalanalysisis motivatedby a theoreticalframeworkin the spiritof
Holmstromand Tirole (1997). A basic assumptionin their model is that firms
requireinvestigation and monitoringby an "informed"lender before "uninformed"lenders invest in the firm. I use this frameworkto obtain predictions
on the effects of the introductionof loan ratings. I assume that loan ratings
reduce the total certificationcost paid by borrowers.With this assumption,the
frameworkpredictsthat borrowerswho obtain loan ratings gain access to the
capital of less-informed investors, and are therefore able to raise more debt
financing.In turn,borrowers'increasedaccess to capital leads to an increase
in investment.
In orderto empiricallyevaluatethis framework,I first examine the ex ante
characteristicsof firmsthat obtaina loan rating.I focus, in particular,on firms
thatdo not have public debt before 1995 and subsequentlyobtaina loan rating
by 1998. The findings suggest that the introductionof loan ratings in 1995
offered this particularset of firms a unique opportunityto obtain third-party
debt certification.More specifically, as of 1994, these firms are in the middle
of the size and firm age distribution,being smaller and younger than firms
that alreadyhave access to public debt markets,but largerand older thanfirms
without public debt that do not subsequentlyobtain a loan rating.In addition,
these firms are smaller and younger in the year prior to obtaining the loan
rating than firms in the year prior to initially accessing public debt markets.
Given priorevidence thatfirm age and firm size are importantdeterminantsof
whetherfirmsare able to access public debt markets(Faulkenderand Petersen,
2006), these results suggest thatthe smallerand youngerfirmsthatobtainloan
ratingswould have been less likely to access public debt marketsif loan ratings
had not been introduced.
In the second set of results, I examine whetherthe introductionof loan ratings increases firms' use of debt and firm investment.Identifying the causal
impact of loan ratings is difficult given that firms choose whether or not to
obtain a loan ratingaftertheir introductionin 1995. If omitted variables(such
as a firm-specifictechnological shock or an increase in investmentdemand)
simultaneouslycause firmsto obtaina loan ratingandincreaseinvestment,then
a positive correlationis not reflectiveof a causal relationship.I employ an empirical strategywith two importantcharacteristicsthat help identify the causal
effect of loan ratings on outcomes. First, the identificationstrategyemploys
firm fixed effects and exploits the introductionof the technology after 1995.
Therefore,identificationof key parameterscomes from comparingoutcomes
for the same firm before and after the technology is introduced.Any omitted
variablethat drives both the decision to obtainthe ratingand outcomes would
have to explainchangesin outcomesfor the samefirm beforeandafterobtaining
the loan rating.The fixed effects specificationalso controlsfor observablemeasures of investmentdemand such as earnings, the market-to-bookratio, debt


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Reviewof Financial Studies/ v 22 n 4 2009

levels, and year by two-digit SIC industryindicatorvariables.The latterset of

indicatorvariablescapturesany general trends over time in industry-specific
Second, I exploit a direct prediction of the theoretical frameworkabove
to identify the causal impact of loan ratings on outcomes: obtaining a bank
loan rating should have the largest impact on financial and real outcomes of
borrowerswhere the services of a certificationagency have the most value. I
use two measuresof ex antecertificationvalue.First,I exploitthe fact thatsome
of the firmsthatobtaina loan ratingalreadyhave an issuer creditratingbefore
loan ratingsareintroduced,while otherfirmsdo not have an issuercreditrating.
As the first set of results suggests the lattergroup is a particularlycompelling
treatmentgroup, given that they would have been less likely to access public
debt marketsin the absence of the introductionof loan ratings.Second, in the
absence of a third-partyratingagency, borrowersof lower quality would find
it more difficult to raise debt finance. I use the initial loan rating obtainedby
the borroweras a proxy for borrowerquality:junk borrowers(with a ratingof
BB+ or worse) should benefit more from third-partydebt certification.
Using this empiricalstrategy,the evidence suggests that the introductionof
loan ratingsincreasesthe supply of availabledebt financing.Firmsthat obtain
a loan ratingexperiencean increaseboth in the changein debt scaled by lagged
assets ratio and in their leverage ratio. The result is significantlystrongerin
magnitudeamong firmsthatdo not have an existing issuer creditratingbefore
the introductionof loan ratings.For example, unratedfirms that obtain a loan
rating experience an increase in their leverage ratio by 0.11 more than rated
firms that obtain a loan rating, which is 40% at the mean. Moreover,among
unratedfirmsthat obtain loan ratings,firmsof lower credit qualityexperience
largerincreases in their leverage ratio when they obtain a loan rating.In other
words, firmsfor whom the value of debtcertificationis highestarepreciselythe
firms that are able to obtainmore debt financingas a result of the introduction
of loan ratings.
While existing researchfinds that issuer credit ratings are associated with
higherleverageratios(FaulkenderandPetersen,2006), a uniquecontributionof
this articleis to documentthe exact channelthroughwhich third-partyratings
increasethe availabilityof debtfinance.Morespecifically,the evidencesuggests
thatloan ratingsallow borrowersto expandthe set of creditorsbeyonddomestic
commercialbanks towardless-informedinvestors, such as foreign banks and
nonbankinstitutionalinvestors. A loan-level analysis shows that the average
firm's first rated loan has 4.5 more lenders on the syndicate comparedto the
previousunratedloan by the same firm.In particular,the increasein the number
of lendersis drivenby an increasein the numberof foreignbanksand nonbank
institutionalinvestorson the syndicate;these relativelyless-informedcreditors
account for almost 60% of the increase in the numberof lenders on a firm's
first rated loan. A nonbankinstitutionalinvestor is 13% more likely to be a
syndicate member on a firm's first rated loan relative to the firm's previous


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification

unratedloan. The evidence is consistent with the model by Boot, Milbourn,

and Schmeits (2006), who arguethat credit ratingsare criticalto attractfunds
from investorswithout specialized monitoringor screeningskills.
In the final set of results, I show that the introductionof bank loan ratings
has real effects. More specifically,firmswithoutan issuer creditratingin 1995
that subsequently obtain a loan rating experience a doubling of both asset
growthand cash acquisitionsrelativeto theirrespectivemeans.While the loan
ratings do not affect capital expenditures,they do lead to a slight increase in
investmentin workingcapital.The increasein asset growth,cash acquisitions,
and working capital is particularlystrong among borrowerswho do not have
a previousissuer credit ratingand have their initial loan ratedBB+ or worse.
These findings suggest that loan ratings affect the allocation of credit in the
economy: the borrowerswho have the most to gain from certificationby a
third-partycreditratingagency arepreciselythe borrowerswho experiencethe
largestincreasesin investment.
Overall, the results are in line with the hypothesis that loan ratings afford
informational-opaqueborrowersincreasedaccess to the capitalof uninformed
investors. In turn,this increasedaccess leads to more investment.Alternative
omitted variables,such as demandshocks, are difficultto reconcile with these
patterns-they cannotexplainwhy lower-qualityunratedborrowersexperience
the sharpestincreasein real outcomes. In the last section of the article,I show
that the effect of obtaining a loan ratingon leverage ratios and investmentis
significantly larger than the effect of firms' obtaining either an issuer credit
ratingor an unratedsyndicatedloan.
This article makes contributionsto three areasof existing research.First,it
contributesto a growing body of researchon the importanceof information
asymmetryin the syndicatedloan market,which representsover $3.5 trillion
of debt financing (Dennis and Mullineaux,2000; Lee and Mullineaux,2004;
Ivashina,2005; Moerman,2005; Sufi, 2007). The paperby Mullineauxand Yi
(2006) is the only existing paper,to my knowledge, that examines syndicated
bank loan ratings in any context. They find that loan rating downgradesare
associatedwith dropsin equityprices.This articleis the firstto provideevidence
thatsyndicatedloan ratingslead to an increasein the availabilityof financingin
the syndicatedloan market.It is also the firstto evaluatethe effect of syndicated
loans on investment.
Second, this articlecontributesto an increasingbody of literaturethatargues
thatthe supplyof debtfinancingis an importantdeterminantof capitalstructure
and investment (Leary, 2005; Faulkenderand Petersen, 2006; Tang, 2006).
Faulkenderand Petersenfind that firms with issuer credit ratingshave higher
leverageratios,andthatfirmsincreaseleveragein the year in which they obtain
an issuer credit rating.Tang exploits Moody's decision to furtherrefine credit
ratingsin 1982, andfindsthatupgradedborrowersissue moredebt andincrease
investmentrelative to downgradedborrowers.This article adds to this line of
research by focusing on bank loan ratings, which representone of the most


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Reviewof Financial Studies/ v 22 n 4 2009

importantratinginnovationsin the last 15 years. In addition,this articleis the

firstto documentthatratingslead to an increasein availabledebtfinancingfrom
less-informedcreditors,suchas noncommercialbankinstitutionalinvestorsand
foreign banks.
Finally, this articleis relatedto the literatureon the importanceof banks in
the macroeconomy(Bernanke, 1983; Bernanke,Lown, and Friedman, 1991;
Kashyap,Stein, andWilcox, 1993). In the last 15 years,bankshave increasingly
shed the credit risk of loans they originatethroughsecuritizations,credit default swaps, and syndications(Loutskina,2005; Loutskinaand Strahan,2005;
Schuermann,2004). This trendin disintermediationrepresentsa departurefrom
banks' traditionaldual role of holding an originatedloan and monitoringthe
borrowerwho obtainsthe loan. This articleshows thatthe introductionof bank
loan ratings has played an importantrole in disintermediationtrends,as they
have increasedthe abilityof banksto syndicatecorporateloans to noncommercial bankinstitutionalinvestorsand foreign banks.
The rest of this articleis outlinedas follows. Section 1 providesbackground
on loan ratings and details the data used in the empirical analysis. Section 2
presentsthe theoreticalframework.Sections 3 and 4 presentresults, and Section 5 concludes.

