博弈论2讲义Word下载.docx
《博弈论2讲义Word下载.docx》由会员分享,可在线阅读,更多相关《博弈论2讲义Word下载.docx(14页珍藏版)》请在冰豆网上搜索。
但是在一些特殊的博弈中,一个参与人的最优战略可能并不依赖于其他参与人的战略选择。
也就是说,不管其他参与人选择什么战略,他的最优战略是唯一的,这样的最优战略被称为“占优战略”。
DefinitionStrategysiisstrictlydominatedforplayeriifthereissome
suchthat
foral
Propositionarationalplayerwillnotplayastrictlydominatedstrategy.
抵赖isadominatedstrategy.Arationalplayerwouldthereforenever抵赖.Thissolvesthegamesinceeveryplayerwill坦白.NoticethatIdon'
thavetoknowanythingabouttheotherplayer.
囚徒困境:
个人理性与集体理性之间的矛盾。
ThisresulthighlightsthevalueofcommitmentinthePrisoner'
sdilemma–commitmentconsistsofcrediblyplayingstrategy抵赖.
囚徒困境的广泛应用:
军备竞赛、卡特尔、公共品的供给。
9.2.3IteratedDeletionofDominatedStrategies(重复剔除劣战略)
智猪博弈(boxedpigs)
小猪
按
等待
大猪
3,1
2,4
7,-1
0,0
此博弈没有占优战略均衡。
因为尽管“等待”是小猪的占优战略,但是大猪没有占优战略。
大猪的最优战略依赖于小猪的占略:
---。
大猪会正确地预测到小猪会选择“等待”;
给定此预测,大猪的最优选择只能是“按”。
这样,(按,等待)就是唯一的均衡。
重复剔除的占优均衡:
先剔除某个参与人的劣策略,重新构造新的博弈,再剔除,---。
应用:
大股东监督经理,小股东搭便车;
大企业研发,小企业模仿。
9.2.4Nashequilibrium
性别战博弈(battleofthesexes):
女
足球赛
演唱会
男
2,1
1,2
在上面的博弈中,两个参与者都没有占优策略,每个参与者的最优策略都依赖于另一个参与人的战略。
所以,没有重复剔除的占优均衡。
DefinitionAstrategyprofiles*isapurestrategyNashequilibriumofGifandonlyif
forallplayersiandall
求解Nash均衡的方法。
ANashequilibriumcapturestheideaofequilibrium:
Bothplayersknowwhatstrategytheotherplayerisgoingtochoose,andnoplayerhasanincentivetodeviatefromequilibriumplaybecauseherstrategyisabestresponsetoherbeliefabouttheotherplayer'
sstrategy.
对纳什均衡的理解:
设想所有参与者在博弈之前达成一个(没有约束力的)协议,规定每个参与人选择一个特定的战略。
那么,给定其他参与人都遵守此协议,是否有人不愿意遵守此协议?
如果没有参与人有积极性单方面背离此协议,我们说这个协议是可以自动实施的(self-enforcing),这个协议就构成一个纳什均衡。
否则,它就不是一个纳什均衡。
问题:
纳什均衡与重复剔除(严格)劣战略均衡之间的关系。
9.2.5CournotCompetition(古诺竞争)
Thisgamehasaninfinitestrategyspace.
Twofirmschooseoutputlevelsqi,costfunctionci(qi)=cqi.
marketdemanddeterminesaprice
:
theproductsofbothfirmsareperfectsubstitutes,i.e.theyarehomogenousproducts.
D={1;
2}
S1=S2=R+
u1(q1,q2)=q1f(q1+q2)-c1(q1)
u2(q1,q2)=q2f(q1+q2)-c2(q2)
the'
best-response'
functionBR(qj)ofeachfirmitothequantitychoiceqjoftheotherfirm:
由
,得FOC:
;
又
。
因
Thebest-responsefunctionisdecreasinginmybeliefoftheotherfirm'
saction.
UsingournewresultitiseasytoseethattheuniqueNashequilibriumoftheCournotgameistheintersectionofthetwoBRfunctions.
