【文档说明】责任代理模式分析报告.pptx，共(26)页，134.177 KB，由精品优选上传

转载请保留链接：https://www.ichengzhen.cn/view-281429.html

以下为本文档部分文字说明：

Principal-agentModeling責任代理模式Dr.Chak-TongChau仇澤棠博士U.S.FulbrightProfessor中美交流富布萊特教授1我請您們考慮一些問題◼Asmallmedicalins

urancescenario一個醫療保健的問題Whenyouhaveasmallillness,doyounormallyseeyourdoctor?當你有小病的時候，你會不會自費看醫生？Whatabout,ifyourfirmpayforyou

rexpense?但是，如果是單位付錢呢，那又怎樣？Dr.Chak-TongChau2FulbrightGuestLectureMaterials我請您們考慮一些問題◼Acarmaintenancescenario一個汽車維修的問題Yourcar

isbeingrentedfor2months.Supposedly,itneedsoilingeverymonth.Howlikelyyouwillremembertodoso?你的汽車是租來用兩個月的，它需要每月潤滑上油一次。你會不會依時地去上油？How

aboutifthisisyourowncar?如果這是你自己的汽車，你又會不會去做？Dr.Chak-TongChau3FulbrightGuestLectureMaterials我請您們考慮一些問題◼Amedicalinsuranceproblem自費醫療保險的問題Whenwe

purchasemedicalinsurance,theinsurancecompanyusuallyrequiresthatyoudiscloseyourmedicalhistory.Pre-conditionsareusuallyexcludedfromthecoverage.購買保險的

時候，它們通常要求你列出你的病歷。但是如果你有大病的話，很可能保險公司不愿意受保。Dr.Chak-TongChau4FulbrightGuestLectureMaterials我請您們考慮一些問題Ifyoudoinfacthavesomemajormedica

lproblemsthatrequireexpensivetreatments,wouldyoudisclosetheseproblems?如果你真的有大病,你會不會真實地上報？Whatdoallthesetellusaboutcertainhumanbehavior?這些問題表

明了一些什么的人性行為？Dr.Chak-TongChau5FulbrightGuestLectureMaterialsAgencyProblemsandBehavior代理人的行為与問題•Amoralhazardproblem(道德危机問題)whenan

individualhasanincentivetodeviatefromthecontractandtakeself-interestedactionsbecausetheotherpartyhasinsu

fficientinformationtoknowifthecontractwashonored.醫療保健雖然我知道我与雇主的契約明确列出我不要浪費公司的資源。但是用公司的好過用我的嘛！而且公司又不會知道我未能遵守契約

。Dr.Chak-TongChau6FulbrightGuestLectureMaterialsAgencyProblemsandBehavior代理人的行為与問題•Ahorizonproblem水平界線問題Ifoneparty’sriskorcompensationisno

tthesameastheotherparty’s,theonewithashorterhorizonwilltendtosecretlymaximizetheshort-termbenefits,attheexpenseoftheotherlonger-termparty

.汽車維修我明白汽車不維修壽命不會長。但是，兩個月以后這車子變成怎么樣与我無關了吧。Dr.Chak-TongChau7FulbrightGuestLectureMaterialsAgencyProblemsandBehavior代理人的行為与問題•Anadverseselectionpro

blem逆向選擇問題Thetendencyofindividualswithprivateinformationaboutsomethingthataffectsapotentialtradingpa

rtner’sbenefitstomakeoffersthataredetrimentaltothetradingpartner.自費醫療保險:雖然我知道保險公司需要知道我的病歷從而決定保險費。但是誠實的代价是較高的費用。此外，我不說，誰知道。D

r.Chak-TongChau8FulbrightGuestLectureMaterials誰是代理人？什么是代理成本？◼Anagentissomeonewhohascertainspecialexpertisethatisde

siredbytheprincipaltouseforhis/herbenefits.Theagentisusuallyriskadverse,hasdecisionrightstomanage,butdoesnotown,theorgan

ization’sassets.代理人(agent)是任何人在公司有決策權力，但是并非產權的最終所有者。代理人通常有較佳的專長，更好的資訊，和對風險抱保守的態度(riskadverse)。Dr.Chak-TongChau9FulbrightGuestLectureMaterials誰是

