Diplomacy and Domestic Politics: the Logic of Two-Level Games Author(S): Robert D

Home , Robert Keohane

Diplomacy and Domestic Politics: The Logic of Two-Level Games Author(s): Robert D. Putnam Source: International Organization, Vol. 42, No. 3 (Summer, 1988), pp. 427-460 Published by: The MIT Press Stable URL: http://www.jstor.org/stable/2706785 Accessed: 10-06-2015 17:31 UTC

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 [email protected].

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to International Organization.

http://www.jstor.org

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics: the logicof two-levelgames RobertD. Putnam

Introduction:the entanglementsof domesticand internationalpolitics

Domesticpolitics and internationalrelations are oftensomehow entangled, but our theorieshave not yetsorted out thepuzzling tangle. It is fruitless to debatewhether domestic politics really determine international relations, or thereverse. The answerto thatquestion is clearly"Both, sometimes." Themore interesting questions are "When?"and "How?" Thisarticle offers a theoreticalapproach to thisissue, but I beginwith a storythat illustrates thepuzzle. One illuminatingexample of how diplomacyand domesticpolitics can becomeentangled culminated at theBonn summitconference of 1978.1 In the mid-1970s,a coordinatedprogram of globalreflation, led by the "locomotive"economies of the United States, Germany, and Japan,had been proposedto fosterWestern recovery from the first oil shock.2This proposal

An earlierversion of thisarticle was deliveredat the 1986annual meeting of theAmerican PoliticalScience Association. For criticisms and suggestions, I am indebted to RobertAxelrod, NicholasBayne, Henry Brady, James A. Caporaso,Barbara Crane, Ernst B. Haas, Stephan Haggard,C. RandallHenning, Peter B. Kenen,Robert 0. Keohane,Stephen D. Krasner,Jacek Kugler,Lisa Martin,John Odell, Robert Powell, Kenneth A. Shepsle,Steven Stedman, Peter Yu, membersof researchseminars at theUniversities of Iowa, Michigan,and Harvard,and two anonymousreviewers. I am gratefulto the RockefellerFoundation for enabling me to completethis research. 1. The followingaccount is drawnfrom Robert D. Putnamand C. RandallHenning, "The BonnSummit of 1978: How Does InternationalEconomic Policy Coordination Actually Work?" BrookingsDiscussion Papers in InternationalEconomics, no. 53 (Washington,D.C.: Brookings Institution,October 1986), and RobertD. Putnamand NicholasBayne, Hanging Together: Cooperation and Conflictin the Seven-Power Summits,rev. ed. (Cambridge,Mass.: Harvard UniversityPress, 1987), pp. 62-94. 2. Amonginterdependent economies, most economists believe, policies can oftenbe more effectiveifthey are internationallycoordinated. For relevant citations, see Putnamand Bayne, Hanging Together,p. 24.

InternationalOrganization 42, 3, Summer 1988 ?D1988 by theWorld Peace Foundationand theMassachusetts Institute of Technology

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 428 InternationalOrganization had receiveda powerfulboost from the incoming Carter administration and was warmlysupported by theweaker countries, as wellas theOrganization forEconomic Co-operation and Development(OECD) and manyprivate economists,who argued that it wouldovercome international payments im- balancesand speedgrowth all around.On theother hand, the Germans and theJapanese protested that prudent and successfuleconomic managers should notbe askedto bail out spendthrifts.Meanwhile, Jimmy Carter's ambitious NationalEnergy Program remained deadlocked in Congress,while Helmut Schmidtled a chorusof complaintsabout the Americans' uncontrolled ap- petitefor imported oil and theirapparent unconcern about the falling dollar. All sides concededthat the worldeconomy was in serioustrouble, but it was notclear whichwas moreto blame,tight-fisted German and Japanese fiscalpolicies or slack-jawedU.S. energyand monetarypolicies. At the Bonn summit,however, a comprehensivepackage deal was ap- proved,the clearestcase yetof a summitthat left all participantshappier thanwhen they arrived. Helmut Schmidt agreed to additionalfiscal stimulus, amountingto 1 percentof GNP, JimmyCarter committed himself to decon- troldomestic oil pricesby theend of 1980,and Takeo Fukudapledged new effortsto reacha 7 percentgrowth rate. Secondaryelements in theBonn accordincluded French and Britishacquiescence in theTokyo Round trade negotiations;Japanese undertakings to fosterimport growth and restrain exports;and a genericAmerican promise to fightinflation. All in all, the Bonn summitproduced a balancedagreement of unparalleledbreadth and specificity.More remarkably, virtually all partsof the package were actually implemented. Most observersat the timewelcomed the policiesagreed to at Bonn, althoughin retrospectthere has been muchdebate about the economic wisdomof this package deal. However,my concern here is notwhether the deal was wise economically,but how it became possiblepolitically. My researchsuggests, first, that the key governments at Bonnadopted policies differentfrom those that they would have pursuedin theabsence of internationalnegotiations, but second, that agreement was possibleonly because a powerfulminority within each governmentactually favored on domestic groundsthe policy being demanded internationally. WithinGermany, a politicalprocess catalyzed by foreignpressures was surreptitiouslyorchestrated by expansionistsinside the Schmidtgovern- ment.Contrary to thepublic mythology, the Bonn deal was notforced on a reluctantor "altruistic"Germany. In fact,officials in the Chancellor's Officeand the Economics Ministry, as wellas inthe Social Democratic party andthe trade unions, had argued privately in early 1978 that further stimulus was domesticallydesirable, particularly in view of the approaching1980 elections.However, they had littlehope of overcomingthe oppositionof the FinanceMinistry, the Free Democraticparty (part of the government coalition),and thebusiness and bankingcommunity, especially the leader-

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 429 shipof the Bundesbank. Publicly, Helmut Schmidt posed as reluctantto the end. Onlyhis closestadvisors suspected the truth: that the chancellor "let himselfbe pushed" intoa policythat he privatelyfavored, but would have foundcostly and perhapsimpossible to enactwithout the summit's package deal. Analogously,in Japana coalitionof businessinterests, the Ministryof Tradeand Industry(MITI), theEconomic Planning Agency, and some ex- pansion-mindedpoliticians within the Liberal Democratic Party pushed for additionaldomestic stimulus, using U.S. pressureas one of theirprime argumentsagainst the stubborn resistance of the Ministry of Finance (MOF). Withoutinternal divisions in Tokyo,it is unlikelythat the foreign demands wouldhave been met,but withoutthe externalpressure, it is even more unlikelythat the expansionistscould have overriddenthe powerful MOF. "Seventypercent foreign pressure, 30 percentinternal politics," was the disgruntledjudgment of one MOF insider."Fifty-fifty," guessed an official fromMITI.3 In theAmerican case, too, internalpoliticking reinforced, and was rein- forcedby, the international pressure. During the summit preparations Amer- ican negotiatorsoccasionally invited their foreign counterparts to putmore pressureon the Americansto reduceoil imports.Key economicofficials withinthe administrationfavored a tougherenergy policy, but theywere opposed by the president'sclosest political aides, even afterthe summit. Moreover,congressional opponents continued to stymieoil pricedecontrol, as theyhad under both Nixon and Ford.Finally, in April 1979, the president decidedon gradualadministrative decontrol, bringing U.S. pricesup to world levelsby October1981. His domesticadvisors thus won a postponementof thispolitically costly move until after the 1980presidential election, but in theend, virtually every one ofthe pledges made at Bonnwas fulfilled.Both proponentsand opponentsof decontrolagree that the summit commitment was at the centerof the administration'sheated intramural debate during thewinter of 1978-79and instrumentalin thefinal decision.4 In short,the Bonn accord representedgenuine international policy coordination.Significant policy changes were pledged and implemented by the key participants.Moreover-although this counterfactual claim is neces- sarilyharder to establish-thosepolicy changes would very probably not have been pursued(certainly not the same scale and withinthe same time frame)in theabsence of theinternational agreement. Within each country, one factionsupported the policy shift being demanded of its countryinter- 3. For a comprehensiveaccount of the Japanese story, see I. M. Destlerand Hisao Mitsuyu, "Locomotiveson DifferentTracks: Macroeconomic Diplomacy, 1977-1979," in I. M. Destler and Hideo Sato, eds., Coping with U.S.-Japanese Economic Conflicts(Lexington, Mass.: Heath,1982). 4. For an excellentaccount of U.S. energypolicy during this period, see G. JohnIkenberry, "MarketSolutions for State Problems: The Internationaland DomesticPolitics of American Oil Decontrol," InternationalOrganization 42 (Winter1988).

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 430 InternationalOrganization nationally,but thatfaction was initiallyoutnumbered. Thus, international pressurewas a necessarycondition for these policy shifts. On the other hand,without domestic resonance, international forces would not have suf- ficedto producethe accord, no matterhow balanced and intellectuallyper- suasivethe overall package. In theend, each leaderbelieved that what he was doingwas in his nation'sinterest-and probably in his own political interest,too, eventhough not all his aides agreed.5Yet withoutthe summit accordhe probablywould not (or couldnot) have changed policies so easily. In thatsense, the Bonn deal successfullymeshed domestic and international pressures. Neithera purelydomestic nor a purelyinternational analysis could account forthis episode. Interpretations cast in termseither of domestic causes and internationaleffects ("Second Image"6)or of internationalcauses and do- mesticeffects ("Second ImageReversed"7) would represent merely "partial equilibrium"analyses and would miss an importantpart of the story, namely, how the domesticpolitics of severalcountries became entangledvia an internationalnegotiation. The eventsof 1978illustrate that we mustaim insteadfor "general equilibrium" theories that account simultaneously for theinteraction of domesticand internationalfactors. This article suggests a conceptualframework for understandinghow diplomacyand domestic politicsinteract.

