+ All Categories
Home > Documents > Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando...

Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando...

Date post: 26-Sep-2020
Category:
Upload: others
View: 0 times
Download: 0 times
Share this document with a friend
40
OCCASIONAL PAPER SERIES NO. 16 / JUNE 2004 MARKET DYNAMICS ASSOCIATED WITH CREDIT RATINGS A LITERATURE REVIEW by Fernando Gonzalez, François Haas, Ronald Johannes, Mattias Persson, Liliana Toledo, Roberto Violi, Martin Wieland and Carmen Zins
Transcript
Page 1: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

OCCAS IONAL PAPER S ER I E SNO. 16 / J UNE 2004

MARKET DYNAMICS ASSOCIATED WITH CREDIT RATINGS

A LITERATURE REVIEW

by Fernando Gonzalez,François Haas, Ronald Johannes, Mattias Persson,Liliana Toledo, Roberto Violi, Martin Wieland and Carmen Zins

Page 2: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

In 2004 all ECB publications will feature

a motif taken from the

€100 banknote.

OCCAS IONAL PAPER S ER I E SNO. 16 / J UNE 2004

MARKET

DYNAMICS

ASSOCIATED WITH

CREDIT RATINGS

A LITERATURE

REVIEW *

by Fernando Gonzalez 1,François Haas 2, Ronald

Johannes 3, Mattias Persson 4,Liliana Toledo 5, Roberto Violi 6, Martin Wieland 7

and Carmen Zins 7

This paper can be downloaded from the ECB’s website (http://www.ecb.int).

* This paper represents a joint effort by a group of European central bank experts co-ordinated by F. Haas (Banque de France). It has benefited from comments and suggestions from Ingo Fender and Franck Packer (Bank for International Settlements). The viewsexpressed in this paper do not necessarily reflect those of the European Central Bank or of the respective institutions of the authors.

1 European Central Bank, Kaiserstrasse 29, 60311, Frankfurt am Main, Germany2 Banque de France, 39, rue Croix-des-Petits-Champs, 75049 Paris Cedex 01, France

3 Bank of England, Threadneedle Street, London EC2R 8AH, United Kingdom4 Sveriges Riksbank, Brunkebergstorg 11, 103 37 Stockholm, Sweden

5 Banco de España, Alcalá 50, 28014 Madrid, Spain6 Banca d’Italia, Via Nazionale 91, 00184 Rome, Italy

7 Deutsche Bundesbank, Wilhelm-Epstein-Str. 14, 60431 Frankfurt am Main, Germany

Page 3: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

© European Central Bank, 2004

AddressKaiserstrasse 2960311 Frankfurt am MainGermany

Postal addressPostfach 16 03 1960066 Frankfurt am MainGermany

Telephone+49 69 1344 0

Websitehttp://www.ecb.int

Fax+49 69 1344 6000

Telex411 144 ecb d

All rights reserved. Reproduction foreducational and non-commercial purposesis permitted provided that the source isacknowledged.

The views expressed in this paper do notnecessarily reflect those of the EuropeanCentral Bank.

ISSN 1607-1484 (print)ISSN 1725-6534 (online)

Page 4: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

3ECB

Occa s i ona l Pape r No . 16June 2004

CONTENT S1 EXECUTIVE SUMMARY 4

2 INTRODUCTION 6

3 THE EXPANDING USE OF RATINGS 7

3.1 A tentative definition of ratings 7

3.2 The expanding use of credit ratings 8

4 RATINGS AND MARKET DYNAMICS: DO RATINGSAND CHANGES IN RATINGS INFLUENCE MARKETDYNAMICS? 10

4.1 The information content of ratings andreactions to changes in ratings 10

4.2 The consequences of the morewidespread use of ratings 12

5 THE CHALLENGE: HOW TO RECONCILE THE(HETEROGENEOUS) INTERESTS OF DIFFERENTUSERS? 16

5.1 The methodology: “rating through-the-cycle” vs. “point-in-time” 16

5.2 Recent developments in methodology:structural changes or refinement? 19

6 ACCURACY, STABILITY AND THE “RELATIVE”PROCYCLICALITY OF RATINGS 22

7 POLICY IMPLICATIONS AND ISSUES 24

7.1 Hardwiring vs. flexibility 24

7.2 Implications of the limited short-termaccuracy of ratings 24

7.3 Rating agency initiatives andtransparency 25

7.4 Possible implications of a change inmethodology 25

8 APPENDIX 27

8.1 Default risk and the informationalefficiency of rating changes: evidencefrom capital markets 27

8.2 Credit risk and transition matrices 28

8.3 Credit Migration Estimates: methodsand risk measurement implications 30

9 REFERENCES 32

Page 5: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

4ECBOcca s i ona l Pape r No . 16June 2004

1. Credit ratings produced by the major creditrating agencies (CRAs) aim to measure thecreditworthiness, or more specifically, therelative creditworthiness of companies, i.e.their ability to meet their debt servicingobligations. In principle, the rating processfocuses on the fundamental long-term creditstrength of a company. It is typically based onboth public and private information, except forunsolicited ratings, which focus only on publicinformation. The basic rationale for usingratings is to achieve information economies ofscale and solve principal-agent problems. Partlyfor the same reasons, the role of credit ratingshas expanded significantly over time.Regulators, banks and bondholders, pensionfund trustees and other fiduciary agents haveincreasingly used ratings-based criteria toconstrain behaviour. As a result, the influenceof the opinions of CRAs on markets appears tohave grown considerably in recent years.

2. One aspect of this development is itspotential impact on market dynamics (i.e. thetiming and path of asset price adjustments,credit spreads, etc.), either directly, as aconsequence of the information content ofratings themselves, or indirectly, as as aconsequence of the “hardwiring” of ratings intoregulatory rules, fund management mandates,bond covenants, etc.

When considering the impact of ratings andrating changes, two conclusions are worthhighlighting:

– First, ratings correlate moderately well withobserved credit spreads, and rating changeswith changes in spreads. However, otherfactors, such as liquidity, taxation andhistorical volatility clearly also enter into thedetermination of spreads. Recent researchsuggests that reactions to rating changes mayalso extend beyond the immediately-affectedcompany to its peers, and from bond toequity prices. Furthermore, this pricereaction to rating changes seems to beasymmetrical, i.e. more pronounced fordowngrades than for upgrades, and may be

1 EX E CU T I V E S UMMARYmore significant for equity prices than forbond prices.

– Second, the hardwiring of regulatory andmarket rules, bond covenants, investmentguidelines, etc., to ratings may influencemarket dynamics, and potentially lead to ormagnify threshold effects. The more thatdifferent market participants adopt identicalratings-linked rules, or are subject to similarratings-linked regulations, the more “spiky”the reaction to a credit event is likely to be.This reaction may include, in some cases, theemergence of severe liquidity pressures.Efforts have recently been made, notablywith support from the rating agenciesthemselves, to encourage a more systematicdisclosure of rating triggers and torenegotiate and smooth the possibly moredestabilising forms of rating triggers.However, the lack of a clear disclosureregime makes it difficult to assess how farthis process has evolved. Questions alsoremain as to the extent to which ratings-based criteria introduce a fundamentally newelement into market behaviour, or,conversely, the extent to which they aresimply a variant of more traditionalcontractual covenants.

3. Rating agencies strive to provide creditassessments that remain broadly stable throughthe course of the business cycle (rating“through the cycle”). Agencies and otheranalysts frequently contrast the fundamentalcredit analysis on which ratings are based withmarket sentiment – measured for example bybond spreads – which is arguably subject tomore short-term influences. Agencies areadamant that they do not directly incorporatemarket sentiment into ratings (although theymay use market prices as a diagnostic tool). Onthe contrary, they make every effort to excludetransient market sentiment. However, asreliance on ratings grows, CRAs are beingincreasingly expected to satisfy a wideningrange of constituencies, with different, andeven sometimes conflicting, interests: issuersand “traditional” asset managers will look for

Page 6: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

5ECB

Occa s i ona l Pape r No . 16June 2004

1

Executivesummary

more than a simple statement of near-termprobability of loss, and will stress the need forratings to exhibit some degree of stability overtime. On the other hand, mark-to-markettraders, active investors and risk managers mayseek more frequent indications of creditchanges. Hence, in the wake of majorbankruptcies with heightened credit stress,rating agencies have been under considerablepressure to provide higher-frequency readingsof credit status, without loss of quality. So far,they have responded to this challenge largely byadding more products to their traditional range,but also through modifications in the ratingprocess.

4. The rating process and the range of productsoffered by rating agencies have thus evolvedover time, with, for instance, an increasingemphasis on the analysis of liquidity risks, anew focus on the hidden liabilities of companiesand an increased use of market-based tools. It istoo early, however, to judge whether thesechanges should simply be regarded as arefinement of the agencies’ traditionalmethodology or whether they suggest a morefundamental shift in the approach to credit riskmeasurement. For the same reason, it is notpossible to draw any firm conclusions aboutchanges in the effects of credit ratings onmarket dynamics.

Page 7: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

6ECBOcca s i ona l Pape r No . 16June 2004

This paper summarises the work conducted by agroup of economists from various Europeancentral banks over the summer of 2003. It isintended to add to the ongoing debate on majorrating agencies and their methodologies. Theanalysis and policy considerations proposed arebased on a review of the literature and are thoseof the authors; they do not necessarily reflectthe positions of their respective institutions.

The paper is aimed at contributing to the currentdebate on this topic in two ways: first, byproviding a factual exposition of thesignificance and evolving use of credit ratingsin the financial markets and, second, byidentifying the possible impacts that suchevolving use may have on market dynamics (i.e.the timing and path of asset price adjustments,the dynamics of credit spreads, the potentialmagnifying effects that rating changes cantrigger) and analysing how credit ratingagencies (CRAs) have responded to theincreasing, and sometimes conflicting, demandsthat market participants put on credit ratings.In doing so, the paper also provides acomprehensive review of literature on creditratings and CRAs.

The paper is organised as follows: Section 3explains how the role of ratings has evolved infinancial markets; Section 4 examines differentchannels through which ratings and ratingchanges may impact on market dynamics andcontribute to asset price movements; Section 5discusses rating methodology and the meaningof some recent developments in this field;Section 6 considers the accuracy, stability andthe relative procyclicality of ratings; andSection 7 concludes by pointing to some policyimplications and issues derived from thepreceding developments. An appendix (Section8) surveys some key issues in credit riskmeasurement.

2 I N TRODUCT I ON

Page 8: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

7ECB

Occa s i ona l Pape r No . 16June 2004

3

The expandinguse of ratings

3.1 A TENTATIVE DEFINITION OF RATINGS

RATINGS ARE CREDIT OPINIONS

Ratings provided by CRAs are a measure of thelong-term fundamental credit strength ofcompanies, i.e. their long-term ability andwillingness to meet debt servicing obligations.More specifically, ratings apply either to thegeneral creditworthiness of an obligor or to itsobligations with respect to a particular debtsecurity (senior and subordinated bonds, eithersecured or unsecured, collateralised debtstructures, etc.) or other specific financialobligations.

CRAs base their analyses on a company’sfinancial statements, franchise value,management quality and competitive position inits industry, and seek to predict creditperformance – the servicing of debt obligationsin full and on time – under a range ofmacroeconomic and credit conditions, includingstress situations. This analysis is based notonly on public information, but also on private/confidential information which companies agreeto share with CRAs.

CRAs stress that ratings are opinions.1 Theseopinions, which stem from fundamental creditanalysis, are used to classify credit risk. Inkeeping with their status as opinions, ratingsare determined by a rating committee.2 As such,ratings do not constitute a recommendation tobuy, sell or hold a particular security, and donot address the suitability of an investment for aparticular investor.

Inherent in this definition of ratings is thenotion that they are an ordinal measure of risk,but not necessarily a cardinal one. Accordingly,all CRAs express the outcome of theirassessments in the form of symbols, such asAaa, AAA, etc., which more or less correspondto each other across agencies. The division ofthe rating scale into these buckets, and thesubsequent assignment of debt obligations tothem, essentially reflects the judgement and

experience of rating agency staff, supplementedin some areas by the use of models.

