who saw sovereign debt crises coming?

OECD DEVELOPMENT CENTRE

WhO saW sOVEREigN DEbT CRisEs COMiNg?by

sebastin Nieto Parra

Research area:Latin american Economic Outlook

November 2008

Working Paper No. 274

Who Saw Sovereign Debt Crises Coming?

DEV/DOC(2008)10

DEVELOPMENTCENTREWORKINGPAPERS

This series of working papers is intended to disseminate the Development Centresresearch findings rapidlyamong specialists in the field concerned.ThesepapersaregenerallyavailableintheoriginalEnglishorFrench,withasummaryintheotherlanguage.

Comments on this paper would be welcome and should be sent to the OECDDevelopment Centre, 2, rue Andr Pascal, 75775 PARIS CEDEX 16, France; or [email protected]. Documentsmay be downloaded from: http://www.oecd.org/dev/wp orobtainedviaemail([email protected]).

THEOPINIONSEXPRESSEDANDARGUMENTSEMPLOYEDINTHISDOCUMENTARETHESOLERESPONSIBILITYOFTHEAUTHORANDDONOTNECESSARILYREFLECTTHOSEOFTHEOECDOROFTHEGOVERNMENTSOFITSMEMBERCOUNTRIES

CENTREDEDVELOPPEMENTDOCUMENTSDETRAVAIL

Cette srie de documents de travail a pour but de diffuser rapidement auprs desspcialistesdans lesdomaines concerns les rsultatsdes travauxde rechercheduCentrededveloppement.Cesdocumentsnesontdisponiblesquedans leur langueoriginale,anglaisoufranais;unrsumdudocumentestrdigdanslautrelangue.

ToutcommentairerelatifcedocumentpeuttreadressauCentrededveloppementdelOCDE,2,rueAndrPascal,75775PARISCEDEX16,France;[email protected]:http://www.oecd.org/dev/wpouobtenusvia leml([email protected]).

LESIDESEXPRIMESETLESARGUMENTSAVANCSDANSCEDOCUMENTSONTCEUXDELAUTEURETNEREFLTENTPASNCESSAIREMENTCEUXDELOCDEOUDESGOUVERNEMENTSDESESPAYSMEMBRES

Applicationsforpermissiontoreproduceortranslateallorpartofthismaterialshouldbemadeto:HeadofPublicationsService,OECD

2,rueAndrPascal,75775PARISCEDEX16,France

OECD2008

2 OECD2008

OECD Development Centre Working Paper No. 274

DEV/DOC(2008)10

TABLEOFCONTENTS

ACKNOWLEDGEMENTS..........................................................................................................................4

PREFACE.......................................................................................................................................................5

RSUM........................................................................................................................................................6

ABSTRACT....................................................................................................................................................7

I.INTRODUCTION.....................................................................................................................................8

II.REVIEWOFTHELITERATURE...........................................................................................................9

III.DESCRIPTIONOFTHEDATA..........................................................................................................11

IV.THESOVEREIGNDEBTCRISES......................................................................................................15

V.THEPRIMARYBONDMARKETINSOVEREIGNDEBTCRISES...............................................18

VI.THEUNDERWRITINGFEEANDFINANCIALMARKETSACTORS.......................................31

VII.CONCLUSIONS..................................................................................................................................34

ANNEXES....................................................................................................................................................35

REFERENCES.............................................................................................................................................50

OTHERTITLESINTHESERIES/AUTRESTITRESDANSLASRIE..............................................54

OECD2008 3


DEV/DOC(2008)10

ACKNOWLEDGEMENTS

SebastinNietoParraisaResearchEconomistattheOECDDevelopmentCentreandtheInstitutdtudesPolitiquesdeParis(SciencesPoParis).Email:[email protected]

Financial support from Chaire Finances Internationales, SciencesPo Paris, is gratefullyacknowledgedfortheinterviewscarriedoutinNewYorkwithinvestmentbanksandinvestors.The author is especiallygrateful toMarcFlandreau for illuminating anduseful commentsonpreviousversionsofthispaper.HealsowishestothankThomasDickinsonforexcellentresearchassistance provided during interviews in New York in 2007 with market participants in theprimary sovereign bondmarket (institutional investors aswell as origination departments ofinvestmentbanks)andestablishedwithinthecontextoftheMasterEconomicsofInternationalDevelopment SciencesPo Paris. The author is grateful to Jeff DaytonJohnson, Ugo Panizza,Helmut Reisen and Javier Santiso for valuable feedback. Comments and suggestions fromseminarparticipantsat the15thWorldCongressof the InternationalEconomicAssociation inIstanbul in June 2008 are acknowledged. Errors, shortcomings and views remain theresponsibilityoftheauthor.

4 OECD2008


DEV/DOC(2008)10

PREFACE

The causes and consequences of sovereign debt crises have been widely studied inresearch literature as important items in the stability and credibility of economic policies inemergingcountries.

Followingaseriesofcapitalmarketcrisesstartinginthemid1990s(Mexico199495,Asia1997,Russia1998andArgentina2001),researchers,policymakersandcapitalmarketactorshaveexploreddeterminantsandmorepreciselyearlywarningindicatorsofthesecrises.

Thispapermakesausefuland illuminating contribution to this literaturebyanalysingrecent sovereign debt crises through the structure of primary bond markets. It analyses theconnectionbetweeninvestmentbanks(basedinOECDcountries)andemergingcountries(mostofthemnonOECDcountries)throughthearenassovereignbondmarket.

Havingconstructedtheannualfeepaidbygovernmentstoinvestmentbanksinordertoissuesovereignbonds in internationalcapitalmarkets,theauthorfindsthatthiscommission ishighbeforedebtcriseswithrespect to theaverage feepaidby issuersduring thesample timeperiod.Moreover, the ratio fee/sovereign bond spreads increases considerably before a crisissuggestingthatinvestmentbanksholdaninformationadvantageoverinvestors.

The relevant information obtained from primary bond markets calls for increasedmonitoringofthefeepaidbygovernmentstoinvestmentbanksduringsovereignbondissuesbypolicymakersaswellasbycapitalmarketactors.

JavierSantisoDirector,OECDDevelopmentCentre

Chair,OECDEmergingMarketsNetwork(EmNet)

OECD2008 5


DEV/DOC(2008)10

RSUM

Cet article apour objectifdanalyser les crisesdedette souverainependant lapriode19932006 partir dumarch primaire souverain.Deux principales conclusions ressortent decette tude. Premirement, les banques dinvestissement valuent les risques de dfaut bienavant les crises et avantmmeque les investisseursne les anticipent.Un trois ans avant ledbutde la crise, lespaysquiprsententun risquededfaut souverain lev commencent verser,enmoyenne,1.10pourcentdesmontantsquilsmettentauxbanquesdinvestissement,soitprsdudoubledumontantmoyenqueversent lensembledespaysmergentspendant lapriodede ltude (0.56pourcent).En revanche, ilsbnficient, avant ledbutdes crises,desprimesde risquequinesontque lgrementsuprieurescellesdu restedespaysmergents(385 contre 319pointsde base).Ce rsultat suggre que les banquesdinvestissement ontunavantagedinformationparrapportauxinvestisseursetquellessontlesseulestirerprofitdurisque de crise de dette souveraine. Deuximement, le comportement des banquesdinvestissementdiffreselonletypedecrise.Avantlescrises,ellesprennentunecommissiondesouscription plus leve pour les pays qui prsentent des difficults en matire de financespubliquesquepourlerestedespaysquiontconnuunecrisededettesouveraine.Larobustessede ces rsultats est vrifie partir dune analyse de donnes de panel. Ces rsultats sonttonnants en ce quils indiquent que les investisseurs nutilisent pas linformation utile etpubliqueleurdispositionpouramliorerlallocationdeleursactifsentitresmisparlespaysmergents.Motscls:Information,MarchPrimaireSouverain,CrisesdeDetteSouveraine,CommissiondeSouscriptionClassificationJEL:F34,G11,G14,G15,G24

6 OECD2008


DEV/DOC(2008)10

ABSTRACT

Thispaperstudiessovereigndebtcrisesduringtheperiod19932006throughtheprismofthe primary sovereign bond market. Two conclusions emerge. First, investment banks pricesovereigndefaultriskwellbeforecrisesoccurandbeforeinvestorsdetectdefaultrisk.Betweenthree and one years prior to the onset of a crisis, sovereign default risk countries paid toinvestmentbanksonaverage1.10percentoftheamountissued,closetodoubletheaveragepaidby emerging countries overall in the same period (0.56per cent). In contrast, the level ofsovereign bond spreads prior to crises is on average only slightly higher than for emergingcountries (385 vs. 319 basis points), suggesting that investment banks have an informationadvantage with respect to investors and are the only parties compensated for the risk ofsovereign debt crises. Second, investment banks behaviour differs depending on the type ofsovereigndebtcrisis.Beforecrises,investmentbankschargedonaverageahigherunderwritingfee to countries presenting public finances difficulties than to other sovereign debt crisiscountries.Therobustnessoftheseresultsisverifiedthroughpaneldataanalysis.Theresultsarepuzzling in that they indicate that valuable, publicly available information is not tracked byinvestorstohelpimproveallocationoftheiremergingmarketfixedincomeassets.Keywords:Information,PrimaryBondMarket,SovereignDebtCrises,UnderwriterSpreadJELClassification:F34,G11,G14,G15,G24

OECD2008 7


DEV/DOC(2008)10

I.INTRODUCTION

Thispaperstudiesrecentsovereigndebtcrisesthroughthestructureoftheprimarybondmarket. The evidence generated by viewing debt crises from this perspective reveals thepossibilitythatinvestmentbankshaveaninformationaladvantagewithrespecttoinvestors.

We examine the behaviour and interactions between the three principal actors in thesovereignbondmarketgovernments,investmentbanks/leadmanagers1andinvestors)priortoandfollowingasovereigndebtcrisis.Tothisendweanalyseimportantsourcesofinformationrelated to price formation in the primary market for sovereign bonds, information that isneverthelessgenerallyoverlooked in theresearch literature.Chiefamong thesevariables is thefeegovernmentspay investmentbanks toplacebonds (i.e. theunderwriting spread), and theprimarypriceatwhichinvestorspurchasethesebonds.

Theempiricalevidence thatwepresent in thispapersuggests thatwecannotreject thehypothesis that investment banks were better informed than investors prior to the abovementionedcrises,and inparticular incrises inwhichsovereigndefaultriskwashigh. Inotherwords, investmentbanksdetectbadcountrieswellbefore investorsdoandbetweenoneandthreeyearspriortotheonsetofcrises.Weanalyseinvestorsperceptionofrisk(i.e.bondspreadson theprimary and secondarymarket),which remained relatively stable before the crises, inrelation to the fee investment banks required from governments in order toplace the bonds,which was high over the same precrisis period2. This is a puzzle because it appears thatinvestorsarenotusingpotentiallyuseful(andpublic)informationinordertoallocateefficientlytheirportfoliosofemergingfixedincomeassets.

Theremainderofthisarticleisorganisedasfollows.InSectionII,weprovideareviewofthe literature. Section III describes the datasets used. Section IV discusses the definition andtimingof sovereigndebtcrises.SectionV, thecoreof thepaper,analyses theefficiencyof thesovereignbondmarket:wecomparetheunderwritingfeepaidbygovernmentstoleadmanagersto the price of the sovereign bonds paid by investors on the primary market and prior tosovereigndebtcrises.SectionVIprovidesanoverviewontheinformationavailabletoinvestorsconcerning the primary bond market. Finally, Section VII provides concluding remarks andsketchesthemajorpolicyimplicationthatfollowsfromthisresearch.

