Beruflich Dokumente
Kultur Dokumente
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Sage Publications, Inc. is collaborating with JSTOR to digitize, preserve and extend access to The Journal of
Conflict Resolution.
http://www.jstor.org
Gaddis claimed that internationalrelationstheory failed to predictthe Gulf War,the Soviet Union's
collapse, and the cold war's end. Subsequently,he acknowledgedthat the expected utility model captures
the logic behindcomplex adaptivesystems such as the cold warinternationalsystem. Thatmodel correctly
predicted two of the events to which Gaddis pointed. Here, that model is used to simulate alternative
scenarios to determinewhetherthe cold war's end could have been predictedbased only on information
availablein 1948. The simulationsshow a 68%to 78%probabilitythatthe UnitedStateswould win the cold
warpeacefullygiven the conditionsin 1948 andplausibleshifts in the attentivenessof each stateto security
concerns over time. The analysis demonstratesa rigorousmethod for testing counterfactualhistories and
shows thatthe pro-Americanend to the cold war was an emergentpropertyof the initialpost-WorldWarII
conditions.
The end of the cold war is a turningpoint not only for the world of affairsbut also
for the smaller world of internationalrelationsresearch.Decades of scholarshipby
areaspecialists, historians,philosophers,statisticians,andpunditsfailed to provide a
clear, explicit predictionthat the cold war was aboutto end. In light of this apparent
failure,Gaddis(1992) arguedthatinternationalrelationstheorywas a failure.Gaddis
challenged the core of internationalrelations theorizing, stimulating a lively and
sometimesheatedexchangeaboutwhetherinternationalrelationstheorywas a failure.
That debatechallengedthe very notion thatthe absenceof explicit predictionsabout
the end of the cold war was germaneto the questionshe raised.In light of a portion
of that debate,Gaddisrevised his assessment,not on the relevanceof predictionbut
ratheron the prospect of successful predictionthroughtheory building. In a recent
study, Ray and Russett (1996) reportedthat, in privatecorrespondence,Gaddis has
agreedthatthe so-called expected utility model "strikesme as an importantadvance
over earlier approachesto predictive modeling because it takes into account the
emergentpropertiesof complex adaptivesystems ... [and]there has been a sort of
first demonstratethat the expected utility model has predicted many specific and
difficultto anticipateevents well in advance,includingseveralspecific events Gaddis
pointed to as exemplars of the failure of internationalrelations theory. Then, after
explainingthe intuitionbehindthe model's logic, I go on to show thatno hindsightis
used in constructingthe simulationsof the end of the cold war discussed here.
Gaddis's(1992) articlepointedto the failureto predictthreeevents as evidence for
the inadequacyof currentinternationalrelationstheory.He wrote,
The Iran-IraqWarended in mid-1988, and of course the hot war in the PersianGulf
that Gaddis referredto as unanticipatedbegan in the summerof 1990 with the Iraqi
invasion of Kuwait.So, at least one of the threeevents to which Gaddisreferredwas
publicly predicted,includingthe timingof the event, well in advanceusing the model
appliedhere.
The prediction of the Gulf War was not a solitary, isolated, correct prediction.
Indeed,representativesfrom the U.S. governmentwho use an in-house version of the
expected utility model (called Factions) braggedto the Russians in 1995 about its
effectiveness, as reportedin Izvestiya.Thatarticlenotes,
of one the decision makerwill sacrifice in exchange for more of the other.I briefly
summarizethat decision process by sketchingthe basic intuitionbehindthe analytic
approach.Of course, the actualanalysis is quite a bit more complicated.The mathe-
maticaldetails of the model can be found in the works cited previously.