Domestic-internationalentanglements: the state of the art

Much of theexisting literature on relationsbetween domestic and internationalaffairs consists either of ad hoclists of countless "domestic influences" on foreignpolicy or of genericobservations that national and international affairsare somehow"linked.' 8 JamesRosenau was one ofthe first scholars to call attentionto thisarea, but his elaborate taxonomy of "linkage politics" generatedlittle cumulative research, except for a flurryof workcorrelating domesticand international"conflict behavior."9 A second streamof relevanttheorizing began withthe workby Karl 5. It is not clear whetherJimmy Carter fully understood the domestic implications of his Bonn pledgeat the time.See Putnamand Henning,"The Bonn Summit,"and Ikenberry, "MarketSolutions for State Problems." 6. KennethN. Waltz, Man, theState, and War: A TheoreticalAnalysis (New York: Columbia UniversityPress, 1959). 7. PeterGourevitch, "The SecondImage Reversed: The InternationalSources of Domestic Politics,"International Organization 32 (Autumn1978), pp. 881-911. 8. I am indebtedto StephanHaggard for enlightening discussions about domestic influences on internationalrelations. 9. JamesRosenau, "Toward the Study of National-International Linkages," in his-Linkage Politics: Essays on the Convergence of National and InternationalSystems (New York: Free Press,1969), as wellas his "TheorizingAcross Systems: Linkage Politics Revisited," in Jon- athanWilkenfeld, ed., ConflictBehavior and Linkage Politics (New York: David McKay, 1973), especiallyp. 49.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 431

Deutschand ErnstHaas on regionalintegration. 10Haas, in particular,em- phasizedthe impact of parties and interest groups on theprocess of European integration,and his notionof "spillover"recognized the feedback between domesticand internationaldevelopments. However, the central dependent variablein thiswork was thehypothesized evolution of new supranational institutions,rather than specific policy developments, and whenEuropean integrationstalled, so did thisliterature. The intellectualheirs of thistra- dition,such as JosephNye and RobertKeohane, emphasized interdepend- enceand transnationalism, but the role of domestic factors slipped more and moreout of focus, particularly as theconcept of international regimes came to dominatethe subfield.11 The "bureaucraticpolitics" school of foreignpolicy analysis initiated anotherpromising attack on the problemof domestic-internationalinter- action.As GrahamAllison noted, "Applied to relationsbetween nations, thebureaucratic politics model directs attention to intra-nationalgames, the overlapof whichconstitutes international relations. "12 Nevertheless,the natureof this "overlap" remainedunclarified, and the theoreticalcontri- butionof thisliterature did not evolvemuch beyond the principle that bu- reaucraticinterests matter in foreignpolicymaking. Morerecently, the most sophisticated work on thedomestic determinants of foreignpolicy has focusedon "structural"factors, particularly "state strength."The landmarkworks of Peter Katzenstein and StephenKrasner, forexample, showed the importance of domestic factors in foreign economic policy.Katzenstein captured the essence of the problem: "The mainpurpose of all strategiesof foreigneconomic policy is to make domesticpolicies compatiblewith the international political economy." 13 Bothauthors stressed the crucialpoint that central decision-makers ("the state") mustbe con- cernedsimultaneously with domestic and internationalpressures.

10. Karl W. Deutsch et al., Political Communityin the North AtlanticArea: International Organization in the Light of Historical Experience (Princeton: PrincetonUniversity Press, 1957) and Ernst B. Haas, The Unitingof Europe: Political, Social, and Economic Forces, 1950-1957(Stanford, Calif.: Stanford University Press, 1958). 11. Robert0. Keohane and JosephS. Nye, Powerand Interdependence(Boston: Little, Brown,1977). On theregime literature, including its neglect of domestic factors, see Stephan Haggardand Beth Simmons, "Theories of International Regimes," International Organization 41 (Summer1987), pp. 491-517. 12. Graham T. Allison, Essence of Decision: Explaining the Cuban Missile Crisis (Boston: Little,Brown, 1971), p. 149. 13. Peter J. Katzenstein, ed., Between Power and Plenty: Foreign Economic Policies of AdvancedIndustrial States (Madison:University of WisconsinPress, 1978),p. 4. See also Katzenstein,"International Relations and Domestic Structures: Foreign Economic Policies of AdvancedIndustrial States," International Organization 30 (Winter1976), pp. 1-45; Stephen D. Krasner,"United States Commercial and MonetaryPolicy: Unravelling the Paradoxof ExternalStrength and InternalWeakness," in Katzenstein,Between Power and Plenty,pp. 51-87; and Krasner, Defending the National Interest: Raw Materials Investmentsand U.S. ForeignPolicy (Princeton: Princeton University Press, 1978).

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 432 InternationalOrganization

Moredebatable, however, is theiridentification of "state strength" as the keyvariable of interest. Given the difficulties ofmeasuring "state strength," thisapproach courts tautology,14 and effortsto locate individualcountries on thisambiguous continuum have proved problematic.15 "State strength," ifreinterpreted as merely the opposite of governmental fragmentation, is no doubtof some interest in the comparative study of foreign policy. However, Gourevitchis quitecorrect to complainthat "the strongstate-weak state argumentsuggests that . .. t-heidentity of thegoverning coalition does not matter.This is a veryapolitical argument."16 Moreover, because "state structures"(as conceivedin thisliterature) vary little from issue to issue or fromyear to year,such explanations are ill-suitedfor explaining differences acrossissues or acrosstime (unless "time" is measuredin decadesor cen- turies).A moreadequate account of the domesticdeterminants of foreign policyand international relations must stress politics: parties,social classes, interestgroups (both economicand noneconomic),legislators, and even publicopinion and elections,not simply executive officials and institutional arrangements.17 Some workin the "state-centric"genre represents a unitary-actormodel runamok. "The centralproposition of this paper," notes one recentstudy, "is thatthe state derives its interests from and advocates policies consistent withthe international system at all timesand under all circumstances."'18 In fact,on nearlyall importantissues "central decision-makers" disagree about whatthe national interest and theinternational context demand. Even ifwe arbitrarilyexclude the legislature from "the state" (as muchof this literature does),it is wrongto assume that the executive is unifiedin its views. Certainly thiswas truein none of the statesinvolved in the 1978negotiations. What was "the" positionof the Germanor Japanesestate on macroeconomic policyin 1978,or ofthe American state on energypolicy? If the term "state" is to be used to mean "centraldecision-makers," we shouldtreat it as a plural noun: not "the state, it . . ." but "the state, they . . ." Central ex- ecutiveshave a specialrole in mediating domestic and international pressures preciselybecause theyare directlyexposed to bothspheres, not because

14. For example,see Krasner,"United States Commercial and Money Policy," p. 55: "The centralanalytic characteristic that determines the abilityof a stateto overcomedomestic resistanceis its strengthin relationto itsown society." 15.Helen Milner, "Resisting the Protectionist Temptation: Industry and the Making of Trade Policyin France and the UnitedStates duringthe 1970s,"International Organization 41 (Autumn1987), pp. 639-65. 16. Gourevitch,"The SecondImage Reversed," p. 903. 17.In theirmore descriptive work, "state-centric" scholars are often sensitive to theimpact ofsocial and political conflicts, such as thosebetween industry and finance, labor and business, and export-orientedversus import-competing sectors. See Katzenstein,Between Pqwer and Plenty,pp. 333-36,for example. 18. David A. Lake, "The Stateas Conduit:The InternationalSources of National Political Action,"presented at the 1984annual meeting of the American Political Science Association, p. 13.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 433

theyare unitedon all issues norbecause theyare insulatedfrom domestic politics. Thus,the state-centric literature is an uncertainfoundation for theorizing abouthow domesticand internationalpolitics interact. More interesting are recentworks about the impactof the internationaleconomy on domestic politicsand domesticeconomic policy, such as thoseby Alt,Evans, Gour- evitch,and Katzenstein.19These case studies,representing diverse meth- odologicalapproaches, display a theoreticalsophistication on the international-to-domesticcausal connectionfar greater than is characteristicof comparablestudies on thedomestic-to-international halfof the loop. Never- theless,these works do notpurport to accountfor instances of reciprocal causation,nor do theyexamine cases in whichthe domesticpolitics of severalcountries became entangled internationally. In short,we need to move beyondthe mereobservation that domestic factorsinfluence international affairs and vice versa,and beyondsimple catalogsof instancesof suchinfluence, to seek theoriesthat integrate both spheres,accounting for the areas of entanglementbetween them.