1. Cf.sectionV.LeadManagersareinvestmentbanksinthesovereignbondmarket.InthispaperweusethetermLeadManagerto

refertoagentswhoplacebondsinthemarket.WedonotdifferentiatebetweenUnderwriters,LeadManagersorBookRunnersandweassumethatthesethreeagentshavethesameresponsibilitywithrespecttotheissueduringasovereignbondissue.

2. Weaddress thesequestions throughstudying the interactionbetweengovernments, investmentsbanksand investors.We takeinvestorsasa singlegroup (e.g.noaccount ismade fordifferences thatmight stem from foreignvs.domestic, individualvs.institutional or international vs. local). Differences in investor behaviour have been researched, notably by Calvo (1999),BorenszteinandGelos(2000)andSantiso(2003).

8 OECD2008


DEV/DOC(2008)10

II.REVIEWOFTHELITERATURE

Empirical literature on the structure of the primary sovereign bond market is fairlyscarce3. This is particularly true for research concerning potential information problems inemergingmarketsasevidencedintheprimarymarket.Availableresultsareneverthelessusefuland reveal the presence of information problems between investment banks, acting asunderwritersandinvestors.

Information asymmetries between investment banks and investors are not new; theproblem is clearly illustrated by the Baring crisis case of 1890.According to Flores (2007),Baringinvestmentbank,whichhadacloseandsuccessfulrelationshipwithArgentina,wasthemain sourceof information reaching investorson that country at the time.WhenArgentinassituationdeteriorated,investorsgavemorecredencetoBaringscontinuingpositivesignalsthantoother,publiclyavailablesignsofadownturn.ConcerningtheLatinAmericansovereignbondmarkettoday,NietoParraandSantiso(2007)findthatwhenaninvestmentbankisactingasleadmanager,90percentofitsrecommendationsarepositive.

Moreprecisely,studiesofthisconflictofinterestduringrecentsovereigndebtcrisesarelargely limited to anecdotal evidence. Analysing the Mexican crisis of 1994, Edwards (1997)addresses a number of questions regarding the predictability of the crisis and asymmetricinformationproblemsbetweenWallStreetanalystsandtheirclients.Blustein(2003),describingtheArgentinacrisisof2001,similarlypoints to theconflictof interest facing investmentbanks.Calomiris(2003),referringtoemergingcrises,notesthepossiblecooperationbetweenresearchdepartmentsandoriginationdepartmentswhosetaskistoissuenewdebtofferingsinemergingeconomies4.Although these are important points, they are not accompanied by a systematicanalysisofthestructureoftheprimarybondmarket.Thepurposeofthispaperistofillthisgap,by studying the interaction between investors and investment banks around sovereign debtcrisesandthroughprimarybondmarket.

Thescantresearchliteratureontheprimarysovereignbondmarketstandsincontrasttotheplentifulstudiesdetailingtheprimarycorporatemarket.Inparticular,there isausefulandrich literatureon thedeterminantsof the fee (i.e.underwritingspread)paidby issuers to lead

3. Analysesconcerningthestructureoftheprimarysovereignbondmarketinemergingcountriesarerareandtheyareconcentrated

ontheaccessofemergingcountriestotheinternationalbondmarket(Grigorian,2003;Gelos,SahayandSandleris,2004;FostelandKaminsky,2007)ordescriptiveaspectsrelatedtotheprimarymarket(seeZervos,2004forthecaseofBrazil,ChileandMexico).

4. Moreover,BeimandCalomiris (2001)note thathow tomonitor themonitors isan importantconcernandone thatmustbecarefullytakenintoaccountbyinternationalorganizationsandpolicymakersinassessingthetransparencyofinvestmentbanksanalysis.

OECD2008 9


DEV/DOC(2008)10

managersinthatmarket.Theseanalysestakeintoaccountvariablesrelatedtothecharacteristicsoftheissue(e.g.maturity,amount,underwriters,regulationandcurrencydenomination),oftheissuer(e.g.creditriskmeasuredinmostcasesbythecreditrating,sizeofthefirm,profitabilityindicators and activity sector group) and finally exogenous variables (e.g. secondary marketconditions and volatility of the prices) as factors potentially determining the behaviour ofunderwritingspreads.Arelevantfindingofthecorporateliteratureforthisresearchconcernstherelationship between the underwriting fee and credit risk. These studies generally find aninverserelationshipbetween thequalityof the issuerandthe levelofunderwriting fees (West,1967;HigginsandMoore,1980;RogowskiandSorensen,1985;Lee etal.,1996;LivingstonandMiller, 2000; Kollo and Sharpe, 2002; Melnik and Nissim, 2003; Hua Fang, 2005). This isinterpretedasaconsequenceofthegreatereffortrequiredfromintermediarieswhentheyactasunderwritersoflowerqualityissues(seeAltinkihcandHansen,2000).

10 OECD2008


DEV/DOC(2008)10

III.DESCRIPTIONOFTHEDATA

Theperiodanalysedinthispaperisfrom1993to2006andthefrequencyofobservationsisannual.The29emergingeconomiesincludedinthisstudycorrespondfirsttoeconomiesthatare included intheEMBIGlobal indexandsecondtocountriesforwhichwehave informationonunderwritingfees5.

Themostimportantpieceofinformationpresentedinthispaperconcernsthestructureoftheprimarysovereignbondmarket.For thatpurposeweemploy theDealogicdatabaseDCMAnalytics6, tracking emerging sovereignbond issues from1993 to thepresent.We reduce thenumber of sovereign bond issues in our analysisdataset according to specific criteria,whichallowsustouseonlystandardissuesintheanalysis:i)WetakeintoaccountonlysovereignbondissuesforwhichwehavetheISIN(InternationalSecuritiesIdentifyingNumber)referenceoftheissues aswell as thevalueofboth theunderwriting fee and theprimarybond spread. ii)Weexcludeissueswithfloatingcouponrateswhichalterthetruevalueofthebondspread7.iii)WeexcludeissuesdenominatedincurrenciesotherthanEUR,JPYandUSD,whichareinanycasethe most commonly used currencies in the sovereign bond market8. iv)We exclude issuespartiallyortotallyguaranteedbyinternationalorganisations,suchastheWorldBankorregionaldevelopmentbanks.

TheDealogicdatabase contains 919 sovereign bond issues corresponding to theEMBIglobalcountriespresentedabove.Werestrictthenumberofsovereignbondissuesemployedinthis study to436 issues9.From this sample,149bondsare issuedduring the11yearwindowaroundsovereigndebtcrises(fromyear5priortotheonsetofthecrisestotheyear5afterthe

5. The29emergingeconomiescoveredinthispaperare:Argentina,Brazil,Bulgaria,Chile,China,Colombia,DominicanRepublic,

Ecuador,Egypt,ElSalvador,Hungary, Indonesia,Lebanon,Malaysia,Mexico,Morocco,Pakistan,Panama,Peru,Philippines,Poland,Russia,SouthAfrica,Thailand,Turkey,Ukraine,Uruguay,VenezuelaandVietnam.

6. Formoreinformationseehttp://www.dealogic.com/

7. For thiskind of issues theprimary sovereignbond spread reported is thebasispoints added to thebenchmark rateused todetermine thecouponrate.For instance, theBrazilianGlobalBond21/06/04 (ISINnumberUS105756BC32)hasacouponrate3monthsLibor+575basispoints.Consequentlytheprimarysovereignbondspreadreportedis575basispoints.

8. Weexcludeissuesdenominatedinotherinternationalcurrencies(e.g.DEM,FFR,LIT,SFRandSTG)andlocalcurrencyissuesinordertoavoidtheadditionalcosttoissueincurrenciesthatarenotcommoninthesovereignbondmarket.

9. Thenumberofbondsissuedbyyearfrom1993to2006is(thefirstis1993andlastis2006):14,7,10,19,30,34,56,55,48,27,35,42,37and22.Inparticular,wenotealargenumberofbondsissuedin1999(56)and2000(55).ThisisdueinalargeparttoArgentinawhichwasanactiveplayerintheinternationalbondmarket(17and12bondswereissuedbyArgentinaduringthesetwoyearsrespectively).

OECD2008 11


DEV/DOC(2008)10

onset of the crises, centred on the onset of crises).From these issueswe calculate the annualaverageofthefee10aswellastheprimarysovereignbondspreadofemergingcountries11.

Table1presentsadescriptionbycountryof the sovereignbondsused in thesample12.ThetotalamountofsovereignbondsusedinthissampleishigherthanUSD300billionandtheaverageamountissuedbycountryandperissueisclosetoUSD700million.Inthissample,thetotal income received by underwriting banks is over USD1.5billion (on average close toUSD3.5millionper issue)13.The averagesof the fee andof the sovereignbond spread in thesamplearerespectively0.54percentoftheamountissuedand329basispoints(bp).

10. Fortheremainderofthispaper,weusethetermfeetorefertotheremunerationpaidbyissuerstounderwritingbanksinthe

primarymarket.Othertermsareusedintheresearchliteratureorinthefinancialjargon(e.g.underwritingspread,underwritingfee,grossspreadandunderwritingdiscount).

11. Theprimarysovereignbondspread isdeterminedfromthepricedefinedduringthebookbuildingprocess.Moreprecisely, inordertointroducethesovereignbondstothemarket,leadmanagersrealizethesocalledbookbuildingprocess,inwhichtheybuildupalistofordersataspecifiedprice.

12. 67percentoftheseissuesaredenominatedinUSD,27percentofaredenominatedinEURandtherestaredenominatedinJPY.

13. Thetotalfeeiscalculatedbytheproductofthedealvalueoftheissue(facevalue)andthefee.

12 OECD2008


DEV/DOC(2008)10

Table1.Descriptionofthesovereignbondsusedinthesample(19932006)

CountryArgentina 53 9,2 36.233,6

36.205,3 710.5 842.0

344,2 217,3

1,21 449Brazil 44 12,4

20.8 0,63 459

Bulgaria 3 9,7 220,8 .964,2

2 92.8

8,1 3,9

0,55 340Chile 5 7,7 3

10.634,9 7 590.8

2 34,7

0,30 159China 18 14,7

7.733,6 54,2 1,4 0,48 104

Colombia 32 13,8 1

5 00,0

8

0,66 446Dominican Rep. 1 5,0 500,0

5

2,5

0,50 569Ecuador 1 5,0 497,9

.993,9 497,9 .497,0

2,5 ,0

0,70 470Egypt 2 7,5 2

1

7

0,45 305El Salvador 4 19,0 1.703,9

1.582,6 426,0 91,0

7,7 8,3

0,51 339Hungary 13 7,7 1

8 1.089,8

2

0,36 59Indonesia 4 15,1 4.359,3

5.722,1 6,5

8,5 0,28 278

Lebanon 27 6,0 1

582,3 .427,1

4 7,1

0,50 387Malaysia 5 8,6 7.135,3

34.453,9 1 1.188,1

1 170,1

0,44 220Mexico 30 11,3

0,57 263

Morocco 2 5,0 611,6 49,5

305,8 24,8

2,2 ,6

0,50 142Pakistan 2 5,0 6

6.529,3 3 502,3

2 33,0

0,50 378Panama 14 18,1

.922,1 90,3 0,9 0,55 348

Peru 8 17,4 3 9.169,2

4 67,6

1 1,4

0,28 432Philippines 37 12,1 2

20.825,3 7 1.041,3

6 34,2

0,38 397Poland 20 11,4

.156,9 03,7 0,30 76

Russia 6 9,4 4.627,8 .956,1

1

1

0,96 597South Africa 11 9,0 6

1.597,5 632,4 399,4

3

2,5

0,52 231Thailand 4 5,7

0.971,3 44,9 5,6

95,8 0,45 51

Turkey 55 8,8 4

7 50,8

1

0,61 482Ukraine 4 6,4 1.803,3

4.321,1 4 240,1

9,5 22,2

0,65 528Uruguay 18 11,3

.182,5 71,1 0,2 0,59 263

Venezue laVietnam

12 10,6 5

4 36,7

3

0,70 5251 10,2 7

36,7

7

4

,9

0,65 256

Total 436 10.1 3 1.844.71 692.2 1.548.3 0.54 329

Total fee (US$ millions)

Underwriting spread

(average; % of amount issued)

Bond spread (average;

bp)

Maturity (average;

years)Number of

bonds

Total amount issued (US$

millions)

Amount issued (average; US$

millions)

Notes:Theamountissuedcorrespondstothedealvalueoftheproceeds.Forthetotalsample,theaverageofthematurity,theaverageoftheamountissued,thefeeaswellasthebondspreadsarecalculatedasthesimpleaverage(averageofcountriesinthesample).Source:AuthorscalculationsbasedonDealogicdatabase.

Table 2 shows the investment bankmarket share for the top ten bookrunners of theemergingsovereignbondsusedinthesample.Thesebanksrepresentthemostprestigiousactorsintheemergingsovereignbondmarketasthereputationofinvestmentbanksismeasuredintheresearch literatureaswellas inthecapitalmarketsbythemarketshare14.Thenumberof leadmanagers in theemergingsovereignbondmarket issmall.Similarly to theUScorporatebondmarket (LivingstonandMiller,2000;HuaFang,2005),approximately90percentof the issues

14. Thereisavastresearchliteraturethatusesasproxyofreputationthemarketshare(seeforinstance,MegginsonandWeiss,1991;

LivingstonandMiller,2000;HuaFang,2005).Forthecaseofcapitalmarkets,seeforinstance,BloombergMarkets(2006).

OECD2008 13


DEV/DOC(2008)10

were realized by the top ten bookrunners and more than 75per cent by the seven mostimportantbookrunners.

Table2.MarketShareforTopTenInvestmentBanks(19932006)

Investment Bank Africa Asia Europe Latin America Middle East Total

JP Morgan

Citigroup 9.4 (1)10.0 (2) 19.6 (29) 18.9 (34) 22.4 (65) 2.0 (2) 19.5 (132)

13.9 (16) 15.6 (25) 12.4 (43) 2.4 (1) 12.9 (86)

Morgan Stanley 8.3 (2) 9.1 (14) 15.7 (22) 11.9 (36) 7.3 (6) 12.1 (80)Deutsche Bank 5.0 (2) 12.1 (21) 9.1 (18) 8.9 (36) 6.9 (3) 9.3 (80)

Merrill Lynch 19.6 (5) 6.8 (10) 2.5 (7) 9.8 (32) 13.9 (8) 7.8 (62)Credit Suisse 3.5 (1) 9.9 (14) 5.6 (11) 4.8 (26) 24.0 (11) 6.9 (63)

Goldman Sachs 10.0 (2) 4.6 (5) 2.3 (3) 10.8 (28) 0.0 (0) 6.9 (38)UBS 7.0 (2) 12.2 (16) 6.3 (11) 5.2 (19) 1.3 (1) 6.5 (49)

BNP Paribas 3.0 (1) 1.1 (2) 4.9 (7) 2.7 (11) 30.7 (7) 4.5 (28)Dresdner K.W. 9.4 (1) 0.0 (0) 6.4 (7) 2.1 (8) 0.0 (0) 2.9 (16)

Notes:Themarketshareiscalculatedfromthedealvalueoftheproceeds.Inthecaseofmultiplebookrunnersforanissue,thedealvalueoftheproceedsisdividedbythenumberofbookrunnersintheoperation.Thenumberofissuesunderwrittenforsovereignbondissuesisinparentheses.

Source:AuthorscalculationsbasedonDealogicdatabase.

Concerning the secondarymarket, the annual sovereign bond spread employed is theEmerging Markets Bond Index Global (EMBI Global) spread, calculated by JPMorgan15 andconsideredanimportantindicatorofcreditriskinemergingmarkets.

Finally, inorder toanalyse the relevanceof the information receivedby investors frominvestment banks concerning the primary bond market, we collected from July 199716 thepublicationsof13 important investmentbanks inemergingcountries17. Inweekly,monthlyorquarterly reports, investmentbankspublish theirviews foreachemerging country,providinginput for their clients,namely the buy side:portfolio assetmanagers,mutual funds,hedgefunds,pensionfunds,etc.

15. FormoreinformationseeJPMorgan(2004).

16. No reportsbefore thisdateareavailable,neither in thewebsitesnor investmentbanksdatabases.For theperiod July1997 December1999,weonlyhaveinformationfromCitigroup.

17. ThenameofthepublicationsusedareEmergingMarketsFortnightly(ABNAMRO),LatAmDriversFortnightly(BarclaysCapital),GlobalEmergingMarketsMonthly (Bear Stearns),Economics/Strategy (Citigroup),DebtTradingMonthly (Credit Suisse),EmergingMarketsMonthly(DeutscheBank),EMStrategist(DresdnerKleinwortWasserstein),GlobalInterestRateStrategy(GoldmanSachs),EmergingMarketsOutlookandStrategy(JPMorgan),EmergingMarketsCompass(LehmanBrothers),EmergingMarketsDebtMonthly(MerrillLynch),EMDPerspectivesQuarterly(MorganStanley)andEmergingMarketsDebtStrategyPerspectives(UBS).

14 OECD2008


DEV/DOC(2008)10

IV.THESOVEREIGNDEBTCRISES

Thereisnoconsensusonthedefinitionofsovereigndebtcrises(seePescatorietal.,2007forareviewoftheliteraturedefiningdebtcrises).Wefirsttakeintoaccountadefinitionlargelyusedandcitedintheliteratureontheearlywarningmodels(inparticularseeManasseetal.,2003;Manasse et al.,2005;Ciarlone et al.,2005andFioramanti,2006).A country isdefinedasbeinginadebtcrisisifi)itisclassifiedasbeingindefaultbyStandard&Poors(S&Ps)orii)itreceivesalargenonconcessionalIMFloandefinedinexcessof100percentofquota.

Concerning the firstpartof thedefinition,Standard&Poors ratessovereign issuers indefaultifagovernmentfailstomeetprincipalorinterestpaymentsonexternalobligationsontheduedate.ThereisalargeheterogeneityinthetypesofdefaultincludedinS&Psdefinition(seeAnnex1). As noted by the IMF (2006), there are different cases of debt restructuring. Debtrestructuringsmayhaveeither followedasovereigndefaultorbeenundertakenpreemptivelyinanefforttoavoiddefault18.Consequently,giventheheterogeneityofrestructuringdebts,wedividecountriesindefaultbyS&Psintotwogroupsdependingontherestructuringcase(preemptiveandpostdefault)19.

A potential problemwith the first part of the debt crisis definition is that itmay notcapture neardefaults or coercive debt restructurings that were only prevented through anadjustmentprogrammeand/oralargefinancialpackagefromtheIMF.Takingthisintoaccount,the second part of the definition considers countries thatwould have defaultedwithout thisexogenous factor20. Large IMF loans are often granted to countrieswith balance of paymentproblems,whichcouldberelatedtofinancialdifficultiesoftheprivatesector.Consequently,wedivide countrieswith large IMFpackages into twogroups.This classificationdependson thecountriesvulnerabilitiestotheexternalpublicbonds.Inordertodifferentiateamongcountries,we calculate a debt index that depends on four external debt indicators presented in theliterature and developed by GDF (Global Development Finance)21. Results are presented inAnnex2.

18. ForthecaseofMoodys,Argentina(2001)andRussia(1998)weretheonlydefaultcountriesinthecontemporaneousera.

19. Durationandintensityofdefaultvaryconsiderablyalsoamongcountries.Forinstance,theArgentineandefaultlastsfouryears(from2001until2005)whiletheDominicanRepublic(2005)andUruguayan(2003)defaultslastonlyoneyear.Additionally,therecoveryratesofthesedefaultsarealsodifferent.ConcerningthereductionoftheprincipalofthedebtrestructuredandaccordingtoIMF(2006),Argentinaobtainedareductionof56percent,incontrasttotheDominicanRepublic(0.0percent),Ukraine(0.0percent)orUruguay(1percent).

20. By using data from the IFS (International Financial Statistics)provided by the IMF and from theGDF (GlobalDevelopmentFinance)providedbytheWorldBank,wecalculatetherationonconcessionalloansoverIMFquota.

21. Theseindicatorsare:DebtServiceoverExportsofGoodandServices,InterestPaymentsoverExportsofGoodandServices,DebtoverExports ofGood and Services, InternationalReserves overDebt,Debt overGrossNationalProduct (GNP) and InterestpaymentsoverGNP.Theconstructionofthisindexissimple.First,forthecaseofReservesoverDebtweusedtheinverseofthis

OECD2008 15


DEV/DOC(2008)10

Table3presentsthetypologyofsovereigndebtcrisesstudiedinthispaper.Thistableisconstructedby taking intoaccount three considerations: first, thedataavailable for thispaper(primary sovereignbondmarket since1993) covers countries that,aboveall,areactiveon thesovereignbondmarket;second,thestandarddefinitionofsovereigndebtcrisesemployedintheearlywarningmodels literature; finally, the different types of crises according to the criteriapresentedabove.

A final concern is the connection between sovereign debt crises and currency crises(denotedwithanasteriskinTable3).Inordertodeterminewhichofthesovereigndebtcrisesarecombinedwitha currency crisis,we constructan indexof currencymarket turbulence, in thespiritofEichengreenetal.(1995)22.Wefindthat10ofthe13sovereigndebtcrisesalsoinvolvedcurrency crisesduring the 12months prior to or following thedebt crises23. In particular, in64percentofthecases,asovereigndebtcrisisprecedesacurrencycrisisinthe12monthspriortotheonsetofthecurrencycrisis(Annex4examineshowsovereigndebtandcurrencycrisesarelinked).

Table3.TypologyofSovereignDebtCrises

Brazil (August 2001)

Turkey (Dec. 2000)*Brazil (Dec. 1998)*Mexico (Feb 1995)*

Thailand (Aug.1997)*Indonesia (Nov. 1997)*

SOVEREIGN DEBT CRISES

Pre-emptive No PBVPost default Public Bonds

DEFAULT (S&P's definition) IMF large package

Argentina (Nov. 2001)*Ecuador (Sept. 1999)

Vulnerabilities (PBV)

Ukraine (Sept. 1998)*

Dom. Rep. (Feb. 2005)*

Uruguay (May 2003)*Pakistan (Jan. 1999)

Russia (August 1998)*

Note:*denotescountriesthatexperiencedalsoacurrencycrisisinthe12monthspriorandfollowingthesovereigndebtcrisis.SeeAnnexes 4 and 5 respectively for thedefinition of currency crises and the combination of both crises (currency andsovereigndebtcrises).

Source:AuthorscalculationsbasedonS&Ps,GDFGlobalDevelopmentFinance(WorldBankdatabase),IFSInternationalFinancialStatistics (IMFdatabase),TheEconomist (variouseditions),OECDEconomicSurveysandCRSReports forCongress,2007.

ThesovereigndebtcrisesinTable3canbereclassifiedintwogroups,dependingonthefragilityofthepublicsectororthecapacityofgovernmentstorepaypublicdebt.Thefirstgroupconsistsofallcountriesexhibitsbothof thesecharacteristics: that is,postdefault restructuring

indicator.Second,wegivethesameweighttoeachindicatoraccordingtoitsevolutionforMiddleIncomecountriesduringtheperiod19952002(theperiodthatenclosestheentirecrisessample)andtoitsvalue.

22. SeeAnnex3foradetaileddescriptionoftheconstructionandresultsoftheindex.

23. These crises areArgentina (2001),Brazil (1998),DominicanRepublic (2005), Indonesia (1997), S.Korea (1997),Mexico (1995),Russia(1998),Thailand(1997),Turkey(2000),Ukraine(1998)andUruguay(2003).

