互聯(lián)網(wǎng)金融P2P網(wǎng)絡(luò)借貸外文翻譯文獻

文獻信息：文獻標題：EvaluatingcreditriskandloanperformanceinonlinePeer-to-Peer(P2P)lending（點對點(P2P)網(wǎng)絡(luò)借貸的信用風(fēng)險與貸款績效評估）國外作者：RizaEmekter,YanbinTu,BenjamasJirasakuldech,MinLu文獻出處：《AppliedEconomics》,2015,47(1):54-70字數(shù)統(tǒng)計：英文3063單詞，15818字符；中文5110漢字外文文獻：EvaluatingcreditriskandloanperformanceinonlinePeer-to-Peer(P2P)lendingAbstractOnlinePeer-to-Peer(P2P)lendinghasemergedrecently.Thismicroloanmarketcouldoffercertainbenefitstobothborrowersandlenders.UsingdatafromtheLendingClub,whichisoneofthepopularonlineP2Plendinghouses,thisarticleexplorestheP2Ploancharacteristics,evaluatestheircreditriskandmeasuresloanperformances.Wefindthatcreditgrade,debt-to-incomeratio,FICOscoreandrevolvinglineutilizationplayanimportantroleinloandefaults.Loanswithlowercreditgradeandlongerdurationareassociatedwithhighmortalityrate.TheresultisconsistentwiththeCoxProportionalHazardtestwhichsuggeststhatthehazardrateorthelikelihoodoftheloandefaultincreaseswiththecreditriskoftheborrowers.Finally,wefindthathigherinterestrateschargedonthehighriskborrowersarenotenoughtocompensateforhigherprobabilityoftheloandefault.TheLendingClubmustfindwaystoattracthighFICOscoreandhigh-incomeborrowersinordertosustaintheirbusinesses.Keywords:Peer-to-Peerlending;creditgrade;FICOscore;defaultriskI.IntroductionWiththeadventofWeb2.0,ithasbecomeeasytocreateonlinemarketsandvirtualcommunitieswithconvenientaccessibilityandstrongcollaboration.OneoftheemergingWeb2.0applicationsistheonlinePeer-to-Peer(P2P)lendingmarketplaces,wherebothlendersandborrowerscanvirtuallymeetforloantransactions.Suchmarketplacesprovideaplatformserviceofintroducingborrowerstolenders,whichcanoffersomeadvantagesforbothborrowersandlenders.Borrowerscangetmicroloansdirectlyfromlenders,andmightpaylowerratesthancommercialcreditalternatives.Ontheotherhand,lenderscanearnhigherratesofreturncomparedtoanyothertypeoflendingsuchascorporatebonds,bankdepositsorcertificateofdeposits.OneoftheproblemsinonlineP2Plendingisinformationasymmetrybetweentheborrowerandthelender.Thatis,thelenderdoesnotknowtheborrower'scredibilityaswellastheborrowerdoes.Suchinformationasymmetrymightresultinadverseselection(Akerlof,1970)andmoralhazard(StiglitzandWeiss,1981).Theoretically,someoftheseproblemscanbealleviatedbyregularmonitoring,butthisapproachposesachallengeintheonlineenvironmentbecausetheborrowersandthebuyersdonotphysicallymeet.Fosteringandenhancingthelender'strustintheborrowercanalsobeimplementedtomitigateadverseselectionandmoralhazardproblems.Inthetraditionalbank-lendingmarkets,bankscanusecollateral,certifiedaccounts,regularreporting,andevenpresenceoftheboardofdirectorstoenhancethetrustintheborrower.However,suchmechanismsaredifficulttoimplementintheonlineenvironmentwhichwillincurasignificanttransactioncost.Toreducelendingrisksassociatedwithinformationasymmetry,currentonlineP2Plendinghasthefollowingarrangements