This action might not be possible to undo. Are you sure you want to continue?
Micromotiv¢s and Macrob¢bavior
W.W norLon and Companv, 1978
Mv aoal ls noL so much Lo provlde a summarv of 1homas C.
Schelllna's Mlctomotlves ooJ Moctobebovlot, buL raLher Lo Lrv Lo
clarlfv some of Lhe lssues LhaL Schelllna deals wlLh from a
phllosophlcal polnL of vlew. 1wo quesLlons lmmedlaLelv arlse. llrsL,
WhaL sorL of perspecLlve do we have ln mlnd when we speak of
provldlna a phllosophlcal analvsls of Mlctomotlves ooJ
Moctobebovlot? Second, ls Lhls perspecLlve adequaLe for offerlna an
accounL, as far as posslble compleLe, of a book wlLh such conLenL?
lndeed, Mlctomotlves ooJ Moctobebovlot seems noLhlna buL an
lnformal lnLroducLlon Lo formal meLhods ln soclal sclences. Such
formal meLhods are Lhe Lools used Lo analvse Lhe maln sub[ecL
maLLer of Lhe book: Lhe lnLerrelaLlonshlp beLween Lhe lndlvlduals
and Lhe soclal aaareaaLe Lhev comprlse. ln parLlcular, Schelllna's
pro[ecL ls Lhe LheoreLlcal explanaLlon of how lndlvldual purposlve behavlour (mlcromoLlves)
can deLermlne aroup equlllbrlum or a dlsequlllbrlum process (macrobehavlor). ln oLher words,
Lhe book «explores Lhe relaLlon beLween Lhe behavlour characLerlsLlcs of Lhe loJlvlJools who
comprlse some soclal aaareaaLe, and Lhe characLerlsLlcs of Lhe oooteoote».
8oLh Lhese concepLs, lndlvlduals and aaareaaLe, deserve some words of clarlflcaLlon. Cn Lhe
one hand, Schelllna's lndlvlduals are aoal-dlrecLed, namelv, Lhev behave ln a wav LhaL we
mlahL call purposlve: «Lhev have preferences, pursulna aoals, mlnlmlzlna efforL or
embarrassmenL or maxlmlzlna vlew or comforL».
8arelv do people care abouL Lhe ouLcome of
Lhe aaareaaLe, Lvplcallv Lhelr own declslons and Lhelr own behavlour are moLlvaLed bv Lhelr
own lnLeresL. ln Lhls perspecLlve, Lhe enLlre book can be read as a sorL of aenerallzed Þrlsoner's
ullemma: behlnd anv acLlon Lhere ls a uLlllLv-based reason Lo acL. 1hls ls a purelv aame-
LheoreLlcal assumpLlon. Schelllna araues LhaL, even lf lL ls someLlmes mlsleadlna, lL ls
lndlspensable for Lhe analvsls of Lhe relaLlonshlp beLween lndlvldual aoals and aaareaaLe
Cnce Lhe aame's scenarlo ls flxed, some lnLeresLlna consequences follow. Cne ls LhaL
prlvaLelv opLlmlzed declslons mlahL lead ln aaareaaLe Lo ouLcomes LhaL are undeslrable for
evervone. As Schelllna savs aL Lhe bealnnlna, «Lhere are several reasons we mlahL lnLeresL
ourselves ln whaL lL ls LhaL Lhose people [are] dolna, or [Lhlnk] Lhev [are] dolna, or [are] Lrvlna
Lo do [...]. Cne ls LhaL we do noL llke Lhe resulL».
Cn Lhe oLher hand, Lhe aaareaaLe, whlch resulLs from aoal-dlrecLed lndlvldual behavlour, ls
noL merelv an exLrapolaLlon from Lhe lndlvldual. Cf course, Lhere are some easv cases ln whlch
Þhu SLudenL, unlverslLv of llorence
1homas C. Schelllna, Mlctomotlves ooJ Moctobebovlot, W. W. norLon and Companv, 1978. p.13. lLallcs
1homas C. Schelllna, clt., p. 17.
1homas C. Schelllna, clt., p. 12.
Bumana.Nente - Issue 1u - Iuly 2uu9
Lhls happens, e.a. Lhe case ln whlch l choose an acLlon no maLLer whaL Lhe oLhers declde Lo do.
lor lnsLance, lf we know LhaL everv drlver Lurns hls llahLs on aL sunseL, lL ls easv Lo auess whaL
Lhe feaLures of Lhe aaareaaLe are: all Lhe car llahLs ln a local area wlll Lurn on aL abouL Lhe same
Llme. 8uL Lhe mosL lnLeresLlna slLuaLlons are Lhose ln whlch people's behavlour depends on Lhe
behavlour of oLher people, LhaL ls, when Lhelr behavlour ls noL (or noL onlv) consLralned bv
some klnd of exLernal resource buL also bv Lhe behavlour of oLher lndlvlduals. lor lnsLance, ln
Lhe example above, lL ls no lonaer Lrlvlal Lo derlve Lhe feaLures of Lhe aaareaaLe lf we know
LhaL everv drlver Lurns hls llahLs on onlv when Lhe oncomlna cars have alreadv done so. 1he
aaareaaLe ls now a sorL of svsLem of lnLeracLlons ln whlch lndlvlduals respond Lo an
envlronmenL, whlch lnclude oLher people's responses Lo Lhelr own envlronmenL.
AnoLher lmporLanL feaLure of Lhe mlcromoLlves underlvlna lndlvldual acLlons ls Lhe facL LhaL
Lhev are noL markeL moLlvaLlons. Schelllna wrlLes LhaL qulLe a number of acLlvlLles operaLe
almosL enLlrelv ouLslde Lhe purvlew of Lhe markeL, so Lhev cannoL even be expecLed Lo recelve
markeL-llke beneflLs or LreaLmenL. Some examples are Lhe lanauaaes we choose Lo speak, Lhe
person we declde Lo marrv, whom we llve near, whlch aames we llke Lo plav, eLc.
Schelllna clalms LhaL Lhe behavlour he ls concerned wlLh ls noL economlcal sttlcto seoso. Pls
lnLeresL ls ln aaareaaLe behavlour when markeL mechanlsms are noL operaLlna, namelv, ln all
Lhose slLuaLlons ln whlch Lhose who are lnvolved are noL volunLarv parLlclpanLs: Lrafflc [ams,
sex-dlsLrlbuLlon ln campus dlnlna hall, sendlna of ChrlsLmas cards, membershlp ln brldae or
Lennls clubs, raclal composlLlon of nelahbourhoods, and so on.
A verv lllumlnaLlna example of Lhls klnd of behavlour ls Lhe one Schelllna bealns hls book
wlLh: Lhe commonlv observed seaLlna paLLern ln a publlc lecLure hall, where Lhe audlence ls
packed lnLo Lhe back rows leavlna four, flve or slx rows ln fronL of Lhe lecLure hall compleLelv
empLv. Anv aLLempL Lo explaln such a slLuaLlon uslna Lhe Lechnlques of economlcs wlll fall, noL
because Lhese Lechnlques are lnadequaLe, buL because Lhe explanaLlon ls, ln a sense,
1he reason for Lhls fallure ls LhaL Lhe lndlvlduals are noL lnvolved ln LhaL slLuaLlon ln vlrLue
of Lhe facL LhaL Lhev wanL Lo be lnvolved, buL lL slmplv so happens LhaL Lhev are. Schelllna's alm
ls Lhen Lo enlarae Lhe scope of soclal sclence bevond Lhe analvsls of economlc behavlour and
Lo show us LhaL «economlcs [ls] a larae and lmporLanL speclal case, raLher Lhan a model for all
LeL me skeLch ln shorL Lhe sequence of Lhe book's Loplcs.