THE ECONOMIC RATIONALE FOR THE USE OFRATINGS

In economic terms, the rationale for usingratings, and their growing “popularity” stemsfrom their ability to provide informationeconomies of scale on the one hand, and fromtheir contribution to solving principal-agentproblems on the other.

– Information economies of scale

Creditors and investors have found it efficientto use ratings opinions in initiating andmonitoring their transactions because of theeconomies of scale achieved in gathering andanalysing information. This, in turn, hasfacilitated the access of borrowers to debtmarkets, by widening the investor pool andreducing adverse selection problems resultingfrom information asymmetries betweeninvestors and issuers of debt,3 and has providedsignificant impetus to the development offinancial markets.

– Principal-agent problems

Another way in which the use of ratings affectsthe market is the pervasive “hardwiring” ofrules and guidelines to ratings. In all cases, theprincipal motivation for hardwiring to ratings isthe same: to formulate a simple and verifiablerule with low transaction costs, so as to be ableto monitor and constrain the actions of agents.

3 TH E E XPAND I NG U S E O F R AT I NG S

1 “Understanding Moody’s Corporate Bond Ratings and RatingsProcess”, Moody’s Investors Service, May 2002, New York.

2 Mara Hilderman, “Opening the Black Box: The RatingCommittee Process at Moody’s”, Moody’s Investors Service, July1999, New York.

3 The general adverse selection problem was introduced byGeorge Akerlof in 1970 (see “The market for lemons”,Quarterly Journal of Economics, Vol. 54, pp. 488-500). In thecase of bond markets, the problem implies that non-ratedmarkets are characterised by very low spreads between theinterest rates paid by strong issuers and those paid by weakissuers. An objective assessment of credit risk significantlyincreases this spread, benefiting strong issuers but harmingweak issuers.

Page 9: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

8ECBOcca s i ona l Pape r No . 16June 2004

In economic terminology, ratings are used tosolve principal-agent problems, that is, aprincipal’s problem of maximising incentivesfor agents to perform well when it is hard toobserve or directly control their actions.

3.2 THE EXPANDING USE OF CREDIT RATINGS

Indeed, precisely due to the aforementionedqualities of ratings, interest in credit ratingservices and the demand for a wider range ofratings, beyond the credit assessment oftraditional corporate bonds, have significantlyincreased4 over the past three decades.Regulators (in regulations), banks andbondholders (in loan and bond covenants),pension fund trustees and other fiduciary agents(in investment guidelines, insurance companycharters, etc.) have made increasing use ofratings-based constraints in their rules. As aresult, the use of ratings and the influence of theopinions of CRAs on securities markets havegrown significantly, to the extent that ratingsare now ubiquitous in financial markets,and increasingly act as benchmarks orcreditworthiness standards, far beyond theirinitial purpose. This role can be highlighted inthe area of regulation as well as in debt issuanceand portfolio management.

“CORE CLIENTS”: ISSUERS AND FIXED-INCOMEINVESTORS

– Debt issuers

From the outset, debt issuers have been amongthe “natural” users of ratings, and increasinglyso, given that, although ratings may notdetermine their ability to enter financialmarkets, they do at least contribute todetermining their financing costs and thequality of their investor bases. Hence,preserving or achieving a desired rating isfrequently incorporated into corporate goalsand represents an integral part of the financingstrategy of companies. Indeed, through the useof specific services offered by CRAs, such asMoody’s Rating Assessment Service or

Standard & Poor’s (S&P’s) Rating EvaluationService, companies are able to “monitor” thebehaviour of their ratings under differentscenarios. The use of rating triggers, that is tosay covenants that imply a change in thecharacteristics of an existing financinginstrument, should the rating of the issuer/borrower change, are an example of howinvestors can use ratings to tailor theirinvestments by issuer. Such ratings-basedtriggers were initially mostly found in bankloan covenants. They became increasinglypopular, however, and took diversified forms,in bond issuance in the mid to late 1990s (seebelow).

– Bond investors and portfolio managers

Ratings provided by recognised CRAs play acentral role in portfolio governance, especiallyfor small to medium-sized asset managers wholack the resources to develop reliable internalcredit assessment systems. However, even formajor asset managers, the use of internal creditassessment systems is frequently limited tosupplementing external ratings when the latterare not available or when they providediverging signals. The use of ratings inportfolio governance and investment mandatesappears to be twofold: (i) ratings-basedguidelines contribute to determining theuniverse of eligible assets – within thisuniverse ratings (in conjunction with maturityconstraints) are also used to determine themaximum, and sometimes minimum, proportion

4 Short-term rating of commercial papers was first introduced inthe 1970s. Bank ratings in the form of financial strength ratingsmeasure credit risk for a bank in the absence of any assumedsupport from governmental authorities. Fitch also produces bankratings which factor in the likelihood of support. Asset-backedratings measure credit risk on structured products in which therated liabilities are backed by a dedicated set of assets, e.g.asset-backed commercial papers backed by trade receivables,or collateralised debt obligations (CDOs) backed by commercialbank loans. Liquidity rating assessments, which were formallyintroduced by Moody’s in 2002 for speculative grade borrowers,are designed to supplement short-term ratings and to give anassessment of vulnerability to sudden loss of market access.Additionally, there are sovereign ratings, both for industrialisedcountries and emerging market borrowers. This type of rating isvery different from the others in its conception of credit risk,and lies beyond the scope of this paper.

Page 10: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

9ECB

Occa s i ona l Pape r No . 16June 2004

3

The expandinguse of ratings

of authorised holdings – and (ii) these ratings-based guidelines also shape the reactions ofasset managers when faced with changes in thecredit quality of their holdings.5

OTHER MARKET PARTICIPANTS

Market participants, be they investors, marketmakers or broker-dealers, also rely extensivelyon external ratings for the assessment of theirtrading counterparties (selection of creditcounterparties and definition of credit limits).This is especially true with regard to theshort-term management of liquidity (repotransactions, for instance) and over-the-counterderivative transactions (swaps, options, etc.):the creditworthiness of market participants, asassessed by CRAs, determines either theconditions (costs) under which thoseparticipants can access the market (thefrequency of margin calls, the magnitude ofcollateralisation that they will be asked toprovide) or even their very access to markets.

In the management of their portfolios and theimplementation of monetary policy, centralbanks also frequently rely on ratings providedby CRAs, in ways similar to that of othermarket participants, i.e. in the definition ofeligible assets, either for the investment of ownfunds and foreign exchange reserves or asmonetary policy collateral. In the latter case, thecredit quality of eligible assets impacts on therequired level of collateralisation and riskcontrols.

REGULATORS

The importance of ratings-based regulations hastraditionally been particularly visible in theUnited States, where it can be traced back to the1930s. These regulations affect not only banks,but also insurers, pension funds, mutual fundsand broker-dealers, restricting or prohibitingthe purchase of bonds with “low” ratings(usually below BBB), imposing variable capitalcharges depending on the rating of the holdingsor easing the issuance conditions or disclosurerequirements for securities carrying a

“satisfactory” rating. While ratings-basedregulation appears to be less common overall inEurope, a similar approach can be found in theCapital Adequacy Directive in the area of bankregulation. More generally, the Basel II project,in its “standardised approach to credit risk”establishes fixed credit risk weights for eachsupervisory category and relies explicitly on“external credit assessments”.

5 See Report of the CGFS “Incentives structures in InstitutionalAsset Management and their Implications for FinancialMarkets”, March 2003.

Page 11: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

10ECBOcca s i ona l Pape r No . 16June 2004

When considering the possible impacts ofratings (and, more specifically, changes inratings) on market dynamics (i.e. the behaviourof asset prices and spreads), it is necessary todistinguish between a direct impact, resultingfrom the information content of a rating change,and an indirect impact, stemming from the“hardwiring” of regulations and guidelines toratings.

4.1 THE INFORMATION CONTENT OF RATINGSAND REACTIONS TO CHANGES IN RATINGS

For bond ratings to have a direct, information-related impact on spreads and spread dynamics,they must contain relevant pricing informationthat investors cannot obtain from other sourcesat comparable cost. The question of theinformation content of ratings has beenaddressed (1) by analysing the relation betweenbond yields and ratings and (2) by studyingprice reactions to rating changes.

RELATIONSHIP BETWEEN BOND YIELDS ANDRATINGS

The various studies that have tried to answer thequestion of the information content of ratings ingeneral come to the conclusion that ratings dohelp explain cross-sectional differences inyield spreads.6 In these studies, ratings may,however, be a proxy for omitted publiclyavailable variables that affect the spreads.7

Indeed, even if ratings and rating changes dopartly explain observed spreads and theirdynamics, there remains a large part of thesespreads that ratings cannot explain. Additionalfactors, whose relative importance variesaccording to the different studies, thereforeneed to be included:

• Taxation. While ratings are found by Gabbiand Sironi (2002) to be effectively the mostimportant factor determining primary yieldspreads between corporate bonds and theequivalent Treasury securities, other factors,such as expected tax treatment for bonds,

are also important. However, the structuralefficiency8 of the market and liquidityvariables do not appear to be significant inexplaining the cross-sectional variability ofspreads.9

• Systematic risk. According to Elton, Gruber,Agrawal and Mann (2001), losses stemmingfrom expected defaults come last among thethree factors that can explain (i.e. breakdown) corporate spreads – expected lossesare found to explain only 17.8% of thevariation in the spread. Differential taxesappear to be more important and explainabout 36% of the spread. The remainingportion of the spread (more than 46%) isfound to be closely related to the factorscommonly accepted as explaining riskpremia for common stocks, i.e. the Fama-French factors.10 Hence, a large portion ofthe spread seems to be compensation forsystematic risk, that cannot be diversifiedaway.

• Volatility. Campbell and Taksler (2003)analysed the effects of equity volatilityon corporate bond yields, and showedthat idiosyncratic volatility was directlyrelated to the cost of borrowing for

4 RAT I NG S AND MARKE T DYNAM I C S : DO R AT I NG SAND CHANGE S I N R AT I NG S I N F L U ENC E MARKE TDYNAM I C S ?

6 See, for example, Liu and Thakor (1984) and Kao and Wu(1990). Ederington et al. (1987) find that, conditional oneconomic and company-specific variables, ratings do haveexplanatory power for bond yields.

7 See Galil (2002). See Appendix for more details and morereferences to the academic literature.

8 Such as fees charged to the issuer, the number of managers in thebond issuing syndicate, and the issuance process (privateplacement versus public issue and fixed-priced versus open-priced issues).

9 The study was conducted on the primary corporate eurobondmarket and analysed spreads on eurobond issues completed byalmost 600 major corporations from 15 industrialised countriesduring the 1991-2001 period. In addition to the above-mentionedresult, the study also showed that bond investors’ reliance onrating agencies’ judgements increased over time during thesample period. Also, empirical evidence shows that ratingagencies adopt a “through-the-cycle” evaluation approach toobligors’ creditworthiness that is different to the forward-looking approach used by bond investors. Finally, ratingagencies’ discordance, as measured by a different numericvalue of the assigned rating notch, appears to be perceived bybond investors as a sign of (or simply reflects) a higher degreeof uncertainty concerning the issuer’s default risk.

10 The excess return on the market (RM) factor, the “small minusbig” (SMB) factor, and the “high minus low” (HML) factor.

Page 12: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

11ECB

Occa s i ona l Pape r No . 16June 2004

4

Ratings andmarket dynamics:

do ratings andchanges in ratings

influence marketdynamics?

corporate issuers. Furthermore, the resultsalso suggest that volatility can explain asmuch cross-sectional variation in yields ascredit ratings.

• Supply and demand. Using dealers’ quotesand transaction prices for industrial bonds,Collin-Dufresne, Goldstein and Martin(2001) investigated the determinants ofcredit spread changes. Their results showthat variables, which in theory determinecredit spread changes, have limitedexplanatory power. Rather, using principalcomponent analysis, they show that most ofthe residuals are driven by a single factor.Monthly credit spread changes appear to bedriven principally by local supply/demandshocks that are independent of both credit-risk factors and liquidity.

• Liquidity. Chen, Lesmond and Wei (2002)find liquidity to also be an important factorexplaining corporate bond spreads, aftercontrolling for credit ratings, maturity,amounts outstanding and volatility. Theseresults indicate that liquidity is indeed pricedinto corporate bonds.