Two-levelgames: a metaphor fordomestic-international interactions

Overtwo decades ago RichardE. Waltonand RobertB. McKersieoffered a "behavioraltheory" of social negotiationsthat is strikinglyapplicable to internationalconflict and cooperation.20 They pointed out, as all experienced negotiatorsknow, that the unitary-actor assumption is oftenradically mis- leading.As RobertStrauss said of the Tokyo Round tradenegotiations: "Duringmy tenure as Special TradeRepresentative, I spent as muchtime negotiatingwith domestic constituents (both industry and labor)and members of the U.S. Congressas I did negotiatingwith our foreigntrading partners."21

19.James E. Alt,"Crude Politics: Oil andthe Political Economy of Unemployment inBritain andNorway, 1970-1985," British Journal of Political Science 17 (April 1987), pp. 149-99;Peter B. Evans, Dependent Development: The Alliance of Multinational,State, and Local Capital in Brazil (Princeton:Princeton University Press, 1979);Peter Gourevitch, Politics in Hard Times: Comparative Responses to InternationalEconomic Crises (Ithaca, N.Y.: Cornell Uni- versityPress, 1986); Peter J. Katzenstein,Small States in WorldMarkets: IndustrialPolicy in Europe(Ithaca, N.Y.: CornellUniversity Press, 1985). 20. Richard E. Waltonand RobertB. McKersie, A Behavioral Theoryof Labor Negotiations: An Analysis of a Social InteractionSystem (New York: McGraw-Hill, 1965). 21. RobertS. Strauss,"Foreword," in JoanE. Twiggs,The TokyoRound of Multilateral Trade Negotiations: A Case Study in BuildingDomestic Supportfor Diplomacy (Washington, D.C.: GeorgetownUniversity Institute for the Studyof Diplomacy,1987), p. vii. Former Secretaryof Labor JohnDunlop is said to have remarkedthat "bilateral negotiations usually requirethree agreements-one across the table and one on each sideof the table," as citedin Howard Raiffa,The Art and Science of Negotiation (Cambridge,Mass.: Harvard University Press,1982), p. 166.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 434 InternationalOrganization

The politicsof many international negotiations can usefullybe conceived as a two-levelgame. At the nationallevel, domesticgroups pursue their interestsby pressuring the government to adoptfavorable policies, and pol- iticiansseek powerby constructingcoalitions among those groups. At the internationallevel, national governments seek to maximizetheir own ability to satisfydomestic pressures, while minimizing the adverseconsequences offoreign developments. Neither of the two games can be ignoredby central decision-makers,so longas theircountries remain interdependent, yet sov- ereign. Each nationalpolitical leader appears at bothgame boards. Across the internationaltable sit his foreign counterparts, and at hiselbows sit diplomats and otherinternational advisors. Around the domestic table behind him sit partyand parliamentaryfigures, spokespersons for domestic agencies, representativesof key interest groups, and theleader's own political advisors. The unusualcomplexity of this two-level game is thatmoves that are rational fora playerat one board(such as raisingenergy prices, conceding territory, or limitingauto imports)may be impoliticfor that same player at theother board.Nevertheless, there are powerfulincentives for consistency between the two games. Players(and kibitzers)will toleratesome differencesin rhetoricbetween the two games, but in theend eitherenergy prices rise or theydon't. The politicalcomplexities for the players in thistwo-level game are stag- gering.Any key player at theinternational table who is dissatisfiedwith the outcomemay upset the game board, and conversely,any leaderwho fails to satisfyhis fellowplayers at thedomestic table risks being evicted from hisseat. On occasion,however, clever players will spot a moveon oneboard thatwill triggerrealignments on otherboards, enabling them to achieve otherwiseunattainable objectives. This "two-table"metaphor captures the dynamicsof the 1978negotiations better than any modelbased on unitary nationalactors. Otherscholars have notedthe multiple-gamenature of internationalrelations.Like Waltonand McKersie,Daniel Druckmanhas observedthat a negotiator"attempts to builda packagethat will be acceptableboth to the otherside and to his bureaucracy."However, Druckman models the domesticand internationalprocesses separately and concludesthat "the in- teractionbetween the processes . . . remains a topic for investigation."22 RobertAxelrod has proposeda "Gamma paradigm,"in whichthe U.S. presidentpursues policies vis-a-vis the SovietUnion with an eye towards maximizinghis popularity at home.However, this model disregards domestic

22. Daniel Druckman,"Boundary Role Conflict:Negotiation as Dual Responsiveness,"in I. William Zartman,ed., The Negotiation Process: Theories and Applications (Beverly Hills: Sage, 1978),pp. 100-101,109. For a reviewof the social-psychological literature on bargainers as representatives,see Dean G. Pruitt,Negotiation Behavior (New York: AcademicPress, 1981),pp. 41-43.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 435 cleavages,and it postulatesthat one ofthe international actors-the Soviet leadership-caresonly about international gains and faces no domesticcon- straintwhile the other-the U.S. president-caresonly about domestic gains, exceptinsofar as hispublic evaluates the international competition.23 Prob- ablythe most interesting empirically based theorizingabout the connection betweendomestic and internationalbargaining is thatof GlennSnyder and Paul Diesing.Though working in theneo-realist tradition with its conven- tionalassumption of unitary actors, they found that, in fully half of the crises theystudied, top decision-makerswere not unified.They concludedthat predictionof international outcomes is significantlyimproved by understand- inginternal bargaining, especially with respect to minimallyacceptable com- promises.24 Metaphorsare not theories,but I am comfortedby Max Black's obser- vationthat "perhaps every science must start with metaphor and end with algebra;and perhapswithout the metaphorthere would never have been any algebra."25Formal analysis of any game requireswell-defined rules, choices,payoffs, players, and information, and even then, many simple two- person,mixed-motive games have no determinatesolution. Deriving analytic solutionsfor two-level games will be a difficultchallenge. In whatfollows I hope to motivatefurther work on thatproblem.

Towards a theoryof ratification: the importanceof "win-sets"

Considerthe following stylized scenario that might apply to any two-level game. Negotiatorsrepresenting two organizationsmeet to reachan agreementbetween them, subject to theconstraint that any tentative agreement mustbe ratifiedby theirrespective organizations. The negotiatorsmight be heads of governmentrepresenting nations, for example, or laborand man- agementrepresentatives, or partyleaders in a multipartycoalition, or a financeminister negotiating with an IMF team,or leaders of a House-Senate conferencecommittee, or ethnic-groupleaders in a consociationaldemoc- racy.For themoment, we shallpresume that each side is representedby a singleleader or "chiefnegotiator," and thatthis individual has no indepen-