CPAÞ1L8 CnL ls a sorL of svnopsls of Lhe enLlre monoaraph and lL ls deflnlLelv Lhe mosL
lnLeresLlna parL for Lhe purpose of alvlna, as we sald above, a phllosophlcal accounL of
Schelllna's work. 1he ChapLer presenLs whaL ls Lo be consldered Lhe maln quesLlon of Lhe book:
lf Lhe lndlvlduals acL purposlvelv buL Lhe behavlour of oLhers elLher consLralns whaL Lhe
lndlvlduals can do or ls parL of Lhe lndlvlduals' preferences, whaL klnd of aaareaaLe behavlour
CPAÞ1L8 1WC develops Lhe models of nlne slLuaLlons where Lhe behavlour of Lhe aaareaaLe
ls a loalcal necesslLv, unaffecLed bv Lhe speclflc aoals or behavlour of Lhe lndlvlduals. lndeed,
Lhev do noL even merlL Lhe appellaLlve of ºmodels", Lhev are merelv ºdeflnlLlonal sLaLemenLs":
Lhelr belna Lrue does noL depend on Lhe wav lndlvlduals acL or on whaL Lhelr aoals are, Lhe
LruLh of such sLaLemenLs derlves dlrecLlv from Lhe deflnlLlon of Lhe Lerms LhaL occur wlLhln
1homas C. Schelllna, clt., p. 27.
Commentaiy -Niciomotives anu Naciobehavioi
lor lnsLance, Lelephonlna ls a Lvplcal acLlvlLv Lhrouah whlch one can lnfluence Lhe
behavlour of oLher people. now, surelv lL ls Lrue LhaL none of us ever makes [usL as manv calls
as he recelves, buL lL ls also Lrue LhaL lf Lhe svsLem of calllna people on Lhe phone presupposes
a sender and a recelver, Lhen Lhe number of lncomlna calls ls equal Lo Lhe number of
ouLcomlna ones. 1herefore, ln Lhe aaareaaLe, lL ls Lrue LhaL Lhe LoLal number of recelved calls
equals Lhe LoLal number of made calls, alLhouah lL could happen (and ln a larae number of
cases lL reallv does happen) LhaL we lndlvlduallv recelve more calls Lhan we make, or vlce
1hls example lllusLraLes an lmporLanL feaLure of a larae class of sLaLemenLs, whlch Lhe
economlsLs are verv lnLeresLed ln: «proposlLlons LhaL are Lrue ln Lhe aaareaaLe buL noL ln
deLall, and are Lrue lndependenLlv of how people behave».
1he reason of Lhe emphasls on
such sLaLemenLs ls LhaL economlcs ls manlv concerned wlLh exchanaes of equlvalenL values.
AlLhouah we don'L need anv maLhemaLlcal Lool Lo recoanlse LhaL Lhe example of
Lelephonlna ls Lrue, Schelllna polnLs ouL LhaL even ln Lhls case one flnds surprlses ln Lhe
relaLlonshlp beLween Lhe mlcromoLlves level and Lhe macrobehavlor level. «Conslderlna how
banal Lhese proposlLlons sound, lL ls asLonlshlna how manv hours of commlLLee meeLlnas have
been spenL on proposals Lo mlx men and women ln dormlLorles, or blacks and whlLes, or
freshmen and sophomores, ln wavs LhaL vlolaLed Lhe slmple arlLhmeLlc prlnclple LhaL no
maLLer how vou dlsLrlbuLe Lhem, Lhe numbers ln all dormlLorles have Lo add up Lo Lhe
numbers LhaL Lhere are».
When Lhe soclal slLuaLlons become even a llLLle more complex,
lnLulLlon ls noL enouah Lo reallse LhaL Lhere are verv slmple and lnescapable maLhemaLlcal
laws, llke Lhe one menLloned above, whlch underlle Lhese slLuaLlons.
CPAÞ1L8 1P8LL lnLroduces some models, or famllles of models, ln whlch lndlvldual aoals and
behavlours do affecL aaareaaLe behavlour. Schelllna's analvsls prlmarllv focuses on such
Cne famllv ls LhaL of Lhe ºcrlLlcal mass" models. CrlLlcal mass models descrlbe a larae
number of phenomena LhaL become self-susLalnlna once Lhe level of acLlvlLles passes some
mlnlmum level. «WhaL lL ls common Lo all [crlLlcal mass slLuaLlons] ls Lhe wav people's
behavlor depends on bow moov are behavlna a parLlcular wav, or how much Lhev are behavlna
Schelllna alves Lhe academlc example of a ºdvlna semlnar", l.e. Lhe semlnar whose
popularlLv depends on lLs popularlLv, lL wlll llve or dle, dependlna upon wheLher Lhe sufflclenL
condlLlon ls saLlsfled, namelv, upon wheLher Lhose Laklna parL ln lL conslder Lhe number of
oLher parLlclpanLs sufflclenL, relaLlve Lo some sLandards. ln oLher words, lf enouah oLher
people are dolna someLhlna, we wlll make Lhe same cholce. no one wanLs Lo be Lhe onlv
person aL a club. So we ao Lo a club onlv lf we expecL enouah oLher people Lo be Lhere Lo make
1he mosL lmporLanL subclasses of crlLlcal mass models are ºlemons" and ºLlpplna" models.
1he former lnvolve buvers and sellers who respond ln dlfferenL wavs Lo Lhe acLual proporLlon
of bad used car, l.e. Lhe ºlemons", ln Lhe Lermlnoloav of Akerlof.
lf lemons are rare, Lhe
markeL wlll be sLable because of exchanaes wlll muLuallv advanLaaeous for boLh buvers and
sellers. 8v conLrasL, when Lhe number of lemons lncreases and when lL exceeds Lhe crlLlcal
number, Lhe enLlre markeL mav dlsappear. 1he laLLer models descrlbe Lhe lnsLablllLles ln Lhe
raclal composlLlon of nelahbourhoods. «lL was observed LhaL Lhe enLrance of a few members
1homas C. Schelllna, clt., p. 49.
1homas C. Schelllna, clt., p. 39.
1homas C. Schelllna, clt., p. 94. lLallcs ln orlalnal.
Ceorae A. Akerlof, 1he markeL of 'lemons': CuallLv uncerLalnLv and Lhe markeL mechanlsm, Ooottetlv
Iootool of £cooomlcs, 84:3, 1970.
Bumana.Nente - Issue 1u - Iuly 2uu9
of a mlnorlLv lnLo a nelahborhood ofLen causes some amona Lhe formerlv homoaeneous
populaLlon Lo leave, or Lo slan of leavlna. 1helr deparLure lefL openlnas, so more member of
Lhe mlnorlLv could enLer, Lhe lncrease ln new resldenLs lnduced more of Lhe old Lo leave [...]
Some of Lhe deparLure mlahL be moLlvaLed bv Lhe mlnorlLv enLranLs who have alreadv arrlved,
some bv Lhe bellef LhaL Lhe process, would conLlnue [...]».
1wo oLher classes of models LhaL are sLrlcLlv relaLed Lo crlLlcal mass models refer Lo,
respecLlvelv, Lhe ºcommons"-Lvpe of processes and Lhe expecLaLlon-deLermlned processes.
More speclflcallv, Lhe former class refers Lo all Lhose slLuaLlons ln whlch Lhe lndlvlduals acL
purposlvelv and, bv dolna so, produce an aaareaaLe LhaL ls consldered undeslrable bv
evervone. 1he Þrlsoner's ullemma ls Lhe mosL famous and clear example of one slLuaLlon of
Lhls klnd. !. Plrshlelfer
has polnLed ouL LhaL Lhe [uxLaposlLlon of commons processes wlLh Lhe
crlLlcal mass phenomena ls qulLe mlsleadlna. lndeed Þrlsoner's ullemma slLuaLlons show
cvcllcal sLrucLures LhaL are Lvplcal of Lhe class of phenomena descrlbed bv crlLlcal mass models.