.First,theLendingClubscreensoutanypotentialhigh-riskborrowersbasedontheFICOscore.TheminimumFICOscoretobeabletoparticipateis640.Second,thetypicalsizeoftheloansproducedinthismarketissmall,whichisunder$35000attheLendingClub.Therefore,theseloansareessentiallymicroloanswhichposearelativelysmalllossincaseofdefault.Third,themarketmakeroffersmatchmakingsystemswhichcanbeusedtogenerateportfoliorecommendationsandminimizelendingrisks.Fourth,ifaborrowerfailstopay,themarketmakerwillreportthecasetoacreditagencyandhireacollectionagencytocollectthefundsonbehalfofthelender.AlthoughtherearecertainstructuresimposedintheonlineP2Pthathelptominimizetherisk,thisformoflendingisinherentlyassociatedwithgreateramountofriskcomparedtothetraditionallending.ThepurposeofthisarticleistoevaluatethecreditriskofborrowersfromoneofthelargestP2PplatformsintheUnitedStatesprovidedbytheLendingClub,whichhelplenderstomakemoreinformeddecisionsabouttheriskandreturnefficiencyofloansbasedontheborrowers'grade.Therearetworelatedresearchquestionsthisarticlewilladdress:(1)Whataresomeoftheborrowers'characteristicsthathelpdeterminethedefaultrisk?and(2)Isthehigherreturngeneratedfromtheriskierborrowerlargeenoughtocompensatefortheincrementalrisk?Lenderscanallocatetheirinvestmentsmoreefficientlyiftheyknowwhatcharacteristicsoftheborroweraffectthedefaultrisk.EachborrowerisclassifiedbycreditgradewithcorrespondingborrowingrateassignedbytheLendingClub.Tomakeanefficientallocation,alendershouldknowwhetherthehigherinterestratessetforhigh-riskborrowersaresufficienttocompensatethelendersforthehigherprobabilitiesofapotentialloss.OurfindingssuggestthatborrowerswithhighFICOscore,highcreditgrade,lowrevolvinglineutilizationandlowdebt-to-incomeratioareassociatedwithlowdefaultrisk.ThisfindingisconsistentwiththestudiesbyDuarteetal.(2012)whoreportthatborrowerswithatrustworthycharacteristicwillhavebettercreditscoresbutlowprobabilityofdefault.Thisresultalsosuggeststhatbesidestheloanapplicants'socialtiesandfriendshipasreportedbyFreedmanandJin(2014)andLinetal.(2013),thefourfactorsdiscussedabovearealsoimportantinexplainingthedefaultrisk.WhencomparingwithUSnationalborrowers,theresultsshowthattheLendingClubshouldcontinuetoscreenouttheborrowerswithlowerFICOscoreandattractthehighestFICOscoreborrowersinordertosignificantlyreducethedefaultrisk.Inrelatingtherisktothereturn,itshowsthathigherinterestratechargedfortheriskierborrowerisnotsignificantenoughtojustifythehigherdefaultprobability.OurfindinghereisconsistentwiththestudybyBerkovich(2011)whoreportsthathighqualityloansofferexcessreturn.II.LiteratureReviewThreemainstreamsofresearchhaveemergedinresponsetothegrowingpopularityofP2Plending.ThefirststreamofresearchexaminesthereasonsfortheemergenceofonlineP2Plending.Thesecondstreamofresearchfocusesondeterminingthefactorsthatexplainthefundingsuccessanddefaultrisk.ThelaststreamofresearchinvestigatestheperformanceofonlineP2Ploanforagivenleveloftherisk.Peergrouplendinghasbeenemerginginlocalcommunitiesandhasattractedtheresearchinthisarea.Conlin(1999)developsamodeltoexplaintheexistenceofpeergroupmicro-lendingprogrammesintheUnitedStatesandCanada.Hefindsthatpeergroupsenablefixedcoststobeimposedontheentrepreneurswhileminimizingtheprogramme'soverheadcosts.AshtaandAssadi(2008)investigatewhetherWeb2.0techniquesareintegratedtosupporttheadvancedsocialinteractionsandassociationswithlowercostsforP2Plending.HulmeandWright(2006)studyacaseofonlineP2Plendinghouse,Zopa,intheUnitedKingdom.TheysuggestthattheemergenceofonlineP2Plendingisadirectresponsetosocialtrendsandademandfornewformsofrelationshipinfinancialsectorunderthenewinformationage.Thereisextantliteraturethatidentifiesthefactorsdeterminingthefundingsuccessanddefaultrisk.UsingtheCanadianmicro-creditdata,GomezandSantor(2003)findthatgrouplendingofferslowerdefaultratesthanconventionalindividuallendingdoes.StudybyIyeretal.(2009)showsthatlenderscanevaluateonethirdofcreditriskusingbothhardandsoftdataabouttheborrower.Linetal.(2013)analysetheroleofsocialconnectionsinevaluatingcreditriskanddiscoverthatstrongsocialnetworkingrelationshipisanimportantfactorthatdeterminestheborrowingsuccessandlowerdefaultrisk.Linetal.(2013)furtherreportthatapplicants'friendshipcouldincreasetheprobabilityofsuccessfulfunding,lowerinterestratesonfundedloans,andtheseborrowersareassociatedwithlowerexpostdefaultratesatProsper.TheimportanceofsocialtiesindeterminingloansfundedisalsoexaminedbyFreedmanandJin(2014).Theresultshowsthatborrowerswithsocialtiesaremorelikelytohavetheirloansfundedandreceivelowerinterestrates.However,theyalsofindevidenceofriskstolendersregardingborrowerparticipationinsocialnetworks.Severalotherstudiesexaminewhethercertainborrowers'characteristicsandpersonalinformationdeterminethesuccessofloanfundinganddefaultrisk.Herzensteinetal.(2008)showthatborrowers'financialstrength,theirlistingandpublicizingefforts,anddemographicattributesaffectlikelihoodoffundingsuccess.StudybyDuarteetal.(2012)furtherarguesthatborrowerswhoappearmoretrustworthyhavebettercreditscorewithhigherprobabilitiesofhavingtheirloansfundedanddefaultlessoften.Larrimoreetal.(2011)demonstratethatborrowerswhouseextendednarratives,concretedescriptionsandquantitativewordshavepositiveimpactonfundingsuccess.However,humanizingpersonaldetailsorloanjustificationshavenegativein?uencesonfundingsuccess.Qiuetal.(2012)furtherrevealthatinadditiontopersonalinformationandsocialcapital,othervariables,includingloanamount,acceptablemaximuminterestrateandloanperiodsetbyborrowers,significantlyin?uencethefundingsuccessorfailure.Galaketal.