23. RobertAxelrod, "The GammaParadigm for Studying the Domestic Influence on Foreign Policy,"prepared for delivery at the 1987Annual Meeting of theInternational Studies Asso- ciation. 24. Glenn H. Snyderand Paul Diesing, ConflictAmong Nations: Bargaining,Decision Mak- ing,and SystemStructure in InternationalCrises (Princeton:Princeton University Press, 1977), pp. 510-25. 25. Max Black,Models and Metaphors(Ithaca, N.Y.: CornellUniversity Press, 1962),p. 242,as citedin DuncanSnidal, "The Game Theoryof InternationalPolitics," World Politics 38 (October1985), p. 36n.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 436 InternationalOrganization dentpolicy preferences, but seeks simplyto achievean agreementthat will be attractiveto his constituents.26 It is convenientanalytically to decomposethe process into two stages: 1. bargainingbetween the negotiators, leading to a tentativeagreement; call thatLevel I. 2. separatediscussions within each groupof constituentsabout whether to ratifythe agreement; call thatLevel II. This sequentialdecomposition into a negotiationphase and a ratification phase is usefulfor purposes of exposition,although it is notdescriptively accurate.In practice,expectational effects will be quite important.There are likelyto be priorconsultations and bargainingat Level II to hammer outan initialposition for the Level I negotiations.Conversely, the need for Level II ratificationis certainto affectthe Level I bargaining.In fact,ex- pectationsof rejection at Level II mayabort negotiations at Level I without anyformal action at Level II. For example,even though both the American and Iraniangovernments seem to have favoredan arms-for-hostagesdeal, negotiationscollapsed as soon as theybecame public and thusliable to de facto "ratification."In manynegotiations, the two-levelprocess may be iterative,as the negotiatorstry out possibleagreements and probe their constituents'views. In morecomplicated cases, as we shall see later,the constituents'views may themselves evolve in the course of the negotiations. Nevertheless,the requirement that any Level I agreementmust, in theend, be ratifiedat Level II imposesa crucialtheoretical link between the two levels. "Ratification"may entail a formalvoting procedure at Level II, suchas theconstitutionally required two-thirds vote of the U.S. Senatefor ratifying treaties,but I use the termgenerically to referto any decision-processat Level II thatis requiredto endorseor implementa Level I agreement, whetherformally or informally.It is sometimesconvenient to thinkof rat- ificationas a parliamentaryfunction, but that is notessential. The actorsat Level II mayrepresent bureaucratic agencies, interest groups, social classes, or even "publicopinion." For example,if labor unions in a debtorcountry withholdnecessary cooperation from an austerityprogram that the govern- menthas negotiatedwith the IMF, Level II ratificationof the agreement maybe said to have failed;ex ante expectationsabout thatprospect will surelyinfluence the Level I negotiationsbetween the government and the IMF. Domesticratification of internationalagreements might seem peculiar to democracies.As theGerman Finance Minister recently observed, "The limit of expandedcooperation lies in thefact that we are democracies,and we 26. To avoidunnecessary complexity, my argument throughout is phrased in terms of a single chiefnegotiator, although in manycases someof his responsibilities may be delegatedto aides. Laterin thisarticle I relaxthe assumption that the negotiator has no independentpreferences.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 437 need to secureelectoral majorities at home."27However, ratification need notbe "democratic"in any normalsense. For example,in 1930the Meiji Constitutionwas interpretedas givinga specialrole to theJapanese military in theratification of theLondon Naval Treaty;28and duringthe ratification of any agreementbetween Catholics and Protestantsin NorthernIreland, presumablythe IRA wouldthrow its power onto the scales. We need only stipulatethat, for purposes of counting"votes" in theratification process, differentforms of politicalpower can be reducedto somecommon denom- inator. The onlyformal constraint on the ratificationprocess is thatsince the identicalagreement must be ratifiedby bothsides, a preliminaryLevel I agreementcannot be amendedat Level II withoutreopening the Level I negotiations.In otherwords, final ratification must be simply"voted" up or down;any modificationto theLevel I agreementcounts as a rejection, unlessthat modification is approvedby all otherparties to theagreement.29 CongresswomanLynn Martincaptured the logic of ratificationwhen ex- plainingher supportfor the 1986tax reformbill as it emergedfrom the conferencecommittee: "As worriedas I am aboutwhat this bill does, I am evenmore worried about the current code. The choicetoday is notbetween thisbill and a perfectbill; thechoice is betweenthis bill and the deathof tax reform."30 Giventhis set of arrangements,we maydefine the "win-set"for a given Level II constituencyas theset of all possibleLevel I agreementsthat would "win"-that is, gainthe necessary majority among the constituents-when simplyvoted up or down.31For twoquite different reasons, the contours of the Level II win-setsare veryimportant for understanding Level I agreements. First,larger win-sets make Level I agreementmore likely, ceteris paribus.32 By definition,any successfulagreement must fall within the Level II win- 27. GerhardtStoltenberg, Wall Street Journal Europe, 2 October1986, as citedin C. Randall Henning,Macroeconomic Diplomacy in the 1980s: Domestic Politics and InternationalConflict Among the UnitedStates, Japan, and Europe, AtlanticPaper No. 65 (New York: Croom Helm, forthe Atlantic Institute for International Affairs, 1987), p. 1. 28. Ito Takashi,"Conflicts and Coalitionin Japan,1930: Political Groups and theLondon Naval DisarmamentConference," in Sven Groenningset al., eds, The Studyof Coalition Behavior(New York: Holt, Rinehart,& Winston,1970); Kobayashi Tatsuo, "The London Naval Treaty, 1930," in James W. Morley, ed., Japan Erupts: The London Naval Conference and the Manchurian Incident, 1928-1932 (New York: Columbia UniversityPress, 1984), pp. 11-117.I am indebtedto WilliamJarosz for this example. 29. Thisstipulation is, infact, characteristic ofmost real-world ratification procedures, such as House and Senateaction on conferencecommittee reports, although it is somewhatviolated by theoccasional practice of appending "reservations" to theratification of treaties. 30. New York Times, 26 September 1986. 31. For theconception of win-set,see KennethA. Shepsleand BarryR. Weingast,"The InstitutionalFoundations of Committee Power," American Political Science Review 81 (March 1987),pp. 85-104.I am indebtedto ProfessorShepsle for much help on thistopic. 32. To avoid tedium,I do notrepeat the "other things being equal" provisoin each of the propositionsthat follow. Under some circumstances an expandedwin-set might actually make practicablesome outcomethat could triggera dilemmaof collectiveaction. See VincentP. Crawford,"A Theoryof Disagreement in Bargaining,"Econometrica 50 (May 1982),pp. 607- 37.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 438 InternationalOrganization

sets of each of theparties to theaccord. Thus, agreement is possibleonly ifthose win-sets overlap, and thelarger each win-set,the more likely they are to overlap.Conversely, the smaller the win-sets, the greater the risk that thenegotiations will break down. For example,during the prolonged pre- warAnglo-Argentine negotiations over the Falklands/Malvinas, several ten- tativeagreements were rejectedin one capitalor the otherfor domestic politicalreasons; when it became clear that the initial British and Argentine win-setsdid notoverlap at all, warbecame virtually inevitable.33 A brief,but importantdigression: The possibilityof failedratification suggeststhat game theoretical analyses should distinguish between voluntary and involuntarydefection. Voluntary defection refers to renegingby a ra- tionalegoist in the absence of enforceablecontracts-the much-analyzed problemposed, for example, in theprisoner's dilemma and otherdilemmas of collectiveaction. Involuntary defection instead reflects the behavior of an agentwho is unableto deliveron a promisebecause of failed ratification. Even thoughthese two typesof behaviormay be difficultto disentanglein someinstances, the underlying logic is quitedifferent. The prospectsfor international cooperation in an anarchic,"self-help" worldare oftensaid to be poorbecause "unfortunately,policy makers gen- erallyhave an incentiveto cheat."34However, as Axelrod,Keohane, and othershave pointed out, the temptation to defect can be dramaticallyreduced amongplayers who expectto meetagain.35 If policymakersin an anarchic worldwere in factconstantly tempted to cheat,certain features of the 1978 storywould be veryanomalous. For example,even though the Bonn agree- mentwas negotiatedwith exquisite care, it containedno provisionsfor temporalbalance, sequencing, or partialconditionality that might have pro- tectedthe parties from unexpected defection. Moreover, the Germans and theJapanese irretrievably enacted their parts of thebargain more than six monthsbefore the president's action on oil pricedecontrol and nearlytwo yearsbefore that decision was implemented.Once theyhad done so, the temptationto thepresident to renegeshould have been overpowering,but in factvirtually no one on eitherside of the decontroldebate within the administrationdismissed the Bonn pledge as irrelevant.In short,the Bonn "promise"had politicalweight, because renegingwould have had highpo- liticaland diplomaticcosts.

33. The SundayTimes InsightTeam, The FalklandsWar (London: Sphere,1982); Max Hastingsand Simon Jenkins, The Battle for the Falklands (New York: Norton, 1984); Alejandro Dabat and Luis Lorenzano, Argentina: The Malvinas and the End of MilitaryRule (London: Verso,1984). I am indebtedto Louise Richardsonfor these citations. 34. MatthewE. Canzoneriand Jo Anna Gray,"Two Essays on MonetaryPolicy in an InterdependentWorld," International Finance Discussion Paper 219 (Boardof Governorsof theFederal Reserve System, February 1983). 35. RobertAxelrod, The Evolution of Cooperation(New York:Basic Books, 1984);Robert 0. Keohane, AfterHegemony: Cooperation and Discord in the World Political Economy (Princeton:Princeton University Press, 1984),esp. p. 116; and the specialissue of World Politics,"Cooperation Under Anarchy," Kenneth A. Oye, ed., vol. 38 (October1985).

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 439

On the otherhand, in any two-levelgame, the credibilityof an official commitmentmay be low,even if the reputational costs of reneging are high, forthe negotiatormay be unable to guaranteeratification. The failureof Congressto ratifyabolition of the "AmericanSelling Price" as previously agreedduring the Kennedy Round trade negotiations is one classicinstance; anotheris the inabilityof JapanesePrime Minister Sato to deliveron a promisemade to PresidentNixon duringthe "TextileWrangle.' '36 A key obstacleto Westerneconomic coordination in 1985-87was the Germans' fearthat the Reagan administration would be politicallyunable to carryout any commitmentit mightmake to cut the U.S. budgetdeficit, no matter how well-intentionedthe president. Unlikeconcerns about voluntary defection, concern about "deliver-ability" was a prominentelement in theBonn negotiations. In thepost-summit pressconference, President Carter stressed that "each ofus has beencareful not to promisemore than he can deliver."A majorissue throughoutthe negotiationswas Carter'sown ability to deliveron hisenergy commitments. The Americansworked hard to convincethe others, first, that the president was undersevere domesticpolitical constraints on energyissues, which limitedwhat he could promise,but second,that he could deliverwhat he was preparedto promise.The negotiatorsin 1978 seemedto followthis presumptionabout one another:"He willdo whathe has promised,so long as whathe has promisedis clearand withinhis power." Involuntarydefection, and thefear of it, can be just as fatalto prospects forcooperation as voluntarydefection. Moreover, in somecases, it maybe difficult,both forthe otherside and foroutside analysts, to distinguish voluntaryand involuntary defection, particularly since a strategicnegotiator mightseek to misrepresenta voluntary defection as involuntary.Such be- havioris itselfpresumably subject to somereputational constraints, although it is an importantempirical question how farreputations generalize from collectivitiesto negotiatorsand vice versa.Credibility (and thusthe ability to strikedeals) at Level I is enhancedby a negotiator's (demonstrated) ability to "deliver"at Level II; thiswas a majorstrength of Robert Strauss in the TokyoRound negotiations.37 Involuntarydefection can onlybe understoodwithin the framework of a two-levelgame. Thus, to returnto theissue of win-sets, the smaller the win- sets,the greater the risk of involuntarydefection, and hencethe more ap- plicablethe literature about dilemmas of collectiveaction.38