1. Loan Ratings: Background and Data

1.1 Background
In Marchand Decemberof 1995, respectively,Moody's and S&P began rating
syndicatedbank loans. Before describingthe introductionof loan ratingsand
the rating process, I first provide a brief description of the syndicated loan
market(Dennis and Mullineaux,2000; Sufi, 2007). The syndicatedloan market is one of the largest sources of corporatefinance worldwide, with over
$3.5 trillion of syndicated loans originatedannually.A syndicated loan is a
loan made to a firmjointly by more than one financial institution.The lead
bankin the syndicatebegins the loan originationby signing a preliminaryloan
agreement("mandate")with the borrowingfirmthat specifies covenants,fees,
and collateral. The preliminaryloan agreementalso specifies a loan amount,
and a rangefor the interestrate.Once the preliminaryloan agreementis signed,
the lead arrangerthen turnsto potentialparticipantlendersto fund partof the
loan. The lead arrangerprovides potential participantswith an information
memorandumon the borrowingfirm. Once the participantsagree to fund part
of the loan, the loan agreementis signed by all parties.During the life of the
loan, the lead arrangertypically also acts as the "agent"bankthatmonitorsthe
firm,governsthe termsof the loan, administersthe drawdownof funds, calculates interestpayments, and enforces financialcovenants (Sufi, 2007). Pieces
of the loan may tradein the secondaryloan market(see Moerman,2005, for a
descriptionof the secondaryloan market).


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification

Moody's and S&P began offering syndicated loan ratings to help expand
the initial base of investors during the syndicationprocess and to encourage
secondarymarketliquidity.Anecdotal evidence suggests that the introduction
of loan ratings was driven primarilyby the desire of nonbank institutional
investorsto participatein the market.As Bavaria(2002, p. 3) notes,
These [nonbankinstitutional]investors,madeup primarilyof retailmutual
funds (primarilyso-called"primerate"funds),specializedloan investment
vehicles (CDOs and similarentities),andtraditionalinstitutionalinvestors
like insurancecompanies,andothermoneymanagerswho havediscovered
the loan asset class, are accustomedto, and take for granted,the existence
of the traditionalsecurities marketinfrastructure.They must have thirdparty research... This has provided a powerful demand for ratings on
syndicatedloans that are targetedfor distributionto these entitities.
In addition, an American Banker (1996, p. 17) article' reportsthat "The increase [in the prevalenceof bankloan ratings]underscoreseffortsby commercial banks and their corporatecustomers to cater to the growing number of
institutionalinvestorswho want a piece of the bank loan syndicationmarket.
Standard& Poor's ratings ... are used by the investorcommunityto evaluate
debt."The fact that nonbankinstitutionalinvestorshelped spurthe creationof
syndicatedloan ratingshelps supportthe notion that the introductionof loan
ratingswas supply-driven.While many anecdotes suggest that ratingagencies
introducedloan ratingsas a responseto increasedwillingness to supplyfinancing by institutionalinvestors,there is little evidence that loan ratings were a
response to increasedborrowerdemandfor financing.
The evidence suggests that less-informedinvestorsdemandcertificationof
a third-partyratingagency in the syndicatedloan marketgiven the importance
of informationasymmetrybetweenthe lead arrangerandparticipantlendersin
the original syndication,and between banks and institutionalinvestors in the
secondarymarket.As noted earlier,a large body of empiricalresearchon syndicatedloans documentsthe problemsassociatedwith informationasymmetry
in this market.The ratingagencies claim that a loan ratingcan help to reduce
these problems:
[Ratings]are most useful when obtainedbefore the loan is syndicated,to
allow syndicatorsand investorsto incorporateourratingand analysis into
their pricing, distribution,and due-diligence decisions' (Bavaria,2004,
p. 2).
A borrowerwithout a previous security rated by the rating agency goes
through a two-step rating process.2 First, a borroweris assigned an issuer
credit rating,which measuresthe purerisk of default,andfocuses solely on the
I thankan anonymousrefereefor pointingme to this article.
2 This process describesS&P's
ratingprocedure.Moody's procedureis similar.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReviewof Financial Studies/ v 22 n 4 2009

overall creditworthinessof the issuer. Second, the loan rating focuses on both
the riskof defaultandthe likelihoodof ultimaterecoveryin the eventof default.
The loan ratingcontains informationabout the loan terms that is not revealed
in the issuer credit rating;as Bavaria(2005, p. 23) notes, "in assigning a loan
rating,Standard& Poor's creditanalystslook to see whetherthereis collateral
security or otherenhancementthat would enable investorsto achieve ultimate
recoveryif the loan defaults."A loan ratingcan be "notched"up-that is, given
a higher ratingthan the issuer credit rating-if holders of the syndicatedloan
can expect 100%recoveryratesin case of default.3
The borrowerdecides whether or not to obtain a rating during the syndication process. An importantelement in understandingwhether a firm obtains a loan rating is the fee schedule. The fee schedule for Moody's and
S&P is quite similar; I describe the S&P fee schedule for brevity. S&P
charges a transactionfee of 2.9 basis points for the first $1 billion of a
loan with maturityof one year or more, and 1.45 basis points for the amount
above $1 billion. For short-termfacilities of 364 days or less, the transaction
fee is 1 basis point for the first $1 billion, and 0.5 basis points thereafter.In
addition,a borrowerwith no previousissuer credit ratingfrom S&P must pay
an initial $50,000 fee. Finally, any borrowerwith an issuer credit ratingmust
pay a $39,000 annualfee to maintainthe rating.
A finalnote concernswhetherfirmschoose to obtaina loan rating,or whether
the loan ratingagency assigns a ratingregardlessof a requestby the firm.When
loan ratings were introduced,S&P provided them only at the request of the
borrower,whereas Moody's sometimes rated loans regardlessof whether a
borrowerrequestedthe rating (AmericanBanker, 1996). As I discuss below,
the most problematicidentificationissue is that borrowerschoose to obtain a
loan rating, and this choice may be correlatedwith increasesin unobservable
investmentdemand.Given thata ratingagency most likely has less information
on the unobservableinvestmentopportunitiesof the borrowerthanthe borrower
itself, it is beneficialin terms of identificationif ratingagencies assigned loan
ratings without the borrower'spermission. Regardless, all results presented
below are robustto the complete exclusion of loans ratedonly by Moody's.

1.2 Data and summarystatistics

1.2.1 Data. The main data set used in this article is drawn from S&P's
Compustat@;it includesthe universeof Compustatfirmsthatexist immediately
before and immediately after the introductionof bank loan ratings in 1995,
and the sample period covers firms from 1990 through 1998. I supplement
the Compustatdata with data on the universeof initial syndicatedbank loan
ratings;these data have been graciously providedby Moody's and S&P. The
final data set covers 3407 firmsfrom 1990 through1998, for a total of 25,538
3 Beginning in December 2003, S&P introducedrecoveryratings, which are on a scale of 1+ to 5, that measure
expected defaultrecovery.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification

firm-yearobservations.Over 15%of the firmsin the sampleobtaina loanrating

at some point afterthe inceptionof the programin 1995.
I examine firmsonly through1998 for two reasons.First,almost 75% of the
firmsthatobtaintheirinitialloan ratingbetween 1995 and2004 obtainthe rating
by 1998 (523 out of 746). In other words, the main impact of the loan rating
programswas realized shortly after their introduction.Second, the empirical
analysis rests on the assumptionthat loan ratingsrepresentan exogenous and
previouslyunavailableservice for firms.For firmsthat wait until 1999 or later
to obtain a loan rating, it is difficult to assert that loan ratings exogenously
provided a previously unavailable service. All core results are robust to an
alternativesample period of 1990 through1997 or 1990 through1999.
The rest of this section provides more detail on the sample construction.
Using Compustat,I begin with the universe of publicly traded nonfinancial
US businesses with strictly positive assets in existence between 1990 and
1998. I eliminatethe initial firm-yearaftera given firm'sinitial public offering
(proxied by the first year that item 25 and item 199 are both availablefor the
firm). I eliminate these firm-yearsgiven that one-year lagged values of key
financialitems for these observationsrepresentvalues for the firmswhile they
are privatelyheld. The resultingsample includes 7030 firms(39,046 firm-year
observations).I then requirethat key financial variablesare available for the
firm in 1994, 1995, and 1996. More specifically,I only include in the sample
firms that have Compustatitems 1, 8, 12, 25, 181, 199, 10, 35, 79, 13, 9, 34,
and 60 available for fiscal years 1994, 1995, and 1996. These data items are
necessary for the constructionof key financialstatistics.I focus only on firms
thatarein existence immediatelybefore andimmediatelyafterthe introduction
of loan ratingsto abstractfrom (potentiallyimportant)effects of loan ratings
on firms'decisions to go public or stay private.The resultingsampleafterthese
eliminations contains 3407 firms, and I include all firm-yearsfor these firms
with availabledatafrom 1990 to 1998 (25,538 firm-years).
I use the bank loan ratingdata providedby Moody's and S&P to determine
which firmsobtaineda bankloan ratingsubsequentto 1995 and when they first
obtainedthe rating.The only potentialmatchingvariablein the data provided
by Moody's and S&P is the borrowername. I use a name-matchingalgorithm
as a firstpass throughthe bankloan ratingdata, where the historicalnames of
Compustatfirms come from names from the CRSP database.I then manually
attemptto match all remainingborrowersin the bank loan rating datasets to
firms in the Compustatuniverse.Using this procedure,I am able to match approximately1200 borrowersto a Compustatfirm.Forthe sampleof Compustat
firmswith the restrictionsmentionedabove, I am able to match 803 borrowers
who obtaina bankloan ratingfrom Moody's or S&P or bothbetween 1995 and
2004.4 Out of these 803 borrowers,57 borrowersobtaineda loan rating after
4 There are almost 400 borrowerswho receive loan ratings that I am able to match to Compustatbut are not
included in the analysis. All of these borrowerswere not public in 1994, which is why they are not included.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReviewof FinancialStudies/