In the model, informationaboutresources,salience, andpolicy preferencesis used
to calculate the perceived relationshipbetween each stakeholderand every other
stakeholder.The model estimates what proposal,if any, is best to make to the other
participantsin the decision process. Sometimes, it is best to live with the status quo
and propose nothing.Othertimes, a compromisecan be struckthatmakes each party
to the deal betteroff. One stakeholdermight gain policy satisfactionat the expense of
political satisfaction,and anothermight gain political satisfactionat the expense of
policy satisfaction.So long as both perceive a net gain, the deal can be made. At still
other times, a stakeholderconcludes that it has the ability,either on its own or with
the help of others,to coerce an opponentinto acceptingits point of view. Sometimes
efforts at coercion result in the desired outcome, but other times coercion leads to a
fight, such as the Gulf War,in which both sides thinkthey are strongenough to secure
benefits from the otherside at acceptablecosts (Haselkorn1998).
In the model, as in real life, I do not assume that people always make the best
proposalsthey can. Rather,I assume thatpeople can make mistakesbut that they do
so while tryingto be as careful as possible with the limited informationthey have at
the time when choices aremade.In thatsense, I tryto takeinto accounthumanfrailties,
differences in perception,and shortsightedness,each of which can lead to errorsin
judgmentandto undesiredor unanticipatedoutcomes.I focus on the strategicchoices
stakeholdersmake before they know how things will turn out. I assume that each
stakeholdermakes what it believes is the best choice given what it knows at the time
a proposal or action must be taken. Hindsight cannot be allowed to color analytic
judgmentsif predictionsand explanationsaboutfuturedecisions are to be accurate.
To apply the expected utility model retrospectivelyto the cold war withouttaking
advantageof ex post facto knowledge, I have undertakennumeroussimulations,all
based on the same initial data. The initial data, drawn from the Correlatesof War
project,include a list of the 36 most powerfulsovereigncountriesas of 1948. For each
country,there is an estimate of its 1948 nationalcapabilities and its position on a
securitydimensionscaled from -100 to 100.
The securitydimension is measuredin termsof the similarityin militaryalliance
portfoliosfor each pairof countriesas of 1948, using Correlatesof Waralliance data,
as explained in Bueno de Mesquita (1975, 1981); that is, each state can have any
possible mix of defense pacts, nonaggressionpacts (or neutralitypacts), ententes,and
no alliance relationshipwith each of the other states in the internationalsystem. A
security portfolio or military alliance portfolio is defined as the array of military
alliance agreementsa state has with all other states. The degree of shared security
3. Given thatthe model does not attachcalendartime to its roundsor iterations,this is one place where
hindsightcan be said to have been used. It is a convenienceto speak of each roundas if it were of 2 years
duration,but the model is not informativeon this. Given the natureof the data,we can be confidentthata
roundreflects no less than 1 year but might reflect 2 or more years.
Before turningto the analysis, let me say a word about how the results are to be
interpreted.Eachof the 100 simulationsthatwererundevelops an alternativeplausible
scenarioor counterfactualhistoryof the post-1948 years. The methodI use does not
TABLE 1
InitialData Conditions, 1948: Pro-Americanor Pro-SovietPolicy Preferences
providean ex antefacto way to choose one scenarioover anotheror say thatthe chosen
scenario is the predictedpath of the cold war.Of course, we can, ex post facto, find
scenarios that look very much like actualhistory,and I will point to such scenarios
(andothersthatdo not look like actualhistory)as I go throughthe analysis.These are
interestingcases thatsuggest a varietyof alternativeways in which the cold warcould
have ended or been prolonged.They are instructiveaboutforeign policy making for
the future,but they are not centralto the predictionsof interest.
The centralpredictionsare aboutthe likelihood that the cold war would continue
or would end with anAmericanor a Soviet victory.Predictionsaboutthese likelihoods
can be made with confidence by examiningthe relative probabilityof each of these
threecore possible outcomes.If the distributionof simulationoutcomes thatsupport,
for instance,an Americanvictoryand a Soviet victoryareaboutequal,then the model
will have failed to predict the centralfeatureof the end of the cold war, that is, a
peaceful victoryby the UnitedStates.Such an approximatelyequaldistributionwould
falsify my claim that this model could predict the end of the cold war, as would a
preponderanceof predictionsfavoringa Soviet victory.If, however,the vast majority
of simulationssupportthe claim for a peacefulAmericanvictoryin the cold war,then
the model will have succeeded in showing that the outcome of the cold war was a
predictableemergentpropertyof the initial conditionsreflectedin the datain Table 1,
with the actual path taken reflecting "noise" or a small perturbationaround the
preponderantcentraltendencyof the predictions.