36. I. M. Destler,Haruhiro Fukui, and Hideo Sato, The TextileWrangle: Conflict in Japanese- American Relations, 1969-1971 (Ithaca, N.Y.: Cornell UniversityPress, 1979), pp. 121-57. 37. GilbertR. Winham,"Robert Strauss, the MTN, and theControl of Faction,"Journal of World Trade Law 14 (September-October 1980), pp. 377-97, and his InternationalTrade and theTokyo Round (Princeton: Princeton University Press, 1986). 38. Thisdiscussion implicitly assumes uncertainty about the contours of the win-sets on the partof theLevel I negotiators,for if the win-sets were known with certainty, the negotiators wouldnever propose for ratification an agreementthat would be rejected.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 440 InternationalOrganization

The secondreason why win-set size is importantis thatthe relative size of the respectiveLevel II win-setswill affectthe distributionof thejoint gains fromthe internationalbargain. The largerthe perceived win-setof a negotiator,the more he can be "pushedaround" by the other Level I negotiators. Conversely,a smalldomestic win-set can be a bargainingadvantage: "I'd like to accept yourproposal, but I could neverget it acceptedat home." Lamentingthe domestic constraints under which one mustoperate is (inthe wordsof one experiencedBritish diplomat) "the naturalthing to say at the beginningof a toughnegotiation." 39 This generalprinciple was, of course,first noted by ThomasSchelling nearlythirty years ago: The powerof a negotiatoroften rests on a manifestinability to make concessions and meet demands.... When the United States Govern- mentnegotiates with othergoverments . . . if the executive branch ne- gotiatesunder legislative authority, with its position constrained by law, . . . thenthe executive branch has a firmposition that is visibleto its negotiatingpartners.... [Ofcourse, strategies such as this]run the riskof establishingan immovableposition that goes beyondthe ability of theother to concede,and therebyprovoke the likelihood of stale- mateor breakdown.40 Writingfrom a strategist'spoint of view, Schelling stressed ways in which win-setsmay be manipulated,but even whenthe win-setitself is beyond thenegotiator's control, he mayexploit its leverage.A ThirdWorld leader whose domesticposition is relativelyweak (Argentina'sAlfonsin?) should be able to drivea betterbargain with his international creditors, other things beingequal, thanone whosedomestic standing is moresolid (Mexico's de la Madrid?).41The difficultiesofwinning congressional ratification are often exploitedby Americannegotiators. During the negotiationof the Panama CanalTreaty, for example, "the Secretaryof State warned the Panamanians severaltimes .. . thatthe new treaty would have to be acceptableto at least sixty-sevensenators," and "Carter,in a personalletter to Torrijos,warned thatfurther concessions by the United States would seriously threaten chances forSenate ratification.'"42 Precisely to forestallsuch tactics, opponents may demandthat a negotiatorensure himself "negotiating room" at Level II beforeopening the Level I negotiations. The "sweet-and-sour"implications of win-setsize are summarizedin Figure1, representinga simple zero-sum game between X and Y. XMand

39. GeoffreyW. Harrison, in JohnC. Campbell, ed., Successful Negotiation: Trieste1954 (Princeton:Princeton University Press, 1976), p. 62. 40. ThomasC. Schelling,The Strategyof Conflict(Cambridge, Mass.: HarvardUniversity Press,1960), pp. 19-28. 41. I am gratefulto Lara Putnamfor this example. For supportingevidence, see RobertR. Kaufman,"Democratic and AuthoritarianResponses to the Debt Issue: Argentina,Brazil, Mexico," InternationalOrganization 39 (Summer 1985), pp. 473-503. 42. W. MarkHabeeb and I. WilliamZartman, The Panama CanalNegotiations (Washington, D.C.: JohnsHopkins Foreign Policy Institute, 1986), pp. 40, 42.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 441

xm r r 'I r IY

YX Y2 XI Y3

FIGURE 1. Effectsof reducingwin-set size YM representthe maximum outcomes for X and Y, respectively,while X1 and Y1 representthe minimal outcomes that could be ratified.At thisstage any agreementin the rangebetween X1 and Y1 could be ratifiedby both parties.If the win-set of Y werecontracted to, say,Y2 (perhapsby requiring a largermajority for ratification), outcomes between Y1 and Y2 wouldno longerbe feasible,and therange of feasible agreements would thus be trun- cated in Y's favor.However, if Y, emboldenedby thissuccess, were to reduceits win-setstill further to Y3 (perhapsby requiringunanimity for ratification),the negotiatorswould suddenlyfind themselves deadlocked, forthe win-sets no longeroverlap at all.43

Determinantsof the win-set

It is importantto understandwhat circumstances affect win-set size. Three setsof factorsare especiallyimportant:

43. Severalinvestigators in otherfields have recently proposed models of linked games akin to this"two-level" game. KennethA. Shepsleand his colleagueshave used the notionof "interconnectedgames" to analyze,for example, the strategy of a legislatorsimultaneously embeddedin twogames, one in thelegislative arena and theother in theelectoral arena. In thismodel, a givenaction is simultaneouslya move in two differentgames, and one player maximizesthe sum of his payoffs from the two games. See ArthurDenzau, William Riker, and KennethShepsle, "Farquharson and Fenno: Sophisticated Voting and Home Style," American Political Science Review 79 (December 1985),pp. 1117-34; and KennethShepsle, "Cooperation and InstitutionalArrangements," unpublished manuscript, February 1986. This approachis similarto modelsrecently developed by economistsworking in the "rationalexpectations" genre.In thesemodels, a governmentcontends simultaneously against other governments and againstdomestic trade unions over monetary policy. See, forexample, Kenneth Rogoff, "Can InternationalMonetary Policy Cooperation be Counterproductive,"Journal of International Economics18 (May 1985),pp. 199-217,and RolandVaubel, "A PublicChoice Approach to InternationalOrganization," Public Choice51 (1986),pp. 39-57. GeorgeTsebelis ("Nested Games: The Cohesion of French Coalitions," BritishJournal of Political Science 18 [April 1988],pp. 145-70)has developeda theoryof "nestedgames," in whichtwo alliances play a competitivegame to determinetotal payoffs, while the individual players within each alliance contendover their shares. Fritz Sharpf ("A Game-TheoreticalInterpretation of Inflation and Unemploymentin Western Europe," Journal of Public Policy 7 [19881,pp. 227-257)interprets macroeconomicpolicy as thejoint outcome of two simultaneous games; in one, the government playsagainst the unions,while in theother, it respondsto theanticipated reactions of the electorate.James E. Alt and BarryEichengreen ("Parallel and OverlappingGames: Theory and an Applicationto the EuropeanGas Trade," unpublishedmanuscript, November 1987) offera broadertypology of linkedgames, distinguishing between "parallel" games, in which "the sameopponents play against one anotherat thesame time in morethan one arena,"and "overlapping"games, which arise "whena particularplayer is engagedat thesame timein gamesagainst distinct opponents, and whenthe strategy pursued in one gamelimits the strategies availablein theother." Detailed comparison of thesevarious linked-game models is a taskfor the future.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 442 InternationalOrganization

* Level II preferencesand coalitions * Level IL institutions * Level I negotiators'strategies Let us considereach in turn.

1. The size of the win-setdepends on the distributionof power, preferences,and possible coalitions among Level II constituents. Anytestable two-level theory of international negotiation must be rooted in a theoryof domesticpolitics, that is, a theoryabout the power and preferencesof the major actors at Level II. Thisis notthe occasion for even a cursoryevaluation of the relevant alternatives, except to notethat the two- level conceptualframework could in principlebe marriedto such diverse perspectivesas Marxism,interest group pluralism, bureaucratic politics, and neo-corporatism.For example,arms negotiations might be interpretedin termsof a bureaucraticpolitics model of Level II politicking,while class analysisor neo-corporatismmight be appropriatefor analyzing international macroeconomiccoordination. Abstractingfrom the detailsof Level II politics,however, it is possible to sketchcertain principles that govern the size ofthe win-sets. For example, thelower the cost of "no-agreement"to constituents,the smaller the win- set.44Recall thatratification pits the proposedagreement, not againstan arrayof other(possibly attractive) alternatives, but only against "no-agreement."145 No-agreementoften represents the statusquo, althoughin some cases no-agreementmay in factlead to a worseningsituation; that might be a reasonabledescription of thefailed ratification of theVersailles Treaty. Someconstituents may face low costs from no-agreement, and others high costs,and theformer will be moreskeptical of Level I agreementsthan the latter.Members of two-wage-earner families should be readierto strike,for example,than sole breadwinners,and small-townbarbers should be more isolationistthan international bankers. In thissense, some constituents may offereither generic opposition to, or generic support for, Level I agreements, moreor less independentlyofthe specific content of the agreement, although naturallyother constituents' decisions about ratification will be closelycon- ditionedon the specifics.The size of thewin-set (and thusthe negotiating

44. ThomasRomer and Howard Rosenthal, "Political Resource Allocation, Controlled Agen- das, and theStatus Quo," PublicChoice 33 (no. 4, 1978),pp. 27-44. 45. In moreformal treatments, the no-agreement outcome is calledthe "reversion point." A givenconstituent's evaluation of no-agreementcorresponds to whatRaiffa terms a seller's "walk-awayprice," that is, theprice below which he wouldprefer "no-deal." (Raiffa, Art and Scienceof Negotiation.) No-agreement is equivalent to whatSnyder and Diesing term "breakdown,"or theexpected cost ofwar. (Snyder and Diesing,Conflict Among Nations.)