PRICE REACTIONS TO RATING CHANGES

Numerous studies have focused on the pricereactions of bonds and equities to changes inratings. A recent study by Klinger and Sarig(2000), which focuses on the refinement ofMoody’s rating system in 1982, shows thatinvestors do indeed react to changes in ratingsif they are unexpected, in the same way as theyreact to new information. Their test was,however, conducted on a one-off event basisthat does not necessarily reflect theinformational content of ratings in future years.More generally, while research conducted inthis field11 usually yields mixed results, twofindings are worth highlighting:

• Reactions to bond rating downgrades percolatefrom the affected company to its rivals, andfrom the bond market to equity prices. Equityanalysts revise their earnings expectations

downward for both the downgraded companyand its rivals, and the extent of this reactiondepends on the initial rating and the size of thedowngraded debt (Caton and Goh, 2003).

• The price reaction to rating changes, and inparticular the effect on stock returns, isasymmetrical, i.e. the market reacts morestrongly to rating downgrades than to ratingupgrades,12 and ultimately this asymmetryappears less significant for bonds than forstocks. Several studies suggest that abnormalequity returns following bond downgrades arenegative, whereas there is no significantabnormal equity return reaction to upgrades.Holthausen and Leftwich (1986) suggestthat the difference between one-year pre-announcement returns to upgrade anddowngrade is in the order of 20% to 30%.They find no abnormal returns after theannouncement of upgrades, but do findevidence of abnormally low returns in thequarter following a downgrade. Dichev andPiotroski (2001) find negative abnormal stockreturns in the order of 10% to 14% in the firstyear following downgrades. Furthermore, theunderperformance is more pronounced forsmall companies with low credit quality.

The above findings could stem from the way inwhich rating agencies produce their ratings orcould reflect the fact that rating agencies expendmore resources on detecting deteriorations incompany balance sheets than they do ondetecting improvements in earnings. A furtherexplanation is, of course, that stock marketsoverreact to rating downgrades (see Dichev andPiotroski, 2001). It could also be argued thatthe overreaction to downgrades reflects the fact

11 Griffin and Sanvicente (1982), Ingram, Brooks and Copeland(1983), Holthausen and Leftwich (1985), Hand, Holthausen andLeftwich (1992) and Goh and Ederington (1993).

12 Hand, Holthausen and Leftwich (1992) find asymmetricalresults with respect to reactions to rating downgrades andupgrades. They observe significantly negative average excessbond and stock returns for downgrades, and a weaker positiveeffect for upgrades. However, when controlling for expectedrating changes, the asymmetries disappear in bond returns butpersist in stock returns. Similarly, according to Ederington, Gohand Nelson (1996), the stock market reacts to downgradeinformation more quickly than analysts do and, in contrast todowngrades, upgrades do not elicit any market response.

Page 13: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

12ECBOcca s i ona l Pape r No . 16June 2004

that downgrades convey additional information:a downgrade signals that the rated company haseither decided not to or proved unable to avoidthe downgrade. This is consistent with the roleof ratings as coordination mechanisms, seeBoot/Milbourn (2002).

4.2 THE CONSEQUENCES OF THE MOREWIDESPREAD USE OF RATINGS

THE EFFECTS OF THE HARDWIRING OF RULESAND REGULATIONS TO RATINGS

As mentioned above, enshrining ratings intorules and regulations is a possible answer toprincipal-agent problems. At the same time,such hardwiring may fuel specific marketdynamics as it injects a dose of automatism (andpredictability) into the reactions of the affectedmarket participants to the initial rating event,potentially magnifying threshold effects. Forinstance, the point at which probabilities ofdefault have been found empirically to risesharply constitutes one of the most importantdivides in rating scales (between BBB- andBB+), and one of the main thresholds in theworld of asset management, as it separates“investment grade” securities from “speculativegrade” securities, which many investors are notauthorised to hold or may only hold in strictlylimited quantities. Hence, the downgrading of abond issue (or an issuer) to below that levelmay force asset managers to restructure theirportfolios, triggering a forced liquidation ofassets. More generally, the more that differentmarket participants are constrained by identicalratings-linked rules or subject to similarratings-linked regulations, the more theirreactions can be expected to be identical in theevent of a credit event, and the morepronounced the effect of such reactions is likelyto be.

In this regard, it appears that investmentmandates offer different degrees of flexibilityin the management of guideline violationstriggered by changes in ratings (most notablydowngrades). Guidelines imposing an

automatic and immediate liquidation of thedowngraded assets seem to be less commonthan they were, and are increasinglysupplemented by flexible rules that either allowthe fund manager to keep the affected assets inthe portfolio (provided that overall these assetsdo not represent more than a certain percentageof the whole portfolio) or allow him to disposeof the affected assets over a certain period oftime. Such flexibility is welcomed as it limitsthe risks of “fire sales” that can fuel downwardspirals in prices. At the same time, however,expectations of such liquidations, even if theyare anticipated to take place over a “certain”period of time instead of in the periodimmediately following the credit event, arelikely to trigger “front running” behaviour byother market participants, a situation that tendsto bring forward much of the ultimate priceimpact. More generally, this trend towardincreased flexibility in the managing of ratingevents can be seen as paralleling theprogressive changes in portfolio managementfrom an initial “buy and hold” approach to amore active, mark-to-market approach. In thesame vein, the framework of the Basel IIstandard approach could smooth the majordiscontinuity between “investment grade” and“speculative grade” securities, as it spreads theincrease in weights across the rating scale.Indeed, the BBB-/BB+ threshold is not arelevant threshold for the setting of riskweights by banking regulators.13

FROM A DOWNGRADE TO A LIQUIDITY CRISIS:RATING TRIGGERS

Ratings-based triggers are intended to protectlenders against credit deterioration andasymmetric information problems, and lendersare willing to pay for triggers by acceptinglower spreads/coupons. Hence, there is a cleardemand-side reason for issuing debtinstruments with embedded rating triggers.There is, however, also a supply-side reason forrating triggers: i.e. borrowers are willing toinclude such triggers because without them

13 Securities rated from BBB+ to BB- are assigned a 100% riskweight in the new accord standard approach.

Page 14: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

13ECB

Occa s i ona l Pape r No . 16June 2004

4

Ratings andmarket dynamics:

do ratings andchanges in ratings

influence marketdynamics?

lenders would probably demand a higher initialspread on debt contracts. Rating triggersattempt to offer protection to investors, but, dueto the way in which they work, they couldprecipitate a liquidity crisis and/or evencontribute to extreme events such asbankruptcies.

The inclusion of rating triggers in debt contractsis not new. The so-called “super poison putprovisions”, for example, that gained prominencein bonds issued in late 1980s, following the RJRNabisco buyout, contained embedded ratingtriggers.14 A super poison put provision allowsbondholders to sell their bonds to the issuingcompany at par value or at a premium after theoccurrence of a “designated event”15 combinedwith a “qualifying downgrade”. Hence, superpoison put provisions can be viewed asconditional rating triggers, conditional on aspecific event or a set of events. The exactprovisions varied from issue to issue, creatinguncertainty about the strength of the protectionoffered in any particular bond issue. In responseto this uncertainty, S&P began rating the eventrisk protection of bonds with put provisions inJuly 1989.

The designs of ratings-based triggers vary, bothin form and in the identity of the contractingparties. In general, a rating trigger providescreditors and counterparties with certain rightsin the event of a borrower’s credit rating fallingto, or below, a specified level. The rights givento the creditors usually vary from an increase inthe nominal coupon to a put option.

According to a recent survey by Moody’s(2001), out of 771 US corporate issuers ratedBa1 or higher, only 12.5% reported no triggers,while the remaining 87.5% reported a total of2,819 rating triggers.16 Not only did ratingtriggers appear to be widely used, but situationsin which a single issuer was subject to multipletriggers were common at the time of the survey.While there are reasons to believe that the useof such features has since declined, nocomprehensive picture is available that wouldhelp to accurately assess the current situation.

14 In October 1988 the market was surprised by the leveragedbuyout of RJR Nabisco. Credit agencies lowered the creditrating of RJR Nabisco bonds and the price of these bondsdecreased by 17%. This led to investor demand for bondcovenants for such events, and such covenants became known assuper poison put provisions. For a discussion on super poison putprovisions see, for example, Bae, Klein and Padmaraj (1994),Crabbe (1991) and Norton (1992).

15 A designated event is one in which the company’s shareholdersgenerally benefit, to the detriment of bondholders. Such eventsinclude mergers, takeovers, major stock repurchase plans andmajor distributions of assets to shareholders. If, after such anevent, the credit rating agencies downgrade the bond to belowinvestment grade, the put option can be exercised.

16 “The Unintended Consequence of rating triggers”, Moody’sInvestors Service, December 2001.

The table below shows common features ofrating triggers and their frequency.

Source: Moody’s Investors Service (2001).

Trigger Frequency

Collateral, letter of credit,bonding provisions 21.6%

Pricing grid 21.1%Acceleration 29.1%

of which Termination 8.5%Material adverse change 5.4%

Default 5.3%Acceleration 4.0%

Put 3.0%Early amortisation 2.9%

Other 28.2%

Table

As can be seen in this table, contingencyclauses are diverse in nature, and hence theirconsequences, if activated, may be wide-ranging:

• Collateral, L/Cs and bonding provisionsare clauses that are usually written intobank loan agreements. When the clause istriggered, the mechanism does not result in achange in the initial financing conditions butrequires the borrower to pledge assets toguarantee its financing over time. Hence, theimpact of the triggered clause should mainlybe on the opportunity cost of capital.

• Pricing grids or adjustments in interestrates or coupons are features found both inbonds and in bank loans where the initialinterest rate or coupon is revised in the event

Page 15: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

14ECBOcca s i ona l Pape r No . 16June 2004

of a change in the borrower’s rating (or insome of its financial ratios). The impact ofthe exercised trigger is a mechanical increasein the cost of capital.

• Acceleration clauses may have more severe,or sometimes even critical, effects. Forexample, for a loan or bond initially issuedfor a long period, the triggering of the clausemay result in an acceleration of repaymentsor even early termination of credit. Asmentioned above, these types of clause areused both in bond contracts and in bank loanagreements as well as in back-up credit lines.Not only does the triggering of a clauseresult in an increase in the cost of capital, butalso in an immediate need for new capital.

Two major problems associated with ratingtriggers are worth highlighting:

• Rating triggers can contribute to “creditcliff” situations. “Credit cliff” is marketjargon for a situation in which direconsequences, i.e. compounding creditdeterioration, possibly leading to default,may be expected should certain riskscenarios materialise. In this regard, S&Phas stated that “in these cases, if there is arating change, it will necessarily be a verysubstantial change (due to) the entity’sgreater sensitivity to credit quality ora particular occurrence.”17 This can putmaterial pressure on the company’s liquidityor its business. For example, whendowngraded, the position of a company thatis performing poorly will worsen as its costof capital rises. Rating triggers and othercovenants, particularly when combined, cancontribute to the development of such creditcliffs and may speed up the pace at which thecost of capital increases due to creditdeterioration. This is especially the case insituations where multiple triggers are set offsimultaneously, or when the triggering ofone clause leads to an accumulation ofnegative consequences.18 It is not clear howCRAs take these situations into account.Bonds rated at the lowest investment-grade

notch (where traditionally a large proportionof these rating triggers have been found)tend to suffer large price falls when they aredowngraded. Owing to the above mentionedrisks of self-fulfilling effects, the presenceof rating triggers may reinforce the findingthat rating agencies are only willing to decideon a rating action when it is unlikely to bereversed shortly afterwards.19

• Disclosure of ratings-based triggers byissuers has until recently been incompleteand largely ignored by analysts andinvestors. Present accounting standardsleave a significant degree of discretion asto whether triggers need to be disclosed.Under US (GAAP/FAS), UK (FRS) andinternational accounting standards (IAS)there is an obligation to disclose materialtriggers, but material in this context meansnot only that the contingent obligation islarge, but that it potentially has a significantbearing on the company’s financial situation.For instance, these requirements do notappropriately address situations where anissuer/borrower has included many “non-material” triggers in its debt covenants/bondissues. However, if there is uncertainty as towhether the company is a going concern,there should be a clear obligation to disclose.Nonetheless, it has proved difficult to obtaina comprehensive picture of the size of thecontingent liability of triggers, despite thefact that this information is crucial forinvestors as well as analysts and ratingagencies in order to fully apprehend the risksattached to a specific issue or issuer.20

Efforts have been made in this area, notablyunder pressure from rating agencies, toencourage a more systematic disclosure of

17 Standard & Poor’s, “Playing out the Credit Cliff Dynamic”,December 2001.

18 For example, a situation in which a company is downgraded and,at the same time, has to redeem some of its long-term debt early.