I begin the analysis, then, with some summarystatisticsthat speak to the model's
predictionsregardingthe likely evolution of the cold war.Then, I examine selected
simulationsto illustratethe characteristicpaths the model suggests for the evolution
of the cold war.These suggest the main alternativehistoriesthatthe model indicates
were likely. The 100 simulationsallow us to estimate the probabilityof alternative
outcomes.The simulationsalso allow us to see the extentto which the end of the cold
warwas attainedoptimallyormight have been achievedearlierwith the same or fewer
costs. The simulationsalso allow us to speak of counterfactualstrategiesfor ending
the cold war and to comparethem to the course actuallyfollowed.
The 1948 data reflect a system of nations highly centralizedaroundneutrality,
uncertainof whetherto invest theirfutureswith the Soviet Union or the UnitedStates.
The initial locus of power was centeredat 2.4 on the policy scale. This locus reflects
the position of greatestnationalsecurityin the sense that at position 2.4, power was
evenly balancedbetweenthose morepro-Americanand those morepro-Soviet.From
a structuralrealist perspective, this can be described as the policy that maximized
security;it reflects a decision to balance one side against the other.From a spatial
modeling perspective, 2.4 representsthe location of the median power unit on the
securityissue dimensionin 1948 andis analogousto Black's (1958) medianvoter.The
1948 structurereflectsa stablesystemin which neitherthe UnitedStatesnorthe Soviet
Union had enough supportto defeat the neutralcenterso long as everyone remained
focused exclusively on security.Thus, a neorealistview, in which securityremainsthe
focus, could not predictthe end of the cold war.The cold war's end dependedon at
least some states being diverted from security concerns by other issues, whether
domestic or foreign.
The randomizationof the salience variablein the 100 simulationswas used exactly
to see what happensto the powerful neutralcenter as the securityconcerns of states
waned and waxed in responseto fluctuatinghypotheticaldomestic or otherpressures.
By allowing salience to vary on the security dimension, my analysis departsfrom
neorealistor structuralrealistprecepts.In thatview of the world, which is tested as a
separatesimulation(the 101st), securityalways is the paramountconcernof states, so
TABLE2
Distributionof Policy Outcomes:The EmergentEnd of the Cold War
Round5/1958 14 15 29 42
Round 15/1978 24 5 20 51
instance, economic might, might have unintentionallyprolonged the cold war. The
long peace, then, might have been due to poor strategizingratherthanany stabilizing
effects of bipolarity.In this regard,it is interestingto note that alreadyby the fifth
iterationof the model, the United Statesis predictedto gain a strongvictory over the
Soviet Union in 42% of the cases. How these more rapidvictories might have been
achieved and how the Soviet Union might have won are topics discussed in the next
section.
Table2 depicts a coarse view of the distributionof likely outcomes of the cold war
as estimatedthroughsimulation.Now we may considera somewhatmorerefinedview,
particularlywith regard to defining what constitutes a U.S. victory. The model
simulationslead to five generic sets of predictions:a slow Americanvictory,a quick
Americanvictory,a continuationof the cold war, a quick Soviet victory, and a slow
Soviet victory.Victoryis definedfor these purposesas arisingwhen the policy stance
of the United States, the Soviet Union, and the majorityof the remainingpowers in
the world converge (to, of course, a pro-Americanor a pro-Sovietposition). The cold
war is said to continue if such convergence does not take place, with the two
superpowersremainingon oppositesides of the neutralpositionof zero on the security
policy scale. Another view of victory is that the preponderanceof global powers
endorsethe Americanor Soviet perspective.Here,the notionis thateven if the Soviet
Union and the United States failed to converge,if enough powers favoredone or the
otherviewpoint,then the remainingsuperpowercould be viewed as marginalizedand
isolated. Figure 1 depicts such a perspective. It shows the cumulative probability
distributionfor the policy stance of the balanceof power in the global setting across
the 100 simulations.Figure1 displaysthe likelihoodof a pro-Americanor apro-Soviet
world by the final iterationof each simulation.