(2011)furthershowthatlenderstendtofavourindividualovergroupborrowersandborrowerswhoaresociallyproximatetothemselves.Theyalsofindthatlendersprefertheborrowerswhoaremorelikethemselvesintermsofgender,occupationandfirstnameinitial.Moreinterestingly,GonzalezandLoureiro(2014)havesimilarfindings:(1)whenperceivedagerepresentscompetence,attractivenesshasnoeffectonloansuccess;(2)whenlendersandborrowersareofthesamegender,attractivenessmightleadtoaloanfailure(i.e.,the‘beautyisbeastly'effect)and(3)loansuccessissensitivetotherelativeageandattractivenessoflendersandborrowers.Herzensteinetal.(2011)findthatherdingintheloanauctionispositivelyrelatedtoitssubsequentperformance,thatiswhetherborrowerspaythemoneybackontime.III.DataInthissection,theloanapplicants'dataisfirstdescribed,followedbyloandistributionbasedonloanpurposes,creditgradeandloanstatusanditendswiththedetaileddescriptivestatisticsoftheloanapplicants.Thisstudyuses61451loanapplicationsintheLendingClubfromMay2007toJune2012obtainedfrom.Overthestudyperiod,theLendingClublentabout$713milliontoborrowers.Toaddresstheborrowers'behaviourinonlineP2Plending,wefirstexaminethemainreasonsforborrowingmoneyfromothers.Table1liststheborrowers'self-claimedreasonssummarizedintheLendingClub.Almost70%ofloanrequestedarerelatedtodebtconsolidationorcreditcarddebtswithatotalloanamountrequestedofapproximately$387millionand$108million,respectively.Thenumberofloanapplicationsforeducation,renewableenergyandvacationcontributelessthan1%oftotalloanswiththetotalloanrequestedrangingfrom1to3million.TheborrowersstatethattheirpreferencestoborrowfromtheLendingClubarelowerborrowingrateandinabilitytoborrowenoughmoneyfromcreditcards.Thesecondpurposeforborrowingistopayhomemortgageortore-modelhome.Table1.Loandistributionsbyloanpurpose(May2007–June2012)Theloan-seekingpersonsareaskedtoprovidethereasonsforrequestingloans.TheLendingClubusestheborrower'sFICOcreditscoresalongwithotherinformationtoassignaloancreditgraderangingfromA1toG5indescendingcreditrankstoeachloan.Thedetailedprocedureisasfollows:afterassigningabasescorebasedonFICOratings,theLendingClubmakessomeadjustmentsdependingonrequestedloanamount,numberofrecentcreditinquiries,credithistorylength,totalopencreditaccount,currentlyopencreditaccountsandrevolvinglineutilizationtodeterminethefinalgrade,whichinturndeterminestheinterestrateontheloan.Table2reportstheloandistributionbycreditgrade.ThemajorityofborrowingrequestshavegradesbetweenA1andE5.TheHighestloanamountsrequestedarefromborrowerswith‘B'creditgrade,whichcontribute29.56%oftotalamountofloansrequested.Thetotalnumberofapplicantsforthis‘B'creditgradegroupis18707,whichrepresentstotalloansofapproximately$210million.Thelowestloanamountsrequestedarefromborrowerswiththelowest‘G'creditgradewhichaccountsfor1.53%oftotalloans.Thereareonly608loanapplicantsforthislowestcreditrating‘G'groupanditrepresentsapproximately$11millionintotalloanvalue.AccordingtotheLendingClub'spolicy,aloancreditgradeisusedtodeterminetheinterestrateandthemaximumamountofmoneythataborrowercanrequest.Thehighertheloangrade,thelowertheinterestrate.Aborrowingrequestwithalowgraderendersahigherinterestrateasacompensationforahighriskheldbylenders.Table2.Loansdistributionbycreditgrades(May2007–June2012)Notes:TheLendingClubusestheborrowers’FICOcreditscoresalongwithotherinformationtoclassifyaloanfromGradeA1toG5indescendingcreditrisk.Therefore,A1creditgraderepresentsthehighestcreditquality/low-riskborrowers,whereasG5creditgraderepresentsthelowestcreditquality/high-riskborrowers.Totalamountofloansrequestedasapercentageoftotalloanis19.35%forcreditgradegroup‘A’,29.56%for‘B’,19.94%for‘C’,14.84%for‘D’,10.15%for‘E’,4.59%for‘F’and1.53%for‘G’.Finally,PanelAofTable3showstheloanstatusforalltheloanrequestson20July2012.Overall,thedefaultrateis4.60%withtotallossesofapproximately$29million.Another2.45%oftotalloanrequestswhichconstitute$18.6millioncouldbepotentiallylostbecausetheborrowersarelateinmakingpaymentwithin30daysor120daysandnotpayingthenormalinstalments.17.98%oftheloansarefullypaidwithanapproximatevalueof$108million.The$557millionloansareincurrentstatusaccountfor74.91%oftotalloans.Naturally,loanswithalowergradedemonstrateahigherdefaultrate.Therefore,studyonriskmanagementonP2Plendingisrelevantforthelenderstooptimizetheirinvestmentportfolios.PanelBofTable3reportstheloanstatusforthematuredloans.Theoveralllossrateismuchhigherformaturedloans.Among4904maturedloans,914loansarecharged-off,whichrepresent18.6%.Thetotallossis$5.5millionwhichrepresents13%ofallmaturedloansamount.Lessthan1%ofthematuredloansarelateintermsofmakingpaymentwiththeunpaidbalanceofapproximately$27000.80.77%or$33millionofmaturedloansarefullypaid.Table3.Loandistributionbytheloanstatus(May2007–June2012)Table4reportsthegeneralcharacteristicsandcredithistoryoftheonlineP2PloanapplicantsfromtheLendingClub.Basedonoursampleof61451loanapplicants,theaveragemonthlyinterestchargedonaloanis12.34%.Onaverage,471dayspassedfromtheissuedateoftheloan.Theaveragecreditgradeofaborroweris25,whichcorrespondstocreditcategorybetweenBandC.Theaveragesizeofatypicalloanis$11604andtheaveragemonthlypaymentis$351.Theborroweringeneralpaysback$4384amonthandhas$7873lefttobepaid.Theaverageratiooftheremainingbalancetototalloansis63%.Examiningtheborrowers'characteristics,itshowsthatthemeanincomeofaborrowerfromtheLendingClubis$5796withthedebtstoincomeratioof0.1381.Onaverage,aborrowerhas9.56opencreditlinesand22totalcreditlines,carries$14315averagerevolvingcreditbalanceandalmosthalf(51.6%)ofhisorhercreditlimit.Inthelastsixmonths,thereis1creditinquiryrequestedbyanaverageborrower.AverageFICOscorecategoryofatypicalborroweris3.48,whichcorrespondstoaFICOscorebetween680and750.Table4.Descriptivestatistics(May2007–June2012)Notes:CreditGradeisthegradeassignedbytheLendingClubbasedontheFICOranocreditratinginformationalongwithotherinformation.CreditGrade‘1’istheloancategoryof‘G’whichistheriskiestclassofloans.CreditGrade‘7’istheloancategoryof‘A’whichisthelowestriskborrowers.FICOranoisthecreditratingoftheborrowersratedbycreditcardcompanies.FICO6correspondstoborrowerswiththeFICOscoreabove780,FICO5correspondstoFICOscorebetween750–779,FICO4=714–749,FICO3=679–713,FICO2=660–678andFICO1=640–659,respectively.IV.ConclusionsCreditriskisanimportantconcernfortheP2Ploans.ThisstudyemploysthedatafromtheLendingClubtoevaluatethecreditriskoftheP2Ponlineloans.Wefindthatcreditscore,debt-to-incomeratio,FICOscoreandrevolvinglineutilizationplayanimportantroleindeterminingloandefault.ThecreditcategorizationusedbytheLendingClubsuccessfullypredictsthedefaultprobabilitywithoneexceptionofnextlowestcreditgrade‘F'.Ingeneral,highercreditgradeloanisassociatedwithlowerdefaultrisk.Themortalityriskalsoincreaseswiththematurityoftheloans.Loanswithlowercreditgradeandlongerdurationareassociatedwithhighmortalityrate.TheCoxProportionalHazardTestresultsshowthatasthecreditriskoftheborrowersincreases,sodoesthelikelihoodofloanbeingdefault.However,thehigherinterestratecurrentlychargedfortheriskierborrowerisnotsignificantenoughtojustifythehigherdefaultprobability.Thissuggeststhatthelenderswouldbebetterofftolendonlytothesafestborrowersinthehighestgradecategoryof7orGradeA.Increasingspreadsonriskierborrowermayleadtoamoresevereadverseselectionresultinginhigherdefaultrisk.TheLendingClublendersshouldeitherextendcreditsonlytothehighestgradeborrowerortrytofindmorecreativewaystolowerthedefaultrateamongcurrentborrowers.WhencomparingwiththeUSnationalconsumers,borrowerswithrelativelyhigherincomeandpotentiallyhigherFICOscoresdonotparticipateintheP2Pmarket.Creatingincentivestoattractthesetypesofborrowerswouldhaveasignificantpotentialtodecreasethedefaultriskinthismarket.中文譯文：點對點（P2P）網(wǎng)絡(luò)借貸的信用風(fēng)險與貸款績效評估摘要近年來點對點（P2P）網(wǎng)絡(luò)借貸開始興起。這種小微貸款市場可以為借款人和貸款人提供一定的收益。本文利用受歡迎的P2P網(wǎng)絡(luò)社交借貸平臺之一的借貸俱樂部的數(shù)據(jù)，探討了P2P貸款的特征，評估了其信用風(fēng)險和貸款績效。我們發(fā)現(xiàn)，信用等級、負債收入比、FICO評分和循環(huán)貸款利用率在貸款違約中起著重要的作用。信用等級較低、期限較長的貸款往往與高死亡率聯(lián)系在一起。這一結(jié)果與Cox比例風(fēng)險模型測試的相一致，這表明貸款違約的風(fēng)險率或可能性隨著借款人的信用風(fēng)險而增加。最后，我們發(fā)現(xiàn)，對高風(fēng)險借款人收取較高利率，并不能夠降低貸款的高違約率。借貸俱樂部需要找到吸引高FICO評分和高收入借款人的方法，以維持其業(yè)務(wù)。關(guān)鍵詞:P2P網(wǎng)絡(luò)借貸；信用等級；FICO評分；違約風(fēng)險1.引言隨著Web2.0時代的到來，創(chuàng)建方便快捷、協(xié)作性強的在線市場和虛擬社區(qū)已經(jīng)不是一件難事。新興的Web2.0應(yīng)用程序之一是點對點(P2P)網(wǎng)絡(luò)借貸市場，在那里貸款人和借款人幾乎可以完成貸款交易。將借款人引薦給貸款人，這種市場提供了平臺服務(wù)，可以為借款人和貸款人提供一些優(yōu)勢。借款人可以直接從貸款人那里獲得小額貸款，并且其支付的利率比商業(yè)貸款要低。另一方面，與任何其他類型的貸款如公司債券、銀行存款或存單相比，貸款人可以賺取更高的回報率。借款人與貸款人之間的信息不對稱，是P2P網(wǎng)絡(luò)借貸的問題之一。也就是說，貸款人不了解借款人的信譽，同樣借款人也不了解貸款人的信譽。這種信息不對稱可能導(dǎo)致逆向選擇（阿克洛夫，1970）和道德風(fēng)險（斯蒂格里茲和溫斯，1981）。從理論上講，這些問題可以通過定期監(jiān)測來得到緩解，但這種做法在網(wǎng)絡(luò)環(huán)境下很難實施，因為借款人和貸款人沒有實際接觸。促進和提高貸款人對借款人的信任也可以實施，以減輕逆向選擇和道德風(fēng)險問題。在傳統(tǒng)的銀行貸款市場中，銀行可以使用抵押品、認證賬戶、定期報告，甚至出席董事會來增強對借款人的信任。然而，這樣的機制將產(chǎn)生巨大的交易成本，因此難以在網(wǎng)絡(luò)環(huán)境中實現(xiàn)。為了減少由信息不對稱所引起的貸款風(fēng)險，目前的P2P網(wǎng)絡(luò)借貸平臺有以下安排。首先，借貸俱樂部根據(jù)FICO評分篩選出潛在的高風(fēng)險借款人，能夠參與平臺借貸的最低FICO評分為640。第二，這個平臺產(chǎn)生的貸款規(guī)模很小，借貸俱樂部的貸款不到35000美元。