代理人？什么是代理成本？◼Therearethree(3)typesofagencycosts.代理成本有三類：◆設計限制性契約的成本(bondingcosts)◆建立監督制度的成本(monitoringcosts)◆剩餘的損耗(resid

ualloss)◼Notethatsomecostsareborntbytheprincipalbutsomeareborntbytheagent.注意的是，有時這些成本是由委托人(principal)負擔。不過有時這些成本是由代理人自己負擔的。Dr.Chak-

TongChau10FulbrightGuestLectureMaterialsAgencyCosts◼Bondingcosts–costsincurred,beforeenteringthecontract,toconvincetheprincipaltha

tsuchagencyrelationshipwillnotresultintheabove-mentionedagencyproblems.Examplesare:reputationbuilding,3rdpartyguarantor,etc.D

r.Chak-TongChau11FulbrightGuestLectureMaterialsAgencyCosts◼Monitoringcosts–costsincurred,afterenteri

ngthecontract,toensurethatsuchagencyproblemswillnotarise.Examplesincludeauditingandinspectioncosts.Dr.Chak-TongChau12Fulbright

GuestLectureMaterialsAgencyCosts◼Residualloss–lossunavoidablyarise,despitethebondingandmonitoringcosts,thecontrac

tstillcannotyieldtheutmostbenefits,because:◆theagencyproblemsdoarise,or◆duetothesuspicionoftheagencyproblems,the

principalrefusestopaytheagentcompensationsthatfullyreflecthis/herefforts.Dr.Chak-TongChau13FulbrightGuestLectureMaterialsExamplesofthePri

ncipal-agentModelEffortlevelProbabilitiesandpayoffsfor4differenteventsS1=0.3S2=0.3S3=0.2S4=0.2E1=6$55,000$55,000$55,000$40,000E2=5$55,000$55,

000$40,000$40,000E3=4$55,000$40,000$40,000$40,000Dr.Chak-TongChau14FulbrightGuestLectureMaterialsExamplesofthePr

incipal-agentModelAgent’sUtilityFunction:Xa½-e2100where:Xa=agent’scompensationse=theeffortlevelusedbytheagentQuestion1:Ifyouwere

theprincipalinenteringthecontract,whichlevelofeffort(e1,e2,ore3)wouldyoudemand?Question2:Ifyou,theprincipal,cancloselymonitorandobser

vetheagentatalltime,whataretheamountandconditionofpayment?And,whatistheexpectedpayofffortheprincipal?Dr.Chak-TongChau15FulbrightGuestL

ectureMaterialsNow,let’sassumethatyoucannotmonitorandobservetheagentdirectly.Whatwouldyou,astheagent,do?Now,canyouseetheagen

cyproblemshere?EffortlevelExpectedutilityoftheagentE1=618,496½-62=100E2=518,496½-52=111E3=418,496½-42

=112Isitlikelytohavethe“adverseselection”problem?Howaboutthe“moralhazard”problem?And,thehorizonproblem?Residualloss?Dr.Chak

-TongChau16FulbrightGuestLectureMaterialsWhatcanwesay,uptothispoint?◼Underconditionofunobservability(

incompleteinformation),fixedpaymentstoagents(i.e.workers,employees)mostlikelydonotwork.◼Whatarethenth

ealternatives?◼Wecangivetheprincipalafixedpaymentinstead.◼Or,wecancomeupwithan“incentivecompatible”conditionalcontra

ct.Dr.Chak-TongChau17FulbrightGuestLectureMaterialsFixedPaymenttothePrincipalConsiderthisnewcontractunder

whichtheprincipalgets$32,750nomatterwhathappensandtheagentkeepstherest.Willthiswork?EffortlevelExpectedpay

offtotheagentE1=6[(55,000½x0.8+40,000½x0.2)-32,750]-36=100.36E2=5[(55,000½x0.6+40,000½x0.4)-32,750]-25=98.