16 OECD2008


DEV/DOC(2008)10

typeaswellascountriesinsidetheIMFpackagetypewithpublicbondsvulnerabilities.Fortheremainder of this paper, we classify these countries as sovereign default risk countries orpublic finances difficulties (PFD) countries. In the second group (no PFD countries),sovereign debt crises are caused above all by liquidity, banking and/or balanceofpaymentproblems.ThecountriesinvolvedinthesecondgrouparepreemptivedefaultcountriesaswellasnopublicbondsvulnerabilitiescountriesreceivingalargeIMFpackage.

Frompublicinformationwecanderivethehighriskofdefaultoftheinternationalpublicdebt (and in particular of the international public bonds) of PFD countries prior to crises.Annex5presents the structuraldebt stateof crisis countriesprior to the crisis itself.Standarddebt indicators constructed from (at least expost) publicly available information show thatbetweenoneandthreeyearspriortotheonsetofacrisis,PFDcountriespresentedcharacteristicsofhighercreditriskthanthoseofothermiddleincomecountries24.

Inthatcontext,goodknowledgeofthepublicsectorandinparticularofthecreditriskofgovernments to pay international debt issued in the capital markets provide sufficientinformationfordeterminingtheprobabilityofdefaultofthepublicdebtofemergingcountries.The next section argues that given the strong connection between investment banks andgovernments in the sovereign bond market, investment banks were better informed thaninvestorsaboutsovereigndebtcrisesandinparticularincrisesthatoriginatedinwhatwehavecalledPFDcountries.

24.

Formiddleincomecountrieswetaketheaverageoftheperiod19952002.Theresult ispracticallythesame ifweusethecrisisdate(seeAnnex2).

OECD2008 17


DEV/DOC(2008)10

V.THEPRIMARYBONDMARKETINSOVEREIGNDEBTCRISES

Inthissection,wetesttheefficiencyofthesovereignbondmarketthroughthestructureoftheprimarybondmarketandduringsovereigndebtcrises.Inperfectlycompetitivemarkets,itisassumed thatpricesarepubliclyknownandallagentsareprice takers. In thatcontext, it issupposed thateachagentknowscharacteristicsofutility functionsandconsequentlyoutcomesforallagentspresentinthemarket.Oneoftheimplicitassumptionsisthatallgoodsareperfectlyobservabletoallmarketparticipants.Itfollowsthatmarketinefficienciesarisewheninformationis asymmetrically available tomarket participants. This inefficiency commonly takes place incasesofadverseselection(Akerlof,1970).Inthisevent,theprincipal(buyer)doesnotperfectlyknowallthecharacteristicsofthegoodofferedbytheagent(seller).

Inthecaseofthebondmarket(seeTirole,2006foranindepthreviewoftheliterature),informationasymmetriescouldbepresent intheevent inwhichfinalbuyers(investors)donothave all existing information concerning the quality of the sovereign bonds sold by issuers(governments). Italsoariseswhen there isasymmetric informationamong investorsaswellasbetween issuers and underwriters (seeAmaro deMatos, 2001 for a review of the theoreticalliterature).

Inthispaper,weanalysemarketinefficienciesinthesovereignbondmarketthroughtheinformationaladvantagethatinvestmentbanks,actingasunderwriters,mayhaveoverinvestors.Todothisweusetheprimarybondmarket,wherewehaveinformationregardingboththepricepaid by investors for sovereign bonds aswell as the feepaid by governments to investmentbankstoplacethebonds.Moreprecisely,duringtheissueofasovereignbond(similartothatofequity or corporate debt issuance); governments pay a fee (i.e. the underwriting spread orunderwritingfee)toinvestmentbankstoplacethebondsinthemarket.

Weusethesovereignbondspreadasaproxyofthepricepaidbyinvestors.Inthisway,we may infer the sentiment of investors regarding the economic and financial situation ofemergingcountries.Furthermore, thefeemaycontain important informationabout investmentbanksperception of sovereign risk.Acting asunderwriters, investmentbanks are facedwithimportant risks in the event of government default. These risks are related to the banksreputationandthetransactionitself.

Investmentbanks could incur a lossof reputation in the eventofgovernmentdefault.Viewing the sovereignbondmarkethistorically,FlandreauandFlores (2007) show the roleofunderwriters reputations inguiding investorsportfolioallocations.Concerning the corporatemarket,MichelandShaked(1990)arguethatprestigiousunderwritersmaychargehigherfeestocompensateforthepotentialdamagetotheirreputationincasetheissuerdefaults.Similarly,by

18 OECD2008


DEV/DOC(2008)10

controllingfortheendogenousmatchbetweenissuersandunderwriters25,HuaFang(2005)findsthat reputable banks charge higher fees, which can be interpreted as economic rents onreputation26.

With regard to the transaction, investment banks are exposed to the issuers liquidityproblems.Indeed,eveninabesteffort(financialintermediariespledgetohelpfindcustomerstopurchaseasmanybondsaspossible,butdonothavetoacquireasinglebondiftherearenobuyers)distributionsystem,investmentbankscanincursomerisks.Firstly,theyhavetobuytheissuebeforesellingitontotheinvestorandarefacedwithasettlementrisk27.Secondly,theyhavetheresponsibilitytoplacethebondsinthemarketandmakeanefforttostabilizethepriceofthebondsinthesecondarymarketforanunspecifiedtime.Aswritteninthebondprospectus,although theunderwriterbank isnotmandated tomakea secondarymarket for thebonds, itplans tomakeone28.According to interviews inoriginationdepartmentsof investmentbanks,marketmakingactivitieson thesecondarymarketcanevenextenduntil thematurityof thebond29.

AsimpleversionofthestructureofthesovereignbondmarketissummarisedinFigure1,which illustrates the interactions among actors in the sovereign bondmarket throughout theexecutionof financial transaction.Theparticipationof investorson theprimarymarketand/ortheir participation on the secondarymarket (by buying or selling securities) are all done viainvestment banks.Most of investment banks income is derived from these transactions30. Inparticular, the fee isdeducted togovernmentsby investmentbanks from thepriceoffered toinvestorsintheprimarymarketanditisfixedbetweeninvestmentbanksandgovernmentspriortothedeterminationofthepriceofthebondintheprimarymarket31.

25. ThisresearchiscriticaltowardsLivingstonandMiller(2000)andJames(1992),inthecontextinwhichittakesintoaccountthat

reputablebanksmayhavechosen(selfselected)tounderwritehigherquality issuespreciselyoutofreputationconcerns.Thus,thisarticlearguesthatfailingtocontrolforthistypeofselfselectioncouldleadtoincorrectconclusions.

26. Fromatheoreticalandempiricalanalysisoftheunderpricing inthecorporatemarket,CarterandManaster(1990)showthatprestigiousunderwriterschargehigherunderwriterspreadsandareassociatedwithlowerriskofferings.

27. Interviewswithmarketparticipantssuggestedthatbesteffortsisthedistributionsystemusedintheemergingsovereignbondmarket.However,insomeinterviews,itappearsthatsomeissuesarepartiallybackstoppedandconsequentlyitcouldproducealossofcapitalforinvestmentbanks.InterviewswereachievedtotheOriginationdepartmentsofthefollowinginvestmentbanksinWallStreet:BearStearns,Citibank,DeutscheBank,GoldmanSachs,JPMorgan,LehmanBrothers,MerrillLynchandMorganStanley.

28. For instance, see theProspectus supplementofUS$750,000,000.FederativeRepublicofBrazil.10.5percentGlobalBondsDue2014.July7,2004:Brazil(theissuer)hasbeenadvisedbytheunderwritersthattheunderwritersintendtomakeamarketintheglobalbondsbutarenotobligatedtodosoandmaydiscontinuemarketmakingatanytimewithoutnotice.Noassurancecanbegivenastotheliquidityofthetradingmarketfortheglobalbonds.

29. ThequestionaskedtooriginationdepartmentsofinvestmentbankswasHowlongdoesaleadmanagermakeamarketinthesecondary?.Formorethanahalfofthemanagersinterviewedthemarketmakingactivitiescanremainforallthedurationofthebond.

30. Moreprecisely,investmentbanksincomeisderivedfromunderwritingactivities(fee)aswellasfromtradingactivitiesinthesecondarymarket(commissionschargedtoinvestorsbetweentheissuedateandthematuritydate).

31. Forinstance,intheGlobalbondissuedbytheRepublicofTurkeyon12July2006(91/2percentNotesdueJanuary15,2014),thepublicofferingpricewas108.62percentoftheamountissued,thefeewas0.125percentoftheamountissuedandconsequentlytheproceedstotheRepublicofTurkeywas108.495percent.Formoreinformationabouttheprospectusofsovereignbondsseehttp://www.secinfo.com

OECD2008 19


DEV/DOC(2008)10

Figure1.StructureofthePricesintheSovereignBondMarket

Governments Inve Banks

Investors

stment

feePt

TP

TP tP+

+

+

tT

tT

C

P

Where:

Maturitydate T :

Issuedate :t

Commissionpaidbyinvestorstoinvestmentbanks:C

Priceofthesovereignbond(financialtransactionthatP :

isthecounterpartyofthetransferofthesecurity)

Underwritingspreadpaidbygovernmentstoinvestmentbanks:fee

Aside from the role of intermediaries between issuers and investors, one of themostimportantresponsibilitiesofinvestmentbanks,actingasunderwriters,concernsthetransmissionof information tocapitalmarkets.Before thebond is introduced to themarket, leadmanagersplay an essential role in publicising a bond, promoting activities such as presentations,conference calls andpublications32.This informationgiven to investors is crucial to achieve asuccessfulissueintheeyesofgovernments,whichassumeahighprimarypricewithrespecttotheintrinsicrisk.

Underwritinginvolvesaclose,regular,andoftenprivileged,relationshipwithimportantactors ingovernment.Suchaclose relationshipwith the treasuryandministryof financeofa

32. Presentationsareprepared for investors inmajor financialplacesand theyconcentrateon themacroeconomic situationof the

countryand themaincharacteristicsof thebond.Thesepresentations,known in the jargonasroadshows,arenot formallyrequired.

20 OECD2008


DEV/DOC(2008)10countryconfersaprivilegedvantagepointoverbotheconomicaspectsaswellasthebehaviouralpatternsofthegovernmentthatmanagesit33.

One could therefore argue that investorsdo not hold all information concerning risksattributedtosecuritiesissuedbygovernments.Moreprecisely,wecouldassumethatinvestmentbanks could have an informational advantage over investors concerning the risks of thesovereignbonds issuedby emerging countriesgiven thedirect and strict link theyhavewithissuers34.

In particular, this information problem can arisewhen the credit risk of the issuer isconsiderable.ThisisthecaseofPFDcountriesinwhichtheriskofdefaultofthepublicbondsishigh.Althoughpublic information isavailable (at least expost) todistinguish these countriesfrom other emerging countries (Annex5), investment banks may be better informed thaninvestors about these risks.This can be explained by the strict and close connection betweeninvestmentbanksandgovernmentsthroughthesovereignbondmarketandbyweakincentivesfor international investors to learn about individual countries (for this last point, see Calvo,1998).

Inordertotesttheefficiencyofthesovereignbondmarket,weanalysethebehaviourofmarketparticipantspriortostresstestingscenarios.Indeed,bystudyingpricesbeforesovereigndebtcrises,weobservetheperceptionofinvestmentbanksandinvestorsconcerningthequalityof bad countries. Empirically, this is related above all to bad countries in which publicfinancedifficultiesof the issuerscouldbeobservedprior tocrises35.Thisempiricalanalysis isrestrictedtobadcountriesthatwereactiveintheinternationalprimarybondmarketpriortosovereigndebtcrises.