19 See, for example, Johnson (2003) and Löffler (2003).20 Rating agencies cannot force an issuer to disclose the nature or

extent of its use of rating triggers. If an issuer deems that publicdisclosure is not required by securities laws or after inquiriesmade by investors or rating agencies, the issuer’s credit prof ileassessment cannot be completed.

Page 16: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

15ECB

Occa s i ona l Pape r No . 16June 2004

4

Ratings andmarket dynamics:

do ratings andchanges in ratings

influence marketdynamics?

rating triggers and to renegotiate and smooththe more dangerous ones. A survey by S&Pin 2002 among more than 1,000 US andEuropean investment-grade debt issuersrevealed that about half of these issuers wereexposed to some sort of ratings-linkedcontingent liability. However, fewer than 3%exhibited serious vulnerability to ratingtriggers or other contingent calls on liquiditywhich could turn a moderate decline in creditquality into a liquidity crisis.21

Transparency and disclosure are importantfeatures that could help mitigate some of thenegative aspects of rating triggers and othercontingency clauses. It is unlikely thatsystematic (mandatory) disclosure of ratingtriggers and greater transparency with regard toexposure to rating triggers could prevent ratingevents from disturbing markets once thetriggers are activated, but it could increase theawareness of the situation in the market andpromote a longer-term view on the part ofmarket participants. The same holds true forcovenants based on balance sheet ratios.Furthermore, the present context of incompletetransparency and disclosure of rating triggersmay be seen as impacting on the price discoverymechanism of fixed income products (and, byextension, equities) as it results in an additionalrisk premium associated with this “ratingtrigger” uncertainty. This in turn may lead to ahigher cost of capital and higher yields thanwould have been the case under a moretransparent framework. Thus, the “benefits” ofthese clauses are not fully exploited. However,if rating triggers were systematically disclosedfrom their inception, this information would bepriced in from the start in bond issues (andstocks) and the number of triggers used in debtissues of any single borrower would probablybe more limited. Moreover, it could also beargued that the expected benefits (for issuers)from these devices would prove illusory, as therelative prices of the various debt instrumentsof an issuer/borrower and its equity price wouldadjust to reflect the existence of rating triggersin some debt instruments, and that the benefits(in terms of favourable financing conditions)

stemming from trigger-carrying instrumentswould be offset by deteriorating financingconditions (and increased volatility) for“unprotected” instruments. It is, of course,unlikely that all rating triggers could bedisclosed, since there are private placementsand bank loan agreements with embeddedoptions. Still, greater transparency should haveboth direct and indirect positive effects oncredit markets.

21 “Survey on Rating Triggers, Contingent Calls on Liquidity”,Standard and Poor’s, 2002.

Page 17: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

16ECBOcca s i ona l Pape r No . 16June 2004

5.1 THE METHODOLOGY: RATING “THROUGH THECYCLE” VERSUS “POINT-IN-TIME”

THE TIME HORIZON OF RATINGS: RATINGS AREORDINAL IN THE SHORT TERM AND CARDINAL INTHE LONG TERM

From an operational standpoint, the purpose ofratings is to measure credit risk in terms ofprobability of default, expected losses orlikelihood of timely payments in accordancewith contractual terms. CRAs are careful tostress, however, that these estimates should notbe seen as a short-term outcome (i.e. one to twoyears), but rather should be considered over alonger horizon.22 Long term generally appearsto mean at least one or two business cyclesAgencies will say that the time horizon isindefinite, but may be thought of as 5 to 10years. The reason for using an indefinitehorizon is that, for a given constant rating, theprobability of default varies with different timehorizons. While agencies have been criticised,and at times rightly so, for being vague as to thetime horizon over which they are rating, itwould appear that if it is assumed that they havealways used a horizon of several years, thentheir various statements are consistent.23

Ratings are a cardinal measure of credit risk ifused over an unspecified long horizon (Keenan,1999, and Brand and Bahar, 1999). Indeed, overthe long term, ratings are found by academicstudies to be an accurate and unbiased estimatorof default probabilities. Thus, while ratings areordinal in their design, associations can bedrawn with cardinal probabilities of defaultin the long term.24 For shorter horizons, there isnot necessarily a stable mapping from ratings toprobability of loss, and the rating becomes anordinal measure of risk. Different agencies usedifferent concepts of loss, although in practicethe differences do not appear to affect theratings outcomes significantly.

FUNDAMENTAL CREDIT ANALYSIS VERSUS MARKETINDICATORS

CRAs compare the fundamental credit analysison which ratings are based with market

sentiment, which is measured by quantitativeindicators such as the market price of corporatebonds and equities, price volatility, thesubordinated debt price, and the credit defaultswap price. CRAs are adamant that they do notdirectly incorporate market sentiment intoratings, although they may use market prices asa diagnostic tool. On the contrary, they takecare to exclude transient market sentiment. Thereason for this is that their clients, in particularportfolio managers, have expressed concernthat the use of market sentiment as an inputwould give rise to greater ratings volatility.Portfolio managers prefer ratings to exhibitstability, i.e. a degree of inertia in ratingschanges. Furthermore, the value of ratingsbased solely on market sentiment/prices wouldbe questionable, as they would not incorporateany information that is not already available tomarket participants.

The graph below summarises the way in whichvarious credit assessment systems differ interms of both the time horizon for theassessment and the resulting ratings volatility.The traditional approach of CRAs is consideredto be closest to a pure through-the-cycleapproach, whereas Merton-type, structuralmodels are closest to a pure point-in-timeapproach. The internal ratings based (IRB)approaches of commercial banks are probablysomewhere between the two latter approaches.Although banks traditionally use a one-yeartime horizon in their probability of defaultestimations, in accordance with the Basel IIConsultative Paper 3, they have to use longer

5 THE CHA L L ENGE : HOW TO R E CONC I L E THE( H E T E ROGENEOU S ) I N T E R E S T S O F D I F F E R ENTU S E R S ?

22 See “The evolving meaning of Moody’s bond ratings”, August1999, “Corporate rating methodology” FitchRatings, June 2003,and “Rating methodology: evaluating the issuer”, Standard &Poor’s, September 2001.

23 In times past, agencies stated that their aim was to rate “throughthe cycle”. This aim was achieved by examining the ability of thecompany to continue servicing its debt under a range of stressfulcredit conditions, both in the macro economy at large and in thespecif ic industry. More recently, they have tended to downplaythe through-the-cycle notion, arguing that business cycles havebecome more irregular.

24 Keenan, S. C. (1999), “Historical default rates of corporatebond issues, 1920-1998” Special Comment, Moody’s InvestorsService, and Brand, L. and R. Bahar (1999), “Ratingsperformance 1998”, Standard and Poor’s Corporation.

Page 18: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

17ECB

Occa s i ona l Pape r No . 16June 2004

5

The challenge:how to reconcile

the (heterogeneous)interests of

different users?

time horizons when assigning ratings. Hence,their probability of default estimates shouldmove away from the estimates of pure Merton-type models. Ultimately, the positioning of eachIRB system on this “through-the-cycle” to“point-in-time” scale will depend on thespecific characteristics chosen by the individualbanks. Rating agency “proxies”25 (“quantitativecredit scoring models”) are positioned betweenIRBs and traditional ratings since, althoughtheir credit assessments should vary more thantraditional agency assessments owing to theirmore frequent revisions, their methodology isdesigned to replicate traditional through-the-cycle ratings (see sub-section 5.2).

THE LIMITS OF “THROUGH-THE-CYCLE”METHODOLOGY

• Increased volatility and downwardmomentum

Several academic studies have examined thebehaviour of credit ratings over time, forinstance through the analysis of credit upgradesand downgrades. Altman and Kao (1992) forexample analyse the stability of newly issuedS&P ratings for two sub-periods (1970 to 1979and 1980 to 1988). They show that for everyrating and time horizon (one to five years)newly-rated issues from the earlier periodexhibit greater stability.26 Lucas and Lonski(1992) examined the credit ratings of more than4,000 rated US and international debt issuesfrom 1970 to 1990 and found that corporatecreditworthiness became more volatile over the

period and that this increased volatility wasaccompanied by a downward trend in ratings.Carty and Fons (1994), using Moody’sdatabase of over 4,700 long-term issues and2,400 short-term issues, found that trends inoverall corporate credit quality, as measured bythe percentage of upgrades and downgrades ofone or more letter, have changed over time.27

They also noted a degree of predictabilityin changes in credit ratings over time, andin particular that rating changes tended toexhibit serial correlation. More specifically, adowngrade is more likely to be followed by asubsequent downgrade than by an upgrade, i.e.credit ratings exhibit downward momentum,which is evident for all grades.28

The existence of momentum in rating changesimplies that the history of past rating actions ofagencies should help predict their futureactions, which may suggest that agency ratingsdo not fully reflect available information. Asexplained by Löffler (2001), the “through-the-

25 These “proxies” are products developed by international ratingagencies (for example, Riscalc by Moody’s, Creditmodel by S&Pand CRS by Fitch) to replicate their traditional through-the-cycle ratings in order to support credit risk management. Theseproducts enable customers to buy the software and run analysesat any time.

26 It should be noted, however, that the composition and size of theratings universe changed over this time period.

27 For example, during the period from 1950 to 1980, on average4.77% of issues changed ratings, while drift averaged a mere-0.07%. During the period from 1980 to 1993, however, theaverage number of issuers experiencing ratings changes rose to12.43%, while drift became more negative at -4.97%.

28 There is some evidence, though less pronounced, of upwardmomentum (see Lando and Skødeberg, 2002, and Bangia et al.,2002).

Chart Point-in-time versus through-the-cycle assessments

SHORTY TIME HORIZON LONG TIME HORIZONHIGH RATING VOLATILITY LOW RATING VOLATILITY

Pure point-in-time

Merton-typemodel

IRB system

Rating agencyproxies

Rating agencies

Pure through-the-style

Page 19: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

18ECBOcca s i ona l Pape r No . 16June 2004

cycle” method of rating, while able to explainimportant stylised facts such as ratingsstability, fails to account for the predictabilityof rating changes. Furthermore, infrequentreviews of ratings cannot explain serialdependence in rating changes. Rating policy orshortcomings in information processing (slowreactions, biases, etc.) can, in principle, be putforward as factors underlying the relativelyweak information content of credit ratings.Differentiating between these alternativeexplanations, however, is a daunting task.Löffler (2003) shows that overlapping ratinggrades in terms of default probabilities, whicharise as a result of the discreteness of the gradesand efforts to avoid “rating bounces” (e.g.fulfilling the market’s “expectation for stableratings”) (Cantor, 2001), would suffice togenerate momentum in rating changes. Blurreddifferences in terms of default probabilitiesbetween adjacent rating categories result fromstickiness in ratings. The ratings overlap thentriggers the subsequent gradual adjustment(momentum).

• Rating changes exhibit a certain degree ofprocyclicality

As shown by Nickell et al. (2000), defaultprobabilities depend strongly on the stage in thebusiness cycle, and transition matrices tend toexhibit a higher frequency of downgradesduring a recession and a higher occurrence ofupgrades during booms. However, withoutfurther conditioning on measures of trueunderlying default risk, which may in part bepro-cyclical, it is not possible to conclude, byconsidering rating transitions in terms of thestate of the business cycle, that ratings areassigned in a procyclical manner, but only thatratings move procyclically. Such evidence mustnevertheless be squared with the claim by themajor CRAs that they rate “through theexpected cycle”.