56E3=4[(55,000½x0.3+40,000½x0.7)-32,750]-16=88.35Dr.Chak-TongChau18FulbrightGuestLectureMaterialsFixedPaymenttothePrincipa

l◼Thus,numericallythiswillworktoensurethattheagentgivesthehighesteffort.◼However,thereisnonethelessalosstotheprincipal(33,504-

32,750)=754whichisinasenseamonitoringcost(maximumcosttopayforaninformationsystemtorevealtheagent’seffortleve

l).◼Butthemostfundamentalproblemisthatthistypeofcontractsviolatesthe“riskadverse”natureoftheagent.Nowtheagentbecomestheprincipal!Dr.Chak-TongC

hau19FulbrightGuestLectureMaterialsIncentiveCompatibleContract–ProblemSetupMaximize(55,000–R55)Φ55(e1)+(40,000-R40)Φ40(e1)Subj

ectto:R55½Φ55(e1)+R40½Φ40(e1)-e12=100R55½Φ55(e1)+R40½Φ40(e1)-e12R55½Φ55(e2)+R40½Φ40(e2)–e22R55½Φ55(e1)+

R40½Φ40(e1)-e12R55½Φ55(e3)+R40½Φ40(e3)–e32Dr.Chak-TongChau20FulbrightGuestLectureMaterialsIncentiveCompatibleContract–

SpecificSolutionsMaximize(55,000–R55)0.8+(40,000-R40)0.2Subjectto:R55½(0.8)+R40½(0.2)-36=100R55½(0.8)+R40½(0.2)-36R55½(0.6)+R40½Φ40(0.4)–25R

55½(0.8)+R40½(0.2)-36R55½(0.3)+R40½(0.7)–16Solutions:R55=21,609R40=8,464Expectedpayoffs:Principal=33,020Agent=18,980Dr.Chak-To

ngChau21FulbrightGuestLectureMaterialsSummaryofDifferentContractsEventundere1Principal’sPayoffsAgent’sPayoffObservableFixedRenttoPri

n.IncentiveCompat.ObservableFixedRenttoPrin.IncentiveCompat.55,000(p=0.8)36,50432,75033,39118,49622,25021,6094

0,000(p=0.2)21,50432,75031,53618,4967,2508,464ExpectedPayoffs33,50432,75033,02018,49619,25018,980Dr.Chak-TongChau22FulbrightGuestLectur

eMaterialsWhatdoweknowfromthese?◼Thebestcasescenariofortheprincipaliswhenhecanobservetheagent’seffortleveldirectly.◼Theworstcasesc

enariototheprincipalappearstobesimplychargingafixedrent.◼Thedifferencebetweenthetwo($754)representsthemaximumamountto

payforaninformationsystemtorevealtheagent’seffort.◼Themiddle,2ndbestsolution(incentivecompatiblecontract)maynotalwaysbethenextbest

thingthough!Dr.Chak-TongChau23FulbrightGuestLectureMaterialsLet’ssaythatwesetthetwovariables,R55andR40,tobe18,769and11,449r

espectively.EffortlevelExpectedutilityoftheagentE1=6(18,769½)0.8+(11,449½)0.2-6½=95E2=5(18,769½)0.6+(11,449½)0.4-5½=100E

3=4(18,769½)0.3+(11,449½)0.7-4½=100Now,theprincipalistellingtheagentNOTtoworkhard!The$33,159isactuallybetterthanthe$33,020

under“incentivecompatible”contract!EffortlevelExpectedutilityoftheprincipalE1=6Notafeasiblesolution,agent’s

utility<100n/aE2=5(55,000-18,769)0.6+(40,000-11,449)0.4=33,159E3=4(55,000-18,769)0.3+(40,000-11,449)0.7=30,855Dr.Chak-TongChau24Fulb

rightGuestLectureMaterialsAFewCautionaryRemarks◼Thismodelpresentedhereisasingle-periodmodel.Multiple-period(repeatedgames)cangiveverydifferentan

swers.◼Therecanbemultipleprincipalsaswellasmultipleagentsinthemodel.Suchmodels,however,becomeextremelycomplex.◼Informationsystemsarenotconsi

deredhere.Dr.Chak-TongChau25FulbrightGuestLectureMaterialsConcludingRemarks◼ThePrincipal-agentmodelistheoreticalelega

ntbutmathematicallytedioustouse.◼Empirical(real-life)evidenceseemstosupportthemodelwell.◼Thechallenges,inmyopinion,include:◆tocomeupwithuse

ful,testablehypotheses;◆toextendthemodeltomorecomplex,butrealbusinesssituations;◆toencourageresearcherstoteachnewcomersthebasicskillinunder

standingthemodelratherthansimplytopublishin“ivory-tower”typeofjournals.Dr.Chak-TongChau26FulbrightGuestLect

ureMaterials