In order to test market inefficiency in the emerging sovereign bond market, the nullhypothesesusedinthispaperarethefollowing:

Hypothesis1:Prior to sovereigndebt crises, investors arenotperfectly informed regarding thequalityofthesovereignbonds issuedbyriskycountries.Bycontrast, investmentbanksobservethisriskpriortotheonsetofcrises.(H1).

Hypothesis 2: This information problem is above all present in PFD countries (i.e. sovereigndefaultriskcountries)(H2).

Bothhypothesesareconditionaltoriskycountriesbeingactiveintheprimarymarketpriortodebtcrises(i.e.issuingsovereigndebt).

33. Oneoftheinvestmentbanksinterviewedpointedoutthat:theinformationyougetfromunderwritingisveryimportantnot

insiderinformation,butalotofknowledgeonwhatasovereigntendstodo.

34. Notethatthiseventcontrastswiththestandardframeworkforexplainingtheraisondtreofbanks.Asmonitorsandperformersofscreenings of the borrowers, banks reduce informationproblems betweendepositors and borrowersprior to the selection ofborrower(adverseselectionproblem)andduringtheloan(moralhazardproblem).SeeFreixasandRochet,1997foralargereviewofthemicroeconomicsofbanking.

35. Empirically,thisispresenttocountriesbelongingtopostdefaultevents(bycontrasttopreemptiveeventsinwhichliquidityriskisconsiderable)aswellastothelargeIMFpackagegiventopublicfinancesdifficultiescountries.

OECD2008 21


DEV/DOC(2008)10

Thehypothesesmentionedabovearevalidatedwhenpriortosovereignbondcrisesthefee of bad countries is higher than itwould have been by analysing only sovereign bondspreads (hypothesis1).Moreover,bydifferentiatingamongsovereigndebtcrises,wenote thiseffectisaboveallpresentincountriesthatpresentsovereignriskdifficulties(hypothesis2).

Thesehypothesesarguethatthere isvaluable informationprovidedbythecostpaidbysovereigndebtorsto investmentbanksandthatsovereignbondspreadsdonotcaptureallthisinformation36. By contrast, under perfectly competitive markets (i.e. alternative hypothesis),sovereignbondspreadsofbadcountriescanexplain thebehaviourof the feeprior tocrises.Indeed, in thatevent, investorsaswellas investmentbanksobserveperfectlybadcountriesandweareinanefficientmarketcontext.

StylisedFacts

Inorder to analyse thebehaviourof investors and investmentbanksduring sovereignbond crises,weproceed as follows.First,we analyse thepath of the fee and sovereignbondspreadspriortoandaftertheonsetofcrises.Second,wecomparethecrisiscountriesfee(priorto the onset of the sovereign debt crises) with respect to those observed for other countriesduringtheperiod19932006.

InFigure2,wepresentannualdatatrackingtheevolutionofthefeesandthesovereignbondspreadsaroundthesovereigndebtcrisesandthecombinationofbothsovereigndebtcrisesandcurrencycrises(twincrisesinthefigure).Priortotheonsetofsovereigndebtcrisis(T


DEV/DOC(2008)10a fee lower than inT


DEV/DOC(2008)10

priortotheonsetofthecrisestoyear5aftertheonsetofthecrises)centredontheonsetofcrises(dateT).TheseresultsarepresentedinAnnex637.

Concerningthefee,onlythethree,twoandoneyeardummiespriortocrisesarepositiveandsignificantat1percent.Indeed,whatitisreflectedinthefeessurroundingcrisesisaboveallthehighremunerationpaidbygovernmentstoinvestmentbankspriortocrises.Bycontrast,thesovereignbondspreadreactstocrisesafterthefee.Thereactionofthesovereignbondspreadislargeandsignificantat10percentduringandoneyearaftertheonsetofcrises.

Itisusefultodifferentiatemarketreactiontothedifferentkindsofcrises.Asmentionedabove, sovereign bond crises may be classified into four different types. Countries whichdefaultedfollowedbypostdefaultrestructuringgroupaswellasthosewhereIMFpackagesandpublicbondsvulnerabilitiesarepresent (i.e.PFDcountriesorsovereigndefaultriskcountries)would have had to pay a large amount of money to underwriting banks prior to crises tocompensateforahigherdefaultprobability.Thisresultcontraststotheexperienceofthesecondgroupof countries,whichhaveundertakenpreemptivedebt restructurings inorder to avoiddefaultorcountrieswithoutpublicbonddifficultiesandwhichobtainlargepackageloansfromtheIMF(seeFigure3).

Figure3.Feeandprimarysovereignbondspreadarounddifferenttypesofsovereigndebtcrises(annualbasis)

PublicFinancesDifficulties(PFD)countriesNoPFDcountries

Sovereign bond spread (bp; rhs)

Fee (% of the amount issued)

600

500

400

300

200

100

0T+3T+2T+1TT-1T-2T-3

21.81.61.41.2

10.80.60.40.2

0

Sovereign bond spread (bp; rhs)

Fee (% of the amount issued)

650

600

550

500

450

400

350T+3T+2T+1TT-1T-2T-3

2

1.8

1.6

1.4

1.2

1

0.8

0.6

0.4

Note:Tistheonsetofthecrisis.Nonoverlappingcrisesepisodesareused.StatisticalanalysisisreportedinAnnex7.

Source:Authorscalculations.

Two conclusions may be drawn from this. Firstly, in the case of PFD countries,investmentbanksdemandahighfeepriortocrises,whichdecreasesoncethecrisisisunderway.

37. Wecalculatetheannualaverageofthefeeandprimarysovereignbondspreadfromthetotalsample(436issues).Bothvariables

areregressedindependentlyagainstthetimedummyvariables.Forthatpurposeweuse149issuesrealizedbybadcountriesduringthe11yearwindowaroundcrises.

24 OECD2008


DEV/DOC(2008)10

At the same time, sovereignbond spreads remain stable.Secondly, for the countries inwhichtherearenopublicfinancesvulnerabilities,resultsdiffertothoseabove.Thefeeremainsstablepriortoandattheonsetofthecrisis.

Investment banks and investors behaviour differ depending on the type of crisis.Investmentbanksdemandahighfeefromcountriesthatexhibitaconsiderableriskofsovereigndefault.Thedecrease of the fee at the onset of the crisis could be explained by the fact thatinvestmentbanksno longerholdamonopolyof informationandotherplayers in themarketmustbecompensatedfortheriskofcrisis,throughareductioninthepriceofthebonds.

Incontrast,countriesthatdemonstrateonlyliquidityorbankingbalanceofpaymentriskshavetopayasmallandstablefeepriortotheonsetofcrisis.This isthecaseoftwocontagioncountries (Ukraine 1998 and Uruguay 2003) that suffered economic and financial crisis afterRussianandArgentineandefaultsrespectively,aswellasIndonesiaandThailandwithbalanceofpaymentscrises38.

Resultspresented inFigure3areconfirmedstatisticallyandreported inAnnex7.Usingthesamemethodologypresentedaboveforthecaseofallsovereigndebtcrises,wenotethatforthe case of countries that exhibit considerable sovereign default risk, the fee is high andstatisticallysignificantat1percentbetweenoneandthreeyearspriortocrises.Sovereignbondspreads react to crises after the fee: this effect ishighand significant at1per centduring theonsetofcrises.

Bycontrast,forcountriesthatexhibitonlyliquidityorbankingbalanceofpaymentrisks,fees aswell as sovereign bond spreads are not statistically significant for any of the 11yearwindowdummyvariables.

Whenwe analyse the subset of countries thatpresentpublic financingdifficulties andcurrency crises (Annex7), the outcome is the same as in the case of only PFD countries.Investment banks demand a high fee three, two and one years prior to crises. This result issignificantat1percent.Thisoccursinmostcasesofsovereigndebtcrisesduetopublicfinancesdifficultiesfollowedbycurrencycrisesinthelastyears(Argentina2001,Brazil1998,Mexico1995andRussia1998).Wenotethatforcountrieswithsovereigndefaultrisk,sovereignbondspreadsincrease considerably at the onset of crises and this result is significant at 1per cent. Indeed,investorsreactonlyattheonsetofcrises.

Finally,wecompare feeswith sovereignbond spreadsduring theperiod19932006 forthe total EMBI global countries, by differentiating between countries that encounter publicfinancingdifficultiesandthosethatdonot.Moreprecisely,wecomparefeesandsovereignbondspreadspriortocrisescomparedtothoseforotherevents.

38. In someof theNoPFDcountries (inparticularAsiancountries), firstgenerationandsecondgenerationcrisismodels fail to

explainthesecrisesgiventhefinancialvulnerabilityofthesecountries(seeReisen,1998).Thisresultcontrastswiththeroleoffiscalpolicyandcredibleeconomicpoliciesinthefirstandsecondgenerationmodels(BragadeMacedoandReisen,2003)thatcouldexplain(atleastexpost)mostofthecrisesinPFDcountries.Forareviewoftheliteratureoncrisesinemergingcountriesandinparticularonthegenerationsofcrisismodels,seeKrugman(2003).

OECD2008 25


DEV/DOC(2008)10

Figure4showstheaverageannualfeeandprimarysovereignbondspreads.Bluesquaresindicatethefee/bondspreadbetweenT3andT1priorattheonsetofcrisis39.Aswecanobservefromthisfigure,feesaresubstantiallyhigher(giventhebondspread)forPFDcountriesrelativetoother emerging countries.On average, sovereigndefault risk countrieshad topay 1.10percentoftheamountissuedtoinvestmentbanksbetweenoneandthreeyearspriortotheonsetofcrisis,almosttwicetheemergingcountriesaverageduringthesampleperiod(0.56percent).Bycontrast,whenwecomparethelevelofprimarysovereignbondspreadsbetweenoneandthreeyearspriortocrisiswithrespecttothetotalforemergingcountries,wefindthattheformerisonaverageonlyslightlyhigherthanthelatter(385bpvs.319bp)andwelllowerthantheprimarysovereignspreadattheonsetofthiscrisis(603bp).

Figure4.Feesandprimarysovereignbondspreads(19932006,annualbasis)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

0 100 200 300 400 500 600 700 800 900

Argentina (T-3)

Argentina (T-2)

Argentina (T-1)

Brazil (T-2)

Brazil (T-1)

Mexico (T-2)

Russia (T-2)Russia (T-1)

Turkey (T-3)

Turkey (T-1)

Turkey (T-2)

Sovereign Bond Spread (basis points)

Fee

(as

a pe

rcen

tage

of t

he a

mou

nt is

sued

)

Average Fee

Note:Thefeeandsovereignbondspreadforcountriesbetweenoneandthreeyearspriortosovereigndebtcrises(dateT)arehighlightedinbleu.TheArgentineancrisiswasin2001,theBrazilianin1998,theMexicanin1995,theRussianin1998andtheTurkishin2000.

Source:AuthorscalculationsbasedonDealogic,2007.

EconometricAnalysis

InordertotesthypothesesH1andH2presentedabove,webuildasimplecrosssectionaltime seriesanalysis.The study covers theperiod starting in1993and ending in2006and thecountriesarethoseincludedintheEMBIGlobalindex.WebeginwithastandardOLSestimation

39. Moreprecisely,inthisfigureweexposebyredsquares,countriesthatpresentedpublicfinancesdifficultiesandcurrencycrisis.

26 OECD2008


DEV/DOC(2008)10using fees, primary bond spreads and a set of dummy variables to differentiate betweencountries that do or do not encounter sovereign debt crises during the years around crises.Consequently,thefollowingpaneldataregressionmodelisused:

itK

itKitit KTSBSfee ++++= =

5

521 )(

where correspondstotheunderwritingspreadreceivedbyinvestmentbanksfrom

country i inperiod t and isdenoted inpercentageof the amount issued; represents the

primary sovereign bond spread, is a dummy variable that takes the value of 1 forcountriesplacedattheyearKwithrespecttotheonsetofcrisis(T)and0otherwise.

itfee

itSBS

itKT +

Inordertotestthefirsthypothesis,weobserveifpriortosovereigndebtcrisesthefeeofbadcountriesishigherthanitwouldhavebeenbyanalysingonlysovereignbondspreads.Inthe equation presented above this hypothesis is validatedwhen some (or all) of the dummyvariable coefficients prior to crises (from beta 5 to beta 1) are positive and statisticallysignificant. Depending on which coefficients are significant, we can first conclude if thishypothesis isvalidatedandseconddeterminehow longprior to theonsetofcrisis investmentbanksobservebadcountries.Thealternativehypothesisisthatcoefficientalpha2issignificantanddummyvariablecoefficientspriortocrises(frombeta5tobeta1)arenotsignificant.Inthatevent,investmentbanksobservebadcountriesatthesametimeinvestorsdo.