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 443 roomof the Level I negotiator)depends on therelative size of the "isola- tionist"forces (who oppose internationalcooperation in general)and the "internationalists"(who offer"all-purpose" support). All-purpose support forinternational agreements is probablygreater in smaller,more dependent countrieswith more open economies,as comparedto moreself-sufficient countries,like the United States, for most of whose citizens the costs of no- agreementare generallylower. Ceteris paribus, more self-sufficient states withsmaller win-sets should make fewer international agreements and drive harderbargains in thosethat they do make. In some cases, evaluationof no-agreementmay be the onlysignificant disagreementamong the Level II constituents,because theirinterests are relativelyhomogeneous. For example,if oil importsare to be limitedby an agreementamong the consuming nations-the sort of accordsought at the Tokyo summitof 1979,for example-then presumably every constituent wouldprefer to maximizehis nation's share of the available supply, although someconstituents may be morereluctant than others to pushtoo hard,for fearof losingthe agreement entirely. Similarly, in mostwage negotiations, theinterests of constituents(either workers or shareholders)are relatively homogeneous,and themost significant cleavage within the Level II consti- tuenciesis likelyto be between"hawks" and "doves," dependingon their willingnessto riska strike.(Walton and McKersierefer to theseas "boundary" conflicts,in whichthe negotiator is caughtbetween his constituency and the externalorganization.) Other international examples in whichdo- mesticinterests are relativelyhomogeneous except for the evaluationof no-agreementmight include the SALT talks,the Panama Canal Treatyne- gotiations,and the Arab-Israeliconflict. A negotiatoris unlikelyto face criticismat homethat a proposedagreement reduces the opponents' arms too much,offers too littlecompensation for foreign concessions, or contains toofew security guarantees for the other side, although in each case opinions maydiffer on how muchto riska negotiatingdeadlock in orderto achieve theseobjectives. The distinctivenature of such "homogeneous" issues is throwninto sharp reliefby contrastingthem to cases in whichconstituents' preferences are moreheterogeneous, so thatany Level I agreementbears unevenly on them. Thus,an internationallycoordinated reflation may encounter domestic op- positionboth from those who thinkit goes too far(bankers, for example) and fromthose who thinkit does notgo farenough (unions, for example). In 1919,some Americans opposed the Versailles Treaty because it was too harshon thedefeated powers and othersbecause it was too lenient.46Such patternsare even morecommon, as we shallshortly see, wherethe negotiationinvolves multiple issues, such as an armsagreement that involves tradeoffsbetween seaborne and airborne weapons, or a laboragreement that 46. Thomas A. Bailey, Woodrow Wilson and the Great Betrayal (New York: Macmillan, 1945), pp. 16-37.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 444 InternationalOrganization involvestradeoffs between take-home pay and pensions. (Walton and McKersie termthese "factional" conflicts, because thenegotiator is caughtbetween contendingfactions within his own organization.) The problemsfacing Level I negotiatorsdealing with a homogeneous(or "boundary")conflict are quite different from those facing negotiators dealing witha heterogeneous(or "factional")conflict. In theformer case, themore thenegotiator can winat Level I-the higherhis national oil allocation,the deeperthe cuts in Sovietthrow-weight, the lower the rent he promisesfor theCanal, and so one-the betterhis odds of winningratification. In such cases, the negotiatormay use the implicitthreat from his own hawksto maximizehis gains(or minimizehis losses) at Level I, as Carterand Vance did in dealingwith the Panamanians. Glancing over his shoulderat Level II, the negotiator'smain problem in a homogeneousconflict is to manage thediscrepancy between his constituents'expectations and thenegotiable outcome.Neither negotiator is likelyto findmuch sympathy for the enemy's demandsamong his own constituents, nor much support for his constituents' positionsin theenemy camp. The effectof domesticdivision, embodied in hard-lineopposition from hawks, is to raisethe risk of involuntary defection and thusto impedeagreement at Level I. The commonbelief that domestic politicsis inimicalto internationalcooperation no doubtderives from such cases. The task of a negotiatorgrappling instead with a heterogeneousconflict is morecomplicated, but potentially more interesting. Seeking to maximize thechances of ratification, he cannotfollow a simple"the more, the better" ruleof thumb;imposing more severe reparations on theGermans in 1919 wouldhave gained some votes at Level II butlost others, as wouldhastening thedecontrol of domesticoil pricesin 1978.In some cases, theselines of cleavagewithin the Level II constituencieswill cut across the Level I di- vision,and the Level I negotiatormay find silent allies at his opponent's domestictable. Germanlabor unionsmight welcome foreign pressure on theirown governmentto adopta moreexpansive fiscal policy, and Italian bankersmight welcome international demands for a moreaustere Italian monetarypolicy. Thus transnationalalignments may emerge,tacit or ex- plicit,in whichdomestic interests pressure their respective governments to adoptmutually supportive policies. This is, of course,my interpretation of the 1978Bonn summitaccord. In suchcases, domesticdivisions may actually improve the prospects for internationalcooperation. For example,consider two different distributions of constituents'preferences as betweenthree alternatives: A, B, and no- agreement.If 45 percentof the constituents rank these A > no-agreement> B, 45 percentrank them B > no-agreement> A, and 10 percentrank them B > A > no-agreement,then both A and B are in thewin-set, even though B wouldwin in a simpleLevel-Il-only game. On theother hand, if 90 percent rankthe alternativesA > no-agreement> B, while 10 percentstill rank themB > A > no-agreement,then only A is in thewin-set. In thissense,

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 445

a governmentthat is internallydivided is morelikely to be able to strikea deal internationallythan one thatis firmlycommitted to a singlepolicy.47 Conversely,to imposebinding ex ante instructionson the negotiatorsin such a case mightexclude some Level I outcomesthat would, in fact,be ratifiablein bothnations.48 Thus far we have implicitlyassumed that all eligibleconstituents will participatein theratification process. In fact,however, participation rates varyacross groupsand across issues,and thisvariation often has implica- tionsfor the size ofthe win-set. For example,when the costs and/or benefits of a proposedagreement are relativelyconcentrated, it is reasonableto expectthat those constituents whose interests are mostaffected will exert specialinfluence on theratification process.49 One reason why Level II games are moreimportant for trade negotiations than in monetarymatters is that the "abstentionrate" is higheron internationalmonetary issues thanon trade issues.50 The compositionof the active Level II constituency(and hencethe char- acterof the win-set)also varieswith the politicization of theissue. Politi- cizationoften activates groups who are less worriedabout the costs of no- agreement,thus reducing the effective win-set. For example,politicization ofthe Panama Canal issue seemsto have reducedthe negotiating flexibility on bothsides of the diplomatic table.51 This is one reasonwhy most profes- sionaldiplomats emphasize the value of secrecyto successfulnegotiations. However,Woodrow Wilson's transcontinentaltour in 1919 reflectedthe oppositecalculation, namely, that by expandingthe active constituency he could ensureratification of the VersaillesTreaty, although in theend this strategyproved fruitless.52 Anotherimportant restriction of our discussionthus far has been the

47. Raiffanotes that "the morediffuse the positions are withineach side,the easier it might be to achieveexternal agreement." (Raiffa, Art and Scienceof Negotiation,p. 12.) For the conventionalview, by contrast, that domestic unity is generallya precondition for international agreement,see Michael Artisand Sylvia Ostry,International Economic Policy Coordination, ChathamHouse Papers:30 (London:Routledge & KeganPaul, 1986),pp. 75-76. 48. "Meaningfulconsultation with other nations becomes very difficult when the internal processof decision-making already has someof the characteristics ofcompacts between quasi- sovereignentities. There is an increasingreluctance to hazarda hard-wondomestic consensus in an internationalforum." Henry A. Kissinger,"Domestic Structure and ForeignPolicy," in James N. Rosenau, ed., InternationalPolitics and Foreign Policy (New York: Free Press, 1969),p. 266. 49. See JamesQ. Wilson,Political Organization (New York:Basic Books,1975) on howthe politicsof an issue are affectedby whetherthe costs and the benefitsare concentratedor diffuse. 50. Anotherfactor fostering abstention is thegreater complexity and opacityof monetary issues;as GilbertR. Winham("Complexity in International Negotiation," in Daniel Druckman, ed., Negotiations: A Social-Psychological Perspective [Beverly Hills: Sage, 1977], p. 363) observes,"complexity can strengthenthe hand of a negotiatorvis-a-vis the organization he represents." 51. Habeeb and Zartman,Panama Canal Negotiations. 52. Bailey, Wilson and the Great Betrayal.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 446 InternationalOrganization assumptionthat the negotiationsinvolve only one issue. Relaxingthis as- sumptionhas powerfulconsequences for the play at bothlevels.53 Various groupsat Level II arelikely to have quite different preferences on theseveral issues involvedin a multi-issuenegotiation. As a generalrule, the group withthe greatest interest in a specificissue is also likelyto holdthe most extremeposition on thatissue. In the Law of the Sea negotiations,for example,the DefenseDepartment felt most strongly about sea-lanes,the Departmentof the Interior about sea-bed mining rights, and so on.54If each groupis allowedto fixthe Level I negotiatingposition for "its" issue, the resultingpackage is almostsure to be "non-negotiable"(that is, non-ratifiablein opposingcapitals).55 Thus,the chief negotiator is facedwith tradeoffs across different issues: howmuch to yieldon miningrights in orderto getsea-lane protection, how muchto yieldon citrusexports to geta betterdeal on beef,and so on. The implicationof thesetradeoffs for the respective win-sets can be analyzedin termsof iso-vote or "politicalindifference" curves. This technique is analogous to conventionalindifference curve analysis, except that the opera- tionalmeasure is voteloss, notutility loss. Figure2 providesan illustrative Edgeworthbox analysis.56The most-preferredoutcome for A (theoutcome whichwins unanimous approval from both the beef industry and thecitrus industry)is the upperright-hand corner (AM), and each curveconcave to pointAM represents the locus ofall possibletradeoffs between the interests ofranchers and farmers, such that the net vote in favorof ratification at A's Level II is constant.The bold contourA1-A2 represents the minimal vote necessaryfor ratification by A, and the wedge-shapedarea northeastof A1-A2represents A's win-set.Similarly, B1-B2 represents the outcomes that are minimallyratifiable by B, and thelens-shaped area betweenA1-A2 and B1-B2represents the set of feasibleagreements. Although additional sub- tleties(such as thenature of the "contract curve") mightbe extractedfrom thissort of analysis,the central point is simple:the possibility of package deals opensup a richarray of strategicalternatives for negotiators in a two- levelgame. One kindof issue linkageis absolutelycrucial to understandinghow domesticand internationalpolitics can become entangled.57Suppose thata majorityof constituentsat Level II oppose a givenpolicy (say, oil price