v22 n 4 2009

ceasing to be a publicly tradedcorporation.I investigatethese borrowers,and

find that the grandmajorityobtaineda loan ratingeither after being acquired
by a privateequity firm or after emerging as a privatecompany following a
bankruptcy.Finally,I augmentthe 1990-1998 Compustatsamplewith the loan
rating data, which leaves a sample of 523 firms that obtained a loan rating
between 1995 and 1998.
The data from Moody's and S&P also contain informationdescribing the
initial loan. More specifically,the datarecordthe amountof the total loan deal,
the numberof tranchesin the loan deal, and the loan rating itself. The loan
ratingscale is identicalto public bond ratings,with the best ratedloans earning
AAA (Aaa), and the worst being ratedCCC (Caa) or lower.
Core financialvariablesare calculatedfrom Compustatdataand are defined
as follows. The book debt-to-assetsratiois short-termdebt plus long-termdebt
(item 34 + item 9), all divided by total assets (item 6). The market-to-book
ratio is defined as total assets less the book value of equity plus the market
value of equity, all divided by total assets. The book value of equity is defined
as the book value of assets (item 6) less the book value of total liabilities
(item 181) and preferredstock (item 10) plus deferredtaxes (item 35). The
market value of equity is defined as common shares outstanding(item 25)
multipliedby shareprice (item 199). The market-to-bookratio also represents
Tobin's Q. A measure of asset tangibility is defined as tangible assets (item
8) divided by total assets. A measure of firm profitabilityis earnings before
interest, taxes, depreciation,and amortization(EBITDA) (item 13), divided
by total assets. Following the investment-cash flow literature(see Fazzari,
Hubbard,and Petersen, 1987; Kaplanand Zingales, 1997; and Rauh, 2006), I
define cash flow as income before extraordinaryitems plus depreciationand
amortization[(item 14 + item 18)/data6]. Finally,in the investmentregressions,
capitalexpenditures(item 128) and cash portionof acquisitions(item 129) are
scaled by lagged total assets. I referto the cash portionof acquisitionsas "cash
acquisitions"throughout.Investmentin workingcapitalis definedas the change
in net working capital assets [(item 4 - item 10 - (item 5 - item 34)] scaled
by lagged total assets. In orderto reducethe influence of outliers,I follow the
literatureand WinsorizeCompustatvariablesat the 1st and 99th percentile.
While the majorityof the dataanalysisemploys the Compustatdata,the analysis presentedin Section 4.2.2 employs data from Loan PricingCorporation's
(LPC's) Dealscan. In orderto find loans by firms in the sample, I begin with
the sample of 344 borrowerswho did not have a credit ratingas of 1994 and
subsequentlyobtained a loan ratingby 2004. For these borrowers,I examine
annual 10-K SEC filings in the fiscal year before and after the day on which
the borrowerin question obtaineda loan rating.I examine the 10-Ks to assess
whether I can identify, for a given firm, a new loan or an amendmentto an
The majorityof these borrowersreceivedtheirinitial bankloan ratingbefore going public (and thus, before they
appearin Compustat).I haveconductedthe empiricalanalysis includingthese 400 borrowerswith almostexactly


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification

existing loan that increased the amount in the same fiscal year that the loan
rating is obtained.An examinationof the 10-K filings reveals that 213 of the
334 borrowersobtainan identifiablenew loan or an increasein the existing loan
in the same fiscal year as the loan ratingis obtained.I then searchfor all loans
by these borrowersin Dealscan from 1990 up to and including the first rated
loan.5I am able to find loans for 161 of the 213 borrowers.I include only loans
to borrowerswho have Compustatdata available, and only loans made from
1990 to 1998. The final sample includes 508 deals for these 161 borrowers.

1.2.2 Issuer credit ratings and bank loan ratings. An importantdataclarificationissue concerns the difference between issuer credit ratings and
bankloan ratings,and how they arerelated.Compustatitem 280 representsthe
issuercreditratingof a borrower,which is S&P's opinionof the issuer'soverall
creditworthiness.S&P provides an issuer credit ratingfor every borrowerfor
which it ratesany security,regardlessof whetherthe securityis registeredwith
the SEC. As mentionedabove, when S&P assigns a new loan rating,they also
assign an issuer credit rating.Consistentwith this practice,almost 90% of the
firms in the sample that do not have an issuer credit ratingbefore they obtain
a loan ratingfrom S&P subsequentlyobtainan issuer creditratingin the same
year they obtainthe loan rating.In otherwords, item 280 becomes nonmissing
in the same year as a firm obtainsa loan rating.
Extant researchuses the existence of item 280 as a proxy for whetherthe
firmhas outstandingpublic debt (Faulkenderand Petersen,2006; Cantilloand
Wright, 2000). Before 1995, this proxy is accurategiven that S&P primarily
rated instrumentsthat were registeredwith the SEC. However, as the above
informationsuggests,afterthe introductionof loanratingsin 1995, the existence
of item 280 is no longer an accurateproxy for whetherthe firmhas outstanding
public debt. A firmthatobtainsa loan ratingalso obtainsan issuer creditrating
even if the firm does not issue public debt.6 After 1995, if a firm obtains an
issuer credit ratingwithout having obtaineda loan rating,it has likely issued
publicly registereddebt securitiesduringthe fiscal year.
Throughoutthis article,I follow the extantliteratureand classify firms with
an issuer creditratingbefore 1995 as firmswith outstandingpublic debt. After
1995, I classify a firm as having outstandingpublic debt if the firm obtains an
issuer creditratingwithouthaving obtaineda loan rating.
1.2.3 Summary statistics. Figure 1 plots the increasein the fractionof firms
thatobtaina bankloan ratingfrom 1995 through2004. The rise is mostdramatic
from 1995 through1998, when the fractionof firmsrises from0 to almost20%.
By the end of 2004, almost 30%of firmshave obtaineda bankloan rating.The
5 In unreportedrobustnessanalyses, I have found similarresultswhen isolating the sample to 1994 through 1998.
6 Examplesof firmsthatobtainan issuer creditrating(andloan rating)butdo not issue new publicdebt are Stryker
Corpin 1998, Arch Coal in 1998, and Kennametalin 1998.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Reviewof Financial Studies/ v 22 n 4 2009




1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
SAll firms

01Firmswithno issuercreditratingrating


Figure 2
Numberof firms that obtain their initial bank loan ratingby year and by whetherthe firm has an issuer credit
ratingas of 1994.

rise in the asset-weighted fraction of firms with a bank loan rating is even
more dramatic.From 1995 to 1998, the asset-weightedfractionof firms that
obtaineda bankloan ratingwent from 0 to 50%. By the end of 2004, over 70%
of the assets of public firmsin the sampleobtaineda bankloan rating.Figure 1
demonstratesthat bank loan ratingsare an importantelement in the corporate
financedecisions of a largenumberof US firms.In addition,the sharpincrease
in the numberof firmsobtainingloan ratingsshortlyaftertheirintroductionis
promisingin orderto identifythe causaleffect of loanratingson firmoutcomes.
Figure 2 plots by year the numberof firms that obtainedtheir initial bank
loan rating. It also splits firms by their issuer credit rating status as of 1994,
which is measuredusing whetherCompustatitem 280 is availablein 1994 for
the firm.The evidence suggests a sharprise in the numberof firmsadoptingthe
bank loan ratingtechnology in 1995 through 1997. By 1998 and 1999, fewer
firmsobtainedan initialbankloan rating,andeven fewer firmsobtaineda rating
each year after 1999. Like Figure 1, Figure 2 suggests that the main impact
of bank loan ratingsis realized in the years directly after their introductionin
1995. Figure 2 also suggests that both unratedand rated borrowersobtained
loan ratingsin almost equal proportion.
Table 1 presentssummarystatistics for the unbalancedpanel of 3407 firms
from 1990 to 1998 (25,538 firm-yearobservations).Overall, 15%of the firms
that are ever in the sample obtain a bank loan rating from Moody's, S&P,
or both by 1998. Slightly more firms obtained their initial bank loan rating
from Moody's than S&P (7.0% versus 5.4%). About 3% of firms obtained
initial bank loan ratingsfrom both Moody's and S&P in the same quarter.The
majorityof firmsobtainedtheirinitial bankloan ratingfrom eitherMoody's or
S&P, which underscoresthe importanceof using data from both of the rating


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification



Fractionthatobtain loan rating

Initialratingby Moody's
Initialratingby S&P
Initialratingby both
Unratedbefore 1995
Initialloan junk rated
Both unratedand initial loanjunk rated







Conditionalon obtainingloan rating

Amountof firstratedloan ($M)
Numberof tranchesof firstratedloan
Loan rating(1 = AAA, 2 = AA, etc.)
Amount of first ratedloan/totalassets




Full sample
Total assets ($M)
Sales ($M)
Firmage (years since IPO)
Book debt to assets
Market-to-bookratio (Q)
Tangibleassets to total assets
Cash flow/assetst




Asset growth
Cash acquisitions/assets,_I
Change in NWC/assets,_




This table contains summarystatistics for an unbalancedpanel of 3407 firms from 1990 through 1998 (25,538

agencies to determinethe quarterin which a firm first obtained a bank loan

rating. Of the firms that obtain a loan ratingat some point after 1994, almost
half have no issuer credit ratingin 1994. Over half of firmsthat obtaina loan
ratinghave their original loan ratedBB+ or worse ("junk").A key treatment
group in the empirical analysis is the set of firms that are both unratedin
1994 and subsequentlyobtain a loan that is rated BB+ or lower: these firms
representalmost a thirdof firms that obtain a loan rating.Initially ratedloans
in the sample are in aggregate$355 billion and are $55 billion for firms that
were unratedin 1994 and subsequentlyobtaineda junk-ratedloan.
Table 1 also presentssummarystatisticsfor the initialloans ratedby Moody's
and S&P. The average (median) loan is $678 million ($300 million) and represents 28% (22%) of the firm'stotal assets. The medianloan has a numerical
rating of 5, which translatesto BB (S&P) or Ba (Moody's). Overall, 52% of
the firms that receive a loan rating obtain an initial rating of BB+ or worse.
As a comparisongroup, only 35% of firms that have issuer credit ratings in
1994 have an issuer credit ratingof BB+ or worse. Finally, Table 1 presents


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReviewof Financial Studies/ v 22 n 4 2009

summarystatistics for Compustatvariables,which coincide with statistics in

other studies.