Weuse thesameprocedure to test thesecondhypothesis.Weonlydifferentiateamongsovereigndebtcrises inorder to test that the informationadvantageof investmentbanksoverinvestors is present above all in PFD countries.To that end, the set ofdummy variables aredifferentiatedby typeof crisis: i)Sovereigndebt crises thatarenot causedbypublic financesdifficulties(preemptivedefaultsandIMFpackageinordertoavoidbalanceofpaymentcrises).ii) Sovereign debt crises that are caused by public finances difficulties (debt restructuringfollowing default and IMF large package for countries with public debt vulnerabilities).iii)Sovereign debt crises that are caused by public finances difficulties and accompanied bycurrencycrisesaswell.

Resultsarepresented inAnnex8.Themost important findingsareas follows.First,as

expected, there is a positive relationship between and primary sovereign bond spread.Howevertheimpactofprimarysovereignbondspreadonfeeislimited.Anincreaseof100bpofthebondspreadimpliesanincreaseofonly0.03percentofthefee.

fee

Second, by including the dummyvariables,hypothesis 1 is validated.Prior tocrises, investment banks charge a high fee to bad countries and this is not captured byinvestors. More precisely, three, two and one year dummy variables prior to the onset ofsovereigndebtcrises(

itKT +

3T , 2T and 1T )arehigh,significantat1percentandimprovethefitoftheregression.Forinstance,oneyearpriortocrises,badcountriespaid0.47percentofextra fee and this result is statistically significant at 1per cent. We obtain similar results byanalysingtwincrises(sovereigndebtandcurrencycrises).Betweenoneandthreeyearspriorto

OECD2008 27


DEV/DOC(2008)10

theonsetofsovereigndebtcrises,investmentbankschargeahighfee.Fortheseyears,dummyvariablesarealsosignificantat1percent.

Finally,bydifferentiatingamongsovereigndebtcrises,hypothesis2 isvalidated.Whenwedividethesovereigndebtcrises intotwogroupsthatdepend ifcountriespresentsovereigndefaultriskornottheresultsdifferconsiderably.Indeed,whenwestudynoPFDcountries,wefind thatdummyvariablesprior to crisesarenot statistically significantand thevaluesof thecoefficientsaresmall.Bycontrast,whenwetakeintoaccountonlyPFDcountries,thefixedcostpaid to investmentbanksprior tocrises ishigh. Investmentbanksbegincharginghigh fees tobadcountriesthreeyearspriortocrises.Dummyvariablesforthree,twoandoneyearpriortocrises( ,3T 2T and 1T )arehighandsignificantat1percent.Moreover,whenwerestrictthe analysis to countries where sovereign default risk is associatedwith currency crises, theadditionalcostofthefeeishighandagainsignificantat1percentone,twoandthreeyearspriorto sovereigndebt crisesand improves the fitof the regression considerably.For instance,oneyearprior to theonsetofdebtcrises,bankschargea fixedcostof0.68percentof theamountissuedtobadcountries.

From the results presented above we can conclude firstly, that investment banks areinformedaboutsovereigndebtcrisesbeforeinvestorsandbetweenoneandthreeyearspriortocrises(between and3T 1T ).Secondly,thisresultisparticularlyevidentforcrisespresentingsovereigndefaultrisk.

InordertoverifytherobustnessoftheseresultsandinparticularthehypothesesH1andH2,firstwecheckthetimeeffectbyconstructingafixedtimeeffectsmodelthatinvestigateshowtime affects the intercept using time dummy variables. Second, we introduce an interactivedummy variablewith thepurpose of testingwhether the slope of thedummy crisis variable(whichtakesthevalueof1forbadcountriesbetween 3T and 1T yearpriortoasovereigndebtcrisisand0otherwise) isequaltothatofthesovereignbondspreadofthetotalemergingcountries40.

We introduce this interactivedummyvariable inorder todetermine the impactofboththeinterceptandinteractiveeffectintheregressions.Inthiscontext,ifshiftsintheregressionareattributable only to the interactive dummy variable, then the fee for bad countries can betotallyexplainedby the sovereignbond spread. In theopposite case (i.e.only the intercept issignificant),itsignifiesthatbadcountrieshavetopayanadditionalfixedcosttounderwritingbanks(thiscostisofcoursenotexplainedbysovereignbondspreads).

40. Moreprecisely,thefollowingpaneldataregressionmodelisused:

itiitititit TDSBScrisiscrisisSBSfee +++++= 54321where corresponds to theunderwritingspreadreceivedby investmentbanks fromcountry i inperiod tand isdenoted in

percentageoftheamountissued, representsthesovereignbondspread, isadummyvariablethattakesthevalueof

1forcountriesplacedpriortotheonsetofasovereigndebtcrisis(betweenT3andT1)and0otherwise. isdefinedas

theproductof and andfinally isatimedummyvariableforeachyear(from1993to2006).

itfee

itSBS itcrisis

itSBScrisis

itSBS itcrisis iTD

28 OECD2008


DEV/DOC(2008)10

Results are presented as follows. First, we analyse the relationship between fees andsovereignbondspreadsbyusingjustthesovereignbondspreadasanexplanatoryvariableandbyintroducingtimedummyvariablesforeachyear(from1993to2006).

Second,weincludetimedummyvariablesinorderanalysethebehaviourofinvestmentbanksandinvestorspriortosovereigndebtcrises.

Third, in order to test if the crisis dummy variable thatwe have introduced into theregressionmodel is for shifts in the interceptof the regression,we introduce the slopeof thisdummyvariable(i.e.interactivedummyvariablebetweencrisisandsovereignbondspreads).

Finally, in order to test hypothesis 2, we replicate the analysis presented above bydifferentiatingbetweenthedifferenttypesofsovereigndebtcrisis.

Results arepresented inAnnex9.First,we find that thepositive relationshipbetween

the andsovereignbondspreadislimited.Anincreaseof100bpofthesovereignbondspreadamplifies the feebyonly0.041percentof theamount issued.Additionally,usinga fixed timeeffectsmodelconsiderably improves the fitof theregressions. Indeed, from1999 to2006, timedummyvariablesarestatisticallysignificantand the impactof timeon the fee isnegativeandincreases over time suggesting that during the last years the reduction of the fee has beennoteworthygiventheprimarybondspread.

fee

Second,byusingafixedtimeeffectsmodelweconfirmthemainresultsexposedaboveandwecanvalidatehypothesis1. Inparticular,prior tosovereigndebtcrises,badcountriespaidalargerfeetoinvestmentbanksthanotheremergingcountries(0.18percentoftheamountissued).Whenweaddtheinteractivedummyvariabletothefixedtimemodelwecannotrejectthehypothesisthatthetworegressions(withandwithouttheinteractivedummyvariable)havethesameslope.Indeed,thecrisisdummyvariablealoneisresponsibleforshiftsintheinterceptoftheregression(significantat10percenthowever)andthefixedcostpaidbybadcountriestoinvestmentbanksremainshighpriortotheonsetofcrises.Wefindthesameresultsforthecaseofthetwincrises(acombinationofbothsovereigndebtandcurrencycrises).

Finally,whenwedifferentiatewithrespecttothetypesofcrisesandbyusingafixedtimeeffectsmodel,noPFD countriespaid a fee similar than the restof emerging countries (the tstatistic isnot significantand the coefficient is close tozero).By contrastandasexpected, forcountrieswith sovereigndefault risk, the fixed cost to issue sovereigndebtbetween one andthreeyearspriortocrisis ishigh(0.24percentoftheamount issued)and issignificantat1percent.Includingonlycountriesexhibitingsimultaneoussovereigndefaultriskandcurrencycrises,weobtainthatitsimpactonfeeissignificant(tstatisticat1percent)andconsiderable(0.28percentoftheamountissued).

Byincludingtheinteractivedummyvariableforeachofthesovereigndebtcriseseventspresentedabove,wefindthatitisnotsignificantandconsequentlytheimpactofcrisesonthefeeisonlythroughtheintercept(howeverforthecaseofcountrieswithsovereigndefaultrisk,thedifferential slope is paradoxically negative and significant at 10per cent). We confirmhypothesis2:PFD countries have topay a high fee to investment banks (0.54per cent of theissue)andthisresultissignificantat1percent.

OECD2008 29


DEV/DOC(2008)10

Themost important results from this empirical study are as follows. First, investmentbanks are informed about sovereign debt criseswell before investors (whose perceptions aremeasuredthroughthesovereignbondspread),andbetweenoneandthreeyearspriortothesecrises. Investment banks behaviour stands in stark contrast to that of investorsduring crisisperiods.Priortotheonsetofacrisis,resultssuggestthatinvestmentbankshaveaninformationadvantage, and are practically the only ones compensated to cover the risk of crisis(hypothesis1).

Second, investment banks do not deal equally with all crisis countries. Indeed, bydifferentiatingamong the typesofsovereigndebtcrises,we find thatbeforecrises investmentbanks chargedonaverage ahigher fee toPFD countrieswith respect toother sovereigndebtcrisescountries(hypothesis2).

30 OECD2008


DEV/DOC(2008)10

VI.THEUNDERWRITINGFEEANDFINANCIALMARKETSACTORS

Asunderlinedintheprevioussection,thefeeholdsimportantinformationoninvestmentbanks sentiments regarding the credit risk of issuers. In this section,wepresentpreliminaryresults on investors perceptions and investment banks analysis concerning the investmentvalueoftheunderwritingfee.

FeescanbeconsultedincapitalmarketsinformationdatabasessuchasBloomberg.Theseholdvaluable information onprimarybondmarkets, such as fee,name of leadmanagers (orbookrunners),amountoutstanding,thesovereignbondrating,andbondsissuecharacteristics.Moreover, as noted before, there is also a database dedicated exclusively to the analysis ofprimarymarketissues(DealogicsDCMAnalytics).

Although investment banks have no obligation to submit deal information andconsequently fees for issues, they have the incentive to provide this information as suchdatabasescompilerankingsofprimarybondmarketleaders(LeagueTables)throughthedealsinvestmentbanksmake.Thisinformationisanimportantbenchmarkformarketmakers,issuers,analysts and financialmedia alike, in ahighly competitive fieldwhere the reputationof eachinvestmentbankismeasuredinmarketshare41.

Thus,while informationon the fee isavailable, it isnotaccessibleatpricingdayor thedayof the issue.In fact, there isa lagbetweenan issuesannouncementdateandtheday thatfinancialdatabasesreleaseinformationonthefee.Onaverage,thisinformationisreleasedafewdaysafter thedateof issue42.According toa teammemberofDealogicat theendof2007forabout80percentoflargedeals(morethanUSD200millionequivalent)weshouldhavethefeewithin1day.Fortheremaining20percentwouldbeonaveragewithinoneweek.Forsmallerdeals(e.g.,smallMTNs,etc)itmaytake23weeksuntilwereceivethepricingsupplement.