53. I am gratefulto ErnstB. Haas and Robert0. Keohanefor helpful advice on thispoint. 54. Ann L. Hollick, U.S. Foreign Policy and the Law of the Sea (Princeton: Princeton UniversityPress, 1981), especially pp. 208-37,and JamesK. Sebenius,Negotiating the Law of theSea (Cambridge,Mass.: HarvardUniversity Press, 1984), especially pp. 74-78. 55. Raiffa,Art and Science of Negotiation, p. 175. 56. I am indebtedto Lisa Martinand KennethShepsle for suggesting this approach, although theyare notresponsible for my application of it. Note that this construction assumes that each issue,taken individually, is a "homogeneous"type, not a "heterogeneous"type. Constructing iso-votecurves for heterogeneous-type issues is morecomplicated. 57. I am gratefulto HenryBrady for clarifying this point for me.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 447

A1 Am B1

BM B2

FIGURE 2. Political indifferencecurves for two-issuenegotiation decontrol),but that some members of that majority would be willingto switch theirvote on thatissue in returnfor more jobs (say, in exportindustries). If bargainingis limitedto Level II, thattradeoff is nottechnically feasible, butif the chief negotiator can brokeran internationaldeal thatdelivers more jobs (say, via fastergrowth abroad), he can, in effect,overturn the initial outcomeat the domestictable. Such a transnationalissue linkagewas a crucialelement in the 1978Bonn accord. Note thatthis strategyworks not by changingthe preferencesof any domesticconstituents, but rather by creatinga policyoption (such as faster exportgrowth) that was previouslybeyond domestic control. Hence, I refer to thistype of issue linkageat Level I thatalters the feasible outcomes at Level II as synergisticlinkage. For example, "in the Tokyo Round . .. nationsused negotiationto achieveinternal reform in situationswhere con- stituencypressures would otherwiseprevent action without the pressure

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 448 InternationalOrganization

(and tradeoffbenefits) that an externalpartner could provide.'" 58 Economic interdependencemultiplies the opportunities for altering domestic coalitions (and thuspolicy outcomes) by expandingthe set of feasiblealternatives in thisway-in effect,creating political entanglements across national bound- aries. Thus, we shouldexpect synergistic linkage (which is, by definition, explicableonly in termsof two-levelanalysis) to becomemore frequent as interdependencegrows.

2. The size of the win-setdepends on the Level II political institutions. Ratificationprocedures clearly affect the size ofthe win-set. For example, ifa two-thirdsvote is requiredfor ratification, the win-set will almost cer- tainlybe smallerthan if only a simplemajority is required.As one experi- encedobserver has written:"Under the Constitution, thirty-four ofthe one hundredsenators can block ratificationof any treaty.This is an unhappy and uniquefeature of our democracy.Because of theeffective veto power of a smallgroup, many worthy agreements have been rejected,and many treatiesare neverconsidered for ratification."59 As notedearlier, the U.S. separationof powersimposes a tighterconstraint on theAmerican win-set thanis truein manyother countries. This increasesthe bargaining power of Americannegotiators, but it also reducesthe scope forinternational cooperation.It raisesthe odds for involuntary defection and makes potential partnerswarier about dealing with the Americans. The Trade ExpansionAct of 1974modified U.S. ratificationprocedures in an effortto reducethe likelihood of congressional tampering with the final deal and henceto reassureAmerica's negotiating partners. After the Amer- ican SellingPrice fiasco, it was widelyrecognized that piecemeal congres- sionalratification of any new agreementwould inhibit international negotiation.Hence, the 1974 Act guaranteed a straight up-or-down vote in Congress. However,to satisfycongressional sensitivities, an elaboratesystem of pri- vate-sectorcommittees was establishedto improvecommunication between the Level I negotiatorsand theirLevel II constituents,in effectcoopting the interestgroups by exposingthem directly to the implicationsof their demands.60Precisely this tactic is describedin thelabor-management case byWalton and McKersie:"Instead of taking responsibility for directly per- suadingthe principals [Level II constituents]to reducetheir expectations, [theLevel I negotiator]structures the situation so thatthey (or theirmore immediaterepresentatives) will persuade themselves."61

58. GilbertR. Winham,"The Relevanceof Clausewitzto a Theoryof InternationalNego- tiation,"prepared for delivery at the 1987annual meeting of theAmerican Political Science Association. 59. JimmyCarter, Keeping Faith: Memoirs of a President(New York: Bantam Books, 1982), p. 225. 60. Winham(see note37); Twiggs,The TokyoRound. 61. Walton and McKersie, Behavioral Theoryof Labor Organizations,p. 321.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 449

Not all significantratification practices are formalized;for example, the Japanesepropensity for seeking the broadest possible domestic consensus beforeacting constricts the Japanese win-set, as contrastedwith majoritarian politicalcultures. Other domestic political practices, too, can affectthe size of the win-set.Strong discipline within the governingparty, for example, increasesthe win-setby wideningthe rangeof agreementsfor which the Level I negotiatorcan expectto receivebacking. For example,in the 1986 House-Senateconference committee on tax reform,the final bill was closer to theSenate version, despite (or rather,because of) CongressmanRosten- kowski'sgreater control of his delegation,which increased the House win- set. Conversely,a weakeningof partydiscipline across the major Western nationswould, ceteris paribus, reduce the scope forinternational cooperation. The recentdiscussion of "state strength"and "state autonomy"is rel- evanthere. The greaterthe autonomy of central decision-makers from their Level II constituents,the larger their win-set and thusthe greater the like- lihoodof achievinginternational agreement. For example,central bank in- sulationfrom domestic political pressures in effectincreases the win-set and thusthe odds forinternational monetary cooperation; recent proposals for an enhancedrole forcentral bankers in internationalpolicy coordination reston thispoint.62 However, two-level analysis also impliesthat, ceteris paribus,the strongera stateis in termsof autonomyfrom domestic pressures,the weakerits relativebargaining position internationally. For example,diplomats representing an entrencheddictatorship are less able than representativesof a democracyto claimcredibly that domestic pressures precludesome disadvantageousdeal.63 This is yetanother facet of thedis- concertingambiguity of thenotion of "state strength." For simplicityof exposition, my argument is phrasedthroughout in terms ofonly two levels. However, many institutional arrangements require several levelsof ratification,thus multiplying the complexity (but perhaps also the importance)of win-set analysis. Consider, for example, negotiations between the UnitedStates and the EuropeanCommunity over agriculturaltrade. Accordingto theTreaty of Rome, modifications ofthe Common Agricultural Policyrequire unanimous ratification by the Councilof Ministers,repre- sentingeach ofthe member states. In turn,each ofthose governments must, in effect,win ratification for its decisionwithin its own nationalarena, and in coalitiongovernments, that process might also requireratification within each of theparties. Similarly, on theAmerican side, ratification would (informally,at least) necessitatesupport from most, if not all, of the major agriculturalorganizations, and within those organizations, further ratification by key interestsand regionsmight be required.At each stage,cleavage patterns,issue linkages,ratification procedures, side-payments, negotiator

62. Artis and Ostry,International Economic Policy Coordination. Of course, whetherthis is desirablein termsof democraticvalues is quiteanother matter. 63. Schelling,Strategy of Conflict,p. 28.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 450 InternationalOrganization

strategies,and so on wouldneed to be considered.At some pointin this analyticregress the complexityof furtherdecomposition would outweigh theadvantages, but the example illustrates the need for careful thought about thelogic of multiple-levelgames.

