1、外文翻译EARLYWARNINGMODELWITHSTATISTICALANALYSISPROCEDURESINTURKISHINSURANCECOMPANIESMATERIALSOURCEAFRICANJOURNALOFBUSINESSMANAGEMENTAUTHORISSEVEROGLUGULSUN,GUCENMEUMITINSURANCECOMPANIESARESOCIALENTITIESORGANIZEDTOOFFLOADTHEFINANCIALRISKSENCOUNTEREDBYINDIVIDUALSORFIRMSINDIVIDUALSPAYAPREMIUMANDGIVENTHEOC
2、CURRENCEOFSPECIFICEVENTS,RECEIVEREMUNERATIONFORLOSSESINCURREDINADDITION,INSURANCECOMPANIESCONTRIBUTESUBSTANTIALLYTOTHENATIONALECONOMYBYUSINGCAPITALGATHEREDTHROUGHPREMIUMSFORINVESTMENTTHEHIGHPOTENTIALFORINSURABILITYANDTHERAPIDIMPROVEMENTOFTHEINSURANCEANDPENSIONSECTORHAVEINCREASEDTHEVISIBILITYOFTHETUR
3、KISHINSURANCEMARKETTOFOREIGNINVESTORSSINCETHEINSURABILITYRATEHASREACHEDASATURATIONPOINTINOTHERCOUNTRIES,FOREIGNINVESTORSBEGANINCREASINGTHEIRINVESTMENTSINDEVELOPINGCOUNTRIESBEGINNINGIN2006,WITHTHISINVESTMENTCONTINUINGTOGROWIN2007TOFURTHERINCREASETHEEFFICIENCYANDEFFICACYOFTHENATIONALINSURANCESECTOR,IT
4、ISVERYIMPORTANTTOEXAMINEANDANALYSETHESECTORDATACOMPANYPROFITABILITYTRENDSATTRACTNOTONLYTHEATTENTIONOFTHESHAREHOLDERS,BUTALSOTHEATTENTIONOFINVESTORS,CREDITORSANDAUDITINGFIRMSBUSINESSFAILUREISANUNFAVOURABLEANDCOSTLYEVENTFORSOCIETYINANENVIRONMENTWITHLIMITEDRESOURCES,BUSINESSFAILURECANMEANTHEMISALLOCATI
5、ONOFRESOURCESANDRESULTINENORMOUSECONOMICANDSOCIALCOSTSWHENAFIRMGOESBANKRUPT,SHAREHOLDERSAREUSUALLYTHEBIGGESTLOSERSOPERATIONSSUCHASMEASURING,EVALUATINGANDRATINGSHOULDBEPERFORMEDINORDERTOBETTERASSESSANINSURANCECOMPANYSFINANCIALPOTENTIALATRUSTWORTHYFLOWOFINCOMETOTHEMARKET,WHICHMIGHTALLOWFORINCREASESIND
6、ECISIONSPEEDANDQUALITY,MAYBEPOSSIBLEWITHTHEIMPLEMENTATIONOFRATINGAPPLICATIONSINADDITION,THEDEVELOPMENTOFINCREASINGLYSOPHISTICATEDCOMPUTERTECHNOLOGIESHASENABLEDTHEUSEOFSTATISTICALMETHODSINSEVERALRESEARCHFIELDSINASIMILARSTUDY,ISSEVEROGLU2005,AIMEDTODEVELOPAMULTIVARIABLEMODELTOPREDICTTHESTARTINGPERIODO
