关闭

关闭

关闭

封号提示

内容

首页 The great surge in mortgage defaults 2006–2009.…

The great surge in mortgage defaults 2006–2009.pdf

The great surge in mortgage def…

上传者: fengqiyaoye 2013-10-10 评分 0 0 0 0 0 0 暂无简介 简介 举报

简介:本文档为《The great surge in mortgage defaults 2006–2009pdf》,可适用于高等教育领域,主题内容包含sganiverersity,explosinationcconomthoffintomfthegulatngfrothatattempttodeco符等。

sganiverersity,explosinationcconomthoffintomfthegulatngfrothatattempttodecomposethecausesofthegreatsurgeindefaultsfromtousingalargedatasetwithloanlevelinformationOneoutgrowthofthisresearchisasetofmetricsthatcanbeusedtotrackthequalityofmortgagewritingcharacteristicsthatwereknowntoinvestorsinthepoolsWiththesedatawecandecomposedefaultsintothreeparts:thoseduetoworseningofeconomicconditions,thoseduetoobservablechangesinunderwritingstandards,andasetoftimevaryingfixedeffectsthatweattributetomoralhazardWeidentifymoralhazardbylookingfordiscontinuitiesor‘‘notches’’indefaultbehaviorTheunderlyingeconomics$seefrontmatterPublishedbyElsevierIncCorrespondingauthorEmailaddress:rvogwuedu(RVOrder)JournalofHousingEconomics()–ContentslistsavailableatScienceDirectJournalofHousiweldoi:jjheandhousinggyrationsofthelastfewyearsForexample,inOctober,nearthepeakofthehousing‘‘bubble’’FederalReserveChairmanBenBernankewasstillarguingthathousingprices‘‘largelyreflectstrongeconomicfundamentals’’andwerenocauseforconcernGiventhefirestormthateruptedshortlythereafterandthatspreadfromthemortgageandhousingmarketstothenationalandworldeconomy,itisundoubtedlymeiosistosaythatbettermethodsandmetricsformonitoringmortgagemarketsareurgentlyneededTowardthisend,thisresearchpresentsinitialestimateswhichtheycanonlymeasureindirectly,usingfixedeffectsforobservationyearTheirresultssuggestthattherewasatrendtowardlowerunderwritingstandardscharacterizedbytwomajorperiodsofdeterioration,oneinthemiddleandlatesandoneafterAfterthefavorableeconomicconditionsthathadmaskedunderwritingdeteriorationchanged,anddefaultsincreasedsharplyThedatainthisresearchcontainsloanlevelinformationonmortgagesinnonagency(notFannieMae,FreddieMac,orGinnieMae)mortgagebackedsecuritiesthatwereoriginatedfromtoThisdatasetincludestheunderIntroductionandoverviewPerhapsthemostimportanttasksincethegreatdepressionhasbeenonlyofindividualbanks,butalsooie,systemicriskUnfortunately,retoanticipatethesystemicrisksarisiancialregulatorsonitorrisksnotbankingsystem,orswereunablemthemortgageportfoliosandseparatethesourcesofdefaultsintoindicesforunderwriting,moralhazardandeconomicconditionsOuranalysisisanextensionoftheearlierworkinAndersonetal(),ACVhenceforthACVuseaggregateddata(foreclosureratesbystate)thatextendbacktothestodecomposedefaultsintothosecausedbyeconomicconditionsandthosecausedbyunderwriting,conditionsPublishedbyElsevierIncThegreatsurgeinmortgagedefaultofeconomicconditions,underwritinDennisRCapozzaa,RobertVanOrderb,aDaleDykemaProfessorofBusinessAdministration,RossSchoolofBusiness,UbOliverTCarrProfessorofFinanceandRealEstate,GeorgeWashingtonUnivarticleinfoArticlehistory:AvailableonlineMayKeywords:CreditriskCreditcrunchMoralhazardabstractInthisresearchwenatedfromtoeconomicconditionobservetheinformaswellaseconomihazardEstimatesfrjournalhomepage:ww–:ThecomparativerolesndmoralhazardsityofMichigan,USAUSAitthepowerofalargeandrichsampleofindividualloansorigitostudytherelativerolesofunderwriting,moralhazardandlocaltheGreatSurgeinmortgagedefaultsWiththesedatawecanavailabletoinvestorsandcontrolforobservableunderwritingditionsWecanalsousethedatatoinfertheshareduetomoralesedatasuggestthatmuchofthevariationwasduetoeconomicngEconomicsseviercomlocatejhecthroughThereisarisingtrendwithoccasionalleveluarterDRCapozza,RVOrderJournalofHousingEconomics()–suggestthatdefaultshouldbeacontinuousfunctionofunderlyingvariableslikeloantovalueratio(LTV)andcreditscore,aswellother,hardertoobservevariablesHowever,pricingandscreeningtendtobedoneoverdiscreteintervalsinpricingandunderwritingmatricesTheconditionsformoralhazardoccurwhenloansellersandsecuritizershaveaccesstobetterinformationthanthatavailabletoinvestorsThequestionis:whatsortsofmortgageswillbedeliveredbytraderswithasymmetricinformationAnLTVandacredit(FICO)scoreappeartobecriticalminimumstandardsofqualitybecauseanLTVaboverequiresinsurance,andatleastaFICOisgenerallyrequiredforagencypurchaseOurhypothesisisthatloansellerswhopossesssuperiorinformationrelativetoinvestorswilltendtodeliverloansthatjustmeettheseminimumstandardsBecauseofthesecutoffpointsinthestandardsforloanquality,weexpecttoseenotchesinthedefaultfunctionatoraroundthesepointsWedoindeedfindnotches,particularlyatLTVHowever,wefindthattheyarenotespeciallyimportantinexplainingthesurgeindefaults,whichappearstobedueprimarilytodeterioratingeconomicconditions,particularlyhousepricedeclines,whichhadpreviouslybeenFigAllforeclosuresstarted:US–quarterlydataFourqDelinquencySurveyveryfavorableBecauseeconomicconditionsweresofavorable,theymasked,orperhapsevencontributedto,atrendofdeterioratingunderwritingconditionsInthenextsectionwedocumenttheGreatSurge,itsimplicationsforbothprimeandsubprimeloans,andtheconcurrenteconomicconditions,especiallyhousepricesThethirdsectiondevelopsandestimatesmodelsofdefaultusingtheloanleveldatasetThemodelisthenappliedtoextracttwotypesofestimatesoftheimportanceofeconomicconditions,underwritingandmoralhazardinthegreatsurgeBackgroundandsummarydataThelongruntrendinforeclosuresFiggraphsthetimeseriesofforeclosuresstartedasapercentoftheoutstandingnumberofloansfromingoffBetweenandforeclosureratesquadrupledfromtoIntheimpressiveGreatSurgefromto,foreclosureratesalmostquadrupledagainfromtoinjustthreeyearsOurpurposewiththisresearchistoanalyzetheavailabledatatoenableadeeperunderstandingofboththetrendandthesurgeThedeteriorationinmortgageperformancehasvariedbyloantypeFigpresentsdatafor–onforeclosuresstartedbymajorproducttypeTheverticalaxisistheannualizedpercentofloansthatentertheforeclosureprocessovereachfourquartersNoteinparticularthehistoryofsubprimeForeclosuresfellaftertherecessionbutthenincreasedsharplyafterAsimilarpattern,butonasmallerscaleandwithaboutaoneyearlag,occurredintheprimemortgagedata,suggestingthatthereisacommonfactoraffectingbothprimeandsubprimeandthatthesurgeinforeclosuresisnotjustasubprimeissueThelagbetweenprimeandsubprimemakessubprimethecanaryinthecoalmineSubprimeborrowersrespondmorequicklytofinancialstressthanprimeborrowersaveragesannualizedSource:MortgageBankersAssociationNationalThetrendsinhousepricesThedataintheprevioussectionsuggestthattheremaybeanimportantrolethateconomicconditionsplayinthepatternofdefaultsModerncontingentclaimsbasedtheoriesofmortgagevaluationtreattheborrower’spositionaslongaputonthecollateralOneimplicationofthisapproachisthattheputoptionshouldbesensitivetothevalueofthecollateralWhencollateralpricesarerising,afinanciallystressedborrowerwithequitywillrationallychoosetosellthecollateralratherthandefaultCorrespondingly,whencollateralpricesarefallingandequitybecomesnegative,thestressedborrowerismorelikelytochoosedefaultTheMBAdatadonottrackhowmanyactuallywentthroughforeclosuretoREO,realestateownedbylendersSubprimerespondsearlierbut,theeventualtotalresponseissmallerinpercentagetermsthanforprimeloans,albeitmuchlargerinabsolutetermsTheearlyexampleisFindlayandCapozza()DRCapozza,RVOrderJournalofHousingEconomics()–Therefore,itisimportanttounderstandbothcollateralpricesandfinancialstressastheyrelatetoforeclosuresFigplotsrealandnominalhousepricessinceThereisacyclicalpatterntorealhousepricesfromtowithinanarrowrangeHowever,duringtheboomyearsfromto,realhousepricesroseaboutabovewhathadbeenthelongrunleveluntilthattimeSuchsteepincreasesshouldbeexpectedtogreatlyreducetheneedforstressedborrowerstodefaultSincemanylendersdevelopunderwritingmodelsbyevaluatingrecentloanperformancedata,anyunderwritingmodelscreatedusingdatafromthisboomperiodwouldunderestimatetheriskstolendersinmoreaveragetimesFigRealandnominalhousepriceindices,FigRateofforeclosuresstartedbyloantype,–()SourcFigprovidesanotherperspectiveontherecentperiodbyplottingtherealappreciationrateofhousepricesThelongrunevidence(eg,Eichholz,)isthatrealhousepricesappreciateatratesclosetozerooverdecadesandcenturiesThusthe–annualappreciationratesofthelastdecadeareextraordinaryFigillustratesthevariationinhousepricesforselectedmetroareasItshowstheextremeupsanddownofsomethemetroareaslikeSanDiegoandMiamithathadprice‘‘bubbles’’TheseareascurrentlyhavehighforeclosureratesfollowingpricedeclinesOthermetroareashadsmallerincreases(BostonandDetroit)withslightlydisplacedpeaksandtroughsDetroitdidnotexperience–(=)Source:FHFAe:MortgageBankersAssociationNationalDelinquencySurveythe‘‘bubble’’levelofpriceincreases,butneverthelesshasbeenexperiencingelevateddefaultratesHighandpersisThesenonAgencydataextendonlytobutcontainalargeamountofdataattheloanlevelonborrowerandloancharacteristicsDRCapozza,RVOrderJournalofHousingEconomics()–DefaultmodelsACVusethedatasetfromtheMortgageBankersAssociationwhichislonger,butaggregatedbystateHereweestimateusingtheloanleveldatafromnonAgencypoolsTheUFAindicesareavailabletoacademicresearchersatwwwufanetcomAsummarystatisticforeconomicconditionsTheFigsabovehighlightthepossibleeffectsofeconomicconditionsonforeclosuresIntheanalysisthatfollowswesplitforeclosureratesintocomponentsarisingfromunderwritingpolicyandchangesineconomicconditionsAsasummarymeasureofeconomicconditionsweusethequarterly‘‘ForeScore’’DefaultRiskindicesbystatecompiledbyUniversityFinancialAssociates(UFA),whichtracktheeffectoflocalandnationaleconomicconditionsontheprobabilityofaconstantqualityloaneverdefaultingTheUFAindicesarederivedfromamodelofdefault,usingbothlocalconditionsandloancharacteristicstoexplaindefaultTheindicesholdstheloancharacteristicsconstantandprojecttheimpactofeconomicconditionsondefaultHousepricechangesarethemostimportantdriveroftheindiceswithothereconomic,demographic,politicalandtopographicvariablesexplainingthebalanceTheindicesenableparsimoniousestimationoftheequationsthatfollowAmoredetailedexplanationiscontainedinACV()FigillustratesoneuseoftheindextotracknationwidedefaultrisksovertimeInthiscasetheconstantqualityloanismovedthroughtimeandspacetocreateanationalindexforeachvintagebyaveragingacrosslocationseachyearTheIndexhasvariedbetweenand,ie,thevariationineconomicconditionshasbeensufficienttocauseaquadruplingofdefaultratesfromtroughtopeakduringthelastdecadeonaconstantqualityloanTherehavebeentwotrendsintheDefaultRiskIndex:improvementfromuntilaroundandthenasharpdeteriorationItshouldbenotedthattheIndexisaforwardlookinglifeofloanpredictionforloansoftheindicatedvintageTheprojectionsinthefigureuseactualdatatotheextenttheyareavailable,andthen(forrecentindices)useforecasts,eg,ofhouseprices,overthelifeofeachloanvintageWhentheindexbeginstoincreasefromon,itisnotnecessarilybecausethemodel‘‘expects’’theindicatedvintagetodefaultathighratesimmediatelyAnyincreaseduringthelifeoftheloanwillaffectthelifeofloanindexvalueforthatvintageThefiguresuggeststhatindeedeconomicconditionscouldbeamajorfactorinexplainingrecenthistorytentlevelsofunemploymentinDetroitcreatehighlevelsoffinancialstressforborrowersthatinteractwiththedecliningcollateralpricesACVexploitthistimeandspatialvariationinforeclosuresintheiranalysisoftherateofforeclosuresintheMBAservicedportfoliodataThedefaultmodelisasfollows:lettheconditionalprobabilityofdefaultforaloantoborroweri,originatedattimevinregionr,observedattimetbe:dvitraðtvÞebXðrtÞþcYiðrvÞþdGðrÞðÞwhereX(r,t)isavectoroftimevaryingcovariatesthatdescribetheeconomyinregionrattimetYi(r,v)isavectorofcharacteristicsofloansinregionrattimeoforigination,vG(r)isavectorofvariablesthatarenottimevaryinganddescriberegionra(tv)isthebaselinehazardforloanagetvb,candarevectorsofcoefficientsTheACVModelACVestimateEq()fromtheMBAdatabutdonotobserveindividualloans,nordotheyknoworiginationyear,soonlydtr,theshareofloansinregion(state)rthatgointoforeclosureattimet,isobservedTheACVmodelaggregatesacrossindividualsandoriginationyearsdtrebXþdGðrÞXvXiaðtvÞecYiðrvÞ=nrtðÞwherenrtareisthenumberofloansoriginatedpriortotimetinregionrthatarestillaliveattimetThisiswhatweestimatefirstTakinglogarithmsofbothsidesof():logðdtrÞbXðrtÞþdGðrÞþlogXvXiaðtvÞecYiðrvÞ=nrt!ðÞwhichcanbesimplifiedtologðdtrÞbXðrtÞþdGðrÞþetmtrþdrfrþetðÞwherefrisafixedeffectforregionrandmtristheForeScoreindexthatappliestoloansoriginatedinstaterattimet,andetisanerrortermTheerrortermisquitecomplicatedItisaweightedaverageofunderwritingcharacteristicsofthepoolofloansacrossthedifferentvintagesACVdecomposetheerrortermin()intotimefixedeffectsandeverythingelsetoget:logðdtrÞmtrþdrfrþdtftþuðrtÞðÞwhereftisasetoffixedeffectsfortimeanduisagaincomplicatedUseofthetimeeffect,ft,asaproxyforcreditstandardsmeanstheycannotdistinguishchangesinloanqualitythataredeliberatechangesintheYvectorfromotherunobservedchangesinloancharacteristicsAshortcomingofthisaggregationacrossvintagesisthatitrisksconfusingchangesinstandardswithchangesinthehistoricdistributionofloansbyvintageandtheirsurvivalratesWealsodonotconsiderthepossibilitythatthetimefixedeffectsmightbeduetochangesinborrowerbehavior,suchasanincreasedwillingnesstodefaultandincludestatefixedeffectspurchDRCapozza,RVOrderJournalofHousingEconomics()–ACVestimateequationsoftheformlogðdtrÞXltatmðrtÞþdrfrþdtftþXqtctutþetðÞwherebarsovervariablesindicateafourquartermovingaverageofthevariable,andlandqarelaglengthsFrom()weshouldexpectthesumofthecoefficientsofmðrtÞin()tobeclosetooneResultsforMBAdata:therelativerolesofeconomicconditionsandunderwritingSimulationsACVusetheirestimatedequationsalongwiththefixedBecausethemtraretheprobabilityofeverdefaultingtheydonotapplytothesametimeperiodasdtrandbecausetherearelagsinadjustmentofdtrtochangesinmtr,ACVestimateversionsof()wherebothdtrandmtrarefourquartermovingaveragesandtherighthandsidehaslagsACVallowutobeanautoregressiveprocess,FigUSrealhousepriceappreciation,–seasonallyadjusted,effectstodecomposeforeclosureratesintoapartduetotheeconomicmultipliersandapartduetotheyearfixedeffectsTheyearfixedeffectsconditionalonthemultipliersaretheestimatesoftheunderwritingcomponent,ie,ofdefaultratesaftercontrollingforeconomicconditionsWhennormalized,thefittedvaluesfromtheregression,ie,differencebetweentheunconditionalyearindicators(ie,theactualyearlyforeclosurerates)andtheyearindicatorsconditionaloneconomicconditionsisanestimateoftheeconomiccomponentByconstructionthetwoadduptotheactuallevelofforeclosuresFigpresentsresultsusingtheACVModelforallloansTheyellowlinegivesthepartduetoeconomicconditions(howforeclosureswouldhavemovedhadunderwritingnotchanged),whichpromoteddecliningforeclosuresuntilThepinklineshowsthecontributionofunderwriting(howforeclosureswouldhavechangedhadeconomicconditionsnotchanged),whichwaspositiveearlyintheperiod,negativelaterandsharplypositiveintoForexample,inFigtheredcurveforunderwritinginiswhileactualdefaultsareandtheeconomicconditionsindexisTheinterpretationisthatwhileactualdefaultratesrosefromto,ifeconomicconditionshadnotbeensofavorable,foreclosuresstartedwouldhaverisenbyinsteadofStateddifferently,underwritingqualityerodedenoughtodoublethelevelofforeclosuresstartedbybutonlyaincreasewasrealizedbecausefavorableeconomicconditionsoffset=ofthepotentialincreaseDuringthisperiodhousepricesappreciatedsteadilyinmostofthecountryNotethattheunderwritingeffectsrefertotheyearinwhichtheloansareobserved,nottheyearinwhichtheywereoriginatedThepoorunderwritingresultsinandareforloansthatwereoriginatedearlierThefiguresuggeststhatthepostincreaseinforeclosurescanbeapportionedaboutequallybetweentheunderwritingandeconomicconditionsThespectacularincreaseinforeclosuresafterisunprecedentedinthedataEconomicconditionsandunderwritingqualitytypicallymovedinoppositedirectionsinthesThisnegativecorrelationisconsistentaseonlyindex,quarterlydata,fourquarterpricechangesSource:FHFAwithlendersbecomingmoreconservativewheneconomicconditionsareweakHowever,after–,economicconditionsandqualitybothdeteriorated,breakingtheearlierpatternandsuggestingapossiblestructuralbreakorregimeshiftinthismarketthatisconsistentwithamoralhazardstoryThedatasuggestthatthepostincreaseinforeclosurescanbeapportionedaboutequallybetweentheunderwritingandeconomicconditionsexplanationsLoanlevelmodelInthissectionwepresentresultsfortheloanlevelmodel,whichusesthenonAgencydatatoestimateafullversionofEq()Thedataincludeprime,subprimeandAltAloansWhilethatdatasetisrich

职业精品

用户评论

0/200
    暂无评论

精彩专题

上传我的资料

热门资料

资料评价:

/11
1下载券 下载 加入VIP, 送下载券

意见
反馈

返回
顶部

Q