Þerhaps Lhls ls Lhe reason whv Schelllna ls lnLeresLed Lhese slLuaLlons ln Lhls ChapLer.
Cn Lhe oLher hand, expecLaLlon-deLermlned processes refer Lo a class of processes LhaL has
Lhe sLrucLure of self-fulfllllna prophecles: «Lhe aeneral ldea ls LhaL cerLaln expecLaLlons are of a
such characLer LhaL Lhev lnduce Lhe klnd of behavlor LhaL wlll cause Lhe expecLaLlon Lo be
fulfllled. [...] lor example, lf a parLlcular mlnorlLv ls consldered lncapable of holdlna responslble
poslLlons, Lhev wlll noL be hlred for responslble poslLlons, Lhev wlll have no opporLunlLv for
responslble poslLlons, and lacklna anv such experlence, Lhev mav lndeed be lncapable».
CPAÞ1L8 lCu8 and CPAÞ1L8 llvL examlne Lhe appllcaLlon of Lhe models presenLed above for
Lwo klnds of problems: seareaaLlon and lnLearaLlon. ln parLlcular, ChapLer four deals wlLh
seareaaLlon and lnLearaLlon wlLh respecL Lo sex and race, whlch are consldered dlscreLe
varlables. lnsLead, ChapLer flve ls dedlcaLed Lo an analvsls of seareaaLlon and lnLearaLlon when
Lhe varlables of Lhe svsLem are conLlnuous: aae and lncome.
CPAÞ1L8 Slx ls qulLe lndependenL from Lhe oLher chapLers. lL was publlshed before Lhen
and lL could be read lndependenLlv of Lhe enLlre book. lL conslders a posslble slLuaLlon - one
LhaL mlahL soon become a reallLv, ln Lhe auLhor's oplnlon - where parenLs would have Lhe
opporLunlLv of selecLlna Lhelr chlldren's aenes.
CPAÞ1L8 SLvLn ls abouL Þrlsoner's ullemma wlLh coallLlons, Lhls ChapLer aeLs raLher
Lechnlcal, wlLh Lhe help of a loL of maLhemaLlcs. As Þleraloralo Cdlfreddl polnLed ouL ln hls
lnLroducLorv noLe Lo Lhe flrsL lLallan edlLlon of Mlctomotlves ooJ Moctobebovlot
dlfference beLween aame Lheorv as flrsL ploneered bv nash, SelLen and Parsanvl and Lhe verv
same Lheorv as developed bv Schelllna and Aumann ls LhaL whlle Lhe flrsL economlsLs llmlLed
Lhemselves Lo Lhe analvsls of confllcLs ln whlch Lwo or more plavers have no posslblllLv of
maklna coallLlons, Schelllna and Aumann sLudled confllcLs wlLh coallLlons. 1hls ls Lhe reason
whv Schelllna and Aumann recelved Lhe nobel Þrlze ln Lconomlcs.
1he resulL of a baslc Þrlsoner's ullemma slLuaLlon ls LhaL prlvaLelv opLlmlzed declslons
(mlcromoLlves) lead Lo an aaareaaLe (macrobehavlor) LhaL ls undeslrable for evervone. ls Lhls
resulL lnescapable? 1haL ls, ls lL alwavs Lrue LhaL, lf we plav a Þrlsoner's ullemma-based aame,
1homas C. Schelllna, clt., p. 101.
!. Plrshlelfer, 8evlew of Mlctomotlves ooJ Moctobebovlot bv 1homas C. Schelllna, Iootool of £cooomlc
lltetotote, Sol 18, no. 3 (SepL., 1980), pp. 1092-1094.
1homas C. Schelllna, clt., p. 113.
Chooslna our chlldren's aenes, ln Mack Llpkln and ÞeLer 1. 8owlev (eds), Ceoetlc kespooslbllltv,
Þlenum Þress, 1974.
Þleraloralo Cdlfreddl, Noto lottoJottlvo a 1homas C. Schelllna, Mlctomotlvozlool Jello vlto oootlJlooo,
Commentaiy -Niciomotives anu Naciobehavioi
Lhe cooperaLlve cholce ls sLrlcLlv domlnaLed bv Lhe non-cooperaLlve one? Schelllna's answer ls
neaaLlve. Pls concern now ls Lo exLend Lhe deflnlLlon of Lhe baslc verslon of Lhe aame (l.e. Lhe
Lwo-plaver verslon), ln order Lo «[caLch] Lhe splrlL of Lhe Þrlsoner's ullemma»
, when lL ls
plaved bv more Lhan Lwo persons. Schelllna makes Lwo hvpoLheses: flrsL, an lndlvldual ls
alwavs beLLer off, Lhe more oLher people choose Lhelr unpreferred alLernaLlve, second, Lhe
lndlvldual's own preference ls consLanL no maLLer how manv amona Lhe oLhers choose one
wav or Lhe oLher. 1hese assumpLlons lead Lo Lhe deflnlLlon of unlform MulLl-Þersons Þrlsoner's
ullemma (MÞu), namelv a slLuaLlon ln whlch:
1. 1here are people, each wlLh Lhe same blnarv cholce and Lhe same pav-offs.
2. Lach person has a preferred cholce reaardless of whaL Lhe oLhers do and
evervbodv prefers Lo make Lhe same cholce.
3. Whlchever cholce a person makes, Lhe more people amona Lhe oLhers choose
Lhelr unpreferred alLernaLlve, Lhe beLLer off LhaL person ls.
4. 1here ls a number , areaLer Lhan 1, such LhaL lf Lhe number of Lhe lndlvlduals
chooslna Lhelr unpreferred alLernaLlve ls equal or hlaher Lhan , Lhen Lhese
lndlvlduals are beLLer off Lhan lf Lhev had all chosen Lhelr preferred alLernaLlve,
buL lf Lhe number of Lhe lndlvlduals chooslna Lhelr unpreferred alLernaLlve ls
less Lhan , Lhen Lhls ls noL Lrue anvmore.
ln Lhls scenarlo, Lhe cenLral lssue ls Lhe parameLer . WhaL ls lL, exacLlv? lL represenLs «Lhe
mlnlmum slze of anv coallLlons LhaL can aaln bv absLalnlna from Lhe preferred cholce. lL ls Lhe
smallesL dlsclpllned aroup LhaL, Lhouah resenLful of Lhe free rlders, can be proflLable for Lhose
who [oln (Lhouah more proflLable for Lhose who sLav ouL)».
1he resL of Lhe ChapLer ls a sorL of dlaarammaLlc varlaLlon on Lhls baslc Lheme, namelv Lhe
posslblllLv of coallLlons ln Þrlsoner's ullemma conLexLs.
LeL me now Lrv Lo offer Lhe phllosophlcal analvsls LhaL l have menLloned aL Lhe bealnnlna.
1he flrsL conslderaLlon ls a meLhodoloalcal one. Schelllna's approach, as he hlmself sLaLes, ls
noL a markeL approach. 1he aLLenLlon Lo markeL-free conLexLs leads lmmedlaLelv Lo a serles of
meLhodoloalcal reflecLlons abouL Lhe consequences of such an analvsls, l.e. Lhe dlveraence of
Lhe 'equlllbrlum' from Lhe 'opLlmum'.