2. TheoreticalFramework
How shouldthe introductionof bankloanratingsaffectfirms?Inorderto answer
this question,I develop a theoreticalframeworkthat is motivatedby the model
of HolmstromandTirole(1997). In theirframework,therearethreeparticipants
in debt markets:firms, informedlenders, and uninformedlenders. There is a
moralhazardproblemat the firmlevel given the existence of a projectwith large
privatebenefits(the "bad"project).In the absenceof a monitorthatcan prevent
the firm from taking on the bad project, uninformedlenders will not lend to
the borrower.Informedlendersexert monitoringeffort, which enables them to
prevent firms from taking on the bad project. However, effort exerted by the
informedlender is costly and unobservable.The unobservabilityof informed
lendereffort is a crucialfeatureof the model. Given thateffort of the informed
lendercannotbe contractedupon, informedlendersare forced to commit their
own capital to the project. Only a lender with a stake in the firm's project
can credibly commit to exertingthe necessary monitoringeffort. Uninformed
lenders commit funds to firms only after an informed lender has made the
commitmentto monitorby investingin the project.Informedlender capitalis
thereforea crucialelement of the ability of firmsto invest. As Holmstromand
Tirole (1997, p. 669) note, "Moralhazardforces intermediariesto inject some
of theirown capitalinto firmsthatthey monitor,makingthe aggregateamount
of intermediary(or 'informed')capital ... one of the importantconstraintson
I informallyextend this frameworkto analyze how the introductionof loan
ratingsaffects borrowers.First,I assume thatthe duties of the informedlender
can be split into two different functions, both of which involve costs. The
first function is certification.The certificationfunction involves certifying the
reputationof the borrowerand evaluatingthe projectbeing undertaken.In the
context of a syndicatedloan, certificationalso involves evaluatingcollateral,
relaying the existing debt structureof the firm to potential participants,and
determiningthe interestrate and fees. Certificationcan take place both before
the loan is originatedand afterwardsthroughupdatinginformationon default
risk.The second functionis monitoring.The monitoringfunctioninvolvesmore
direct managementof the firm's project.Examples include deciding whether
to enforce or waive a violation of a financial covenant, seizing or changing
collateral,or disallowing continuationof projectsprogressingpoorly.
The second assumptionI makeis thatthe introductionof a third-partyratings
technology reduces the total cost of certificationborne by the borrower.In
other words, a loan rating by Moody's or S&P provides certificationfor the
borrowerat a cheapercost thancertificationby the lead arrangeralone. While
the empirical analysis does not attempt to discern why Moody's and S&P


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReal Effectsof Debt Certification

provide certificationat lower cost, I offer three reasons here.7First, consistent

with a regulatoryadvantage,many institutionalinvestors are restrictedeither
internallyor externallyfrominvestingin securitiesthatarenotratedby Moody's
or S&P.Second, Moody's and S&P specialize in certificationfunctions across
many different types of securities for many different firms. If learning by
doing is an importantelement of certification,Moody's and S&P will provide
this service at a lower cost than any given bank. Third, a lead arrangeron
a syndicatedloan has obvious incentive problems when relaying information
to participantlenders and secondarymarketinvestors.Lead arrangersreceive
a fee for originatingthe loan and typically want to hold as little of the loan
as possible, which leads to adverse selection problems.Of course, participant
lenders and secondarymarketinvestorsunderstandthese problems,and either
force the lead arrangerto hold a higher fraction of the loan or compensate
them with higher fees. Ultimately,the certificationcosts for the borrowerare
higher.A third-partyratingagency,which relies uniquelyon reputationcapital,
mitigatesthese agency problemsand certifiesthe borrowerat a lower cost.
The assumptionthat third-partyrating institutionsprovide certificationat
lower cost than the lead arrangeris consistent with the anecdotal evidence
that certain institutionalinvestors will not, under any circumstances,provide
financing for a syndicatedloan that is unrated.S&P (2005, p. 1) specifically
pitch their productas importantto investorsthat "needratings or do not have
large internalcredit staffs."Bavaria (2002) argues that nonbankinstitutional
investors "must have ratings"in order to participatein the syndicated loan
market.In addition,a largebody of empiricalevidence supportsthis assumption
by showing that credit ratings contain new informationthat affects market
prices (Hand,Holthausen,andLeftwich, 1992;MullineauxandYi, 2006; Tang,
A reductionin the cost of certificationhas a directeffect on real outcomes in
the Holmstromand Tirole (1997) framework.The total income that a borrower
can pledge is decreasingin the certificationcost. The more income that a borrowercan pledge to uninformedinvestors,the moreinvestmentit can undertake.
The intuitionis straightforward.Informedlendercapitalis necessaryto reduce
moral hazard,but expensive relative to uninformedcapital given monitoring
and certificationcosts. By relying on a third-partyratingagency that certifies
the borrowerat lower cost, the borrowerrelies less on informedlendercapital.9

7 See Boot, Milbourn,and Schmeits (2006) for theoreticalresearchon this subject.

8 Mullineauxand Yi (2006) find thatloan ratingdowngradesaffect stock prices, but they do not find any evidence
thatloan ratinginitiationsaffect stock prices. This is partlydue to the fact thatloan ratingsare typically initiated
simultaneouslywith changes in investmentpatternsor acquisitions, which makes it difficult to measure the
unique effect of the ratingitself on stock prices.
9 A third-partyrating agency cannot completely eliminate the need for informed capital given that it cannot
undertakemonitoringduties; that is, a third-partyratingagency can relay informationex post throughupdates
of the credit rating, but cannot dictate whether firms continue or discontinue a project. As Diamond (1991)
hypothesizes, only reputablefirmscan obtaindebt without any degree of monitoring.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReviewof Financial Studies/ v 22 n 4 2009

Whoobtainsa bankloanrating?

Totalassets ($M)
Sales ($M)
Firmage (years since IPO)
Book debt to assets
Tangibleassets to total assets

ratingand has
no public debt
(N = 234)

Does not obtain
loan ratingand
has no public debt
(N = 2495)

ratingand has
public debt
(N = 289)

Does not obtain
loan ratingand
has public debt
(N = 389)





This table presents 1994 mean characteristicsof firmsby whetherthey obtain a loan ratingsubsequentto 1995
and whetherthey have public debt as of 1994. Whethera firmhas public debt in 1994 is measuredby whether
the firm has an S&P issuer credit ratingas of 1994 (Compustatitem 280). * indicatesstatisticallydistinct from
column I at the 5% level.

This leads to a lower total certificationcost, allowing more capital from uninformed investorsand thus a higher investmentrate. Therefore,the framework
suggests that increasedborrowerinvestmentas a result of debt certificationis
financedprimarilyby less-informedlenders.

3. Who Obtainsa BankLoanRating?

In this section, I examine the ex ante characteristicsof the firms that obtain
bank loan ratingsbetween 1995 and 1998. This analysis serves two purposes.
First,it contributesto existing researchthatdescribesthe firmsthatobtaindebt
ratings. Second, it isolates the set of firms for which loan ratingsrepresenta
uniqueopportunityto obtainthird-partydebtcertification.It thereforeserves as
a preview of the identificationstrategyemployed in Section 4, where I attempt
to demonstratea causal effect of loan ratingson leverage and investment.
Table2 presentssummarystatisticsfor four groupsof firms,measuredas of
1994. Column 1 focuses on the key treatmentgroupin the empiricalanalysis:
the set of borrowerswho do not have public debt in 1994 and subsequently
obtain a loan rating between 1995 and 1998. These firms lie in the middle
of the distributionon almost every characteristic.They are larger,older, and
more levered than firms without public debt that do not obtain a loan rating
(Column2), andthey are smaller,younger,andless leveredthanfirmsthathave
public debt in 1994 (Columns3 and 4).
One of the key determinantsof the value of loan ratings is the question of
whether borrowerswho obtain loan ratings would have been able to access
public debt marketsif loan ratings were not introduced.Results in Table 2,
together with the findings of Faulkenderand Petersen (2006), suggest that
borrowerswithout public debt in 1994 that obtain loan ratingsbetween 1995
and 1998 would have been less likely to be able to access public debt markets


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification

Table 3
Comparing firms in year before obtaining loan rating to firms in year before obtaining public debt

Totalassets ($M)
Total sales ($M)
Firmage (years since IPO)

Yearbefore obtaining
loan rating
(N = 193)

Yearbefore initially
obtainingpublic debt
priorto 1995 (N = 188)

Yearbefore initially
obtainingpublic debt
1995 or after (N = 181)




This table comparesthe size and age of firmsin the year beforeobtaininga loan rating(conditionalon not having
public debt) and the year before obtainingpublic debt. Before 1995, the year a firm initially obtains public debt
is the year it obtains an issuer creditrating.For 1995 and after,the year a firminitially obtainspublic debt is the
year it obtains an issuer credit rating without having previously or concurrentlyobtained a loan rating. Assets
and sales are measuredin 2000 dollars. * indicatesstatisticallydistinctfrom column I at the 5% level.

in the absence of loan ratings.In particular,Faulkenderand Petersen suggest

thatsize and firmage are two criticaldeterminantsof whetherfirmsare able to
access public debt markets.Size is critical because there are large fixed costs
associated with bond issues relative to bank debt, and bond amountsmust be
sufficientlylargefor marketliquidity.Firmage is criticalgiven the necessity of
informationin the market,and the fact thatthe marketknows more aboutfirms
with longer track records. The statistics in Table 2 demonstratethat unrated
firms that obtain loan ratings are significantly smaller and younger than the
averagefirm with access to public debt markets.
Table 3 presentsa sharpertest of the hypothesis that firms that obtain loan
ratingswould not have been able to access public debt marketsif loan ratings
were not introduced.It comparesthe size (totalassets andtotal sales) andage of
firmsin the year before they obtaina loan ratingto the size and age of firmsin
the year before they firstobtainan issuer creditrating.As discussedearlier,for
firms that obtain an issuer credit ratingbefore 1995, the first year they obtain
the ratingproxies well for the firstyear they access public debt markets.After
1995, the firstyearin which a firmobtainsan issuercreditratingwithouthaving
obtained a loan ratingproxies well for the first year the firm accesses public
debt markets.As Table3 shows, firmsin the yearpriorto accessing public debt
marketsare larger and older than firms in the year prior to obtaining a loan
rating.This is trueboth for firmsthatinitially access publicdebtmarketsbefore
1995 (Column 1 versusColumn2) andfor firmsthatinitiallyaccess publicdebt
marketswithout obtaininga loan ratingafter 1995 (Column 1 versus Column
3). The differences in total assets and total sales are statisticallysignificantat
the 1 to 15% level, although the differences in firm age are not statistically
significantat a reasonableconfidencelevel.
The results in Tables 2 and 3 suggest that the introductionof loan ratings
offereda uniqueopportunityfor youngerandsmallerfirmsto obtainthird-party
debt certification.Based on theirsize and age, these firmswould have been less
likely to access public debt marketsin the absenceof loan ratings.The analysis
in Section 4 assesses whetherthis unique opportunityenabled firms to access


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Reviewof Financial Studies/ v 22 n 4 2009

the capital of less-informed investors, and whether these firms subsequently

experiencedchanges in leverageand investment.
An additionalquestion raised by Table 2 is why the smallest and youngest
firmsare unlikelyto obtaina loan rating.Therearethreepossible explanations.
First, there is a large fixed cost component to the pricing of a loan rating,
which suggests that the loan ratings are relatively expensive for firms with
smaller financing needs. Second, the frameworkby Holmstrom and Tirole
(1997) suggests thatborrowerswith small investmentprojectsrely on informed
capital alone. According to this explanation,smaller borrowersdo not have
investmentprojectsthatarelargeenoughto requireuninformedcapital.Finally,
Dennis and Mullineaux (2000), Lee and Mullineaux (2004), and Sufi (2007)
suggest that lead arrangersmust hold large percentagesof syndicated loans
when borrowersare informational-opaque.These findings suggest that very
young and small firms may be unable to access the capital of uninformed
lenders, even with third-partydebt certification.Distinguishing among these
three explanationsis materialfor futureresearch.