Becausemarket actors have access to fee information after the issue date, it could beconvenienttostudysecondarymarketpricesinordertodeterminetheadvantageofinformationthatinvestmentbankscouldhaveoverinvestors.

To that end, we employ secondary sovereign bond spreads in order to determineinvestorsbehaviour.Moreprecisely,wereplicatetheanalysispresentedaboveandweuseEMBI

41. ForinstanceinBloomberg(2006),thereisadetailedanalysisofthemajorinvestmentbanksintheprimarymarkets,bycomparing

feesandmarketsharesamonginvestmentbanks.

42. Forinstance,fortheColombianGlobalBond09/08/06(ISINnumberXS0213272122),weobtainedfromBloombergtheinformationconcerning the feeoneweekafter the issuedate.Moreover, thispieceof informationwasneitherdisclosedby theColombianGovernmentnortheinvestmentbanksintheirexternaldocumentsorwebsitesduringthedayoftheissue.

OECD2008 31


DEV/DOC(2008)10

Globalspread1day (EMBI1)aswellas10daysaverage (EMBIAV10)after thesovereignbondissuesusedinthesample.

ResultsarepresentedinAnnex10forthecaseofEMBI1andinAnnex11forthecaseofEMBIAV10.Byusinga fixed timeeffectsmodelwe confirm themain resultspresentedaboveandwe canvalidatehypothesis1 forboth secondary sovereignbond spreads (i.e.EMBI1andEMBIAV10).Inparticular,wenotethatthedummycrisisvariable(between and3T 1T priortotheonsetofcrisis T )ispositiveandsignificantat1percent.Whenweincludetheinteractivedummyvariabletothefixedtimemodel,wefindthatthecrisisdummyvariableisagainpositiveandsignificantat1percentforEMBI1andEMBIAV1043.Thefixedcostpaidbybadcountriestoinvestmentbanksremainshighpriortotheonsetofcrises(closeto0.60percentoftheamountissuedforbothkindofsecondarybondspreads).

Whenwedifferentiatewithrespecttothetypesofcrisesandbyusingafixedtimeeffectmodel, tstatistic is not significant for countries that do not have sovereign default risk. Bycontrast, for thecaseofcountrieswithsovereigndefault risk, the fixedcost to issuesovereigndebt is high and is significant at 1per cent. As we further restrict attention to countriessimultaneouslyexhibitingPFDandcurrencycrises,weobtainresultssimilartothecaseofPFDcountries. By including the interactive dummy variable for each of the sovereign debt criseseventspresentedabove,theresultsdifferwithrespecttotheregressionfeeprimarysovereignbondspread.Indeed,forallsovereigndebtcrises(excludingnoPFDcountries),thetstatisticsoftheslopearesignificantat1percent.Themostimportantresultisthatthefixedcostrelatedtothe fee paid to investment banks for crises countries is even higher than in the previousregressionsandissignificantat1percent.Forinstance,whenwestudyEMBIAV10spread,PFDcountriespaidasurplusinthefeepriortocrisesequivalentto0.90percentoftheamountissued.Weconfirmhypothesis2.PFDcountrieshavetopayahighfixedfeetoinvestmentbanks.

Aquestionremains:do investorsmakeuseof fees indeterminingportfolioallocations?This question can be answered bydirectly questioning investmentmanagement firms on theprimarybondmarket.Weundertookasurveythatanalysedinstitutionalinvestorsperceptionsof thestructureof theprimarysovereignbondmarket44.Anaspectof thesurveyconcerns therelevance of the fee to investors. Seven investors of the eight interviewed argue thatunderwritingfeesareofnoconcernininvestmentdecisions.Inparticular,mostoftheinvestorsinterviewed argue thatbecause thisvariable is formedby the connectionbetween investmentbanksandgovernments, investorsdonothave topayattention to this. Indeed, theyattributeimportance above all to market variables in which investors play a crucial role (e.g. bondspreads,yieldcurvesorCDSspreads).Finally,investorsdonotperceivefeesasagoodindicatorofcreditrisk.Indeed,onlyoneinvestorattributesanyplausibilitytothatargument45.

43. Bycontrasttotheregressionpresentedinannex9(primarybondspreadandfee),theslopecoefficientissignificantat5percent

andparadoxicallyitisnegative.

44. Thisprojectwas financedby theChaireFinances Internationales,SciencesPoParis.This surveywas carriedout inNewYorkduring 2006 and the investmentmanagement firms interviewedwereAllianceBernstein,AllianceCapital,Fidelity,GEAssetManagement,GMO,Goldman,InvescoandWesternAsset.

45. Moreprecisely,thequestionsaskedregardingtherelevanceofthefeeforinvestorswere:Areunderwritingfeesofanyrelevancetoaninvestor?andIsunderwritingfeeagoodindicatorofcreditrisk?

32 OECD2008


DEV/DOC(2008)10

According to these preliminary results, there appears to be a disconnection betweeninvestors reaction to information available on the primary bond market, and the outcomespresented in the statisticalanalysis.Thisquestionnaire suggests that investorsdonot take feeintoaccountandaremoreattentivetotheevolutionofothervariables.Itwouldthusappearthatunderwritingfeesarenotusedasatoolindeterminingportfolioallocations.

Also important for investor decision making are investment banks publications onemerging sovereign bond markets. We consider more than 600 publications of thirteeninvestment banks46 covering the period 19972007. This corresponds to research by themostimportantplayersinemergingbondmarketsasLeadManagers(seeTable2).Duringthisperiodandfortheinvestmentbanksanalysed,theleveloffeesofsovereignbondissuesisnotapieceofinformation given by investment banks to institutional investors and consequently is notanalysed intheirreports.Bycontrast, thesepublicationsregularlypresentdetailed informationrelated to theprimarybondmarket. Inparticular,avastdescriptionof the structureofbondsissued (outstanding amount, coupon rate, maturity, primary bond spread, currencydenomination,) is presented, as well as forecasts concerning future public bond issuesdependingonthepublicfinancingneedsofeachemergingcountry(seeAnnex12foranexampleoftheseforecasts).

Why,therefore,doinvestorsnotpayattentiontotheevolutionoffees?Thisispuzzlinginthatuseful,publiclyavailableinformationisnottrackedbyinvestorstohelpimproveallocationoftheiremergingmarketfixedincomeassets.

A possible hypothesis is that the heterogeneity of information contained in the feescomplicates their use as an early warning indicator for investors. Indeed, we have studiedstandard issues inorder toanalyzeunderwriting feeswith respect to sovereignbond spreadsaround sovereigndebt crises. It could be that there is a high fixed cost in order to dissectunderwritingspreadsthatindividualinvestorsareunableorunwillingtoincur.

In that context and for future research, it can be useful to determine whether otherrelevant variables could have an impact on fees. These variablesmay be associatedwith thecharacteristicsofthebondissued(e.g.currencydenomination,maturity,regulationofthebond),withtheissuer(e.g.creditrating),totheinvestmentbankactingasunderwriter(e.g.reputation,recommendations given by the underwriter47) and/or with external financial conditions.

46. ThesebanksareABNAMRO,BarclaysCapital,BearStearns,CitigroupformerSalomonSmithBarney,CreditSuisseformer

Credit Suisse FirstBoston ,DeutscheBank,DresdnerKleinwortWasserstein,Goldman Sachs, JPMorgan,LehmanBrothers,MerrillLynch,MorganStanleyandUBS.

47. Inthecorporatemarket,thereisavastliteraturethatstudiespossibleconflictofinterestproblemsbetweeninvestmentbankingactivitiesand informationgivenby investmentbanks to investors. Inparticular, the relationshipbetween the structureof theprimarymarketandrecommendationsgivenbyunderwritershasbeenanalysedforthecorporatemarketbystudyingtheimpactoffavourablerecommendationsonunderwritingactivities(seeLinandMcNichols,1998;ChenandRitter,2000;Ljungqvistetal.,2006;Bradleyetal.,2003;MichaelyandWomack,1999).

OECD2008 33


DEV/DOC(2008)10

VII.CONCLUSIONS

Thispaperhasanalysed informationproblemsbetween investmentbanksand investorsduringsovereigndebtcrises,bystudyingthestructureoftheprimarysovereignbondmarket.

The findingsof thispaperare the following.First,wecannot reject thehypothesis thatinvestment banks price sovereign default risk well before crises emerge and even beforeinvestorsdo.This result suggests that investment banks hold an information advantage overinvestors, and are the only ones compensated to cover the risk of sovereign debt crises.Onaverage,countrieswithpublicfinancesdifficulties(PFD)hadtopayonaverage1.10percentoftheamountissuedtoinvestmentbanksbetweenoneandthreeyearspriortotheonsetofcrisis,almost twice the emerging countries average during the period 19932006 (0.56per cent). Incontrast,whenwecompare the levelofprimarysovereignbondspreadsbefore thecrisiswithrespect to the total foremergingcountries,we find that the formeronaverage isonlyslightlyhigherthanthelatter(385bpvs.319bp)andwelllowerthantheprimarysovereignspreadattheonsetofacrisis(603bp).Therobustnessofthisresultisverifiedthroughpaneldataanalysis.Wefindthatthereisanadditionalfixedcostforcrisiscountriespriortheonsetofasovereigndebtcrisiswithrespecttootherevents.Onaverage,priortoasovereigndebtcrisis,countriespaidasurplusontheunderwritingfee.

Second, investment banks behaviourdiffersdepending on the type of sovereigndebtcrisis. By differentiating among types of sovereign debt crises, we find that prior to crisesinvestment banks charged on average higher fees to countries with public financesvulnerabilitiescomparedtoothersovereigndebtcrisescountries.

Finally,thereisadichotomybetweeninvestorsperceptionsconcerningtheinvestmentvalueofthefeeandtheresultspresentedabove.Thisisapuzzleinthatitappearsthatinvestorsare not using potentially useful (and public) information in order to allocate efficiently theirportfoliosofemergingfixedincomesassets.

The major policy implication that follows from this research is that fees paid bygovernments to investmentbanksduringsovereignbond issuesshouldbeclosely followedbypolicymakersandactors in capitalmarkets.This informationhasbeenneglectedduringpastsovereign bond crises, while it may have served as an early warning market indicator ofsovereign bond crisis, and one perhaps more relevant than standard indicators such assecondarysovereignbondspreads(Edwards,1984;Grandes,2002;DuffieandSingleton,2003)orCredit Default Swap (CDS) spreads (Andritzky, 2003; ChanLau, 2003; Ciarlone et al., 2005;Remolonaetal.,2007).

34 OECD2008


DEV/DOC(2008)10

ANNEXES

ANNEX1

Debtrestructuringcasesandrecoveryrates

Argentina 2001 2005 33 30 56.0 6482 13.65 PostdefaultDominicanRep. 2005 2005 92 95 0.0 NA 0.39 Preemptive

Ecuador 1999 19992000 44 60 37.3 19 47 1.32 PostdefaultPakistan 1999 1999 65 65 1.0 2932 NA PreemptiveRussia 1998 19982000 18 50 17.2 50 75 6.32 PostdefaultUkraine 1998 19982000 69 60 0.0 2235 NA PreemptiveUruguay 2003 2003 66 85 1.0 520 0.12 Preemptive

Weightinthesecondarymarket****(%totalEmerging

SovereignDebt)

Haircut(accordingtoSturzenegger

andZettelmeyer,

2005)

CountryDefaultyear(accordin tog

S&Ps)

Debtrestructuring

year(accordingtoIMF,2006)

AverageTradingPrice(%ofpar)*

PVRatioofCashFlows**

(%)

NominalPrincipal

Reduction***(%debt

restructured)

Restructuringcase

(accorodingtoIMF,2006)

Notes:*30daypostdefaultpriceorpredistressedexchangetradingprice(Moodys,2006).**Ratioofthepresentvalueofcashflowsreceivedas resultofthedistressedexchangeversusthoseinitiallypromised,discountedusingyieldtomaturityaimmediatelypriortodefault(BankofEngland,2005).***Negativenumbersindicateanincreaseinprincipal(IMF,2006).