7、FFINANCIALFAILURESORDIFFICULTIESFORENTERPRISESBYAPPLYINGMULTIDIMENSIONALSTATISTICALANALYSESTOTHETURKISHINSURANCESECTORANDDEFININGFACTORSCAUSINGFINANCIALFAILUREOURDATASETHASBEENCOMPILEDFROMTHEFINANCIALSTATEMENTSOF45TURKISHINSURANCECOMPANIESWEDEVELOPEDOURMODELUSINGDATAABOUT45DEPENDENTAND17INDEPENDENTV
8、ARIABLESANDEMPLOYEDMULTIPLEREGRESSIONANDMULTIPLEDISCRIMINATORTECHNIQUESTHESAMEMODEL,USINGIDENTICALVARIABLES,WASOBTAINEDASTHERESULTOFTHETWOANALYSESTHEEXACTPREDICTIVEPOWEROFTHEFIVEFINANCIALRATIOSOBTAINEDTHROUGHTHEMULTIPLEREGRESSIONMODELWAS93,89AND87FORTHEFIRST,SECONDANDTHETHIRDYEARS,RESPECTIVELYTHEMUL
9、TIPLEDISCRIMINATORMETHODWASTHENRUNUSINGTHESAMEDATAANDTHERESULTSWERECOMPAREDWITHTHOSEOFTHEMULTIPLEREGRESSIONMETHODANDTHEMODELWASTHENRUNUSINGTHESTEPWISEMETHOD,THESAMEFINANCIALRATIOSOFTHEREGRESSIONMODELWEREALSOOBTAINEDINTHEDISCRIMINATORMODELFINANCIALRATIOS,BEINGTHESAMEINBOTHMODELS,DEMONSTRATEDTHATTHEYH
10、ADANIMPORTANTDIFFERENTIATIONPOWERINCLASSIFYINGENTERPRISESTHEMULTIPLEDISCRIMINATORMODELSPOWERTOPREDICTENTERPRISESFINANCIALSUCCESSESANDFAILURESINTHEPRECEDING1,2ANDTHE3YEARSWAS100,94AND81,RESPECTIVELYTHEVALIDITYOFTHEMODELWEHAVEDEVELOPEDCOULDBETESTEDDURINGTHESTUDYBYINTEGRATINGDATAFROM2003AND2004FOURTEEN
11、COMPANIESEMERGEDWITHEARLYWARNINGSIGNALSINBOTHMODELSBASEDONDATAFROM2003THEANALYSISCARRIEDOUTWITH2004DATASHOWEDSIXCOMPANIESEMERGINGWITHWARNINGSIGNALSINTHESECONDYEARFOURCOMPANIESOUTOFTHESESIXLEFTTHESECTORACCORDINGTOINFORMATIONISSUEDBYTHEUNDERSECRETARIATOFTREASURYTHESTUDYDEMONSTRATEDTHATTHERESULTSOBTAIN
12、EDINTHEDEVELOPEDMODELSAREAPPLICABLETOTHEPRESENTTIMEINTHISSTUDY,THEPERFORMANCEOBTAINEDWITHTHEMULTIPLEDISCRIMINATORANDTHEMULTIPLEREGRESSIONMODELSARECOMPAREDTHEACCURACYOFTHEFORESIGHTMODELSMADEEARLIERBYLOGITANALYSIS,ASTATISTICALMETHOD,WILLBETESTEDONCEMOREITWILLBEDETERMINEDIFASIMILARBENEFITWOULDBEOBTAINE
13、DUSINGTHEMETHODSDESCRIBEDWHENAPPLIEDTO20052006DATAINADDITION,THEEMPLOYABILITYOFTHEFORESIGHTMODELWILLALSOBEASSESSEDBYADDINGTHEDATAFROMTHEYEARS2005AND2006THEFIRSTANDTHEMOSTCITEDRESEARCHINTHELITERATUREONTHISTOPICISTHE1966STUDYBYBEAVER1966BEAVERMEASUREDTHEPOWEROFFINANCIALRATIOSANDCAMETOTHECONCLUSIONTHAT