An equlllbrlum ls a slLuaLlon ln whlch several Lhlnas LhaL have been lnLeracLlna are
evenLuallv ln balance, aL resL. 1vplcallv, and ln splLe of Þrlsoner's ullemma slLuaLlons, from
equlllbrlum analvsls economlsLs derlve Lhe concluslon LhaL self-lnLeresLed mlcromoLlves ofLen
lead Lo surprlslna and soclallv useful coordlnaLlon ln Lhe aaareaaLe, a svsLem ln equlllbrlum ls
Lhus a aood one. 1hls ls Lhe old and fasclnaLlna ldea bv Adam SmlLh of an lnvlslble hand.
Schelllna does noL Lrv Lo assess wheLher a SmlLhlan approach Lo economlc phenomena ls
rlahL or wrona, whaL he does ls slmplv Lo araue LhaL, when a SmlLhlan approach ls applled Lo
non-markeL conLexLs of behavlour, lL ls surelv wrona. Cne can sLlll belleve LhaL someLhlna ln
Lhe ldea of an lnvlslble hand ls Lrue, buL one has also Lo admlL LhaL Lhls ls Lrue onlv ln slLuaLlons
dlfferenL from Lhose LhaL Schelllna conslders.
Schelllna lnslsLs on Lhe facL LhaL equlllbrla are slmplv convenlenL LheoreLlcal resulLs, «Lhere
ls noLhlna parLlcularlv aLLracLlve abouL an equlllbrlum».
noL everv slLuaLlon LhaL ls ln
equlllbrlum ls a [usL or an opLlmal slLuaLlon, Lhe bodv of a hanaed man ls ln equlllbrlum when lL
1homas C. Schelllna, , p. 217.
1homas C. Schelllna, , p. 218.
1homas C. Schelllna, , p. 26.
Bumana.Nente - Issue 1u - Iuly 2uu9
flnallv sLops swlnalna, buL lL ls hard Lo acknowledae LhaL someone could sav LhaL Lhe man ls all
rlahL. lrom Lhls meLhodoloalcal polnL of vlew, Mlctomotlves ooJ Moctobebovlot can be
consldered as an aLLempL Lo provlde a larae number of formal Lools for all Lhe slLuaLlons where
reachlna Lhe equlllbrlum ls noL Lhe aeneral and deflnlLlve purpose LhaL we are Lrvlna Lo
achleve. Powever, ln order Lo recoanlse Lhe asvmmeLrv beLween equlllbrla and deslred
aaareaaLe resulLs, we need Lo look bevond a purelv markeL-orlenLed approach. 1hls ls Lhe aoal
of Schelllna's book.
1he second conslderaLlon concerns Lhe Þrlsoner's ullemma. Schelllna shows how we can
exLend Lhls class of aames ln such a wav LhaL Lo [oln a coallLlon ls no more Lhe klnd of cholce
we do noL prefer Lo make. 1he fallure of cooperaLlve cholce ln Lhe baslc verslon of Þrlsoner's
ullemma ls due Lo Lhe facL LhaL lndlvlduals acL solelv Lo promoLe Lhelr own lnLeresLs and Lo Lhe
facL LhaL Lhls behavlour ls raLlonal. lrom Lhls perspecLlve, lL ls lrraLlonal Lo [oln a coallLlon and
Lo keep Lhe aareemenLs. 1he raLlonal mlcromoLlve ls Lo free rlde. Lven lf evervone recoanlsed
LhaL Lhe aaareaaLe resulLlna from raLlonal mlcromoLlves ls noL Lhe deslrable aaareaaLe, Lhls
acknowledamenL would noL consLlLuLe a sufflclenL reason for chooslna Lhe unpreferred
alLernaLlve of cooperaLlna.
neverLheless, ln observed real-llfe slLuaLlons, coallLlons do work. Pow can we explaln LhaL?
1he scepLlc mav suaaesL LhaL Lhe lnLrlnslc llmlL of aame Lheorv ls Lhe lncapablllLv of alvlna an
adequaLe deflnlLlon of raLlonallLv. lf we adopL a purelv aame LheoreLlc concepL of raLlonallLv,
coallLlons are lrraLlonal. 1he scepLlc derlves Lhe concluslon LhaL whaL we need ls a sorL of
ºconverslon" of raLlonallLv.
lL seems LhaL Lo make Lhe cooperaLlve cholce raLlonal we have Lo
chanae our concepL of raLlonallLv. ln oLher words, lf we wanL Lo capLure Lhe loalc underlvlna
coallLlons we have Lo alve up aame Lheorv. CLherwlse, Lhe expresslon ºÞrlsoner's ullemma
wlLh coallLlon" would be a loalcal conLradlcLlon llke Culne's «round square cupola on 8erkelev
ls Lhe scepLlc rlahL? Schelllna mlahL replv LhaL lL ls cerLalnlv Lrue LhaL Lhe Þrlsoner's
ullemma, ln lLs baslc verslon, does noL capLure Lhe splrlL of all Lhose soclal conLexLs ln whlch
we have a preference Lo [oln a coallLlon raLher Lhan sLavlna ouL. ln oLher words, Lhere ls
cerLalnlv a sense ln whlch Lhe Þrlsoner's ullemma ls noL adequaLe Lo Lhe facLs. Powever, Lhls
does noL mean LhaL cooperaLlve cholces are lrraLlonal or LhaL Lhelr raLlonallLv exceeds Lhe
resources of aame Lheorv. !olnlna a coallLlon can be raLlonal noL ln Lhe llahL of a dlfferenL
concepL of raLlonallLv, buL ln Lhe llahL of Lhe verv same ldea of raLlonallLv LhaL we flnd ln Lhe
Þrlsoner's ullemma. 1hus, Lhere ls no need Lo abandon Lhe aame-LheoreLlc framework. We [usL
have Lo reflne lL.
Schelllna's exLenslon of Lhe Þrlsoner's ullemma ln lLs MÞu verslon has Lo counL as a proof of
Lhe facL LhaL lL ls posslble Lo undersLand Lhe raLlonallLv of coallLlons from Lhe perspecLlve of
Lhe aame LheoreLlc noLlon of raLlonallLv. ln MÞu slLuaLlons people acL purposlvelv llke ln Lhe
baslc verslon of Lhe aame. WhaL ls dlfferenL ls Lhe soclal conLexL ln whlch Lhev acL, and noL Lhe
lndlvldual reasons leadlna Lhem Lo perform ln Lhe wav Lhev do. ln Schelllna's Lermlnoloav,
whaL ls new ln MÞu slLuaLlons ls Lhe aaareaaLe macrobehavlor and noL Lhe lndlvldual
mlcromoLlves. lndlvldual mlcromoLlves are sLlll self-lnLeresLed and raLlonallLv ls sLlll aoal-
orlenLed, buL Lhe aaareaaLe macrobehavlor sLrlcLlv depends on how manv mlcromoLlves
comprlse Lhe aaareaaLe. lrom Lhls perspecLlve we can undersLand beLLer Lhe reason whv
Schelllna conslders Lhe case of ºcommons" (Lvplcallv caLeaorlzed as Þrlsoner's ullemma
slLuaLlons) ln ChapLer Lhree whlle deallna wlLh crlLlcal mass models. 1he parameLer k
!ean PampLon, nobbes ooJ tbe 5oclol coottoct 1toJltloo, Cambrldae unlverslLv Þress, 1986.
W. v. C. Culne, Cn WhaL 1here ls, ln ltom o loolcol lolot of vlew, Parvard unlverslLv Þress, 1933.
Commentaiy -Niciomotives anu Naciobehavioi
menLloned above and represenLlna Lhe mlnlmum number of people who have Lo choose Lhelr
unpreferred alLernaLlve ln order Lo make coallLlon Lhe preferred cholce ls consldered a polnL of
crlLlcal mass. lL seems LhaL Schelllna's ls concerned, amona oLher Lhlnas, wlLh a sorL of new
formulaLlon of Lhe Þrlsoner's ullemma, LhaL ls, wlLh Lhe aLLempL Lo lnLearaLe Lhe Þrlsoner's
ullemma wlLhln Lhe larae famllv of crlLlcal mass models.