4. BankLoanRatings,Leverage,and Investment
4.1 Empiricalmethodology
The theoretical frameworkhypothesizes that the introductionof bank loan
ratingsreducesthe cost of debt certificationand increasesthe pool of investors
willing to provide financingto borrowers.As a result, firms are able to raise
additionaldebt financingand increaseinvestment.The empiricalmethodology
in this section seeks to identify the causal impact of bank loan ratingson the
use of debt financingand on investment.
In order to describe the empirical methodology,I begin by describing the
ideal experimentalenvironment.The ideal experimentalenvironmentinvolves
randomassignmentof bank loan ratingsto firms.If loan ratingsare randomly
assigned to firms,then the specificationthatestimatesthe causal effect of bank
loan ratingson outcome y for a firmi is a basic linearregression:
yi = ot+


LoanRatei+ Ei.


With random assignment, the coefficient estimate of P representsthe causal

effect of the loan ratingon outcome y. Controlvariablesare not necessary,and
the standardorthogonalityconditionis satisfied:E[EiJLoanRatei]= 0.
In the empirical setting of the 1990s, bank loan ratings are not randomly
assigned. Instead,firmschoose whetheror not to obtaina loan rating.The fact
that loan ratings are unavailablebefore 1995 is an advantage,but firms still
choose whetheror not to obtaina bankloan ratingsubsequentto 1995. It is this
choice thatpresentsan identificationproblem.If the same factorsleading firms
to choose a bankloan ratingare also influencingoutcomes,then the coefficient
estimate of P does not capturea causal impact of the loan rating on the firm


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReal Effectsof Debt Certification

outcome. Perhapsthe most worrisome omitted variables are those related to

investmentdemand.If investmentdemandat a given firmincreasesandthe firm
incidentallyobtainsa loan rating,then the firmmay have increasedinvestment
even if bank loan ratingsdid not exist.
There are three elements of the identificationstrategythat mitigateomitted
variableconcerns.First,the empiricalmethodologyrelies on firm fixed effects
regressions and exploits the fact that bank loan ratingsare unavailablefor all
firmsbefore 1995. More specifically,for a sample in which firm i is observed
annuallyat t, I estimate




SICi + Ec i + P x LoanRateit+ 8'Xit + Eit.



The fixed effects specificationexploits within-firmvariationin the existence

of a bank loan rating, where firms are unable to obtain a ratingbefore 1995.
In this specification, Brepresentsthe effect on y of a given firm obtaining a
loan ratingrelativeto the periodwhen the same firmdid not have a loan rating.
The fixed effects specificationremoves all firm-specifictime-invariantomitted
Althougha fixed effects specificationeliminatesfirm-specifictime-invariant
omittedvariables,a remainingconcernis thattime-varyingwithin-firmomitted
variables are simultaneouslyleading firms to choose to obtain a bank loan
ratinganddrivingoutcomes.Forexample,if a given firmreceives a technology
shock in a given year and I cannot properlymeasurethe shock, then the firm
may expandproductionand simultaneouslyobtaina loan rating.This problem
can be viewed in terms of the counterfactual:Would leverage and investment
have increased for a given firm in the year it obtains a loan rating even if
loan ratingsdid not exist? The second and thirdelements of the identification
strategyhelp assure that increases in outcomes are the result of loan ratings.
The second element is the inclusionof a comprehensiveset of controlvariables
that partial out the effect of firm demand for investment or debt financing.
When the outcome of interestis leverageratios,I include in the matrixXit the
market-to-bookratio, asset tangibility,firm size, firmcash flow, and measures
of the firm's advertisingand researchand developmentexpenses.10When the
outcome of interestis investment,I include in the matrixXi, cash flow, lagged
cash flow, and Tobin's Q. In addition, in all specifications, I include SIC
two-digit industryby year indicatorvariables.These interactiontermscapture
industry-wideinvestmentdemandshifts over time. Given the presenceof these
interactionterms,the coefficientestimateof Prepresentsthe changein outcome
y for a given firm i that obtains a loan ratingin year t, relativeto the average
change of all firmsin the same two-digit industryin the same year.

In unreportedresults,I follow Faulkenderand Petersen(2006) andinclude a measureof asset volatilityand stock

returnscomputedusing marketdata.The resultsare almost identicalwith the inclusion of the two variables.I do
not include these variables,because they are missing for almost 10%of observations.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Reviewof Financial Studies/ v 22 n 4 2009

Finally, I rely on the cross-section characteristicsof firms that obtain loan

ratingsto show that the results appearuniquelyconsistent with the theoretical
frameworkoutlinedin Section 2. The idea is straightforward:
among the firms
of Moody's
or S&P more than others.
the theoreticalframeworkpredictsthatthe effect of a loan ratingon investment
and leverage should be lower for firms that have an existing issuer credit
rating.Consistentwith this prediction,the results in Section 3 suggest that the
introductionof loan ratings in 1995 offered a unique opportunityfor unrated
borrowersto access the capitalof uninformedinvestors.Second, firmsof lower
credit quality would have a more difficult time obtainingdebt financingfrom
uninformedlenders in the absence of a third-partyratings institution. This
follows directly from the theoretical frameworkoutlined in Section 2: the
aggregate amount of lead arrangercapital places a strict restrictionon total
available financing when the returnsfrom certification and monitoring are
highest. Given thatcertificationand monitoringare especially importantwhen
there is a high probabilityof default, lower credit quality borrowersshould
benefit more from certificationby Moody's or S&P.I use the initial loan rating
as a measureof creditquality.'"

4.2 Bankloan ratingsand debt financing

4.2.1 Loan ratings and debt. Table4 presentscoefficient estimates where
outcomes are measuresof debt used by the firm. For firms that obtain a loan
rating,the sample for the regressionsin Table4 includes only the observations
up to and including the year that the loan ratingis obtained.(The effects of a
loan ratingafterthe yearof initialratingareexploredin Table9.) The coefficient
estimatesin Column I show thatfirmsthatobtaina bankloan ratingexperience
statisticallysignificantincreasesin net debt issuance. In termsof magnitude,a
firm with a credit ratingin 1994 experiences an increase in net debt issuance
of 0.046 and a firm without a credit ratingin 1994 experiencesan increase of
0.205 (0.046 + 0.159). The specificationreportedin Column2 includescontrol
variablesthat Rajanand Zingales (1995) and Faulkenderand Petersen(2006)
show are correlatedwith the use of debt by firms. The coefficient estimates
when the controlvariablesare includedare almost identical.Column3 reports
a specificationin which the bankloan indicatorvariablesareinteractedwith an
"initialloan ratedjunk"indicatorvariable.Firmswith an existing creditrating
thathave a junk ratingon theirinitial loan experienceno distinctchange in net
debt issuancefromotherfirmswith an existing creditrating(Row 2). However,
among the set of firms that are unratedbefore the introductionof bank loan
ratings, firms that are of lower credit quality experience a larger increase in
net debt issuance. In other words, among firms that are both unratedand of

also use an implied junk ratingbased on lagged firm characteristics(firm size and firm EBITDA scaled by
assets) with qualitativelysimilarresults(unreported).


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Eafeltso/Debt Certification

Table 4
The effect of obtaining a bank loan rating on availability of debt
(A debt/laggedassets)i,
Loan ratingi,





Loan ratingit,*Unratedi.
junk ratedi,
Tangibleassets to total assetsi,t_-I
Numberof firms
Numberof firm-years


Debt to assets ratioi,








Loan ratingi,*Loanjunk ratedit,

Loan ratingi,*Unratedi,1994









This table presentscoefficient estimatesfrom firmfixed effects regressionsrelatingdebt at firmi in year t to the
existence of a bankloan ratingfor firmi in year t. The dependentvariablein columns 1 through3 is the change
in debt levels from t - 1 to t scaled by the book value of assets at t - 1, and the dependentvariablein columns
4 through 6 is the book debt to assets ratio. All regressions include year indicator variables interactedwith
two-digit SIC industrycodes. Standarderrorsare clusteredat the firmlevel. * Coefficientestimateis statistically
distinct from 0 at the 5 and 1%levels, respectively.

lower creditquality,thereis a largerincreasein net debt issuancewhen the firm

obtains a loan rating.
Columns 4 through6 of Table4 presentthe estimates where the dependent
variableis the book debt to book assets ratio.12Firmsthat are ratedbefore the
introductionof bankloan ratingsexperienceno changein leverageratios when
obtaininga loan rating, whereasfirms thatare unratedbefore the introduction
of bank loan ratingsexperience a statisticallysignificantincrease. In terms of
magnitudes,the coefficient estimate in Column 4 implies an increase of 0.12
in the leverage ratio, which is 41% (0.12/0.29) among these firms. Column 6
shows evidence consistent with Column 3: firms that are unratedbefore the
introductionof loan ratingsand of lower credit quality experience the largest
increasein leverageratioswhen they obtaina loan rating.
The results in Table4 suggest that firms are able to raise significantlymore
debt in the year they obtain a loan rating relative to previous years. This is
consistent with the theoreticalframeworkin Section 3, which suggests that a
12 I also replicate all
leverage specificationsreplacing the book debt to book assets ratio with the book debt to

marketvalue of assets ratio.The results, which are unreported,are qualitativelysimilar.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Reviewof Financial Studies/ v,22 n 4 2009


Amount of loan deal ($M)

Numberof trancheson loan deal
Numberof term loan tranches
Interestrate spread(bp + LIBOR)
Numberof lenderson deal
Domestic banks
Fractionof uninformedlenders







This table presents summarystatistics on loan deals obtained by borrowerswho do not have an issuer credit
ratingas of 1994, and subsequentlyobtaina loan ratingin 1995 or after.The sample includesonly borrowersfor
whom the first ratedloan is ratedat origination,only borrowerswhom I am able to matchto LPC's Dealscan,
and only borrowersfor whom lagged book debt, marketto book ratio,EBITDA,tangibleassets, and total assets
from Compustatare nonmissing.The sample includes 508 loan deals by 161 borrowers,and the loan deals are
originatedbetween 1990 and 1998. A domestic bank is any domestic commercialbank, finance company,or

reductionin certificationcosts provides greateraccess to external financing.