Becauseofsymmetryweknowthatq1=q2=q*.
Henceweobtain
Thisgivesusthesolution
将寡头竞争的古诺均衡与垄断企业的最优产量和利润进行比较。
9.2.6BertrandCompetition(伯特兰竞争)
Firmscompeteinahomogenousproductmarketbuttheysetprices.
Consumersbuyfromthelowestcostfirm.
demandcurveq=D(p)
Therefore,eachfirmfacesdemand
WealsoassumethatD(c)>
0,i.e.firmscansellapositivequantityiftheypriceatmarginalcost.
LemmaTheBertrandgamehastheuniqueNE
=(c;
c).
Proof:
Firstwemustshowthat(c,c)isaNE.Itiseasytoseethateachfirmmakeszeroprofits.Deviatingtoapricebelowcwouldcauselossestothedeviatingfirm.Ifanyfirmsetsahigherpriceitdoesnotsellanyoutputandalsomakeszeroprofits.Therefore,thereisnoincentivetodeviate.
Toshowuniquenesswemustshowthatanyotherstrategyprofile(p1;
p2)isnotaNE.It'
seasiesttodistinguishlotsofcases.
CaseI:
p1<
corp2<
c.Inthiscaseone(orbothplayers)makesnegativelosses.Thisplayershouldsetapriceabovehisrival'
spriceandcuthislossesbynotsellinganyoutput.
CaseII:
c<
p1<
p2orc<
p2<
p1.Inthiscasethefirmwiththehigherpricemakeszeroprofits.Itcouldprofitablydeviatebysettingapriceequaltotherival'
spriceandthuscaptureatleasthalfofhismarket,andmakestrictlypositiveprofits.
CaseIII:
c=p1<
p2orc=p2<
p1.Nowthelowerpricefirmcanchargeapriceslightlyabovemarginalcost(butstillbelowthepriceoftherival)andmakestrictlypositiveprofits.
CaseIV:
c<
p1=p2.Firm1couldprofitablydeviatebysettingaprice
.Thefirm'
sprofitsbeforeandafterthedeviationare:
Notethatthedemandfunctionisdecreasing,so
.Wecanthereforededuce:
Thisexpression(thegainfromdeviating)isstrictlypositiveforsufficientlysmall
.Therefore,(p1;
p2)cannotbeaNE.
9.2.7MixedStrategies(混合战略)
猜谜游戏matchingpenniesgame:
儿童B
H(正面)
T(反面)
儿童A
1,-1
-1,1
每一个参与者都想猜透对方的战略,而又不能让对方猜透自己的战略。
Thisgamehasnopure-strategyNashequilibrium.Whateverpurestrategyplayer1chooses,player2canbeathim.如果一个参与人采用混合战略(以一定的概率选择某种概率),他的对手就不能准确地猜出他实际上会选择的战略,尽管在均衡点上,每个人都知道其他参与人在不同战略上的概率分布。
Intuitively,gamesinwhichtheparticipantshavealargenumberofstrtegieswilloftenoffersuffifientflexibilitytoensurethatatleastoneNashEquilibriummustexist.Ifwepermittheplayerstouse“mixed”strategies,theabovegamewillbeconvertedintoonewithaninfinitenumberof(mixed)strategiesand,again,theexistenceofaNashequilibriumisensured.
Supposethattheplayersdecidetorandomizeamongsthisstrategiesandplayamixedstrategy.PlayerAcouldflipacoinandplayHwithprobabilityrandTwithprobability1-r,andplayerBflipacoinandplayHwithprobabilitysandTwithprobability1-s.
Giventheseprobabilities,theoutcomesofthegameoccurwiththefollowingprobabilities:
H-H,rs;
H-T,r(1-s);
T-H,(1-r)s;
T-T,(1-r)(1-s).PlayerA’sexpectedutilityisthengivenby
Oviously,A’soptimalchoiceofrdependsonB’sprobability,s.If
utilityismaximizedbychoosing
.If
Ashouldoptfor
.Andwhen
A’sexpectedutilityis0nomatterwhatvalueofrischoosen.