14、THEYMIGHTBEUSEDINPREDICTINGTHEENTERPRISESFAILURETHEPURPOSEOFHISEARLIERSTUDYWASTODETERMINEHOWWELLFINANCIALRATIOSCOULDPREDICTFAILURERELATIVETORANDOMPREDICTIONTHEFINDINGSOFTHESTUDYDEMONSTRATEDTHATTHEFINANCIALRATIOSPREDICTEDTHEFAILURESTATUSOFFIRMSTOAMUCHGREATEREXTENTTHANWOULDBEEXPECTEDFROMRANDOMPREDICTI
15、ONFOREXAMPLE,ONEYEARBEFOREFAILURE,THECASHFLOWTOTOTALDEBTRATIOMISCLASSIFIEDONLY13OFTHESAMPLEFIRMSFIVEYEARSBEFOREFAILURE,THESAMERATIOMISCLASSIFIEDONLY22SINCETHEREWEREANAPPROXIMATELYEQUALNUMBEROFFAILEDANDNONFAILEDFIRMSINTHESAMPLE,THEEXPECTEDERRORFROMRANDOMPREDICTIONWASAPPROXIMATELY50THEREWASTHUS,ANEXTR
16、EMELYSMALLPROBABILITYTHATRANDOMPREDICTIONCOULDHAVEDONEASWELLASTHERATIOBEAVER,1968THISANALYSISSUGGESTEDTHATFINANCIALRATIOSCOULDBEUSEFULINTHEPREDICTIONOFFAILUREUPTOFIVEYEARSPRIORTOTHEEVENTEDWARDIALTMAN,APRIMARYCONTRIBUTORTOTHEFINANCIALFAILURELITERATURE,CHOSE33SUCCESSFULANDUNSUCCESSFULENTERPRISESBYRAND
17、OMSAMPLINGDURINGTHEPERIODOF1946THROUGH1965FINANCIALDATACOVERINGAPERIODOFFIVEYEARSAND22FINANCIALRATIOSWEREANALYSEDFIVEFINANCIALRATIOSBESTMEASURINGTHEFINANCIALPOWERWEREOBTAINEDASARESULTOFLINEARDIFFERENTIATIONANALYSISADISCRIMINANTFUNCTIONWITHALINEARCOMBINATIONOFFIVERATIOSWASDERIVEDFROMDATAONEYEARPRIORT
18、OBANKRUPTCYCALLEDTHEZMODEL,THISMODELWASDERIVEDUSINGTHEPREVIOUSLYMENTIONEDFIVERATIOSANDTHEDIFFERENTIATIONSCOREITWASCALCULATEDASFOLLOWSZ012X1014X2033X3006X40999X5ASHIGHAS94OFUNSUCCESSFULENTERPRISESANDSUCCESSFULENTERPRISESWITH97ACCURACYRATIOSWERECORRECTLYCLASSIFIEDFORTHEFIRSTYEARPRECEDINGTHEFAILUREUNSU
19、CCESSFULENTERPRISESWERECLASSIFIEDWITH72EXACTITUDEFORTHESECONDYEARBEFORETHEFAILURE,48FORTHETHIRDYEAR,29FORTHEFOURTHAND36FORTHEFIFTHYEARCONSECUTIVELYHOWEVER,THEACCURACYOFCLASSIFICATIONDECLINEDRAPIDLYASTHENUMBEROFYEARSPRIORTOFAILUREINCREASEDTHEMODELHASBEENFOUNDTOBEPREDICTIVE,ALBEITWITHDIMINISHINGEXACTI