Figure 1 shows quitedramaticallythatthe cumulativeprobabilityof a pro-American
worldview or of Americanhegemony was extremelyhigh. Only 22% of the simula-
tions lead to predictedpro-Sovietcold warresolutionsby this definition,whereas78%
yield a pro-Americanworld outlookby the end of the simulations.This is a key view
of the emergent propertyof American hegemony. It is ironic that the simulations
capturethis-which certainlyhas been echoed in the real world-whereas much of
the academiccommunityin the 1980s focused on the apparentdecline of American
hegemony (Gilpin 1981; Keohane 1984; Krasner1981; but for a view consistentwith
the finding here, see Russett 1985; Kuglerand Organski1989).
.9 -
, .8-'
._ .7-
.>
.4
-
0 - i- ...........
.... . ....... .
vi ..
-100 -50 0 50 100
Pro-Soviet<
7.------>Pro-U SA
1 00-s A US Position
75o
75J \
e? ^A /r- --c
EL 25 /
t
o l--. .7v
-3 - Balance of Power Position
Z -I
-25
-75
-100 _-
-100
I-' . ..-. ...... i
. .. . ..
1948 1958 1968 1978 1988 1998
Simulated Year
t I Ir . -
LI I . q-
-100 - _ /
100 j
r
75-
50 - Bs Position
a.,
rPosition
25-
0-
z -25 -
a)
-50 -
0
CO)
-75 -
Soviet Position /
I.. .
-100-
1948 1958 1968 1978 1988 1998
Simulated Year
o 75-
.)
E 50-
6
IL 25 -
-- Balance of Power Position
-
Iu0 0-
-25 -
.1
o
2 -75 -
0-
Soviet Position /
-100 -
1948
148 1958 1968 1978 1988 1998
Simulated Year
100 - I
I
c i
(U
o 75- I
United States Position
C
U)
E 50-
0--
a)
-25-
U) Soviet Position
> -50 -
o
U)
2 -75-
0-
Soviet Position
-100 -
, I - I I I I
So far, we have seen thatthe simulationsshow that the United States had a better
than two thirdschance of winning the cold war.They also highlightthe observation
that the actual path to victory was not the only possible route. Many policies could
have helped foster the end of the cold war and the establishmentof a pro-American
balanceof power.I turnnow to anexaminationof the five hypotheticalworldsdepicted
in the figures. In so doing, I provide additionaldetails of the politics andpolicies that
supportthe counterfactualhistoriesrepresentedby these simulations.These areuseful
to compareto the actualhistoryof events between 1948 andthe fall of the BerlinWall
in 1989 or the collapse of the Soviet Union in 1991.
COUNTERFACTUALHISTORIES
coupled with the HelsinkiAccords, were the realkey to the Soviet collapse (Deudney
and Ikenberry1991; Risse-Kappen1991). The simulationreinforcesboth strandsof
currentargumentsfor the end of the cold war:the Soviet Union implodedbecause of
internalfailings, and it collapsed because of externalpressure.The simulation,then,
may help to reframe what seems to be a fruitless debate between advocates of
alternativepoints of view. Kegley (1994), for instance,cast the debatein a seemingly
either/orcontext while favoringthe liberal,internalimplosion explanation.Yet, there
is no compelling logical reason to view these as competing explanations.Quite the
contrary,with the SovietUnion facingtwo sourcesof high costs for continuingthe cold
war, the pressureto alterits behavioris greaterthanif it faced only one source. This
is a point reinforcedby analysesof others,includinga prescientperspectiveexpressed
by Zbigniew Brzezinskias early as 1963 (Brzezinski1963; George 1991; Greenstein
1996; Lepgold, Bueno de Mesquita,and Morrow 1996).