Amato and Furfine (2002) note that, while theratings of most companies change little,significant evidence, important from both astatistical and economic perspective, points to

ratings exhibiting sensitivity to business cycleconditions. Rating agencies monitor theconditions of companies to a greater or lesserextent at any given time, and generally do notreact to small movements in their risk profiles.This is consistent with the CRAs’ often-statedobjective of taking a rating action only “when itis unlikely to be reversed within a relativelyshort period of time” (Cantor, 2001). However,when rating agencies do make a change, theyoverreact in relation to prevailing conditions,and the nature of this overreaction positivelycorrelates with the state of the aggregateeconomy. This could be the consequence ofexcessive optimism/pessimism during upturns/downturns on the part of rating agencies(Amato and Furfine, 2002).

Empirical models tend to indicate a rise in creditrisk during recessions. For instance, Altman etal. (2002) show that there is a relationshipbetween the correlation of default probabilitiesand loss in the event of default and the businesscycle. These authors argue that models thatassume independence of default probabilitiesand loss-given-default will tend tounderestimate the probability of severe lossesduring economic downturns. A study by Bangiaet al. (2002) demonstrates the empiricalsignificance of the procyclicality of creditquality changes by showing that estimatedcredit losses are much higher in a contractionthan in an expansion. Kavvathas (2001), whomade a systematic study of the variation ofcredit migration (including default risk) overthe business cycle, found that an increase inshort and long term real rates, a lower equityreturn and a higher equity return volatility weregenerally associated with higher conditionaldowngrade probabilities. The accuracy of creditrating transition probability (CRTP) matrixforecasting, thanks to the use of state variableinformation, has generally been improved, bothstatistically and economically, in in-sample andout-of-sample experiments. The statistical andeconomic importance of the term structure andequity return variables give rise to aninterpretation that may defuse some of theabove mentioned criticisms directed at the

Page 20: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

19ECB

Occa s i ona l Pape r No . 16June 2004

5

The challenge:how to reconcile

the (heterogeneous)interests of

different users?

forward-looking nature of the rating activity ofCRAs. The fact that ratings, according to thefindings of Kavvathas (2001), correlatecontemporaneously with market variables in ananticipated fashion, goes some way toaddressing this criticism. Nonetheless,empirical evidence on ratings has to beinterpreted with care, since apparent violationsof informational efficiency could well resultfrom the CRAs’ objectives and constraints.Hence, their performance would need to becompared with an appropriate benchmark.

Recently, literature has also focussed onwhether the severity of the ratings process haschanged over time. Specifically, Blume, Limand MacKinlay (1998) consider whether therecent trends in corporate bond upgrades anddowngrades are the result of the declining creditquality of US corporate debt or whether ratingsstandards have evolved over time. Using ratingsdata from the period 1978-1995, they argue thatrating agencies have become more strict,implying in part that the downward trend inratings is a result of changing standards.

5.2 RECENT DEVELOPMENTS IN METHODOLOGY:STRUCTURAL CHANGES OR REFINEMENT?

International rating agencies strive to providecredit assessments that hold generally steadythrough the course of the business cycle (rating“through-the-cycle”). However, in a “post-Enron” world, rating agencies have been underconsiderable pressure from some investors toprovide more timely and accurate readings ofthe credit outlook (“What’s going on at ratingagencies”, Morgan Stanley, 2003). Perhapspartly induced by this increased pressure onrating agencies, market participants sense achange in rating methodologies with ratingsbecoming more sensitive to the business cycle.

The increased volatility that ratings haveexhibited in the recent period may havecontributed to this perception. Indeed,according to Moody’s, “the last two years(2001 and 2002) have been atypically volatile”

with on average 28% of issuers experiencing arating change of any type, and 8% experiencinga “large” (three or more notches) rating change(the series hit an all-time high in 2001). In anytypical year, the corresponding figures arebelow 25% (rating change of any type) andbelow 5% (large rating change). Consequently,“rating volatility … is currently at the highestlevel observed since 1982”.29

Rating actions can be triggered by two types offactor: changes in rating methodology andchanges in business and economic outlook. Asregards changes in methodology, as the relianceon ratings has been growing, from a broadeningbase of market participants, CRAs are expectedto satisfy a variety of constituencies, withdifferent, if not sometimes conflicting,interests: issuers and “traditional” assetmanagers require more than a simple statementof near-term probability of loss, but stress theneed for ratings to exhibit some degree ofstability over time. As regards the lattercriterion, this desire for stability is alsomotivated by the fact that ratings have becomepervasively embedded in investment guidelinesand bond indices, meaning that volatile orunexpected rating changes can force assetmanagers to sell or buy securities against theirbetter judgement at inopportune times.However, mark-to-market traders, activeinvestors and risk managers require frequentindications of credit changes.

Consequently, the rating process and the rangeof products offered by CRAs have evolved overtime:

• Renewed emphasis on communication

While notches (+/-, 1,2,3) were introducedbetween 1971 and 1982 by the major agencies inresponse to these new demands, outlooks andwatchlists were developed more recently inorder to provide additional signals to ratingusers as to where the balance of probabilitieslies regarding future changes in ratings or to

29 “Measuring the Performance of Corporate Bond ratings”,Moody’s Special Comment, April 2003.

Page 21: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

20ECBOcca s i ona l Pape r No . 16June 2004

attract attention to exceptional rating reviews inthe light of specific developments. Notches andwatchlists may be viewed as attempts to givemore timely indications of changes in creditquality in response to allegations that theagencies lagged market prices and were “behindthe curve”. As such, they are expected tocontribute to smoothing market reactions torating changes by providing early warningsignals. At the same time, rating review periodshave been shortened in response to investorcriticism regarding the lack of timelinessof rating actions. In May 2000 Moody’sannounced an improvement in communicationand transparency of ratings by indicating thelikelihood of future rating changes and theirseverity.

• Emphasis on liquidity risks

Similarly, more emphasis has been placedon liquidity risk, reflecting the recentshortcomings on the part of rating agencies inaccounting for this factor in recent well-publicised defaults. New products have beendeveloped which aim to assess the availabilityof short-term financing for companies and takeinto account the increasing volatility offinancing conditions, especially for speculativegrade issuers: Liquidity risk assessments(LRAs) for issuers of US commercial paperwere introduced in March 2002, the speculativegrade liquidity rating (SGLs) for speculativegrade issuers followed in September 2002.30

SGLs are opinions about an issuer’s ability togenerate cash from internal sources and theavailability of external sources of committedfinance relative to its cash obligations over thecoming 12 months. More specifically, liquidityratings are defined as a measure of the impactthat a loss of access to liquidity would have onan issuer; and the short-term rating is defined asa product of that impact and the probability ofoccurrence of a loss of access.

• Emphasis on “hidden” liabilities

In early 2003 S&P announced changes in theframework for analysing financial measures and

ratios, in order to better reflect the potentialimpact on corporate profitability of pensionliabilities that companies may carry, especiallyin the current environment of low interest ratesand weak stock market performances.

• Increased use of market-based tools

Rating agencies are making more extensiveuse of quantitative and market-based methodsto provide additional perspectives in theircredit risk assessment process. Moody’s, forinstance, is using tools such as KMV to identifymaterial and systematic gaps betweenfundamental ratings and ratings implied bymarket data. Similarly, the renewed focus onevent risks (litigation risks, accountingirregularities, cash/debt-financed M&As andshare repurchase programmes) justifies, inMoody’s view, this increased reliance onquantitative risk models in order to “capture”stock market concerns and better reflect them inratings (see, for instance, “Implications of theacquisition of KMV for Moody’s Ratings”,March 2002).

CRAs have also been developing “agencyproxies”, i.e. quantitative credit scoring modelsthat analyse financial statement data to producedefault probability predictions and/orquantitatively-derived estimates of “traditional”credit ratings. While these quantitativeapproaches are supposed to supplement but notreplace traditional ratings, there appears to besome ambiguity in the definition of the rolethey are expected to play alongside traditionalcredit ratings. For instance, with a view tofacilitating market participants’ comprehensionof the results of the modelling process, theoutput of these models can be expressed usingtraditional rating symbols. Furthermore, asstated by S&P, although scores are not creditopinions, “the scoring models interpret the datain a way that is consistent with how Standard &Poor’s analysts work” and “the models reflect

30 See Moody’s Investors Service, “Moody’s Liquidity RiskAssessments – Q&A”, March 2002, and Moody’s InvestorsService, “Speculative Grade Liquidity Ratings”, September2002.

Page 22: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

21ECB

Occa s i ona l Pape r No . 16June 2004

5

The challenge:how to reconcile

the (heterogeneous)interests of

different users?

Standard & Poor’s specific credit analysisexperience and prospective views of eachindustry.”

All these changes may indicate that creditsignals produced by rating agencies arebecoming both more diverse and generally moreresponsive to current market conditions, whichin turn arouses suspicions that there may be a“post-Enron” regime change in the ratingprocess. At the same time, however, it isinteresting to note that rating agencies arepublicly stating that they are not pursuing activechanges in the way they conduct their ratingprocess. They reaffirm the value of ratingstability and the meaning of ratings as a long-term fundamental credit risk assessment. Ratingagencies believe that the market does not look atratings primarily as buy/sell signals, and doesnot want ratings to be pro-cyclical or to add tomarket volatility.31

Although recent changes and refinements inrating methodologies may have contributed tothe additional volatility in rating actions, theincrease in uncertainty about economic andbusiness events is perhaps the main factorbehind this higher ratings volatility. Ratingagencies recognise that even for a ratingprocess that aims to produce long-term stableratings, periods of heightened credit and eventstress could contribute to a larger number ofrating actions than would have been historicallyexpected on a “normal” basis. On the basis ofthis argument, it could be said that it is not achange in the rating process (i.e. from athrough-the-cycle to point-in-time assessments)that is responsible for the recent higherfrequency of rating actions, but rather theincreased difficulty in seeing through the fog ofeconomic forecasts. It is this difficulty thatmakes the role of rating analysis morechallenging.

31 See, for example, Moody’s Investor Service Special Comment“Understanding Moody’s Corporate Bond Ratings and RatingProcess”.

Page 23: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

22ECBOcca s i ona l Pape r No . 16June 2004

Although the recent increase in rating volatilitymay to some extent have been due to an increasein economic and business uncertainty,questions remain as to whether this uncertaintywill lead the CRAs to adjust the weights theyattach to different objectives – accuracy andstability – in their rating process more actively.Assessing the performance of rating agencieswith regard to these two objectives can be doneeither in relation to the methodology they use(i.e. do ratings provide an accurate and stablepicture of default risk “through the cycle”?)or in relation to alternative credit riskmeasurement techniques (i.e. how do signalsprovided by ratings compare with other signalsof credit risk?). This latter approach isdiscussed below.

Obviously, aiming for accuracy and stability atthe same time involves a trade-off. Moody’sown calculations illustrate that in terms ofstability their ratings outperform implied bondratings (ratings inferred from bond spreads) bya large margin.32 Over the recent period (1999-2002), as a twelve-month average, 25% ofissuers experienced a rating action by Moody’s.However, market-implied rating changesaffected 91% of the issuers. This general resultalso holds for large rating changes (7% against43%) and rating reversals (1% against 76%),which are categories that institutional investorsare particularly concerned with.33

As regards accuracy over a short horizon,however, Moody’s ratings do not seem to matchthose of market-based indicators. For a one-year horizon, for example, bond market-impliedratings are, on average, a better approximationof corporate defaults than Moody’s ratings. Asthe time horizon lengthens, however, the gapbetween these two measures is reduced.

Given their large outperformance with respectto rating stability, it would be interesting toassess whether CRAs are effectively becomingmore concerned with (short-term) accuracy, i.e.their secondary objective. Such behaviour couldbe reinforced by the fact that many marketparticipants and observers criticise rating

agencies for “being behind the curve” or“lagging the market”, which might – although itis the natural outcome of the CRAs’ traditionalapproach – represent a threat to their all-important reputation. In fact Moody’s findsthat, by historical standards, the stability of itsown ratings is currently low. However, it maybe too early to judge at this stage whether this isthe beginning of a trend. This questionnonetheless deserves close monitoring.