Schelllna's revlslLed Þrlsoner's ullemma suaaesLs a flnal conslderaLlon concernlna Lhe
aeneral meLhodoloav of aame Lheorv afLer Lhe appearance of Mlctomotlves ooJ
Moctobebovlot ln Lhe laLe '70s and lLs role wlLhln Lhe Lheorv lLself. 1here are several reasons
whv Mlctomotlves ooJ Moctobebovlot was consldered revoluLlonarv. 1he frlendlv approach Lo
Lhe sub[ecL, whlch makes lL «a book noL [usL for Lhe bookshelf or Lhe docLoral semlnar, buL for
Lhe bedslde Lable», surelv counLs as one of Lhese reasons.
lL also seems Lo me LhaL behlnd
Lhe auLhor's lnLenLlon Lo wrlLe a fun book Lo read, a clear poslLlon abouL Lhe meLhodoloav of
soclal sclences ls acLlve. lndeed, Lhe real novelLv of Mlctomotlves ooJ Moctobebovlot was a
compleLelv dlfferenL aLLlLude Lowards Lhe aeneral meLhods LhaL, ln Schelllna's perspecLlve,
aame Lheorv had Lo emplov. We can hlahllahL Schelllna's chanae of perspecLlve bv asklna Lhe
quesLlon: does lndlvldual behavlour cause svsLems, or do svsLems cause lndlvldual
Cr, ln oLher words: whaL ls Lhe relaLlonshlp beLween facLs and model? Whlch one
of Lhese Lwo elemenLs - facLs or model - has Lhe prlorlLv? uoes aame Lheorv derlve facLs from
Lhe formal model, or ls Lhe model derlved from Lhe facLs?
Schelllna's poslLlon ls ln favour of Lhe prlorlLv of facLs over Lhe model. As 8lchard
Zeckhauser clalms, Schelllna «sLudles a real-world problem and develops a concepLual model.
Pe Lhen Lakes LhaL concepLual model back Lo a dozen real-world problems Lo see how lL
applles, and Lhen rlcocheLs back Lo reflne Lhe model. Pe keeps Lhe process aolna unLll he ls
happv wlLh hls model, and saLlsfled wlLh hls lnslahLs lnLo Lhe problems LhaL mosL lnLeresL
WhaL we wanL Lo do now ls clarlfvlna Lhe phllosophlcal moLlvaLlon leadlna Schelllna Lo
Lhls meLhodoloalcal aLLlLude.
lf we look aL Lhls Loplc from a phllosophlcal polnL of vlew, we see LhaL we face here Lhe
classlcal problem posed bv Lhe dlchoLomv beLween deducLlve and lnducLlve meLhods. 1he
phllosophlcal debaLe abouL Lhe dlsLlncLlon beLween deducLlon and lnducLlon ls vasL. Clven Lhe
llmlLed space avallable, lL wlll noL posslble Lo analvse Lhe maLLer ln areaL depLh. neverLheless
some words of clarlflcaLlon are needed.
1he concepLs of deducLlon and lnducLlon remlnd us lmmedlaLelv of an old dlsLlncLlon. l am
referrlna Lo Lhe dlsLlncLlon LhaL ldenLlfles deducLlon wlLh Lhe lnference from sLaLemenLs
referrlna Lo aeneral concepLs Lo sLaLemenLs referrlna Lo parLlcular cases, and, converselv,
lnducLlon wlLh Lhe lnference from parLlcular sLaLemenLs Lo more aeneral sLaLemenLs. 1he
developmenL of probablllLv Lheorv and sLaLlsLlcs allowed conLemporarv phllosophv of sclence
Lo reflne Lhls dlsLlncLlon
. Accordlna Lo lL, when we Lalk abouL deducLlon and lnducLlon we Lalk
abouL Lwo dlfferenL Lvpes of lnference, namelv, Lwo wavs Lo derlve a concluslon from cerLaln
premlses. lf Lhe concluslon follows from Lhe premlses wlLh Lhe force of loalcal necesslLv, we
1errv Connollv, 8evlew of Mlctomotlves ooJ Moctobebovlot bv 1homas C. Schelllna, AdmlnlsLraLlve
Sclence CuarLerlv, vol. 24, no. 3 (Sep., 1979), p. 304.
uel 1aebel and Charles Llder, uoes Lhe lndlvldual 8ehavlor Cause SvsLems or do SvsLems Cause
lndlvldual 8ehavlor? utboo Affolts Ooottetlv, vol. 13 no. 2, uec. 1979, pp. 229-233.
8lchard Zeckhauser, 1homas Schelllna, 8lcocheL 1hlnker, ln 8oberL uodae, 1be 5ttoteolst. 1be llfe ooJ
1lmes of 1bomos 5cbellloo, ÞurlLan Þress, new Pampshlre, 2006, pp. vll-xll.
See C.C. Pempel, Þ. Cppenhelm, 5toJles lo tbe loolc of £xploootloo, ln «Þhllosophv of Sclence», 13,
pp. 133-173, 1948, W.C. Salmon, loot uecoJes of 5cleotlflc £xploootloo, MlnnesoLa 1989.
Bumana.Nente - Issue 1u - Iuly 2uu9
sav LhaL such an lnference ls deducLlve, oLherwlse, lf Lhe concluslon follows from Lhe premlses
wlLh a cerLaln dearee of probablllLv, Lhe lnference ls lnducLlve.
lL ls mv oplnlon LhaL Schelllna's revoluLlonarv conLrlbuLe Lo aame Lheorv can be appreclaLed
onlv lf we conslder Lhe laLLer wav of concelvlna Lhe dlsLlncLlon beLween deducLlon and
lnducLlon. Maklna Lhls dlsLlncLlon ln Lerms of loalcal deducLlon and deducLlon wlLh a cerLaln
dearee of probablllLv, lnsLead of Lhe classlcal explanaLlon ln Lerm of lnference from parLlcular
Lo aeneral (lnducLlon) and from aeneral Lo parLlcular (deducLlon), alves us Lhe opporLunlLv Lo
reflecL on Lhe role of sLaLlsLlcs wlLhln Schelllna's work. SLaLlsLlcs ls lndeed Lhe sclence LhaL
derlves concluslons wlLh a cerLaln dearee of probablllLv. lf we aeL an lnslahL lnLo Schelllna's
meLhod of analvsls, we see LhaL concluslons abouL Lhe macrobehavlor ln Lhe aaareaaLe are
sLaLlsLlcallv - and noL loalcallv - derlved from lndlvldual mlcromoLlves. 1he presence of
sLaLlsLlcal lnferences ln Mlctomotlves ooJ Moctobebovlot ls ln facL pervaslve. 8efore
Mlctomotlves ooJ Moctobebovlot, aame Lheorv, llke anv maLhemaLlcal Lheorv, was a
deducLlve sclence. AfLer flxlna Lhe models, sLaLemenLs concernlna facLs were Lherebv derlved
ln a purelv deducLlve wav. 1he emplovmenL of sLaLlsLlcal Lools ln Mlctomotlves ooJ
Moctobebovlot radlcallv chanaed Lhe scenarlo. 1he model has now Lo be ad[usLed Lo Lhe facLs,
and noL vlce versa. 1he formal apparaLus ls sub[ecL Lo a conLlnuous process of reflnemenL ln
order Lo adequaLelv descrlbe soclal phenomena LhaL are essenLlallv varlable. 1hus, Schelllna's
sLraLeav ls Lo sLarL wlLh lndlvldual mlcromoLlves and Lo derlve, wlLh a cerLaln dearee of
probablllLv, Lhe resulLlna aaareaaLe macrobehavlor.