因此，這些貸款基本上是小額貸款，在違約的情況下造成的損失相對較小。第三，平臺建設(shè)者提供配對體系，可以用來生成投資組合建議，并盡可能減少貸款風(fēng)險。第四，如果借款人沒有付款，則平臺建設(shè)者將向信貸機構(gòu)報告情況，并聘請收款機構(gòu)代表貸款人收取資金。雖然P2P網(wǎng)絡(luò)借貸中有一些有助于降低風(fēng)險的強化結(jié)構(gòu)，但與傳統(tǒng)貸款相比，這種形式的貸款在本質(zhì)上與更大的風(fēng)險聯(lián)系在一起。本文的目的是評估美國最大的P2P網(wǎng)絡(luò)借貸平臺之一的借貸俱樂部的借款人的信用風(fēng)險，這有助于貸款人根據(jù)借款人的等級對貸款的風(fēng)險和回報率做出更明智的決策。本文將討論兩個相關(guān)的研究問題：（1）有助于確定違約風(fēng)險的借款人的特征有哪些？（2）高風(fēng)險借款人的回報率是否高到能夠彌補增量風(fēng)險？如果貸款人知道借款人的哪些特征影響到違約風(fēng)險，那么貸款人可以更有效地分配他們的投資。每個借款人按照信用等級進行分類，借貸俱樂部分配相應(yīng)的借款利率。為了進行有效的分配，貸款人應(yīng)該知道高風(fēng)險借款人的高利率是否能夠補償貸款人潛在損失的更高概率。我們的研究結(jié)果表明，F(xiàn)ICO評分高、信用等級高、循環(huán)貸款利用率低、負債收入比低的借款人，往往違約風(fēng)險偏低。這一發(fā)現(xiàn)與杜阿爾特等人（2012年）的研究相一致，他們指出具有可信賴特征的借款人信用評分較好，而違約概率低。這一結(jié)果還表明，除弗里德曼和吉恩（2014）、林等人（2013）指出的貸款申請人的社會關(guān)系和朋友關(guān)系外，上述四個因素對解釋違約風(fēng)險也很重要。與美國國家借款人相比，結(jié)果顯示，貸款俱樂部應(yīng)繼續(xù)篩選FICO評分低的借款人，吸引FICO評分高的借款人，從而大幅度降低違約風(fēng)險。在將風(fēng)險與回報相關(guān)聯(lián)時，本研究表明，對高風(fēng)險的借款人收取高的利率并不能對高違約率作出解釋。我們在這里的發(fā)現(xiàn)與別爾科維奇（2011）的研究相一致，他指出高質(zhì)量的貸款提供超額回報。2.文獻綜述隨著P2P網(wǎng)絡(luò)借貸的日益普及，出現(xiàn)了三大研究方向。第一個研究方向是分析P2P網(wǎng)絡(luò)借貸出現(xiàn)的原因。第二個研究方向集中于確定籌資成功和違約風(fēng)險的因素。第三個研究方向是調(diào)查在一定水平的風(fēng)險下P2P網(wǎng)絡(luò)貸款的表現(xiàn)。同儕團體貸款在當(dāng)?shù)厣鐓^(qū)涌現(xiàn)，并吸引了這一領(lǐng)域的研究。康林（1999）開發(fā)了一個模型來解釋美國和加拿大存在同儕團體小額貸款項目。他發(fā)現(xiàn)，同儕團體可以將固定成本強加給企業(yè)家，同時最大限度地減少項目的管理成本。阿什達和亞沙（2008）研究了Web2.0技術(shù)是如何集成的，以支持高級社會互動和交往，降低P2P網(wǎng)絡(luò)借貸成本。休姆和萊特（2006）研究了在英國的P2P網(wǎng)絡(luò)借貸平臺Zopa的案例。他們認為，P2P網(wǎng)絡(luò)借貸的出現(xiàn)是新信息時代的金融業(yè)對社會趨勢的直接反應(yīng)，以及對新形式關(guān)系的需求?，F(xiàn)有的文獻中，有的確定了籌資成功和違約風(fēng)險的決定因素。利用加拿大小額信貸數(shù)據(jù)，戈麥斯和桑托爾（2003）發(fā)現(xiàn)，團體貸款的違約率比傳統(tǒng)的個人貸款要低。伊耶等人（2009）的研究顯示，通過分析借款人的硬數(shù)據(jù)和軟數(shù)據(jù)，貸款人可以評估出三分之一的信貸風(fēng)險。林等人（2013）分析了社會關(guān)系在評估信用風(fēng)險中的作用，發(fā)現(xiàn)強大的社交網(wǎng)絡(luò)關(guān)系是決定借款成功和降低違約風(fēng)險的重要因素。林等人（2013年）進一步指出，申請人的朋友關(guān)系可能會增加籌資成功的可能性，降低貸款利率，而且在Prosper上，這些借款人的事后違約率較低。弗里德曼和吉恩（2014）也研究了社會關(guān)系在確定貸款資金中的重要性。結(jié)果表明，具有社會關(guān)系的借款人更有可能獲得貸款資金，且利率較低。然而，他們也發(fā)現(xiàn)這一跡象，即借款人參與社交網(wǎng)絡(luò)，對貸款人也有風(fēng)險。其他一些研究探討了借款人的某些特征和個人信息是否決定了籌資成功和違約風(fēng)險。赫斯坦恩等人（2008）表示，借款人的財務(wù)實力、上市和宣傳工作以及個人特征，影響到籌資成功的可能性。杜阿爾特等人（2012）的研究進一步認為，看起來更值得信賴的借款人信用評分較高，往往籌資成功可能性更高，違約率更低。拉里莫爾等人（2011）表明，借款人使用擴展敘述、具體描述和定量單詞，對籌資成功有積極的影響。然而，個性化的個人資料或貸款理由對籌資成功有負面影響。邱等人（2012）進一步揭示，除了個人信息和社會資本外，借款人設(shè)定的其他變量，包括貸款金額、可接受的最高利率和貸款期限，都顯著影響了籌資的成敗。加拉克等人（2011）進一步表明，與團體借款人相比，貸款人更傾向于個人借款人，以及社交上接近自己的借款人。他們還發(fā)現(xiàn)貸款人傾向于在性別、職業(yè)和名字的首字母上更像自己的借款人。更有趣的是，岡薩雷斯和洛雷羅（2014）也有類似的發(fā)現(xiàn)：（1）當(dāng)把外表年齡視為其能力的體現(xiàn)時，吸引力對貸款成功沒有影響；（2）當(dāng)貸款人和借款人的性別相同時，吸引力可能導(dǎo)致貸款失?。础凹t顏禍水”效應(yīng)）；（3）貸款成功對貸款人和借款人的相對年齡與吸引力很敏感。赫斯坦恩等人（2011）發(fā)現(xiàn)，貸款拍賣中的羊群效應(yīng)與其后續(xù)表現(xiàn)呈正相關(guān)，即借款人是否按時還錢。3.數(shù)據(jù)本部分首先對貸款申請人的資料進行了描述，然后是根據(jù)貸款目的、信用等級及貸款狀況的貸款分配，最后是貸款申請人的詳細描述統(tǒng)計。本研究利用了借貸俱樂部上的自2007年5月至2012年6月的61451筆貸款申請，這些數(shù)據(jù)是從網(wǎng)站獲取的。在研究期間，借貸俱樂部貸給借款人約7.13億美元。為了研究