CRAs state that they purposely incorporatestability into their ratings in response todemands from their “core” client base. If indeedagencies’ primary client base does not consistof mark-to-market investors but of portfoliomanagers and issuers, then their role is not onlyone of providing economies of scale ininformation but also one of providingmonitoring signals in a principal-agent relation.These two roles, however, necessitate a trade-off. Stability can be seen as a device that affordsthe fund manager greater discretion bysmoothing cyclicality. The signal (the rating)attempts to filter out the noise of everyfluctuation in market sentiment. A comparisonmay be drawn with models that aim to forecastdefault probabilities over a short time horizon(one year). Such models, which draw heavily onmarket prices and exhibit extreme cyclicality,appeal to mark-to-market traders.

The degree of smoothing – the trade-offbetween roles – is therefore essentially a matterof judgement: if changes in credit fundamentalsare considered to be minor or transient, norating change is warranted, and vice versa forlarge and permanent changes. One way tomeasure the degree of smoothing is to assess

6 A C CURA CY, S TA B I L I T Y AND TH E “ R E L AT I V E ”P ROCY C L I C A L I T Y O F R AT I NG S

32 This result is consistent with another line of research whichindicates that credit ratings assigned by the rating agenciesgive rise to much smoother fluctuations in banks’ capitalrequirements within the Basel II framework as compared toMerton-type rating systems like KMV. See, for example, P.Lowe, “Credit risk measurement and procyclicality”, BISWorking Papers No 116, September 2002; E. Catarineu-Rabell etal., “Procyclicality and the new Basel Accord – banks’ choice ofloan rating system”, Bank of England Working Paper No. 181,2003.

33 “Measuring the Performance of Corporate Bond ratings”,Moody’s Special Comment, April 2003.

Page 24: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

23ECB

Occa s i ona l Pape r No . 16June 2004

6

Accuracy, stabilityand the “relative”

procyclicalityof ratings

the volatility of ratings (compared, for instance,to the volatility of market credit spreads). Whatis surprising is that academic testing has, untilrecently,34 virtually ignored this aspect ofratings, preferring to treat them as rivals tomarket prices that are deficient if they do notincorporate every scrap of market information.

34 Löffler, G. (2001), “An Anatomy of Rating through the Cycle”,University of Frankfurt, Frankfurt.

Page 25: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

24ECBOcca s i ona l Pape r No . 16June 2004

7.1 HARDWIRING VERSUS FLEXIBILITY

The paper has argued that ratings and ratingchanges can potentially give rise to specificmarket dynamics. However, it is worthstressing that such dynamics are much morelikely to be triggered by the way ratings areused by market participants, far beyond theirinitial purpose, than by the actual informationcontent of ratings themselves. The hardwiringof rules and regulations into ratings, theproliferation of ratings-based trigger clausesand the extensive use of ratings in assetmanagement have contributed to turning ratingsand rating agencies into structural elements ofmodern financial markets. As a result, ratingchanges may themselves become credit events.In this context, in order to avoid rating changesturning into automatic triggers for portfoliorestructuring and forced sales, it is crucial that,when ratings are enshrined in regulations andrules and hence potentially shape behaviour,enough flexibility is afforded to marketparticipants. The challenge here is to strike theright balance between the benefits ofmonitoring and disciplining that ratings canprovide and the “breathing space” that marketparticipants need in order to conduct theiractivities efficiently. The change depicted in thepractices of the asset management industry, i.e.moving away from strictly rule-based responsesto rating changes, illustrates this search for theoptimum combination.

7.2 IMPLICATIONS OF THE LIMITED SHORT-TERMACCURACY OF RATINGS

Some of the empirical findings presented in thispaper regarding the lack of short-term accuracyof ratings relative to other, market-basedindicators have implications that appear to beproblematic for capital market participants andbank regulators in view of the increasingreliance on external ratings issued by CRAs incredit assessments and in the setting ofeconomic capital by banks and capitalrequirements by the authorities. The largedivide between ratings and market price-based

credit risk measures is manifest in individualcredit spreads, which vary greatly within agiven rating category, with a substantial degreeof overlapping among adjacent categories (somecredit spreads in higher categories are largerthan others in lower categories). As a result,ratings may not be efficient short-termpredictors of default (or credit qualitydeterioration) – and, indeed, they are notdesigned to be. This evidence does notnecessarily contradict the (less robust) findingthat average credit spreads for each ratingcategory increase monotonically going downthe rating scale (i.e. ratings are generallyinformative, as a lower credit ratingcorresponds to a higher probability of default).

It could be argued that market price-basedrating systems may lead to more accurate creditrisk estimates and, in a regulatory capitalsetting, to more timely changes in requiredcapital than systems based on external ratings(or ratings methodologies similar to those usedby the CRAs). This assumes, however, thatfinancial markets offer consistent and reliableleading indicators of the business cycle. Indeed,Moody’s has recently recognised that bondmarket-implied ratings are more powerful thanMoody’s ratings over a one-year horizon.35

However, as some developments documented inthis study have shown, asset prices in general,and credit spreads in particular, incorporate alarge variety of factors in addition to (market)estimates of credit risk. Some of these factorscan be viewed, in the context of the settingof economic capital by banks and capitalrequirements by the authorities, as transientevents or as noise that needs to be “filtered out”.

Ultimately, an appropriate balance must bestruck between the added value thatincorporating relevant market price informationcan bring to the credit assessment process(accuracy) without at the same time contributingto market fluctuations and giving rise to

7 PO L I C Y IMP L I C AT I ON S AND I S S U E S

35 See Moody’s Special Comment report “Measuring theperformance of corporate bond ratings”, April 2003, page 25.

Page 26: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

25ECB

Occa s i ona l Pape r No . 16June 2004

7

Policyimplicationsand issues

unintended fluctuations in capital requirementsin the event of large swings in market sentiment(stability).

7.3 RATING AGENCY INITIATIVES ANDTRANSPARENCY

In the context of the current volatility of ratings,disentangling the impact of economicuncertainty (in itself very hard to assess) frompossible changes in the methodology used byrating agencies, has proved particularlydifficult. The developments reviewed above donot lead to the conclusion that CRAs are in theprocess of changing their methodology infavour of a more “point-in-time” approach.However, CRAs have shown a very “proactive”stance in recent years in refining their approachto credit risk measurement and in adding newproducts and tools to their initial range. From apurely technical point of view, rating agenciesnow possess instruments that could enable themto move towards an increasing use of market-based models, which are currently intended tocomplement but not replace the traditionalapproach to credit rating. Developments in thisarea are of interest to central banks and policymakers.

Faced with numerous criticisms in recent years,rating agencies have made renewed efforts tojustify their actions and make their activitymore transparent and understandable. Resultshave been mixed to date, with communicationon, for instance, the long-term performance ofratings clearly improving. But few insightshave been provided into the rating process itself(i.e. how ratings are arrived at), which is anaspect that rating agencies consider to be to alarge extent proprietary. At the same time ratingagencies have devoted many resources toexpanding the range of rating products andcredit risk modelling tools they offer in order tokeep pace with the needs of market participants.Without additional communication efforts,rating agencies may ultimately be faced with asituation in which market participants willencounter increasing difficulties in

understanding the interconnection betweenthese different rating products, and how theydiffer in the context of an overall consistentapproach. For instance, how do LRAs interactwith CP ratings and issuer ratings? Similarly,the interactions between the agencies’“traditional” ratings approach and their growingpresence in the domain of the quantitativemodelling of credit risk needs to be clarified.Should the perception develop that ratingagencies are sending blurred signals to themarkets (appearing either redundant orincoherent), this would have a negative impacton both the agencies (loss of credibility) and thefunctioning of the market.

7.4 POSSIBLE IMPLICATIONS OF A CHANGE INMETHODOLOGY

All in all, precisely because ratings are widelyused by market participants, a move towards amore market-based methodology would havefar-reaching implications for financial marketsand financial stability, and would be likely toresult in an increase in the risks of extreme pricemovements, especially at the micro level. A keyfeature of ratings is that they contain someinformation that is not publicly known and,furthermore, information which is relevant topricing. Most often, however, ratings seem toincorporate only a small portion of “new”information. Nevertheless, this does not meanthat ratings do not play an important role incorporate bond markets, as they are liable to bevery valuable for less-informed investors giventhat they translate risks into simple letters (i.e. asimple ranking) and offer a long-term analysisbased partly on private information. Moregenerally, because rating agencies provideinformation economies of scale, filtering andextracting noise from market information, it canbe said that they contribute to the informationefficiency of financial markets. In the light ofthe empirical evidence on spreads and spreaddynamics provided above, a change in ratingmethodology towards an exclusively “point-in-time” approach would probably produce ratingswith no pricing-relevant information that was

Page 27: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

26ECBOcca s i ona l Pape r No . 16June 2004

not already provided by market prices. Hence,the challenge facing rating agencies is rather toadequately combine, in their credit assessment,the input that market prices can provide with theprivate information that they gather.Furthermore, should ratings become morepoint-in-time, it is likely that credit spreadswould become more volatile since the marketwould be more frequently surprised.Presumably, however, the increased volatilityof ratings would ultimately lead to changes intheir use. For instance, as regards ratingtriggers, it seems unlikely that the contractsoutstanding and new contracts issued wouldrely solely on ratings if they were state-dependent, i.e. if they changed over thebusiness cycle. Since there is both demand fromlenders and supply from borrowers for option-like features in debt contracts, a more volatilerating environment would probably lead to newsolutions, with less volatile features, whichwould, however, offer lenders the sameprotection and borrowers a lower spread. Undersuch circumstances, it is possible that bothcreditors and borrowers would turn back toconditional rating triggers, such as superpoison put provisions. More generally, shouldratings become more point-in-time, both lendersand borrowers would bear increased costs as aresult of the higher volatility of ratings and adecline in their information content.

Page 28: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

27ECB

Occa s i ona l Pape r No . 16June 2004

8

Appendix8.1 DEFAULT RISK AND THE INFORMATIONALEFFICIENCY OF RATING CHANGES: EVIDENCEFROM CAPITAL MARKETS

A number of academic papers have investigatedthe informational efficiency of ratings inrelation to the level of and changes in defaultrisk. Some of these studies tested theconsistency of ratings across industrialsegments and geographical regions. Ammer andPacker (2000) showed that, in some years, USfinancial companies obtained higher ratingsthan other companies with similar annualdefault risks. Cantor et al. (2001) also examinedinconsistencies across several groups. Thesestudies did not set out to control forinconsistencies across narrower sectors or todetermine any company-specific variables, suchas size or leverage. They only took account ofMoody’s ratings and did not address thequestion of the information provided by creditrating sub-categories.

Galil (2002) examined the quality of corporatecredit ratings in relation to their defaultprediction power. He focused on whetherratings efficiently incorporate publicly-available information at the time of rating, theextent to which rating classifications areinformative and whether rating classificationsare consistent across industries and countries ofincorporation. The results reveal that ratingscould be improved by using publicly-availableinformation such as size, leverage andavailability of collateral. Therefore, combiningsuch public information (industryclassification) with ratings could produce abetter assessment of default risk. Despite thefact that ratings have some undesired qualities,the real informational content of ratings cannotbe disregarded. Ratings provide a betterassessment of default risk than publicinformation alone. This result is consistent withthe findings of Kliger and Sarig (2000) and mayconfirm that CRA activity adds value, eventhough ex-post ratings are not found to beentirely consistent across industries and thenarrowness of rating categories is found to benot particularly informative. Since these results

hold ex-post, the argument that ratings areoptimal ex-ante, even if a sample includesratings over a long period (1983-1993) and theperiod of exposure to default risk is even longer(1983-2000), is hard to refute. For example, thefact that S&P underestimated the risks in someindustries at the time it assigned its ratingsmight have been due to an unexpected shockduring the sample period.

Vassalou and Xing (2003) provide new insightsinto the informational content of bond upgradesand downgrades. They show that default riskvaries too much over time for credit ratings toprovide any useful information about the futuredefault risk of a company. Furthermore, theirresults imply that grouping stocks according totheir credit ratings (A, B, or C) provides almostidentical information about default risk as aclassification of companies into size or book-to-market (BM) tertiles. Using an alternative-to-bond-ratings measure of default risk, theyare able to show that stocks with large increasesin their default risk earn significantly highersubsequent returns than stocks with largedecreases in their default risk. This result isconsistent with economic intuition whichdictates that investors will require a higherreturn to hold stocks with higher (default) risk.They reconcile the two sets of results byintroducing a forward-looking measure ofdefault risk based on the contingent claimsapproach of Merton (1974). This measure,known as the default likelihood indicator (DLI),gives the company’s default probability36 andcan be updated frequently (e.g. every month). Ittherefore stands a chance of providing a betterestimate of the company’s current defaultprobabilities than a bond rating, which istypically not updated more often than once ayear.