20、TUDE,ASONEMOVESEARLIERINTIMETHUS,ADDITIONALREFINEMENTOFALTMANSMODELISREQUIREDTOIMPROVETHEACCURACYOFEARLIERYEARPREDICTIONFINALLY,ALTMANEXAMINEDTHEPREDICTIVEVALIDITYOFTHEMODELONANEWSAMPLE,CONSISTINGOF25BANKRUPTFIRMSAND66NONFAILEDFIRMSINTHESAMEPERIODASTHEINITIALSAMPLETHERESULTINDICATED96OFBANKRUPTFIRMS
21、AND79OFNONBANKRUPTFIRMSWERECORRECTLYCLASSIFIEDONEYEARPRIORTOBANKRUPTCYALTMANWASTHEFIRSTTOUSEMDATODEVELOPAFAILUREPREDICTIONMODELALTMANOBTAINEDTHEZETAMODELBYDEVELOPINGHISFIRSTZMODELIN1993ALTMAN,1993SINCETHATTIME,THEMDAMETHODHASBEENEXTENSIVELYUSEDINBUSINESSFAILUREPREDICTIONSTUDIESHECOMPARED53ENTERPRISE
22、STHATWENTBANKRUPTAND58ENTERPRISESTHATDIDNOTINSTEADOFCLASSIFYINGENTERPRISESASSUCCESSFULORUNSUCCESSFULINTHEZETAMODELFROMTHIS,HEOBTAINEDSEVENSIGNIFICANTFINANCIALRATIOSEXACTITUDERATIOSOF95INTHEFIRSTYEARPRECEDINGTHEFAILURE,87INTHESECOND,75INTHETHIRD,68INTHEFOURTHAND64INTHEFIFTHYEARWEREDEMONSTRATEDALTMANA
23、LSOPROVEDINHISSTUDY,USINGQUADRATICDISCRIMINATORANALYSISANDLINEARDISCRIMINATORANALYSIS,THATTHEREWASNOTAGREATDIFFERENCEINEXACTITUDEINCLASSIFYINGGROUPSDEAKEN,BYANALYZING32FIRMS,MATCHEDEACHFAILEDFIRMWITHANONFAILEDFIRMACCORDINGTOINDUSTRYANDSIZEINTHEPERIOD19641970DEAKINSMODELCORRECTLYCLASSIFIED97,95AND95F
24、ORTHEFIRSTTHREEYEARSPRIORTOFAILURE,RESPECTIVELYDEAKEN,1972INASAMPLEOF63BANKRUPTAND80NONBANKRUPTENTERPRISESDURINGTHEPERIOD19661971,DEAKEN1977EMPLOYEDMULTIPLEDISCRIMINATORANALYSISUSINGBOTHLINEARANDQUADRATICFORMSTOCLASSIFYENTERPRISESWITHA94EXACTITUDERATIOINTHELINEARAND84INTHEQUADRATICMODEL,RESPECTIVELY
25、BLUM1974,CONSTRUCTEDAFAILINGCOMPANYMODELWITHDISCRIMINANTANALYSISTOASSESSTHEPROBABILITYOFBUSINESSFAILUREDURINGTHEPERIODFROM1954TO1968BLUMSSTUDYINCLUDES115SUCCESSFULNONFAILEDAND115UNSUCCESSFULFAILEDENTERPRISESTHEFIRMSWEREMATCHEDONTHEBASISOFINDUSTRY,SALES,NUMBEROFEMPLOYEESANDFISCALYEARBLUMCALCULATEDACU