****Itcorrespondstotheweightofeachcountryonemonthpriortotheonsetofthecrisis(accordingtotheweightoftheEMBIGlobalindexcalculatedbyJPMorgan).

:Source BankofEngland(2005),IMF(2006),Moodys(2006),StandardandPoors(2006)andSturzeneggerandZettelmeyer(2005).

OECD2008 35


DEV/DOC(2008)10

ANNEX2

DebtRiskindexattheonsetofcrisisSample:NonconcessionalIMFloans/quota>100percent

Mexico Indonesia Thailand Brazil Turkey Brazil MiddleInc.1995 1997 1997 1998 2000 2001 19952002

Total Debt Outstanding 10.6 10.9 7.2 17.0 10.2 18.1 6.7/MiddleIncomecountries(9502) 1.6 1.6 1.1 2.6 1.5 2.7 1.0/MiddleIncomecountries(crisisdate) 1.6 1.7 1.1 2.4 1.5 2.7Public Debt 8.6 6.2 2.0 7.0 6.7 8.4 4.7/MiddleIncomecountries(9502) 1.9 1.3 0.4 1.5 1.4 1.8 1.0/MiddleIncomecountries(crisisdate) 1.6 1.3 0.4 1.5 1.5 2.0Public Bonds 4.6 0.2 0.3 3.7 3.0 5.3 1.7/MiddleIncomecountries(9502) 2.7 0.1 0.2 2.2 1.8 3.2 1.0/MiddleIncomecountries(crisisdate) 3.0 0.1 0.2 2.2 1.6 3.1

Notes:1.GDFonlycalculates these indicators for theTotalDebt.Wehaveadapted thesedefinitions to the totaloutstandingdebt,publicdebtandpublicbonds.

2.Theindicatorsusedtocalculatethisindexare:DebtServiceoverExportsofGoodsandServices,InterestPaymentsoverExportsofGoodsandServices,DebtoverExportsofGoodsandServices,InternationalReservesoverDebt,DebtoverGrossNationalProduct(GNP)andInterestpaymentsoverGNP.Inordertoconstructthisindex,wegivethesameweighttoeachindicatoraccordingtoitsvalueforMiddleIncomecountriesduringtheperiod19952002(theperiodthatenclosestheentirecrisessample).

3.MiddleIncomecountriesaredefinedaccordingtotheWorldBank(seeDataandStatisticsathttp://www.worldbank.org/).Allthecountriesstudiedinthispaperareincludedinsidethiscategory.

Source:TheauthorfromGlobalDevelopmentFinance(GDF).

36 OECD2008


DEV/DOC(2008)10

ANNEX3

ExchangeMarketPressureIndex

Theexchangemarketpressure index captures the threemaindefencesa countryhasagainstaspeculativebalanceofpaymentattack(devaluateitscurrency,raiseinterestratesorpayoutreserves).Mathematically, theweighted indexofexchangemarketpressure (EMP)used inthisarticlecanbewrittenasfollows:

( ) ( )[ ]

+

=

1

1

1

111

1

1

Ut

UtUt

nt

ntntUtntUtnt

nt

ntntnt R

RRR

RRiiii

EEE

EMP

where 11

nt

ntnt

EEE

is the nominal exchange rate variation for country ns currency inUS

dollarsbetweentimetandt1(thetimeperiodisinannualbasis). and arerespectivelythe

shortterm interest rate for country n and for the US at time t.

nti Uti

1

nt

nt

RR 1ntR

and 11

Ut

UtUt

RRR

arerespectivelythevariationoftheinternationalreservesofthecountrynandtheUSbetweentimetandt1.

Higher values of the index indicate greater levels of exchangemarket pressure. Eachcomponentoftheindexisalsoweightedbytheinverseofthestandarddeviationforeachseries

(the , and ) inorder toequalizeconditionalvolatilitiesandensure thatnosingleseriesdominatestheindex.

Theperiodused tocalculate this indexstarts in January1990andends inAugust2007.Weuse15countries,namelyArgentina,Brazil,DominicanRepublic,Ecuador,Indonesia,SouthKorea,Malaysia,Mexico,Pakistan,Philippines,Russia,Thailand,Turkey,UkraineandUruguay.

Thedatacollected is fromDatastream48andcompiledonamonthlybasis.Weexcludethe cases in which countries present an annual rate of inflation (with CPI index data fromDatastream)greaterthan100percent.Indeed,asnotedbyKaminskyandReinhart(1999),whilea 100per cent devaluationmay be traumatic for a countrywith lowtomoderate inflation, adevaluationofthatmagnitudeiscommonplaceduringhyperinflations.

The finalstep is to indicate thecriticalvalue for theEMP indexsuch that indexvaluesabovethislevelqualifyasacrisis.WeusethecriteriausedbyEichengreenetal.(1995):

48.The exceptionsare the exchange rateofDominicanRepublic (theDominicanCentralBank is the source)and the

InternationalReservesofPakistan(thesourceistheInternationalInstituteofFinanceIIF).

OECD2008 37


DEV/DOC(2008)10

otherwiseCrisis

EMPifCrisis

nt

EMPEMPntnt

0

5.11

=

+>=

i)

Inordertorealizeaseriesofsensitivitytests,wechangethedefinitionofacrisisusedinthe equation above.More specifically,weuse twomore andone less stringent criteria for aneventtoqualifyacrisis.Wedefineacrisisasanymonthforanycountryinthesamplewhere:

EMPEMPntnt EMPifCrisis 0.21 +>= ii)

EMPEMPntnt EMPifCrisis 0.11 +>= iii)

EMPEMPntnt EMPifCrisis 0.31 +>= iv)

The resultsare listed in the tablebelow for theperiod starting in June1994 toAugust2007.Manyoftheonemonthcrisisperiodsareclearlyasinglecrisisevent(e.g.Malaysia19971998, Russia 199899, Ukraine 199899, Argentina 200102). Therefore, we include any crisismonththatoccurswithintwoyearsofacountrysinitialcrisisaspartofasinglecrisisevent(i.e.acountrycanhaveatmostonecrisisevery24months).Bycontrast,weexcludecrises thatareinferiortoonemonth(thisiscaseofPakistaninOctober1998).

Finally, inorder to find thepeakof the currency crises,we take thedate inwhichweobtain the higher value of the index of exchangemarket pressure during currency crises.Mathematically:

EMPEMPnttntt EMPEMP 5.12323 +> v)Where ntt EMP23 is thehighervalueof theEMP indexofcountryn takenplaceduring

thecrisisstartinginthemontht23andfinishinginthemontht.

38 OECD2008


DEV/DOC(2008)10

Country Peak of the Crisis

Argentina Apr-02 Jul-01, Nov-01, Dec-01, From Apr-02 to Sep-02 Jul-01, Aug-01, From Nov-01 to Feb-02, From Apr-02 to Oct-02

Brazil Mar-99 From Jan-99 to Aug-99, Oct-99 From Jan-99 to Dec-99, From Oct-02 to Mar-03

From Mar-90 to Jun-90, From Aug-90 to Jul-91, Dec-01, Feb-02, Mar-02, Sep-02, From Mar-03 to May-04

Ecuador

Indonesia Jun-98 From Dec-97 to Oct-98 Jan-01, From Mar-90 to May-90, From Dec-97 to Nov-98

S. Korea Dec-97 Jan-90, Apr-90, From Nov-97 to Jun-98 From Jan-90 to Aug-90, From Jun-91 to Nov-91, Fom Nov-97 to Jul-98

Malaysia Jun-98 From Nov-97 to Aug-98, Oct-98, Nov-98 From Oct-97 to Aug-98, Oct-98, Nov-98, Oct-01, Nov-01

Mexico Mar-95, Sep-98 From Dec-94 to Dec-95, Sep-98, Oct-98 Mar-90, From Dec-94 to Dec-95, From Sep-98 to Jan-99

From May-98 to Jul-98, Oct-98, Aug-99, Sep-99, Jun-00, Nov-00,From Jan-01 to Sep-01

Feb-94, Apr-94, From Jul-97 to Jun-98, Aug-98, Sep-98, From Oct-00 to Jan-01, Mar-01, Apr-01

Russia Sep-98 From Sep-98 to Apr-99 From May-98 to Aug-99

Thailand Jan-98 From Jul-97 to Jun-98 From Jul-97 to Jun-98

From Jan-94 to Mar-94, Feb-98, Oct-98, Nov-98, Jan-99, From Apr-99 to Jul-99, Dec-00, From Feb-01 to Nov-01, Jan-02

Ukraine Oct-98 From Aug-98 to Jun-99, Aug-99 From Jul-98 to Aug-99, Nov-99, Dec-99

Uruguay Sep-02 From May-02 to Apr-03 From Jan-02 to Apr-03

Oct-97 Jul-97, From Oct-97 to Jun-98, Aug-98

Feb-01Turkey

Philippines

Feb-94, Mar-94, Feb-01, Apr-01, From Jun-01 to Oct-01, Jan-02

The Currency Crisis Events

Pakistan

Dom. Rep. Feb-04

Apr-01

From Oct-90 to April-91, From Jun-03 to Mar-04

Nov-98, From Apr-01 to Sep-01

EMPEMPntnt EMPifCrisis 0.11 +>=EMPEMPntnt EMPifCrisis 5.11 +>=

Country Peak of the Crisis

Argentina Apr-02 Nov-01, From Apr-02 to Jun-02

Brazil Mar-99 Feb-99, Mar-99

Dom. Rep. Feb-04 Feb-91, Mar-91, From Jul-03 to Aug-03, Jan-04, Feb-04

Ecuador

Indonesia Jun-98 From Jan-98 to Sep-98 From Jan-98 to Jul-98

S. Korea Dec-97 From Dec-97 to May-98 From Dec-97 to Feb-98

Malaysia Jun-98 From Dec-97 to Aug-98, Oct-98, Nov-98 Jan-98, Jun-98

Mexico Mar-95, Sep-98 From Jan-95 to Dec-95 From Jan-95 to Jul-95, Oct-95, Nov-95

Pakistan Apr-01 Apr-01, Jul-01

Philippines Oct-97 Oct-97, From Dec-97 to Apr-98, Jun-98 Oct-97

Russia Sep-98 Sep-98, Oct-98, Dec-98, Jan-99, Mar-99, Apr-99 Sep-98, Oct-98

Thailand Jan-98 From Aug-97 to Jun-98 From Oct-97 to Jun-98

Turkey Feb-01 Feb-94, Mar-94, Feb-01, Apr-01, From Jul-01 to Oct-01 Mar-94, Feb-01

Ukraine Oct-98 From Sep-98 to Nov-99, From Feb-99 to May-99, Aug-99 Sep-98, Oct-99, Mar-99

Uruguay Sep-02 From May-02 to Jan-03 Jul-02, Sep-02

Source: AuthorscalculationsbasedonDatastream

EMPEMPntnt EMPifCrisis 31 +>=EMPEMPntnt EMPifCrisis 21 +>=

OECD2008 39


DEV/DOC(2008)10

ANNEX4

Theconnectionbetweensovereigndebtandcurrencycrises

There is a significant link between balance of payment problems and sovereign debtcrises,whichhasbeenstudiedintheempiricalliterature(e.g.Burnsideetal.,2003;Reinhart,2002;Kaminsky,2006;Herzetal.,2007).

With thepurposeof examininghow the sovereigndebtand currency crisesare linkedduringtheperi

who saw sovereign debt crises coming?

Documents

oecd development centrewho

sovereign debt crises

oecddevelopment centre

rue andr pascal

paris cedex