Vassalou and Xing compare changes in DLIswith changes in credit ratings. In the case ofdowngrades, the results show that the averageDLI for all downgrades starts increasing about

8 A P P END I X

36 Risk measures of default probabilities along the lines of DLIshave become popular among investors and are regularlysupplied by commercial providers such as KMV.

Page 29: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

28ECBOcca s i ona l Pape r No . 16June 2004

two to three years prior to the downgrade, andreaches its peak at time zero, the date of thedowngrade announcement. This result waslargely to be expected, since some substantialchange in the default risk of a company has tooccur in order for a downgrade to take place.What is surprising, however, is the fact that,following the downgrade, the average DLIstarts decreasing at about the same rate at whichit previously increased. Furthermore, it returnsto almost the same level it had three years priorto the downgrade. In other words, the graph ofaverage DLI as a function of time around thedowngrade (plus-minus 36 months) has aninverted V shape, with the peak at theannouncement date of the downgrade.

The above finding implies that equity returnsfollowing a downgrade should be lower, giventhat the company’s default risk is lower. It alsoimplies that it is important to adjust for thevariation in the DLI when calculating abnormalequity returns following a downgrade. Indeed,if equity returns are adjusted not only for sizeand BM but also for DLI, the short-horizonnegative abnormal equity returns found inDichev and Piotroski (2001) disappear. Somenegative abnormal returns are still found in thetwo to three-year horizon. However, about 42%of stocks with a downgrade experiencesubsequent downgrades in the three-year periodfollowing the initial one. When this fact is alsotaken into account, the economically significantnegative abnormal returns disappearcompletely.

The inverted V pattern in the DLI arounddowngrades is most pronounced for companieswith C-grade debt, with the rate of change indefault risk being particularly high during theyear surrounding the announcement of thedowngrade. The change in default risk in theperiod around the downgrade is lesspronounced in the case of companies with gradeB debt, and non-existent in the case of firmswith grade A debt.

These results are consistent with those ofDichev and Piotroski (2001) in the sense that

they explain why the negative returns followinga downgrade are most pronounced for smallnon-investment grade companies. The reason isthat most companies with low-grade debt aresmall, and the reduction in default riskfollowing a downgrade is steeper in their casethan it is for larger, investment gradecompanies. Therefore, in those cases, it is evenmore important to take into account the DLI ofthe companies in calculating their abnormalreturns.

The picture that emerges in the case of upgradesis quite different from that described above. Theline of average DLI for all companies is almostflat, with a slight dip on the announcement dateof the upgrade. This dip is so small, that itcannot possibly be associated with a significantincrease in subsequent equity returns. Weobserve a rapid decrease in default risk forgrade C companies prior to an upgrade, but theincrease subsequent to the announcement date isagain too small to give rise to large positivereturns.

The asymmetry observed in previous studies inthe reaction of equity returns to downgradesand upgrades can be explained by theasymmetric change in average DLI associatedwith credit rating changes, depending on thenature of the event (i.e. upgrade or downgrade).DLI varies a lot around downgrades, but notaround upgrades. Therefore, adjusting for DLIin calculating abnormal equity returnsfollowing downgrades is essential, whereas it isimmaterial in the case of upgrades, since DLIexhibits little, if any, variation in the latter case.

8.2 CREDIT RISK AND TRANSITION MATRICES

The value of most fixed income securities isinversely related to the probability of default.Thus, fixed income investors are veryconcerned about changes in the value of theirinvestments due to changes in the probability ofdefault, even though actual default seldomoccurs. In fact, fixed income investors may bemore concerned with changes in the perceived

Page 30: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

29ECB

Occa s i ona l Pape r No . 16June 2004

8

Appendixcredit quality of their bond holdings than withactual default, because bond spreads react tocredit risk and affect the performance of bondportfolio managers. Rating migrations, whichoffer one reflection of changes in the perceivedquality of bonds, occur much more frequentlythan defaults.37

Rating agencies regularly measure the historicaldefault frequency of both US and non-UScorporate issuers. While these historical defaultfrequencies are of interest, they are notforward-looking. The same argument applies tohistorical transition matrices computed frompast frequencies of rating migrations. Asmentioned above, the DLI measure of defaultprobability, based on option theory andcomputed from stock market data, can provideinformation about expected changes in creditrisk. Corporate bond spreads should also reflectsuch expected changes in credit risk(migration).

Delianedis and Geske (2003) focus on theinformation contained in (risk neutral) defaultprobabilities, derived according to the Merton(1974) and Geske (1977) models. These modelswere used to estimate a monthly time series ofrisk neutral default probabilities overapproximately 12 years, from 1988 to 1999. Inexamining the changes in these defaultprobabilities before the event of a ratingmigration or default, there appears to besignificant leading information about ratingmigrations and about defaults in these forward-looking risk-neutral probabilities of default.The term structure of default probabilities fromthe Geske model appears to contain additionalinformation. The short-term probability ofdefault from the Geske model appears to containsignificant information about both the defaultevent and the shape of the term structure ofdefault probabilities prior to the actual default.It appears that this short term defaultprobability is able to distinguish impendingcash flow problems for the company.Furthermore, rating migrations and defaults donot appear to be a surprise to the market since

they can be detected months in advance byeither model.

Credit migration or transition matrices, whichdepict the past changes in credit quality ofobligors (typically companies), are essentialinputs in many risk management applications,including portfolio risk assessment, themodelling of the term structure of credit riskpremia, and the pricing of credit derivatives.Risk management tools, such as CreditMetrics,specifically utilise credit migration measures asone of their primary inputs. Ratings changesreflect an agency’s assessment that a company’scredit quality has improved (upgrade) ordeteriorated (downgrade).

The issue of credit quality migration is veryimportant for fixed-income investors,institutions, regulators, and managers of creditrisk. Investors are concerned with the migrationof ratings, because it influences the price of abond. Institutions are concerned with ratingschanges because of internal policies limiting thepercentage of below-investment-grade loansthat banks permit themselves to hold.Regulators are concerned with ratings since insome cases they determine investmenteligibility of assets and valuation for capitaldetermination. In the New Basel Accord (BaselCommittee on Banking Supervision, 2001)capital requirements are driven in part byratings migration. Their accurate estimation istherefore critical.

Transition matrices measure the probability of acredit rating being upgraded or downgradedwithin a specific time period. S&P and Moody’sboth look at the rating migration of creditquality in all ratings categories for various timehorizons, including one, five and ten years, andin some cases longer. The transition matrices

37 In the Delianedis and Geske (2003) study of approximately 12years (1988-1999) of US corporations rated by S&P onCompustat, the number of rating migrations other than to default(1,800) was about 100 times greater than the number ofmigrations to default (18), after screening for sufficient data inthe University of Chicago Graduate School of Business Centerfor Research in Securities prices (CRSP) and Compustat.

Page 31: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

30ECBOcca s i ona l Pape r No . 16June 2004

issued by the major rating agencies include allindustrial and transportation companies,utilities, financial institutions and sovereignsthat have issued long-term debt to the public.Transition matrices are calculated by comparingbeginning-of-period ratings to end-of-periodratings. Transition matrices focus on twodistinct points in time, typically the first andlast day of a year, and ignore any interveningchanges.

S&P transition matrices use the implied seniorunsecured rating of each issuer, regardless ofthe size of a particular issue or the number ofshares outstanding from that particular issue.Similarly, Moody’s relies upon an impliedsenior unsecured rating of the issuer, ratherthan the ratings of individual debt instruments.

Several academic studies have taken a slightlydifferent approach to measuring and reportingrating transitions. Altman (1989) and Altmanand Kao (1992) were the first to take anapproach to constructing transition matriceswhich assesses the changes from an initial bondrating, usually at the time of issuance. Theyargue that this distinction is important becauseof an ageing or seasoning effect that isobservable in the early years after issuance andthat such an effect generally disappears withinfour to five years. This result is intuitivelyappealing because, as Altman (1998) notes, astime passes strong companies are able to call orrepurchase their debt and refinance it withlower coupon issues. Thus, the remaining poolsof issuers naturally display higher default/transition rates. Besides cohort or poolconstruction, there are two other importantdifferences in how Altman and Kao (1992)construct transition matrices. First, Altman andKao transitions are based on the ratings ofspecific issues, rather than the implied seniorunsecured rating of issuers. Second, unlike therating agencies, Altman and Kao (1992) do notinclude the ratings category “withdrawn” whenreporting their transition matrices. The primarydifference that arises when comparing the twoways of constructing transition matrices is thatthe pools or cohorts tracked by the major rating

agencies contain portfolios of both seasonedand new-issue bonds. Issues of constructionaside, comparing transition matrices isproblematic because of the different timeperiods covered by the raters’ data. Moreover,changes in the number and types of debt issues,the industries rated, and initial credit qualityover those time periods exacerbate thedifficulties in making direct comparisonsbetween transition matrices.

8.3 CREDIT MIGRATION ESTIMATES: METHODSAND RISK MEASUREMENT IMPLICATIONS

At least two estimation techniques have beensuggested in the literature: the standardfrequentist (cohort) approach and the duration(hazard) approach. The latter, which uses thetransition information from obligor moreefficiently than does the cohort method, alsoenables proper testing for time homogeneity andnon-homogeneity of the transition matrix (thedistance between dates, but not the datesthemselves, influences the transitionprobability).

The frequentist method, which is the currentindustry standard, estimates the transitionprobability as a simple proportion of companiesat the end of the sampling period (horizon) (e.g.at the end of the year for an annual matrix) withrating j having started out with rating i.Typically, any rating change activity whichoccurs within the period (horizon) is ignored,and companies whose ratings were withdrawnor migrate to “not rated” (NR) status areremoved from the sample. In addition, twocritical aspects are ignored in the cohortmethod: (right) censoring, which means thatwhat happens to the company after the samplewindow closes is not known (e.g. does itdefault right away or does it live on until thepresent), and (left) truncation, which means thatcompanies only enter sample if they havesurvived long enough or have received a rating.These issues are addressed by the durationapproach (see Schuermann and Jafry, 2003) inwhich the estimation method varies, accepting

Page 32: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

31ECB

Occa s i ona l Pape r No . 16June 2004

8

Appendixor relaxing the time homogeneity assumption.Statistically significant differences in migrationmatrices estimated by the cohort and durationapproaches do indeed arise for a one-yearhorizon, which is typical in many riskmanagement applications.38 However, suchdifferences are confined to the homogenousduration and cohort methods; relaxing the timehomogeneity assumption would appear to yieldlittle difference. Thus, the non-Markovianbehaviour hypothesis of the rating processwould not be materially significant. Thisconclusion, however, seems to contrastsomewhat with the estimates reported inKavvathas (2001). Similarly, looking at theeconomic significance of such differences, themeasurement of credit risk capital implies adivergence between the cohort and the durationmethod of the same order as that implied bybusiness cycles which, in turn, is about 40%(excess capital that should be held during arecession over the optimal level set during anexpansion – see Bangia et al., 2002). Creditpricing is also affected substantially when theestimated matrices differ significantly; usingthe “wrong” matrix can lead to mis-pricing byover 50%.

38 This difference is likely to decrease for shorter horizonmatrices (e.g. quarterly or monthly), but increase for longer,multi-year horizons.

Page 33: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

32ECBOcca s i ona l Pape r No . 16June 2004

Altman, E. (1989), “Measuring corporate bond mortality and performance”, The Journal ofFinance, September, pp. 909-922.

Altman, Edward I. (1998), “The importance and subtlety of credit rating migration”, Journal ofBanking & Finance, Vol. 22, Nos. 10 to 11, October, pp. 1231-1247.

Altman, E. and D. L. Kao (1992), “Rating drift in high-yield bonds”, Journal of Fixed Income,Vol. 2, pp. 15-20.

Altman, E. I., B. Brooks, A. Resti and A. Sironi (2002), “The link between default and recoveryrates: implications for credit risk models and procyclicality”, paper presented at the BISconference on “Changes in risk through time: measurement and policy options”, 6 March.