26、TOFFPOINTTODISCRIMINATEBETWEENFAILEDFIRMSANDNONFAILEDFIRMSHISMODELHADTHEOVERALLCLASSIFICATIONACCURACYOF94FORTHEFIRSTYEARPRECEDINGTHEFAILURE,80FORTHESECONDYEARAND70FORTHETHIRDYEARLOGISTICREGRESSIONANALYSISISONEALTERNATIVEDEVELOPEDINRESPONSETOTHELIMITSOFMDAMDAISMODESTLYSUCCESSFULINCLASSIFYINGFAILUREAN
27、DNONFAILURE,BUTTHEREAREVALIDITYPROBLEMSWITHTHEAPPLICATIONOFTHISMETHODINFAILUREPREDICTIONSTUDIESONEOFTHESEPROBLEMSISTHATEACHINDIVIDUALVARIABLEMUSTBENORMALLYDISTRIBUTEDUNDERMDAEISENBEIS1977,IDENTIFIEDANOTHERPROBLEMTHECOEFFICIENTSOFINDIVIDUALVARIABLESINTHEDISCRIMINANTFUNCTIONARENOTMEANINGFULANDITISIMPO
28、SSIBLETODETERMINETHESIGNIFICANCEOFEXPLANATORYVARIABLESINTHEMODELLOGITANALYSISHASSEVERALADVANTAGESOVERMDATHEFIRSTOFTHESEISTHATTHEMETHODISMOREROBUSTANDRELIABLESINCETHENORMALITYASSUMPTIONFORRATIOVARIABLESISNOTREQUIREDINADDITION,INSTEADOFACOMPOSITESCOREFORTHEDEPENDENTVARIABLEINMDA,THEDEPENDENTVARIABLEIN
29、ALOGITMODELFALLSWITHINA0,1DISTRIBUTIONUSINGAPROBABILITYDISTRIBUTIONTOEXPLAINPOTENTIALFAILEDFIRMSISGENERALLYTHOUGHTTOBEASUPERIORMETHODOLOGYINFAILUREPREDICTIONRESEARCHMOREOVER,THECOEFFICIENTOFANINDIVIDUALVARIABLEINALOGITFUNCTIONISINTERPRETABLEANDTHESIGNIFICANCEOFAVARIABLECANBETESTEDSTATISTICALLYASARES
30、ULT,EACHFINANCIALRATIOINAFAILUREPREDICTIONMODELISEXAMINEDSOTHATTHEPREDICTIVEABILITYOFTHEMODELCANBEIMPROVEDOHLSONEE1980WASTHEFIRSTTOUSELOGITANALYSISTOASSESSTHEPROBABILITYOFCORPORATEFAILUREHEOBTAINED105PUBLICLYTRADEDINDUSTRIALFIRMSTHATWENTBANKRUPTDURINGTHEPERIOD1970TO1976AND2,058NONBANKRUPTFIRMSDURING
31、THESAMEPERIODOHLSONDEVELOPEDTHREELOGITMODELSFOREACHOFTHETHREEYEARSPRIORTOBANKRUPTCYTHERESULTSINDICATEDTHATTHATTHECOEFFICIENTSOFSIXVARIABLESREACHEDLEVELSOFSTATISTICALSIGNIFICANCEINALLTHREEMODELSTHELOGITFUNCTIONSHOWED84FORTHEMODELONEYEARBEFOREBANKRUPTCYANDWASSTATISTICALLYSIGNIFICANTOHLSONSELECTEDACUTO
32、FFPOINTOF0038OHLSONSMODELMISCLASSIFIED174OFNONFAILEDFIRMSASFAILEDFIRMS,AND124OFTHEBANKRUPTFIRMSASNONFAILEDONEYEARBEFOREBANKRUPTCYPLATTANDPLATTS1990SAMPLEINTHESTUDYCONSISTEDOF57FAILEDFIRMSAND57NONFAILEDFIRMSTHEFIRMSWEREMATCHEDBYASSETSIZEANDINDUSTRYMEMBERSHIPINTHEPERIODOF1972TO1986THERESULTSINDICATEDT