The fact that the result is uniquely strongamong firms that are unratedbefore
the introductionof bank loan ratingsand are of lower credit quality mitigates
concerns with omitted variables, such as borrowerquality or demand shifts.
Overall,these resultssuggest thatloan ratingshave importanteffects on capital

4.2.2 Loanratings,loan characteristics,and syndicatecomposition. The

section abovedemonstratesthatborrowerswho obtainloan ratingssignificantly
increasetheiruse of debt.In this section, I documentthe channelthroughwhich
the expansionof debt financingoccurs. Table5 presentssummarystatisticsfor
the sample of 508 loan deals by 161 borrowers.These are borrowerswho are
unratedin 1994 and subsequentlyobtaina loan rating,and the sample includes
all loans by these firms from 1990 through 1998. The average(median) loan
amountis $212 million ($109 million), and the average(median) interestrate
spreadis 161 (134) basis points above LIBOR.
In termsof the syndicatesize and composition,thereis an average(median)
of 6.5 (4) lenders providingfunds on deals. I classify each lender into one of
five groups: domestic commercial banks, foreign commercial banks, investment banks,financecompanies,and nonbankinstitutionalinvestors.Domestic
commercial banks, investmentbanks, and finance companies are combined
into one grouplabeled"domesticbanks."These banksareconsideredinformed
lenders-lenders who have specialized screeningand monitoringskills. A few
foreignbankswith a largepresencein the US throughsubsidiariesareclassified
as domesticbanks(e.g., ABN AMRO).Giventhatthe classificationis notbased
on SIC codes (which are unavailablein Dealscan), the Appendix provides a
sample of financialinstitutionsand the groupsin which they are classified.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification

As Table 5 shows, domestic banks are the most common type of lenders in
this market,with an average of 4.0 domestic banks in the syndicate.Foreign
banksrepresenton average2.3 lendersin syndicates,and nonbankinstitutional
investors represent only 0.25 lenders. These averages demonstratethat the
syndicatedloan marketis primarilya marketin which banks with specialized
monitoringand screeningskills participate.
The results presentedin Table 6 describehow a firm's firstratedloan compares to its previous unratedloans. Each cell in Table 6 reportsa coefficient
estimate from a separateregression;more specifically, each cell provides the
coefficient estimate on an indicatorvariablethat is equal to one if the loan is
rated. Column 1 reportsfirm fixed effects coefficient estimates, and Column
2 reports first difference estimates. All specifications include year indicator
variablesand control for the firm's lagged book debt ratio, EBITDA to assets
ratio, tangible assets ratio, the market-to-bookratio, and the naturallogarithm
of total assets. As the first row demonstrates,there is a large increase in the
loan amountwhen the borrowerobtainsa loan rating.The fixed effects estimate
implies a 70% increasein the size of the loan, and the first differenceestimate
implies a 47% increase. This finding is consistent with evidence presentedin
the previous section that borrowersexperience large increases in debt in the
year they obtain a loan rating.The increase in the size of the loan is drivenin
partby an increasein the numberof loan trancheson a given deal. In particular,
there is a significantincrease in the presence of term loan trancheson a firm's
first rated loan. Although the loan size increases by almost 70%, the interest
rate spreaddoes not show a statisticallysignificantchange.
Rows 5 through11 of Table6 provideevidence thatthe syndicatecomposition shifts towardless-informedlenderswhen firmsobtainloan ratings.In row
5, the coefficient estimate of the effect of a ratingon the numberof lenders is
5.8 in the fixed effects specification,and4.5 in the firstdifferencespecification.
Moreover,rows 7 and 8 show that between 50 and 60% of the increasein the
numberof lenders representsincreases in foreign bank and nonbankinstitutional investorparticipation.Row 9 shows thatthereis a statisticallysignificant
increase in the fractionof uninformedlenders in the syndicate,where the first
differencecoefficient estimateis significantat the 10%level. The fixed effects
coefficient estimates in rows 10 and 11 imply that a foreign bank or nonbank
institutionalinvestors is 26 and 16% more likely to be on the first rated loan
syndicate,relativeto previousloans by the same firm.
The evidence in Table 6 suggests that loan ratings increase the pool of
investors that are willing to hold part of a syndicatedloan. In particular,the
results suggest thatborrowersgain increasedaccess to less-informedinvestors,
such as foreign banks and nonbankinstitutionalinvestors.13
13 Table 5 also implies a simultaneousincrease in the numberof term loan tranchesand
the numberof nonbank

institutionalinvestors when a loan is rated. This evidence supportsthe model by Kashyap, Rajan, and Stein
(2002), which hypothesizes that commercial banks funded by deposits will specialize in lending via lines of


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReviewof Financial Studies/ v 22 n 4 2009

Table 6
The effect of a loan rating on loan characteristics and syndicate composition




Numberof trancheson loan deal


Numberof term loan tranches


Interestratespread(bp + LIBOR)


Numberof lenderson deal


Numberof domestic banks


Numberof foreign banks


Numberof institutionalinvestors


Fractionof uninformedlenders


Indicatorfor foreign bank


Indicatorfor institutionalinvestor

Fixed effects
Loan rated

Loan rated

This table presentsregressioncoefficients from regressionsrelatingthe characteristicsof a borrower'sloans to

whetherthe borrower'sloan is rated.Each cell representsa separateregression,with each row listing a distinct
dependent variable. The first column presents the fixed effects coefficient estimates, and the second column
presentsthe firstdifferencecoefficientestimates.The coefficienton the "loanrated"indicatorvariablerepresents
the effect of the loan rating on the first rated loan relative to unratedloans by the same firm. Standarderrors
are calculated by clustering at the firm level. The sample includes all loans by borrowersfrom 1990 up to
and including borrower'sfirst rated loan. All regressionsinclude year indicatorvariables,and controls for the
borrowers'lagged book debt ratio,EBITDAto assets, asset tangibility,naturallogarithmof total assets, and the
market-to-bookratio.The samplesize for each regressionis 508 loan deals by 161 borrowers.* and**coefficient
estimate is statisticallydistinctfrom 0 at the 5 and 1%levels, respectively.

4.3 Bankloan ratingsand investment

The results presentedin Tables 4 through6 suggest that loan ratings increase
the availability and use of debt financing. Table 7 examines whether firms
increase investmentas a result. More specifically, it examines the impact of
the availability of bank loan ratings on asset growth, capital expenditures,
cash acquisitions,and workingcapitalinvestment.The coefficientestimatesin
Column 1 show that firms with an issuer credit ratingbefore the introduction
of loan ratings experience an increase in asset growth of 0.078 when they
obtaina loan rating.Unratedfirmsexperiencea statisticallysignificantlylarger
increase in asset growthof 0.274 (0.078 + 0.196). Coefficientestimates from
regressionsusing capital expenditures,cash acquisitions,and workingcapital
investmentare presentedin columns 4, 7, and 10, respectively.The coefficient
estimateson bankloan ratingssuggest a similarpatternfor capitalexpenditures
credit.Afterthefirmobtainsa loanratingandnonbank
arewillingto lend,theyusuallyparticipate
in termtranches.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReal Effectsof Debt Certification

Theeffectof obtaininga bankloanratingon realoutcomes
Panel A

Asset growthit








Loan ratingit,













Loan ratingi,*Loanjunk ratedi,

Loan ratingit*Unratedi,1994
Loan ratingi,*Unratedi,
Loanjunk ratedit,
Cash flowit,/Totalassetsi,t_l
(Book debt/Totalassets)i,t-1

Numberof firms
Numberof firm-years



Panel B



Loan ratingit
Loan ratingi,*Loanjunk ratedi,
Loan ratingi,*Unratedi,
Loan ratingit*Unratedi,1994*
Loanjunk ratedi,



Cash flowit/totalassetsi,tl

(Book debt/totalassets)i,,_
Numberof firm-years










A Net workingcapitalit/Assetsi,,-1









This table presents coefficient estimates from firm fixed effects regressions relating real outcomes at firm i
in year t to the existence of a loan rating for firm i in year t. Columns I through 3 relate asset growth at
time t ([A, -A, _-1]/At-1), columns 4 through 6 relate capital expendituresat t scaled by assets at t- 1,
columns 7 through9 relate cash acquisitions at t scaled by assets at t - 1, and columns 10 through 12 relate
the change in net working capital at time t scaled by assets at t - 1 to the existence of a loan rating at t. All
regressions include year indicatorvariablesinteractedwith two-digit SIC industrycodes; standarderrorsare
clusteredat the firmlevel. *,**indicatecoefficient estimateis statisticallydistinctfrom 0 at the 5 and 1%levels,

and workingcapitalinvestment,butthe coefficients arenot statisticallydistinct

from0 at a reasonableconfidenceinterval.Unratedfirmsthatobtaina bankloan
rating experience an increase in cash acquisitionsof 0.081 (0.014 + 0.067),
which is higherthanthe mean of cash acquisitionsamong this sample (0.067).


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReviewof Financial Studies/ v 22 n 4 2009

The effect of obtaining a bank loan rating is significantlystronger,both in a

statisticalsense andin termsof economic significance,amongfirmsthatdo not
have a creditratingbefore bankloan ratingsbecame available.
Columns 2, 5, 8, and 11 of Table 7 present coefficient estimates from regressions that include a variety of investmentopportunitycontrol variables.
Although these variables are statistically powerful predictorsof investment,
they do not significantlychange the coefficientestimateson bankloan ratings;
in fact, the coefficientestimatesare almostidentical.These results suggest that
unobservableinvestmentopportunitiesarenot drivingthe results-the fact that
coefficient estimateson bank loan ratingsdo not change when adding observable measures of investmentopportunitiessuggests that loan ratings are not
correlatedwith unobservableinvestmentopportunities(assuming, of course,
that observable and unobservablemeasures of investment opportunitiesare
correlated).In addition,all regressionsinclude year indicatorvariablesinteracted with two-digit SIC industrycodes. These interactionterms control for
any technology or demand shifts that occur over time in the firm's two-digit
Columns 3, 6, 9, and 12 of Table 7 examine whether the positive effect
of the loan ratingson real outcomes is differentiallystrongeramong firms of
lower credit quality.The results suggest thatthe increasein asset growth,cash
acquisitions, and working capital investmentis largerfor firms that are both
unratedbefore 1995 and subsequentlyobtaina loan ratingof BB+ or worse.
Consistentwith the theoreticalframeworkdiscussed in Section 3, the results
in Table 7 suggest that the introductionof bank loan ratings increases outcomes for borrowerswho obtainthem. The evidence suggests that third-party
certificationreduces the cost of obtainingfinance from uninformedinvestors,
and it allows firms to raise more capital and grow. While these results appear
consistent with an expansionin the availabilityof debt financing,the fact that
the results are strongestamong unratedlow-credit-qualityborrowersappears
inconsistentwith most omittedvariablestories.For example,it is unlikely that
low-credit-qualityfirmshave betterunobservableinvestmentopportunities.