AL Lhls polnL we sLarL Lo see Lhe consequences of Schelllna's lnducLlve approach. nelLher
Lhe feaLures of formal models are consldered lmmuLable nor soclal phenomena are derlved
from Lhem ln a merelv loalcal wav. lormal models are sub[ecL Lo revlslons. lf Lhe models are
sLaLlsLlcallv deLermlned, lL ls clear LhaL Lhe answer Lo Lhe quesLlon abouL prlorlLv, lf anv,
beLween model and facLs ls LhaL facLs have Lhe prlorlLv over Lhe model. 8uL Schelllna's
preference for Lhe lnducLlve meLhod would be a petltlo ptloclpl lf we dld noL focus on Lhe facL
LhaL Lhls preference ls arounded on and [usLlfled bv Lhe use of sLaLlsLlcs. 1he around for
adopLlna an lnducLlve meLhod ls Lhus Lhe emplovmenL of sLaLlsLlcs.
ln Lhls scenarlo we can beLLer undersLand Lhe role of Schelllna's examples. 1hese are noL onlv a
dldacLlc sLraLeav Lo make Lhe book fun Lo read, buL Lhev are parL of a dlsLlncLlve meLhodoloav.
PlsLorlcallv, Lhe endorsemenL of an lnducLlve meLhod ls Lhe elemenL whlch dlsLlnaulshes
Mlctomotlves ooJ Moctobebovlot from Lhe oLher aame LheoreLlcal approaches Lo soclal
phenomena. 8ecenL conLrlbuLes llke Malcolm Cladwell's 1be 1lpploo lolot
bv uubner and LewlLL cannoL be undersLood wlLhouL Lhe lnLermedlaLe role
of Schelllna's Mlctomotlves ooJ Moctobebovlot.
Malcolm Cladwell, 1be 1lpploo lolot. now llttle 1bloos coo Moke o 8lo ulffeteoce, 8ack 8av 8ooks,
SLeven LevlLL and SLephen !. uubner, lteokecooomlcs. A koooe £cooomlst £xplotes 1be nlJJeo 5lJe of
£vetvtbloo, Wllllam Morrow/Parper Colllns, 2003.
which games we like to play.g." Let me sketch in short the sequence of the book's topics. rather than a model for all social phenomena». etc. but because the explanation is. For instance.Humana. His interest is in aggregate behaviour when market mechanisms are not operating. sex-distribution in campus dining hall. the person we decide to marry. as we said above. The aggregate is now a sort of system of interactions in which individuals respond to an environment. but it simply so happens that they are. whom we live near. Any attempt to explain such a situation using the techniques of economics will fail. the truth of such statements derives directly from the definition of the terms that occur within such statements.Issue 10 . they are merely "definitional statements": their being true does not depend on the way individuals act or on what their goals are. Schelling's aim is then to enlarge the scope of social science beyond the analysis of economic behaviour and to show us that «economics [is] a large and important special case.Mente . Another important feature of the micromotives underlying individual actions is the fact that they are not market motivations. and so on. CHAPTER NEis a sort of synopsis of the entire monograph and it is definitely the most O interesting part for the purpose of giving.. p. not because these techniques are inadequate. in a sense. The reason for this failure is that the individuals are not involved in that situation in virtue of the fact that they want to be involved. they do not even merit the appellative of "models". Schelling claims that the behaviour he is concerned with is not economical strictu sensu. 200 .July 2009 this happens. For instance. if we know that every driver turns his lights on at sunset. membership in bridge or tennis clubs. Schelling. it is easy to guess what the features of the aggregate are: all the car lights in a local area will turn on at about the same time. so they cannot even be expected to receive market-like benefits or treatment. what kind of aggregate behaviour will result? CHAPTER TWOdevelops the models of nine situations where the behaviour of the aggregate is a logical necessity. where the audience is packed into the back rows leaving four. But the most interesting situations are those in which people's behaviour depends on the behaviour of other people. Indeed. 4Thomas C. namely. cit. e. it is no longer trivial to derive the features of the aggregate if we know that every driver turns his lights on only when the oncoming cars have already done so. Some examples are the languages we choose to speak. that is. 27. the case in which I choose an action no matter what the others decide to do. in all those situations in which those who are involved are not voluntary participants: traffic jams. The Chapter presents what is to be considered the main question of the book: if the individuals act purposively but the behaviour of others either constrains what the individuals can do or is part of the individuals' preferences. Schelling writes that quite a number of activities operate almost entirely outside the purview of the market. unaffected by the specific goals or behaviour of the individuals. racial composition of neighbourhoods. five or six rows in front of the lecture hall completely empty. incomplete. A very illuminating example of this kind of behaviour is the one Schelling begins his book with: the commonly observed seating pattern in a public lecture hall. in the example above. sending of Christmas cards. which include other people's responses to their own environment. a philosophical account of Schelling's work. when their behaviour is not (or not only) constrained by some kind of external resource but also by the behaviour of other individuals.
the seminar whose popularity depends on its popularity. like the one mentioned above. «It was observed that the entrance of a few members 5 6 7 8 Thomas C. cit. the numbers in all dormitories have to add up to the numbers that there are». 59. So we go to a club only if we expect enough other people to be there to make it enjoyable. it is astonishing how many hours of committee meetings have been spent on proposals to mix men and women in dormitories. or how much they are behaving that way». Schelling. Schelling points out that even in this case one finds surprises in the relationship between the micromotives level and the macrobehavior level. we will make the same choice. the market will be stable because of exchanges will mutually advantageous for both buyers and sellers. p.7 Schelling gives the academic example of a "dying seminar". 49. Thomas C. Quarterly Journal of Economics. telephoning is a typical activity through which one can influence the behaviour of other people." When the social situations become even a little more complex. in the aggregate. Akerlof. but it is also true that if the system of calling people on the phone presupposes a sender and a receiver. The former involve buyers and sellers who respond in different ways to the actual proportion of bad used car. cit. By contrast. and are true independently of how people behave». when the number of lemons increases and when it exceeds the critical number. Schelling's analysis primarily focuses on such models. upon whether those taking part in it consider the number of other participants sufficient. 94. the entire market may disappear. p. intuition is not enough to realise that there are very simple and inescapable mathematical laws. cit. No one wants to be the only person at a club. The most important subclasses of critical mass models are "lemons" and "tipping" models. or families of models. depending upon whether the sufficient condition is satisfied." If lemons are rare. or vice versa. Italics in original.Commentary . it is true that the total number of received calls equals the total number of made calls. 1970.Micromotives and Macrobehavior For instance.. namely. This example illustrates an important feature of a large class of statements. «What it is common to all [critical mass situations] is the way people's behavior depends on how many are behaving a particular way. relative to some standards. Now. Thomas C. Schelling.e. «Considering how banal these propositions sound. Schelling. In other words. George A. it will live or die. in which individual goals and behaviours do affect aggregate behaviour. which the economists are very interested in: «propositions that are true in the aggregate but not in detail. surely it is true that none of us ever makes just as many calls as he receives. The market of 'lemons': Quality uncertainty and the market mechanism. or freshmen and sophomores. 84:3. the "lemons". then the number of incoming calls is equal to the number of outcoming ones.e. although it could happen (and in a large number of cases it really does happen) that we individually receive more calls than we make. 201 . i. Therefore. Critical mass models describe a large number of phenomena that become self-sustaining once the level of activities passes some minimum level.. or blacks and whites.." The reason of the emphasis on such statements is that economics is manly concerned with exchanges of equivalent values. in the terminology of Akerlof. which underlie these situations. p. The latter models describe the instabilities in the racial composition of neighbourhoods. if enough other people are doing something. Although we don't need any mathematical tool to recognise that the example of telephoning is true. One family is that of the "critical mass" models. CHAPTER THREE introduces some models. in ways that violated the simple arithmetic principle that no matter how you distribute them. i.