33、HATTHELOGITMODELACCURATELYCLASSIFIED90OFTHEFIRMSAYEARBEFOREBANKRUPTCY,INCLUDING93OFFAILEDFIRMSAND86OFNONFAILEDFIRMSCONCLUSIONTHEEARLYWARNINGSYSTEMISANIMPORTANTSTUDYLAYINGTHEFOUNDATIONFORAMORECOMPETITIVETECHNICALANDFINANCIALSTRUCTUREANDOPERATIONALIZINGTHESECTORSAUTOCONTROLMECHANISMTHEEARLYWARNINGSYST
34、EMSHOULDBEUSEDTOIDENTIFYINSURANCECOMPANIESTHATAREFAILINGANDNOTSATISFYINGTHEIROBLIGATIONSINORDERTOSETUPANINSURANCECONSCIOUSNESSTHEEMPIRICALRESULTSOFTHELOGITSTUDYFURTHERINDICATETHATTHEMODELPROVIDEIMPRESSIVEPREDICTIONACCURACYANDOUTPERFORMSOTHERPOPULARMODELSBASEDONFINANCIALRATIOSTHESTUDYREVEALEDTHATTHER
35、ESULTSOBTAINEDTHROUGHDEVELOPEDLOGITMODELAREALSOPOWERFULENOUGHTODESCRIBETHECURRENTSITUATIONASWELL译文土耳其保险公司统计分析预警模型资料来源AFRICANJOURNALOFBUSINESSMANAGEMENT作者ISSEVEROGLUGULSUN,GUCENMEUMIT保险公司是为了分担个人或公司所遭遇的财务风险的一个社会组织。个人缴纳保险费是为了将来由于特定事件的发生而造成的损失是能获得补偿金。此外，保险公司通过保险费投资来聚集资本也大大地促进了国民经济的发展。可保性的巨大潜力和养老保险的不断完善，
36、增加了土耳其保险市场在外国投资者中的知名度。由于一些国家的保率已经达到了饱和点，2006年外国投资者开始增加其在发展中国家的投资，这部分投资在2007年也得到了持续的增长。这是一项检查和分析土耳其进一步提高保险业的效率和效力一项总要数据。公司盈利趋势吸引的不仅是股东的关注，也有投资者、债权人以及审计公司的注意。经营失败对社会来说是一件不利的，要付出代价的。在一个资源有限的环境里，经营失败可能意味着资源配置失衡，产生巨大的经济和社会成本。当一个公司破产时，股东通常是最大的输家。为了更好地评估一家保险公司的财务潜力，应该进行测量、评估和评级等操作。履行评级应用就可以有一个可靠的市场收入流动，这能让
37、决策在速度和质量上都有所增加。此外，发展日益精密的计算机技术也为统计方法在多个领域的应用提供了可能性。在一个类似的研究中，ISSEVEROGLU2005，运用多维统计方法来分析土耳其保险业以及财务危机的决定性因素，旨在建立一个多变量模型，可以在早起预测企业财务失败或财务危机出现的。我们的研究数据来自于45家土耳其保险公司的财务报表。我们用45个相关和17个不相关的数据作为自变量，通过多元回归和多元鉴别技术来构建模型。相同的模型，使用相同的变量，得到两种分析的结果。通过多元回归模型得到的5个财务比率预测能力分别为第一年93、第二年89、第三年87。多元鉴别方法运用相同的数据并将其结果与多元回归方
38、法进行比较，然后用逐步代入方法，代入回归模型相同的财务比率，得到了鉴别模型。结果显示在两种模型中，财务比率同样有判别企业状况的能力。多元鉴别模型预测企业财务成功和失败的准确率在前面的1，2和3年分别为100、94、81。我们改良的模型可有效地测试在研究过程中整合的2003年和2004年的数据。从2003年的数据看，14家公司在两个模型的基础上都出现了早期预警信号。2004年的数据分析显示在第二年六家公司出现了早期预警信号。根据财政部门印发的资料，这六家公司里有四家破产。这项研究表明经改良的模型得到的结果适用于现在。BEAVER（1966）是第一个研究这个主题的，也是被关于这个主题的研究文献引用