ForplayerB,expectedutilityisgivenby
Now,when
B’sexpectedutilityismaximizedbychoosing
A’sexpectedutilityisindependentofwhatsischoosen.
NashequilibriaareshowninthefigurebytheintersectionsofoptimalresponsecurvesforAandB.
Or,wecangettheequilibriumthroughtheFOC
Definition:
LetGbeagamewithstrategyspacesS1,S2,..,SI.Amixedstrategy
forplayeriisaprobabilitydistributiononSi,i.e.forSi,amixedstrategyisafunction
Severalnotationsarecommonlyusedfordescribingmixedstrategies.
1.Function(measure):
and
2.Vector:
Ifthepurestrategiesare
write
e.g.(1/2,1/2)
3.(1/2)H+(1/2)T
Write
(also
)forthesetofprobabilitydistributionsonSi.
Write
for
.Amixedstrategyprofile
isann-tuple
with
Wewrite
forplayeri'
sexpectedpayoffwhenheusesmixedstrategy
andallotherplayersplayasin
Remark:
ForthedefinitionofamixedstrategypayoffwehavetoassumethattheutilityfunctionfulfillstheVNMaxioms.Mixedstrategiesinducelotteriesovertheoutcomes(strategyprofiles)andtheexpectedutilityofalotteryallowsaconsistentrankingonlyifthepreferencerelationsatisfiestheseaxioms.
DefinitionAmixedstrategyNEofGisamixedprofile
suchthat
foralliandall
ThedefinitionofMSNEmakesitcumbersometocheckthatamixedprofileisaNE.Thenextresultshowsthatitissufficienttocheckagainstpurestrategyalternatives.
Proposition:
isaNashequilibriumifandonlyif
foralliand
Example:
Thestrategyprofile
isaNEofMatchingPennies.
Becauseofsymmetryisitsufficienttocheckthatplayer1wouldnotdeviate.Ifheplayshismixedstrategyhegetsexpectedpayoff0.Playinghistwopurestrategiesgiveshimpayoff0aswell.Therefore,thereisnoincentivetodeviate.
9.3完全信息动态博弈
9.3.1Theextensiveformofagame
Theextensiveformofagameisacompletedescriptionof
1.Thesetofplayers.
2.Whomoveswhenandwhattheirchoicesare.
3.Theplayers'
payoffsasafunctionofthechoicesthataremade.
4.Whatplayersknowwhentheymove.
Example:
ModelofEntry
Currentlyfirm1isanincumbentmonopolist.Asecondfirm2hastheopportunitytoenter.Afterfirm2makesthedecisiontoenter,firm1willhavethechancetochooseapricingstrategy.Itcanchooseeithertofighttheentrantortoaccommodateitwithhigherprices.
ExampleII:
StackelbergModel
Supposefirm1developsanewtechnologybeforefirm2andasaresulthastheopportunitytobuildafactoryandcommittoanoutputlevelq1beforefirm2starts.Firm2thenobservesfirm1beforepickingitsoutputlevelq2.Forconcretenesssuppose
andmarketdemandis
.Themarginalcostofproductionis0.
9.3.2DefinitionofanExtensiveFormGame
Formallyafiniteextensiveformgameconsistsof
1.Afinitesetofplayers.
2.AfinitesetTofnodeswhichformatreealongwithfunctionsgivingforeachnon-terminalnode
(Zisthesetofterminalnodes)
theplayeri(t)whomoves
thesetofpossibleactionsA(t)
thesuccessornoderesultingfromeachpossibleactionN(t;
a)
3.Payofffunctions
givingtheplayerspayoffsasafunctionoftheterminalnodereached(theterminalnodesaretheoutcomesofthegame).
4.Aninformationpartition:
foreachnodet,h(t)isthesetofnodeswhicharepossiblegivenwhatplayeri(x)knows.Thispartitionmustsatisfy
Wesometimeswritei(h)andA(h)si