. Piergiorgio Odifreddi. Schelling. 12 CHAPTERIXis quite independent from the other chapters. the difference between game theory as first pioneered by Nash. 1980).July 2009 of a minority into a neighborhood often causes some among the formerly homogeneous population to leave. is it always true that.where parents would have the opportunity of selecting their children's genes. Their departure left openings. Micromotivazioni della vita quotidiana. J. respectively. so more member of the minority could enter. Schelling and Aumann studied conflicts with coalitions. if we playa Prisoner's Dilemma-based game.2008. Schelling.. Selten and Harsanyi and the very same theory as developed by Schelling and Aumann is that while the first economists limited themselves to the analysis of conflicts in which two or more players have no possibility of making coalitions. the former class refers to all those situations in which the individuals act purposively and. Perhaps this is the reason why Schelling is interested these situations in this Chapter.Mente . Hirshleifer. the "commons" -type of processes and the expectation-determined processes. ]». the increase in new residents induced more of the old to leave [. It was published before then S and it could be read independently of the entire book. they may indeed be incapable». Review of Micromotives and Macrobehavior by Thomas C. 12 Choosing our children's genes..one that might soon become a reality. Hirshleifer'" has pointed out that the juxtaposition of commons processes with the critical mass phenomena is quite misleading. In particular. J.. On the other hand. Chapter four deals with segregation and integration with respect to sex and race. 11 Thomas C. pp. This is the reason why Schelling and Aumann received the Nobel Prize in Economics. Sol 18. expectation-determined processes refer to a class of processes that has the structure of self-fulfilling prophecies: «the general idea is that certain expectations are of a such character that they induce the kind of behavior that will cause the expectation to be fulfilled. Genetic Responsibility. Schelling. cit. with the help of a lot of mathematics.. Journal of Economic Literature. produce an aggregate that is considered undesirable by everyone. Instead. It considers a possible situation .9 Two other classes of models that are strictly related to critical mass models refer to.. Is this result inescapable? That is.] Some of the departure might be motivated by the minority entrants who have already arrived. Nota introduttiva Bompiani. Chapter five is dedicated to an analysis of segregation and integration when the variables of the system are continuous: age and income. As Piergiorgio Odifreddi pointed out in his introductory note to the first Italian edition of Micromotives and Macrobenavior". 13 Plenum Press." CHAPTEROUR F and CHAPTERIVE F examine the application of the models presented above for two kinds of problems: segregation and integration. [. 202 . in Mack Lipkin and Peter T. cit. would continue [. which are considered discrete variables. 9 10 Thomas C. 101.] For example. CHAPTER SEVENis about Prisoner's Dilemma with coalitions. The Prisoner's Dilemma is the most famous and clear example of one situation of this kind.. Schelling. p. if a particular minority is considered incapable of holding responsible positions. Rowley (eds). and lacking any such experience. this Chapter gets rather technical. 115. they will not be hired for responsible positions. in the author's opinion . a Thomas C. they will have no opportunity for responsible positions. More specifically. p. 1974.3 (Sept.Humana. The result of a basic Prisoner's Dilemma situation is that privately optimized decisions (micromotives) lead to an aggregate (macrobehavior) that is undesirable for everyone. some by the belief that the process. Indeed Prisoner's Dilemma situations show cyclical structures that are typical of the class of phenomena described by critical mass models. or to sign of leaving. No. 1092-1094..Issue 10 .. by doing so.
then these individuals are better off than if they had all chosen their preferred alternative. second.. This is the old and fascinating idea by Adam Smith of an invisible hand. What is it.e. the two-player version).cit. the divergence of the (equilibrium' from the (optimum'. 4. 217.cit.. what he does is simply to argue that.Commentary . the individual's own preference is constant no matter how many among the others choose one way or the other. Typically. a system in equilibrium is thus a good one. can be profitable for those who join (though more profitable for those who stay out)». The attention to market-free contexts leads immediately to a series of methodological reflections about the consequences of such an analysis. the more people among the others choose their unpreferred alternative. is not a market approach. p. the better off that person is. such that if the number of the individuals choosing their unpreferred alternative is equal or higher than k. but one has also to admit that this is true only in situations different from those that Schelling considers. from equilibrium analysis economists derive the conclusion that self-interested micromotives often lead to surprising and socially useful coordination in the aggregate. namely the possibility of coalitions in Prisoner's Dilemma contexts.. p. then this is not true anymore. the central issue is the parameter k.Micromotives and Macrobehavior the cooperative choice is strictly dominated by the non-cooperative one? Schelling's answer is negative. greater than 1. His concern now is to extend the definition of the basic version of the game (i. exactly? It represents «the minimum size of any coalitions that can gain by abstaining from the preferred choice." The rest of the Chapter is a sort of diagrammatic variation on this basic theme.e. the more other people choose their unpreferred alternative. 26.cit. These assumptions lead to the definition of Uniform Multi-Persons Prisoner's Dilemma (MPD). 15 203 . namely a situation in which: 1. 218. Whichever choice a person makes. The first consideration is a methodological one. Each person has a preferred choice regardless of what the others do and everybody prefers to make the same choice. in order to «[catch] the spirit of the Prisoner's Dilemma»14. when it is played by more than two persons. as he himself states. Schelling. Schelling's approach. There are n people. It is the smallest disciplined group that. Schelling. «there is nothing particularly attractive about an equilibrium». 16 Thomas C. In this scenario. There is a number k. i. An equilibrium is a situation in which several things that have been interacting are eventually in balance. p. an individual is always better off. Schelling insists on the fact that equilibria are simply convenient theoretical results. Thomas C. though resentful of the free riders. 3. it is surely wrong. One can still believe that something in the idea of an invisible hand is true. Let me now try to offer the philosophical analysis that I have mentioned at the beginning." Not every situation that is in equilibrium is a just or an optimal situation. each with the same binary choice and the same pay-offs. 2. at rest. but if the number of the individuals choosing their unpreferred alternative is less than k. the body of a hanged man is in equilibrium when it 14Thomas C. Schelling makes two hypotheses: first. Schelling does not try to assess whether a Smith ian approach to economic phenomena is right or wrong. Schelling. when a Smithian approach is applied to non-market contexts of behaviour. and in spite of Prisoner's Dilemma situations.