39、最多的。BEAVER选定的财务比率和得出的结论可以预测企业的失败。他的早期研究目的是确定与随机预测相比，财务比率预测企业危机的效果如何。这项研究的结果表明，财务比率预测企业危机的效果比随机预测更好。例如，在危机出现前一年现金流量与总负责比率误判的样本公司只有13。在危机出现前五年，这个比例只有22。由于样本公司中危机企业与正常企业的数量接近，随机预测的误差约为50。因此随机预测的效果不太可能比得上比弗。这项分析表明，财务比率可以在危机出现之前长达5年的时间来预测就非常有用。ALTMAN，一个财务危机研究主要贡献者，随机抽样了1946年到1965年33家成功和失败的企业。对这5年的22个财务比率
40、数据进行综合分析。五个判断力最好的财务比率得到一组线性判定的分析结果。判别函数的五个财务比率来自于破产前一年的数据。利用前面提到的五个比率和相对应的系数推导出了Z模型。计算方法如下Z012X1014X2033X3006X40999X5破产前一年失败企业判断准确率高达94，成功企业判断准确率为97。第二年失败企业预测准确度为72，第三年为48，第四年为29，第五年为36。判断准确度随着失败前年数的增加而迅速降低。该模型虽然预测结果准确性降低，但仍然能够提前预测到危机。因此，需要更多的改进ALTMAN的模型进而提高早些时候预测的准确性。最后ALTMAN用同一时期的25家破产企业和66家正常企业组成
41、一个新的样本，用来检测该模型的预测有效性。结果表明破产前一年96的破产企业和79的正常公司预测准确。ALTMAN是第一个使用MDA的构建危机预警模型的。1993年他第一次构建出Z模型。从那时起，MDA技术广泛用于商业失败预测研究。他比较了53家破产企业和58家灰色企业，获得了七个重要的财务比率。经证明其失败前一年准确率为95，第二年为87，第三年为75，第四年为68，第五年为64。ALTMAN也证明了他的研究中，用二元判别分析和线性判别分析，在预测精度上没有很大的差别。DEAKEN对工业企业19641970年间的32个危机企业和32个与其相匹配的正常企业进行研究。DEAKEN的模型前三年预测准
42、确率分别为97、95、95。DEAKEN分别采用多元判别分析法和线性判别分析法，对一组有19661971年的63家破产企业和80家非破产企业组成的样本进行研究，判定准确度分别为94、84。BLUM1974,构建了一个破产公司的判别分析模型来评估公司在1954年到1968年经营失败的概率。研究对象为115家非失败企业和115家产业基础，销售情况，雇员人数和会计年度相匹配的失败企业。该模型失败前一年预测精度为94、第二年为80、第三年为70。逻辑回归分析克服了MDA的限制，MDA可以成功地判定失败和非失败企业，但在危机预测应用研究过程中有效性存在问题。其中一个问题是，每个变量在MDA下必须成正态分
43、布。EISENBEIS（1977年），确定了另一个问题判别函数变量系数本身是没有意义的，这无法解释变量在模型中的意义。逻辑回归模型比多元线性模型有更多的优点。其中第一项就是改方法更加可靠，不需要假设变量服从正态分布。此外，逻辑回归模型属于（0，1）分布，而不是多元线性判定模型是变量的综合数值。使用概率分布来研究失败企业在危机预测上更有优势。逻辑回归模型中的任何一个函数变量系数都是有意义的，任何一个变量都可以被测试统计的。因此，危机预警模型中的每一个财务比率都可以进行审查的，使得该模型的预测能力得以提高。OHLSONEE（1980）是第一个使用逻辑回归分析来评估企业失败的概率的。他获得了1970
44、年至1976年期间内破产105家公开上市企业和2058家在同一期间内的非破产工业企业。他为破产前三年构建了三个逻辑回归模型。结果表明，这六个变量的系数在三个模型中都显著相关。该模型在破产前一年预测精度为84。OHLSONEE确定分界点为0038该模型破产前一年将174的非失败公司误判为失败的公司，124的失败公司的误判为非失败公司。PLATTANDPLATT1990的研究样本包括57个失败企业和57个非失败企业。这些公司是1972年到1986年期间资产规模、行业相配比的。结果表明，逻辑回归模型破产前一年预测准确率为90，其中破产企业准确率为93，非破产企业准确率为86。早期预警系统是建立在技术竞争力、财务结构以及内部控制的基础上的一项重要的研究。保险公司应该用预警系统来预测是否会出现危机进而不能履行义务，以便树立一个保险意识。逻辑回归模型研究结果进一步表明，该模型预测准确率能够让人信服并且优于其他常用的模型。该研究显示，通过构建逻辑回归模型得到的结果足够应用于当前局势。