Simulation 19, of course, is not a replica of the actualevents that ended the cold
war.Yet, in many details, it provides a historyratherclose to actualevents. NATO's
formationis clearlyevidentearlyon, as is its growingstrengthas the systempolarizes.
The WarsawPact emerges as a resilient alliance thatis resistantto externalpressure
but ultimatelydoomed by the superiorcapabilitiesof the United States to attractand
hold allies andto isolate the Soviet bloc as well as by the prospectsof an internalcrisis
in the Soviet Union. The simulationproduces a history in which Soviet-American
relationsare distantbut stable for a long period, followed by a suddenand dramatic
collapse of the Soviet bloc, much as actuallyoccurred.In many otherdetails, simula-
tion 19 reproducesimportantfeaturesof the cold war,includingtheemergenceof India
as a leader of neutralstates and the large swings in policy position of China. Again,
rememberthatthe base year for the data, 1948, is before the Communistvictory over
the Nationalistforces in China,so thatthe model was not informedof actualchanges
in China.
Figure 2 (simulation 8) depicts a much faster transitionfrom the cold war to
Americansuccess. In this simulation,the Soviet Union acceptsthe Americanapproach
to policy within threeiterationsor, roughly,by the end of 1954. The key to the rapid
Americansuccess is thatthe Soviet Union earlyon is tornby domesticconsiderations
thatsuggest internalinstability.Forinstance,in round2 of this simulation,even before
a NATO-likeallianceforms, Czechoslovakia,Romania,and Hungaryshift from their
adamantlypro-Sovietposition to neutrality(1.3 on the scale). This happensbecause
the Soviet Union is not committedto holding its allies because it is distractedby
domestic politics or some issue otherthansecurity.The salience score for securityin
round 2 for the Soviet Union is only 13, leaving about 87% of its effort turned
elsewhere, such as to domestic issues. This contrastswith America's65 salience for
securityat this stage in the simulation.
Why might Soviet salience have been so low in the early 1950s, as reflected by
simulation8? Stalin's deathcould easily have precipitatedan internalpolitical crisis.
Indeed,in importantrespects,it did.Following his death,a power struggleensuedthat
took several years to resolve. During this period, the United States played a fairly
quiescentrole by not takingany bold steps to exacerbatethe internalproblemsin the
CONCLUSION
The end of the cold war did not depend on going down one uniquepath. Conse-
quently,it is not importantto establishthatany one simulationwas morelikely to arise
thanany other.Whatis importantis to see to whatthe preponderanceof possible paths
lead. Depending on how we define victory,between 67% and 78% of the simulated
histories or paths from 1948 forwardproducedpredictionsthat the cold war would
end on pro-Americanterms.The specific pathmighthave beenunique,butmanypaths
or mixes of policies yield the same conclusion. This means that the end of the cold
war certainly was predictable.Predicting the particulardetails of the history that
culminatedin the Americanvictoryin the cold war would, of course,requireconsid-
erablymore data.But, the model also has pointedout thatex antetherewere superior
historiesto the one thatactuallyplayed out;thatis, therewere alternativepolicies that
could have led to a quicker,peaceful end to the cold war on terms favorableto the
United States. The few cases of Soviet victory,by contrast,were considerablymore
likely to arise throughviolence than throughpeaceful transition.The counterfactual
historiessimulatedheregive us insightinto how we mightplanthe resolutionof future
policy disputes.
The end of WorldWarII appearsto haveplantedthe seeds of a cold wartermination
thattook time to germinate.The ultimateend of the cold war appearsto have been an
emergentpropertywhose essentialantecedentscould alreadybe detectedin the world
structureat least as of 1948.
REFERENCES