the expression "Prisoner's Dilemma with coalition" would be a logical contradiction like Quine's «round square cupola on Berkeley College». Individual micromotives are still self-interested and rationality is still goaloriented. Joining a coalition can be rational not in the light of a different concept of rationality. Harvard University Press. The rational micromotive is to free ride. this acknowledgment would not constitute a sufficient reason for choosing the unpreferred alternative of cooperating. but it is hard to acknowledge that someone could say that the man is all right. From this methodological point of view. In other words. Even if everyone recognised that the aggregate resulting from rational micromotives is not the desirable aggregate. in order to recognise the asymmetry between equilibria and desired aggregate results. and not the individual reasons leading them to perform in the way they do. v. Schelling's extension of the Prisoner's Dilemma in its MPD version has to count as a proof of the fact that it is possible to understand the rationality of coalitions from the perspective of the game theoretic notion of rationality. this does not mean that cooperative choices are irrational or that their rationality exceeds the resources of game theory.Issue 10 .Mente . Schelling shows how we can extend this class of games in such a way that to join a coalition is no more the kind of choice we do not prefer to make. coalitions are irrational. From this perspective. but the aggregate macrobehavior strictly depends on how many micromotives comprise the aggregate." It seems that to make the cooperative choice rational we have to change our concept of rationality. in From a Logical Point of View. it is irrational to join a coalition and to keep the agreements. From this perspective we can understand better the reason why Schelling considers the case of "commons" (typically categorized as Prisoner's Dilemma situations) in Chapter three while dealing with critical mass models. Cambridge University Press. Nevertheless. if we want to capture the logic underlying coalitions we have to give up game theory. Thus. The failure of cooperative choice in the basic version of Prisoner's Dilemma is due to the fact that individuals act solely to promote their own interests and to the fact that this behaviour is rational. there is no need to abandon the game-theoretic framework. Micromotives and Macrobehavior can be considered as an attempt to provide a large number of formal tools for all the situations where reaching the equilibrium is not the general and definitive purpose that we are trying to achieve. does not capture the spirit of all those social contexts in which we have a preference to join a coalition rather than staying out. The parameter k 17 18 Jean Hampton. W. there is certainly a sense in which the Prisoner's Dilemma is not adequate to the facts. In other words. we need to look beyond a purely market-oriented approach. We just have to refine it. The second consideration concerns the Prisoner's Dilemma. What is different is the social context in which they act. How can we explain that? The sceptic may suggest that the intrinsic limit of game theory is the incapability of giving an adequate definition of rationality. Otherwise. 204 . If we adopt a purely game theoretic concept of rationality. However. Hobbes and the Social Contract Tradition. However. The sceptic derives the conclusion that what we need is a sort of "conversion" of rationalitv. in observed real-life situations." Is the sceptic right? Schelling might reply that it is certainly true that the Prisoner's Dilemma. Quine. in its basic version. coalitions do work. 1953. but in the light of the very same idea of rationality that we find in the Prisoner's Dilemma. In Schelling's terminology. In MPD situations people act purposively like in the basic version of the game. what is new in M PD situations is the aggregate macrobehavior and not the individual micromotives. 1986. o.Humana. On What There Is.July 2009 finally stops swinging. This is the goal of Schelling's book.
pp. 504. Schelling's revisited Prisoner's Dilemma suggests a final consideration concerning the methodology of game theory after the appearance of Micromotives and Macrobehavior in the late '70s and its role within the theory itself. Oppenheim. conversely. and satisfied with his insights into the problems that most interest him». The concepts of deduction and induction remind us immediately of an old distinction. 24. The philosophical debate about the distinction between deduction and induction is vast.facts or model . 22 See e.Commentary . Studies in the Logic of Explanation. Indeed. when we talk about deduction and induction we talk about two different types of inference. 15 No. Nevertheless some words of clarification are needed. I am referring to the distinction that identifies deduction with the inference from statements referring to general concepts to statements referring to particular cases. The friendly approach to the subject. 205 . a clear position about the methodology of social sciences is active.e. Administrative Cause Systems or do Systems Cause Science Quarterly. He keeps the process going until he is happy with his model. Dec. W. According to it. Minnesota 1989. it will not possible to analyse the matter in great depth. in «Philosophy of Science». or is the model derived from the facts? Schelling's position is in favour of the priority of facts over the model. New Hampshire. There are several reasons why Micromotives and Macrobehavior was considered revolutionary." It also seems to me that behind general the author's intention to write a fun book to read. 2006. Vol. the real novelty of Micromotives and Macrobehavior was a completely different attitude towards the general methods that. 135-175. 20 Del Taebel and Charles Elder. Does the Individual Individual 21 Behavior? Urban Affairs Quarterly. or do systems cause individual behaviour?" Or. We can highlight Schelling's change of perspective by asking the question: does individual behaviour cause systems. induction with the inference from particular statements to more general statements. Puritan Press. If the conclusion follows from the premises with the force of logical necessity. p. Richard Zeckhauser. Schelling «studies a real-world problem and develops a conceptual model. Thomas Schelling. surely counts as one of these reasons.has the priority? Does game theory derive facts from the formal model. two ways to derive a conclusion from certain premises. we 19 Terry Connolly.Micromotives and Macrobehavior mentioned above and representing the minimum number of people who have to choose their unpreferred alternative in order to make coalition the preferred choice is considered a point of critical mass. and. Hempel. in Schelling's perspective. in Robert Dodge. Given the limited space available. The development of probability theory and statistics allowed contemporary philosophy of science to refine this distinction'". 15. 229-235. 1979. pp. The Strategist: The Life and Times of Thomas Schelling. P. but for the bedside table». He then takes that conceptual model back to a dozen real-world problems to see how it applies. in other words: what is the relationship between facts and model? Which one of these two elements .2. 1948. It seems that Schelling's is concerned. which makes it «a book not just for the bookshelf or the doctoral seminar. No. among other things. game theory had to employ. Schelling. Ricochet Thinker." What we want to do now is clarifying the philosophical motivation leading Schelling to this methodological attitude. As Richard Zeckhauser claims. If we look at this topic from a philosophical point of view. that is. we see that we face here the classical problem posed by the dichotomy between deductive and inductive methods. pp. namely.G. with the attempt to integrate the Prisoner's Dilemma within the large family of critical mass models. vii-xii. Four Decades of Scientific Explanation.3 (Sep. with a sort of new formulation of the Prisoner's Dilemma. Salmon. 1979). Review of Micromotives and Macrobehavior Behavior by Thomas e.. Vol. and then ricochets back to refine the model.
These are not only a didactic strategy to make the book fun to read. If we get an insight into Schelling's method of analysis. The formal apparatus is subject to a continuous process of refinement in order to adequately describe social phenomena that are essentially variable.Mente . 2002. The employment of statistical tools in Micromotives and Macrobehavior radically changed the scenario. Neither the features of formal models are considered immutable nor social phenomena are derived from them in a merely logical way. The model has now to be adjusted to the facts. The presence of statistical inferences Micromotives in Micromotives and Macrobehavior is in fact pervasive. 23 Malcolm Gladwell. statements concerning facts were thereby derived in a purely deductive way. Back Bay Books. Freakeconomics: A Rogue Economist Explores The Hidden Side of Everything. otherwise. Before and Macrobehavior. But Schelling's preference for the inductive method would be a petitio principi if we did not focus on the fact that this preference is grounded on and justified by the use of statistics. and not vice versa. If the models are statistically determined. with a certain degree of probability. After fixing the models. the endorsement of an inductive method is the element which distinguishes Micromotives and Macrobehavior from the other game theoretical approaches to social phenomena.and not logically . 206 . if the conclusion follows from the premises with a certain degree of probability. we see that conclusions about the macrobehavior in the aggregate are statistically . instead of the classical explanation in term of inference from particular to general (induction) and from general to particular (deduction).July 2009 say that such an inference is deductive.derived from individual micromotives. The Tipping Point: How Little Things Can Make a Big Difference. gives us the opportunity to reflect on the role of statistics within Schelling's work. but they are part of a distinctive methodology.Issue 10 . like any mathematical theory. was a deductive science. In this scenario we can better understand the role of Schelling's examples. The ground for adopting an inductive method is thus the employment of statistics. between model and facts is that facts have the priority over the model. At this point we start to see the consequences of Schelling's inductive approach. Dubner. if any. 2005. Statistics is indeed the science that derives conclusions with a certain degree of probability.Humana. Recent contributes like Malcolm Gladwell's The Tipping Point23 or Freakeconomics'" by Dubner and Lewitt cannot be understood without the intermediate role of Schelling's Micromotives and Macrobehavior. Schelling's strategy is to start with individual micromotives and to derive. Formal models are subject to revisions. William Morrow/Harper Collins. game theory. 24 Steven Levitt and Stephen J. Thus. it is clear that the answer to the question about priority. the inference is inductive. Making this distinction in terms of logical deduction and deduction with a certain degree of probability. It is my opinion that Schelling's revolutionary contribute to game theory can be appreciated only if we consider the latter way of conceiving the distinction between deduction and induction. the resulting aggregate macrobehavior. Historically.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.