Amato, J. D. and C. H. Furfine (2003), “Are credit ratings procyclical?”, BIS Working Papers,No. 126, Basel.

Ammer, J. and F. Packer (2000), “How Consistent are Credit Ratings? A Geographical andSectoral Analysis of Default Risk”, Board of Governors of the Federal Reserve SystemInternational Discussion Paper No. 668.

Bae, S., D. Klien and R. Padmaraj (1994), “Event-Risk Covenants and Shareholder Wealth: AnEmpirical Investigation”, Financial Management, Winter, pp. 28-41.

Bangia, A., F. X. Diebold and T. Schuermann (2002), “Ratings Migration and the Business Cycle,With Applications to Credit Portfolio Stress Testing”, Journal of Banking & Finance, Vol. 26,Nos. 2 and 3, pp. 445-474.

Barnhill, T. M., F. L. Joutz and W. F. Maxwell (2000), “Factors affecting the yields on noninvestment grade bonds indices: a cointegration analysis”, Journal of Empirical Finance,Vol. 7, pp. 57-86.

Basel Committee on Banking Supervision (2001), “Overview of The New Basel Capital Accord”.

Blume, M., F. Lim and C. MacKinlay (1998), “The declining Credit Quality of U.S. CorporateDebt: Myth or Reality?”, Journal of Finance, August, pp. 1389-1413.

Boot, A. W. and T. T. Milbourn (2002), “Credit ratings as co-ordination mechanisms” DavidsonInstitute, Ann Arbor.

Brand, L. and R. Bahar (1999), “Ratings performance 1998”, Standard and Poor’s Corporation.

Campbell, J. Y. and G. B. Taksler (2003), “Equity Volatility and Corporate Bond Yields”, Journalof Finance, December.

Cantor, R. (2001), “Moody’s Investors Service response to the consultative paper issued by theBasel Committee on Banking Supervision and its implications for the rating agency industry”,Journal of Banking and Finance, Vol. 25, pp. 171-186.

9 RE F ERENCE S

Page 34: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

33ECB

Occa s i ona l Pape r No . 16June 2004

9

ReferencesCantor, R. and F. Packer (1995), “The Credit Rating Industry”, The Journal of Fixed Income,Vol. 5, No. 3, December, pp. 10-34.

Cantor, R., T. Collins, E. Falkenstein, D. Hamilton, C. M. Hu, C. M. Chair, S. Nayar, R. Ray,E. Rutan and F. Zarin (2001), “Testing for Rating Consistency in Annual Default Rates”,Moody’s Investors Service.

Carty, L. V. and J. S. Fons (1993), “Measuring Changes in Corporate Credit Quality”, Moody’sSpecial Report, New York.

Catarineu-Rabell, E. et al., “Procyclicality and the new Basel Accord – banks’ choice of loan ratingsystem”, Bank of England Working Paper No. 181, 2003.

Caton, G. L. and J. Goh (2003), “Are all Rivals Affected Equally by Bond-Rating Downgrades?”,working paper.

Committee on Global Financial Systems, “Incentives structures in Institutional Asset Managementand their Implications for Financial Markets”, March 2003.

Chen, L., D. A. Lesmond and J. Wei (2003), “Bond Liquidity estimation and the Liquidity Effectin Yield Spreads”, working paper.

Collin-Dufresne, P., R. S. Goldstein and J. S. Martin (2001), “The Determinants of Credit SpreadChanges”, Journal of Finance, December, pp. 2177-2208.

Collins, Stephen, Ronald L. Johannes, Ian W. Marsh and Laurence Roberts, “Ratings: A Post-Enron Review”, mimeo, Bank of England, June 2003.

Crabbe, L. (1991), “Event-risk: An Analysis of Losses to Bondholders and “Super Poison Put”Bond Covenants”, Journal of Finance, June, pp. 689-706.

CreditMetrics J.P. Morgan & Co. Incorporated, New York, 1997.

Delianedis, G. and Robert Geske (2003), “Credit Risk And Risk Neutral Default Probabilities:Information About Rating Migrations and Defaults”, mimeo, The Anderson School ofBusiness at UCLA, Los Angeles.

Dichev, I. D. and J. D. Piotroski (2001), “The Long-Run Stock Returns Following Bond RatingsChanges”, Journal of Finance, Vol. 56, No. 1, pp. 173-204.

Ederington, L. H., J. C. Goh and J. Nelson (1996), “Bond Rating Agencies and Stock Analysts:Who Knows What When?”, Journal of Financial and Quantitative Analysis, Vol. 33, pp. 569-585.

Elton, E. J., M. J. Gruber, D. Agrawal and C. Mann (2001), “Explaining the Rate Spread onCorporate Bonds”, Journal of Finance, February, pp. 247-277.

Ferri, G., L. G. Liu and J. E. Stiglitz (1999), “The Procyclical Role of Rating Agencies: Evidencefrom the East Asian Crisis”, Economic Notes, Vol. 28, No. 3, pp. 335-355.

Page 35: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

34ECBOcca s i ona l Pape r No . 16June 2004

Gabbi, G. and A. Sironi (2002), “Which Factors Affect Corporate Bond Pricing? EmpiricalEvidence From Eurobonds Primary Market Spreads”, working paper, Newfin Research Center,Bocconi University.

Galil, K. (2002), “The Quality of Corporate Credit Rating: An Empirical Investigation”, mimeo,Tel-Aviv University, Tel-Aviv.

Geske, R. (1977), “The Valuation of Corporate Liabilities as Compound Options”, Journal ofFinancial and Quantitative Analysis, pp. 541-552.

Griffin, P. A. and A. Z. Sanvicente (1982), “Common Stock Returns and Ratings Changes: AMethodological Comparison”, Journal of Finance, Vol. 37, pp. 103-119.

Hand, John R. M., Robert W. Holthausen and Richard W. Leftwich (1992), “The Effect of BondRating Agency Announcements on Bond and Stock prices”, Journal of Finance, Vol. 47,pp. 733-752.

Hilderman, M. (1999), “Opening the Black Box: The Rating Committee Process at Moody’s”,Moody’s Investors Service, New York, July.

Holthausen, Robert W. and Richard W. Leftwich (1986), “The Effect of Bond Rating Changes onCommon Stock Prices”, Journal of Financial Economics, Vol. 17, pp. 57-89.

Johnson, R. (2003), “An Examination of Rating Agencies’ Actions Around the Investment-GradeBoundary”, working paper, Federal Reserve Bank of Kansas City.

Kavvathas, D. (2001), “Estimating Credit Rating Transition Probabilities for Corporate Bonds”,University of Chicago, Chicago.

Keenan, S. C. (1999), “Historical default rates of corporate bond issues, 1920-1998”, SpecialComment, Moody’s Investors Service.

Kliger, D. and O. Sarig (2000), “The Information Value of Bond Ratings”, Journal of Finance,December, pp. 2879-2902.

Lando, D. and T. M. Skødeberg (2002), “Analyzing rating transitions and rating drift withcontinuous observations”, Journal of Banking and Finance, Vol. 26, Nos. 2 and 3, pp. 423-444.

Levy, S. (2002), “The development of contingency clauses: appraisal and implications for financialstability”, Financial Stability Review, Banque de France, November.

Löffler, G. (2001), “An Anatomy of Rating through the Cycle”, University of Frankfurt,Frankfurt.

Löffler, G. (2003), “Avoiding the Rating Bounce: Why Rating Agencies are Slow to React to NewInformation”, Journal of Banking and Finance.

Page 36: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

35ECB

Occa s i ona l Pape r No . 16June 2004

9

ReferencesLowe, P. (2002), “Credit risk measurement and procyclicality”, BIS Working Paper No. 116,September.

Lucas, D. J. and J. G. Lonski (1992), “Changes in Corporate Credit Quality 1970-1990”, Journalof Fixed Income, pp. 7-14.

Merton, R. C. (1974), “On the Pricing of Corporate Debt: The Risk Structure of Interest Rates”,Journal of Finance, Vol. 29, pp. 449-470.

Moody’s Investors Service, “Measuring the performance of corporate bond ratings”, April 2003.

Moody’s Investors Service, “Understanding Moody’s Corporate Bond Ratings and RatingProcess”, May 2002.

Moody’s Investors Service, “The Unintended Consequence of rating triggers”, December 2000.

Nickell, P., W. Perraudin and S. Varotto (2000), “Stability of Rating Transitions”, Journal ofBanking and Finance, Vol. 24, pp. 205-229.

Resti, A. and A. Sironi (2002), “The Basel Committee Proposal on Risk-Weights and ExternalRatings: What Do We Learn from Bond Spreads?”, working paper, Bocconi University, Milan.

Schuermann, T. and Y. Jafry (2003), “Measurement and Estimation of Credit Migration Matrices”,Federal Reserve Bank of New York, New York.

Standard & Poor’s, “Playing out the Credit Cliff Dynamic”, December 2001.

Standard & Poor’s, “Survey on Rating Triggers, Contingent Calls on Liquidity”, 2002.

Vassalou, M. and Y. Xing (2003), “Equity returns following changes in default risk: New insightsinto the informational content of credit ratings”, mimeo, Graduate School of Business,Columbia University, New York.

Page 37: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto
Page 38: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

37ECB

Occa s i ona l Pape r No . 16June 2004

EUROPEAN CENTRAL BANKOCCASIONAL PAPER SERIES

1 “The impact of the euro on money and bond markets” by J. Santillán, M. Bayle andC. Thygesen, July 2000.

2 “The effective exchange rates of the euro” by L. Buldorini, S. Makrydakis and C. Thimann,February 2002.

3 “Estimating the trend of M3 income velocity underlying the reference value for monetarygrowth” by C. Brand, D. Gerdesmeier and B. Roffia, May 2002.

4 “Labour force developments in the euro area since the 1980s” by V. Genre andR. Gómez-Salvador, July 2002.

5 “The evolution of clearing and central counterparty services for exchange-tradedderivatives in the United States and Europe: a comparison” by D. Russo,T. L. Hart and A. Schönenberger, September 2002.

6 “Banking integration in the euro area” by I. Cabral, F. Dierick and J. Vesala,December 2002.

7 “Economic relations with regions neighbouring the euro area in the ‘Euro Time Zone’” byF. Mazzaferro, A. Mehl, M. Sturm, C. Thimann and A. Winkler, December 2002.

8 “An introduction to the ECB’s survey of professional forecasters” by J. A. Garcia,September 2003.

9 “Fiscal adjustment in 1991-2002: stylised facts and policy implications” by M. G. Briotti,February 2004.

10 “The acceding countries’ strategies towards ERM II and the adoption of the euro:an analytical review” by a staff team led by P. Backé and C. Thimann and includingO. Arratibel, O. Calvo-Gonzalez, A. Mehl and C. Nerlich, February 2004.

11 “Official dollarisation/euroisation: motives, features and policy implications of current cases”by A. Winkler, F. Mazzaferro, C. Nerlich and C. Thimann, February 2004.

12 “Understanding the impact of the external dimension on the euro area: trade, capital flows andother international macroeconomic linkages“ by R. Anderton, F. di Mauro and F. Moneta,March 2004.

13 “Fair value accounting and financial stability” by a staff team led by A. Enria and includingL. Cappiello, F. Dierick, S. Grittini, A. Maddaloni, P. Molitor, F. Pires and P. Poloni,April 2004.

14 “Measuring Financial Integration in the Euro Area” by L. Baele, A. Ferrando, P. Hördahl,E. Krylova, C. Monnet, April 2004.

Page 39: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

38ECBOcca s i ona l Pape r No . 16June 2004

15 “Quality adjustment of European price statistics and the role for hedonics” by H. Ahnert andG. Kenny, May 2004.

16 “Market dynamics associated with credit ratings: a literature review” by F. Gonzalez, F. Haas,R. Johannes, M. Persson, L. Toledo, R. Violi, M. Wieland and C. Zins, June 2004.

Page 40: Market dynamics associated with credit ratings: a literature review · 2004. 6. 22. · by Fernando Gonzalez, François Haas,Ronald Johannes,Mattias Persson, Liliana Toledo,Roberto

Recommended