4.4 Robustnessand extensions

4.4.1 Obtainingpublic debt or an unrated loan. The resultspresented
above suggest that loan ratings increasethe availabilityof debt financingand
also real outcomes, such as asset growthand cash acquisitions.The empirical
methodology contains a numberof techniquesthat mitigate omitted variable
bias concern, and help strengthenthe interpretationthat the introductionof
loan ratings has a real effect. In this section, I conduct two counterfactual
analyses with the goal of answeringthe following question:In the absence of
loan ratings, would firms that obtainedloan ratings still have experiencedthe
same increasesin leverage,asset growth,and cash acquisitions?
PanelA of Table8 examines net debt issuance,leverageratios,asset growth,
and cash acquisitions,and comparesthe coefficient estimates on loan ratings


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification





(A Debt/
Laggedassets)i, assetsratioi,
PanelA: Obtainingloanratingversusobtainingcreditratingwithoutloanrating
(a) Loan ratingit

(b) Loan ratingit*Unratedi,1994
(c) Creditrating,without loan ratingi,
Difference:(a) + (b) - (c)
Panel B: Obtaining loan rating versus obtaining unrated loan
(a) Loan ratingit,
(b) Loan ratingi,*Unratedi,1994
(c) Unratedloani,
(d) Unratedloanit*Unratedit
Difference:(a) + (b) - (c) - (d)


This table presents two counterfactualanalyses. Panel A reportscoefficient estimates from firm fixed effects
regressions relating outcomes to an indicatorfor whether the firm obtains a loan rating and an indicatorfor
whetherthe firminitially obtainsa creditratingwithouta loan rating.Panel B reportscoefficient estimatesfrom
firm fixed effects regressionsrelatingoutcomes to an indicatorfor whetherthe firmobtains a loan ratingand an
indicatorfor whetherthe firm obtains an unratedloan. The regressionspecificationreportedin column 1 also
includes all control variablesin column 2 of Table4, and the regressionspecificationsreportedin columns 2-4
include all control variablesin column 2 of Table 7. All regressionsinclude year indicatorvariablesinteracted
with two-digit SIC industrycodes; standarderrorsare clustered at the firm level. *,** indicate that coefficient
estimate is statistically distinct from 0 at the 5 and 1% levels, respectively.+,++ indicate that difference in
coefficient estimates is statisticallydistinctfrom 0 at the 5 and 1%levels, respectively.

to the coefficient estimate on whether a firm initially obtains an issuer credit

rating. The underlyingassumptionof this analysis is that a reasonablyclose
counterfactualto firmsthat initially obtaina loan ratingare firmsthat initially
obtain an issuer creditratingwithoutobtaininga loan rating.These results are
comparableto the resultsin FaulkenderandPetersen(2006), whereleverageis
regressedon an indicatorof whethera borrowerhas an issuercreditrating.The
specificationsin columns 1 and 2 include all controlvariableslisted in Table4,
and the specificationsin columns 3 and 4 include all controlvariableslisted in
The resultssuggest a significantlylargerincreasein net debt issuance, leverage, asset growth, and cash acquisitions for unratedfirms that obtain a loan
ratingrelative to unratedfirms that obtain an issuer credit rating without obtaining a loan rating.These results togetherwith the results in Tables 2 and 3
are consistent with the following two claims. First, the results in Tables 2 and
3 suggest that unratedfirmsthat obtain loan ratingswould not have been able
to access public debt marketsin the absence of the loan rating.Second, even if
they would have been able to access public debt markets,the resultsin Table8
suggest that the increase in outcomes would have been lower if they obtained
public debt without obtaininga loan rating.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Reviewof Financial Studies/ v 22 n 4 2009

The estimates reportedin Panel A of Table 8 provide a more conservative

magnitude assessment than do the results in Tables 4 and 7. The estimates
suggest thatfirmsincreasetheirleverageby 0.07 more, or 22%evaluatedat the
mean, when obtaininga loan ratingrelativeto obtainingan issuer creditrating
without a loan rating.A firm increases asset growthand cash acquisitionsby
41 and 80% more than it would have had it obtained an issuer credit rating
withouta loan rating,respectively.
Panel B of Table8 conductsa similarcounterfactualexercise that examines
firms that obtain unratedloans. I measurewhether a firm obtains an unrated
loan in a given year using LPC's Dealscan. Although Dealscan omits smaller
loans andloans to smallerfirms,this samplebias shouldbias coefficientson the
unratedloan indicatorvariableupwards,as only the largest loans for a given
borrowerare likely to make it into the dataset. In otherwords, any sample bias
will tend to cause the coefficient estimate on the unratedloan indicatorto be
higher, which would make it more difficult to show that rated loans uniquely
lead to increases in outcomes. The evidence in Panel B suggests that unrated
firms that obtain a rated loan experience significantly larger increases in all
outcomes relative to unratedfirms that obtain unratedloans. The differences
in Panel B are even larger than in Panel A: for example, unratedfirms that
obtain a rated loan in a given year experience asset growth that is 0.22 more
thando unratedfirmsthatobtainan unratedloan in a given year.It is important
to emphasize that all specificationsin Table 8 include firm fixed effects. The
coefficientestimatesthereforeimplythatthe same firmexperiencesmuchlarger
increases in outcomes when obtaininga ratedversus an unratedloan.

4.4.2 Advancementsin the loan salesmarket. An additional
the empirical strategyis that advancementsin the secondaryloan marketoccurredsimultaneouslywith the introductionof loan ratings.For example, the
Loan Syndicationand TradingAssociation (LSTA) was founded in 1995 and
certaintradingconventionswere introducedin 1996. The worryis thatadvancements in the secondaryloan marketare responsiblefor the trendsdocumented
above, as opposed to the introductionof loan ratings.
The evidence does not supportthis view. In fact, the major impact of the
introductionof loan ratings was complete before the secondaryloan market
reachedsignificantvolumes. Accordingto the LPC, the secondaryloan market
had aggregatevolume of only $80 billion in 1998. In contrast,the aggregate
amountof all initiallyratedloans (bothin andnot in my finalsample)from 1995
through1998 was $490 billion, which reflectsthe rapidadoptionof loan ratings
after their introductionin 1995. Even if one assumes that all the loans traded
on the secondarymarketwere initially ratedloans, only 16%of initially rated
loan amountswere tradedon the secondarymarketin 1998. Consistent with
this growth pattern,althoughthe LSTA was founded in 1995, its coverage of
secondaryloan prices was limiteduntil well afterthe introductionof bankloan
ratings.For example, Gupta,Singh, and Zebedee (2006) reportthatonly about


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification

Financialand realoutcomesafterbankloanratingis obtained
Debt to assets ratioit Asset growthit Acqit/Assetsi,t-_

PanelA: Short-runpersistence(1990-1998)
Bankloanratingobtainedin thisyeari,
After bank loan ratingobtainedi,
Bank loan ratingobtainedin this yeari,*Unratedi,1994
After bank loan ratingobtainedi,*Unratedi,1994







PanelB: Long-runpersistence(1990-2004)
Bank loan ratingobtainedin this yearit
After bank loan ratingobtainedi,
Bank loan ratingobtainedin this yeari,*Unratedi,
After bank loan ratingobtainedi,*Unratedi,

0. 108**

This table presentscoefficient estimates from firm fixed effects regressionsrelatingfinancialand real outcomes
at firm iin year t to firm iobtaining a bank loan ratings in year t, and after firm iobtains a bank loan rating.
Panel A includes observationsup to and including 1998, and Panel B includes observationsup to and including
2004. The regressionspecificationreportedin column I also includes all control variablesin column 2 of Table
4, and the regression specifications in columns 2 and 3 include all control variables in column 2 of Table 7.
All regressionsincludeyear indicatorvariablesinteractedwith two-digit SIC industrycodes; standarderrorsare
clustered at the firm level. *,** indicate that coefficient estimate is statisticallydistinct from 0 at the 5 and 1%
levels, respectively.

100 loans were coveredin the LSTAmaindatabasebeforeNovember 1999 (see

also Moerman,2005). The timing of the introductionof loan ratings and the
secondaryloan marketexpansionsuggests thatthe introductionof loan ratings
preceded the developmentof the secondaryloan market,and involved a large
numberof loans that were not traded.Even after 1998, Moerman(2006) finds
a strong positive effect of a loan being rated at originationon the probability
that a loan is tradedin the secondarymarket.This suggests that loan ratings
facilitate secondarymarkettrading,not vice versa.
4.4.3 Persistence. How persistentare the increasesin leverage and real outcomes for borrowerswho obtaina loan rating?Table9 documentshow leverage
and real outcomes change after the firm obtains the initial loan rating. Panel
A uses the sample only through 1998, whereas Panel B examines long-run
persistence by extending the analysis through2004. In both panels, the third
row of coefficients shows that the core results of Tables 4 and 7 are robust
when including the years after a borrowerinitially obtains the loan rating.
Unratedfirms that obtain a loan ratingexperience increases in leverage, asset
growth,and cash acquisitionsin the year they obtainthe bankloan rating.The
fourthrow containsthe key variableof interest;the coefficientestimateson the


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReviewof Financial Studies/ v 22 n 4 2009

After bank loan rating obtained x Unrated variablequantifythe changes in

leverage and real outcomes in the years after the loan ratingis obtained.The
short-runandlong-runresultssuggest thatthe increasein leverageis persistent
but decreasesslightly; these results suggest thatloan ratingslead to permanent
increasesin leverageratios.Columns2 and3 of PanelA show thatasset growth
and cash acquisitionsreturnto their preloanratinglevel in the years after the
bank loan rating is obtained. In the long run (Panel B), the results suggest
that asset growth declines relative to the preloan levels. Overall, these results
suggest thata borrowerwho obtainsa bankloan ratingexperiencesa one-time
increase in real outcomes, followed by a period in which they invest either at
or below preloanratinglevels.

5. Conclusion
The findings suggest that third-partycertificationby ratingagencies increases
the availabilityof debt financingfor firms,and increasesinvestmentand acquisitions. I informallyextendthe theoreticalframeworkof Holmstromand Tirole
(1997) to show how the innovationof a third-partyratingstechnology reduces
the certificationcost borneby the borrower.The technologyincreasesthe future
income streamsthatthe borrowercan pledge to uninformedlenders,andtherefore increasesthe availabilityof debt financing.The increasedparticipationof
uninformedlendersalso allows the firmto increaseinvestment.
I use this frameworkto empiricallyexplore the introductionof loan ratings
by Moody's and S&P in 1995. My firstset of results suggests thatloan ratings
affordunratedfirmsa uniqueopportunityto obtainthird-partydebtcertification.
The second set of resultsshows thatthe firmsthatobtainloan ratingsexperience
significantincreasesin the availabilityof debtfinancing,asset growth,andcash
acquisitions.The identificationstrategyexploits the cross-sectionalpatternof
firmsthatchoose to obtainloan ratings,and shows thatthe resultsare strongest
the hypothesis that the introductionof loan ratingscaused an expansionin the
supply of debt financing for these firms. Several robustnesstests imply that
the increases in outcomes would not have been as large had loan ratings not
been available.Finally, I documenta channel throughwhich third-partydebt
ratingsexpandthe supply of debt financing:I find evidence that less-informed
nonbankinstitutionalinvestorsand foreign banks are more willing to serve on
syndicateswhen a firmobtainsa loan rating.
Researchpresentedhere points to two directionsfor futureresearch.First,
there are additionalresearch avenues to examine the real effects of ratings.
For example, many privatecompanies obtained bank loan ratings after their
introductionin 1995. Given that privatecompanies are even less certifiedex
ante than public companies without existing credit ratings, I expect that the
resultspresentedhere would be even strongeramong privatecompanies.Also,
in December 2003, S&P introducedbank loan recoveryratings, which assess


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification

the liquidationvalue of a bank loan and vary within a given loan rating.The
introductionof recovery ratings may furtherimprove liquidity and access to
capital for borrowers who obtain bank loan ratings. In addition, while the
introductionof bankloan ratingsin the US is arguablyone of the largestrating
technology introductionsin the last 15 years, severalratingagencies in various
countries also began rating borrowersduring this time period. An analysis
of the effects of ratingagencies in less financiallydeveloped economies may
furtherstrengthenthe hypothesisthatthird-partyratingorganizationsincrease
the availabilityof externalfinancefor borrowers.
Second, there is a large and growing body of research documenting the
importanceof creditratingsin corporatepolicy (FaulkenderandPetersen,2006;
Kisgen, 2006; Tang,2006), but thereis little empiricalevidence on the precise
mechanismby which ratingsimproveaccess to capital.In otherwords, why do
ratingsreducecertificationcosts andexpandthe base of creditors?Is it because
of the informationproducedby rating agencies? Do agency problems within
organizationsof institutionalinvestorsmake a ratingnecessaryfor investment?
Is it due to holdingrequirementsimposedby regulatoryauthorities?Answering
these questions will likely necessitate detailed informationon the securities
held by institutionalinvestors.An analysis of why many nonbankinstitutional
investorsmust have ratingswould offer valuableinsight into the importanceof
ratingagencies in the economy.
of Lendersin LPC'sDealscan
Domestic commercialbanks
Bank of New York
CreditSuisse First Boston
Bank One
Seattle First National
First NationalBank of Maryland
Bank of Oklahoma
IndustrialBank of Japan
Banca di Roma
BNP Paribas
Royal Bank of Scotland
Bank Hapoalim
Allied Irish

Police OfficersPension System of the City of Houston
RobertFleming &Co Ltd
MountainCLO Trust
ReliastarLife InsuranceCo
KZH Holding Corp IV
CanpartnersInvestmentsIV LLC
KeyportLife InsuranceCo
NorthAmericanSenior Floating Rate Fund
SankatyAdvisors LLC
ORIX LeveragedFinance [aka Orix Business Credit]
Stein Roe & Farnham
Fidelity & GuarantyLife InsuranceCo
SRF TradingInc
MackayShields OffshoreHedge Fund
Prime Income
Sequils-CumberlandI LLC


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

TheReviewof Financial Studies/

v22 n 4 2009


Bear Steams
Heller Financial
Congress Financial

Debt StrategiesFund III
Stein Roe FloatingRate Ltd
CapitalBusiness Credit
SunAmericaLife InsuranceCo
HartfordLife InsuranceCo
United of OmahaLife Insurance
PPM Spyglass FundingTrust
Pimco Advisors
PilgrimSenior Income Fund

This table presentsa sample of lendersin LPC's Dealscan, and how I classify these lendersinto categories.

Bavaria, S. 2002. Syndicated Loans-A Rated Marketat Last! Standardand Poor's Document, available on
RatingsDirectby S&P.
Bavaria,S. 2004. To Our Clients in the SyndicatedLoan Market.Standardand Poor's Document, availableon
RatingsDirectby S&P.
Bavaria, S. 2005. An Introductionto Rating Secured Loans. Standardand Poor's Document, available on
RatingsDirectby S&P.
Ben-Amos, Omri,"Uptickin S&P Ratingof Bank Loans Hints CrossoversAre on the Rise."AmericanBanker,
December 13, 1996, 17.
Bernanke,B. 1983. NonmonetaryEffects of the FinancialCrisis and the Propogationof the GreatDepression.
Bernanke,B., C. Lown, and B. Friedman. 1991. The Credit Crunch.BrookingsPapers on Economic Activity
Boot, A., T. Milbourn,and A. Schmeits. 2006. Credit Ratings as CoordinationMechanisms. The Review of
Financial Studies 19:81-118.
Cantillo,M., andJ. Wright.2000. How Do FirmsChoose TheirLenders?An EmpiricalInvestigation.TheReview
of Financial Studies 13:155-89.
Dennis, S., and D. Mullineaux.2000. SyndicatedLoans. Journalof Financial Intermediation9:404-26.
Diamond, D. 1991. Monitoringand Reputation:The Choice Between Bank Loans and PrivatelyPlaced Debt.
Journal of Political Economy99:689-721.
Djankov,S., C. McLiesh, and A. Shleifer.2007. PrivateCreditin 129 Countries.Journalof Financial Economics
Faulkender,M., and M. Petersen. 2006. Does the Source of Capital Affect Capital Structure?The Review of
Financial Studies 19:45-79.
Fazzari, S., G. Hubbard,and B. Petersen. 1988. Financing Constraintsand CorporateInvestment.Brookings
Papers on EconomicActivity 1:141-95.
Gupta,A., A. Singh, and A. Zebedee. 2006. Liquidityin the Pricingof SyndicatedLoans. WorkingPaper,Case
WesternReserve University,August.
Hand, J., R. Holthausen,and R. Leftwich. 1992. The Effect of Bond Rating Agency Announcementson Bond
and Stock Prices. Journalof Finance 47:733-52.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions

The Real Effectsof Debt Certification

Holmstrom,B., and J. Tirole. 1997. FinancialIntermediation,LoanableFunds, and the Real Sector. Quarterly
Journalof Economics 112:663-9 1.
Ivashina,V. 2005. The Effects of SyndicateStructureon LoanSpreads.WorkingPaper,HarvardBusiness School,
Kaplan,S., andL. Zingales. 1997. Do Investment-CashFlow SensitivitiesProvideUseful Measuresof Financing
Constraints?QuarterlyJournalof Economics. 112:169-215.
Kashyap,A., R. Rajan,and J. Stein. 2002. Banks as LiquidityProviders:An Explanationfor the Coexistence of
Lendingand Deposit-Taking.Journalof Finance 57:33-73.
Kashyap,A., J. Stein, and D. Wilcox. 1993. MonetaryPolicy and CreditConditions:Evidence from the Composition of ExternalFinance.AmericanEconomicReview83:78-98.
Kisgen, D. 2006. CreditRatingsand CapitalStructure.Journalof Finance, 61:1035-72.
Leary,M. 2005. Bank Loan Supply, LenderChoice, and CorporateCapital Structure.WorkingPaper,Cornell
Lee, S., and D. Mullineaux. 2004. Monitoring,Financial Distress, and the Structureof CommercialLending
Syndicates.Financial Management33:107-30.
Loutskina, E. 2005. Does SecuritizationAffect Bank Lending: Evidence from Bank Responses to Funding
Shocks. WorkingPaper,Universityof Virginia,May.
Loutskina,E., and P. Strahan.2005. Securitizationand the Declining Impactof Bank Finance on Loan Supply:
Evidence from MortgageAcceptanceRates. WorkingPaper,Boston College, July.
Moerman,R. 2005. The Role of InformationAsymmetryand FinancialReportingQuality in Debt Contracting:
Evidence from the SecondaryLoan Market.WorkingPaper,Universityof Pennsylvania,Wharton.
Moerman, R. 2006. The Impact of InformationAsymmetry on Debt Pricing and Maturity.Working Paper,
Universityof Pennsylvania,Wharton.
Mullineaux, D., and H. Yi. 2006. The InformationalRole of Bank Loan Ratings. The Journal of Financial
Rajan,R., and L. Zingales. 1995. WhatDo We KnowaboutCapitalStructure?Some Evidencefrom International
Data. Journalof Finance 50:1421-60.
Rauh,J. 2006. Investmentand FinancingConstraints:Evidence from the Fundingof CorporatePension Plans.
Journalof Finance 61:33-72.
Schuermann,T. 2004. Why Were Banks BetterOff in the 2001 Recession? Federal Reserve Bank of New York
CurrentIssues in Economicsand Finance 10:1-7.
Standardand Poor's. 2005. Bank Loan and Recovery Ratings:2005 Fee Schedule and User's Guide, availableat
Sufi, A. 2007. InformationAsymmetryand FinancingArrangements:Evidence from SyndicatedLoans.Journal
of Finance 62:629-68.
Tang,T. 2006. InformationAsymmetryand Firms'CreditMarketAccess: Evidencefrom Moody's CreditRating
FormatRefinement.WorkingPaper,Universityof Chicago GSB.


This content downloaded from on Mon, 30 Sep 2013 13:05:11 PM

All use subject to JSTOR Terms and Conditions