3. The size of the win-setdepends on the strategiesof the Level I negotiators. Each Level I negotiatorhas an unequivocalinterest in maximizingthe otherside's win-set,but withrespect to his own win-set,his motivesare mixed.The largerhis win-set, the more easily he can concludean agreement, but also the weakerhis bargainingposition vis-a-vis the othernegotiator. Thisfact often poses a tacticaldilemma. For example,one effectiveway to demonstratecommitment to a givenposition in Level I bargainingis to rally supportfrom one's constituents(for example, holding a strikevote, talking abouta "missilegap," or denouncing"unfair trading practices" abroad). On theother hand, such tactics may have irreversible effects on constituents' attitudes,hampering subsequent ratification of a compromiseagreement.f4 Conversely,preliminary consultations at home,aimed at "softeningup" one's constituentsin anticipationof a ratificationstruggle, can undercuta negotiator'sability to projectan implacableimage abroad. Nevertheless,disregarding these dilemmas for the moment and assuming thata negotiatorwishes to expandhis win-setin orderto encourageratifi- cationof an agreement,he may exploitboth conventional side-payments and generic"good will." The use of side-paymentsto attractmarginal sup- portersis, of course,quite familiar in gametheory, as well as in practical politics.For example,the CarterWhite House offeredmany inducements (suchas publicworks projects) to helppersuade wavering Senators to ratify the Panama Canal Treaty.65In a two-levelgame the side-paymentsmay come fromunrelated domestic sources, as in this case, or theymay be receivedas partof theinternational negotiation. The role of side-paymentsin internationalnegotiations is well known. However,the two-level approach emphasizes that the value of an interna- tionalside-payment should be calculatedin terms of its marginal contribution to thelikelihood of ratification,rather than in termsof its overallvalue to therecipient nation. What counts at Level II is nottotal national costs and benefits, but their incidence, relative to existing coalitions and proto- coalitions.An across-the-board trade concession (or still worse, a concession on a productof interestto a committedfree-trade congressman) is less effectivethan a concession(even one oflesser intrinsic value) thattips the balancewith a swingvoter. Conversely, trade retaliation should be targeted,

64. Walton and McKersie, Behavioral Theoryof Labor Organizations,p. 345. 65. Carter,Keeping Faith, p. 172. See also Raiffa,Art and Science of Negotiation, p. 183.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 451 neitherat free-tradersnor at confirmedprotectionists, but at the uncom- mitted. An experiencednegotiator familiar with the respectivedomestic tables shouldbe able to maximizethe cost-effectiveness(to himand his constituents)of the concessionsthat he mustmake to ensureratification abroad, as wellas thecost-effectiveness ofhis own demands and threats, by targeting his initiativeswith an eye to theirLevel II incidence,both at home and abroad. In thisendeavor Level I negotiatorsare oftenin collusion,since each has an interestin helpingthe otherto get the finaldeal ratified.In effect,they are movingjointly towards points of tangencybetween their respectivepolitical indifference curves. The empiricalfrequency of such targetingintrade negotiations and trade wars, as wellas inother international negotiations,would be a crucialtest of therelative merits of conventional unitary-actoranalysis and thetwo-level approach proposed here.66 In additionto theuse of specificside-payments, a chief negotiator whose politicalstanding at home is highcan moreeasily win ratificationof his foreigninitiatives. Although generic good will cannot guarantee ratification, as WoodrowWilson discovered, it is usefulin expandingthe win-setand thusfostering Level I agreement,for it constitutesa kindof "all-purpose glue" forhis supportingcoalition. Walton and McKersiecite membersof the UnitedAuto Workerswho, speakingof theirrevered leader, Walter Reuther,said, "I don'tunderstand or agreewith this profit-sharing idea, but ifthe Red Head wantsit, I willgo along."67 The Yugoslavnegotiator in the Triestedispute later discounted the difficulty ofpersuading irredentist Slov- enes to acceptthe agreement, since "the government [i.e., Tito]can always influencepublic opinion if it wantsto."68 Note thateach Level I negotiatorhas a stronginterest in thepopularity ofhis oppositenumber, since Party A's popularityincreases the size ofhis win-set,and thusincreases both the odds of success and therelative bar- gainingleverage of Party B. Thus,negotiators should normally be expected to tryto reinforceone another'sstanding with their respective constituents.

66. The strategicsignificance of targetingat Level II is illustratedin John Conybeare, "Trade Wars:A ComparativeStudy of Anglo-Hanse,Franco-Italian, and Hawley-SmootConflicts," WorldPolitics 38 (October1985), p. 157: Retaliationin theAnglo-Hanse trade wars did not havethe intended deterrent effect, because it was not (and perhaps could not have been) targeted atthe crucial members of the opposing Level II coalition.Compare Snyder and Diesing, Conflict AmongNations, p. 552:"If onefaces a coerciveopponent, but the opponent's majority coalition includesa fewwavering members inclined to compromise,a compromiseproposal that suits theirviews may cause theirdefection and theformation of a differentmajority coalition. Or if theopponent's strategy is accommodative,based on a tenuoussoft-line coalition, one knows thatcare is requiredin implementingone's own coercivestretegy to avoidthe opposite kind of shiftin theother state." 67. Walton and McKersie, Behavioral Theoryof Labor Negotiations, p. 319. 68. VladimirVelebit, in Campbell,Trieste 1954, p. 97. As notedearlier, our discussion here assumesthat the Level I negotiatorwishes to reacha ratifiableagreement; in cases (alluded to later)when the negotiator's own preferencesare morehard-line than his constituents,his domesticpopularity might allow him to resistLevel I agreements.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 452 InternationalOrganization

Partlyfor this reason and partlybecause of media attention, participation on theworld stage normally gives a headof government a special advantage vis-a-vishis or herdomestic opposition. Thus, although international policy coordinationis hamperedby hightransaction costs, heads of government mayalso reapwhat we mightterm "transaction benefits." Indeed, the recent evolutionof Westernsummitry, which has placedgreater emphasis on pub- licitythan on substance,seems designed to appropriatethese "transaction benefits"without actually seeking the sort of agreementsthat might entail transactioncosts.69 Higherstatus negotiators are likelyto disposeof more side-payments and more"good will" at home,and henceforeigners prefer to negotiatewith a head ofgovernment than with a lowerofficial. In purelydistributive terms, a nationmight have a bargainingadvantage if its chiefnegotiator were a mereclerk. Diplomats are actingrationally, not merely symbolically, when theyrefuse to negotiatewith a counterpartof inferiorrank. America's negotiatingpartners have reason for concern whenever the American president is domesticallyweakened.

Uncertaintyand bargainingtactics

Level I negotiatorsare oftenbadly misinformedabout Level II politics, particularlyon theopposing side. In 1978,the Bonn negotiators were usually wrongin theirassessments of domesticpolitics abroad; for example, most Americanofficials did not appreciate the complex domestic game that Chan- cellorSchmidt was playingover the issue of Germanreflation. Similarly, Snyderand Diesingreport that "decision makers in our cases onlyocca- sionallyattempted such assessments,and whenthey tried they did pretty miserably.... Governmentsgenerally do not do well in analyzingeach other'sinternal politics in crises[and, I wouldadd, in normaltimes], and indeedit is inherentlydifficult. "70 Relaxingthe assumption of perfect infor- mationto allowfor uncertainty has manyimplications for our understanding of two-levelgames. Let me illustratea fewof theseimplications. Uncertaintyabout the size of a win-setcan be botha bargainingdevice and a stumblingblock in two-levelnegotiation. In purelydistributive Level I bargaining,negotiators have an incentiveto understatetheir own win-sets. Since each negotiatoris likelyto knowmore about his own Level II than his opponentdoes, theclaim has someplausibility. This is akinto a tactic

69. Transactionbenefits may be enhancedif a substantiveagreement is reached,although sometimesleaders can benefitdomestically by loudlyrejecting a profferedinternational deal. 70. Snyderand Diesing, Conflict Among Nations, pp. 516,522-23. Analogous mi*perceptions in Anglo-Americandiplomacy are thefocus of RichardE. Neustadt,Alliance Politics (New York:Columbia University Press, 1970).

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 453 thatSnyder and Diesing describe, when negotiators seek to exploitdivisions withintheir own governmentby saying,in effect,"You'd bettermake a deal withme, because thealternative to me is even worse.""71 On the otherhand, uncertainty about the opponent'swin-set increases one's concernabout the riskof involuntarydefection. Deals can onlybe struckif each negotiatoris convincedthat the proposed deal lies withinhis oppositenumber's win-set and thuswill be ratified.Uncertainty about party A's ratificationlowers the expected value ofthe agreement to partyB, and thusparty B willdemand more generous side-payments from party A than would be needed underconditions of certainty.In fact,party B has an incentiveto feigndoubt about party A's abilityto deliver,precisely in order to extracta moregenerous offer.72 Thus,a utility-maximizingnegotiator must seek to convincehis opposite numberthat his own win-setis "kinky,"that is, thatthe proposed deal is certainto be ratified,but that a deal slightlymore favorable to theopponent is unlikelyto be ratified.For example, on theenergy issue in 1978,by sending SenatorByrd on a personalmission to Bonnbefore the summit and thenby discussinghis politicalproblems in a lengthtete-'a-tete with the chancellor, Cartersought successfully to convinceSchmidt that immediate decontrol was politicallyimpossible, but thatdecontrol by 1981was politicallydo- able. Kinkywin-sets may be morecredible if they pivot on whatSchelling calls a "prominent"solution, such as a 50-50 split,for such outcomes may be distinctlymore "saleable" at home. Anotherrelevant tactic is forthe negotiatoractually to submita trialagreement for ratification, in orderto demonstratethat it is notin his win-set. Uncertaintyabout the contoursof the respective"political indifference curves" thushas strategicuses. On the otherhand, when the negotiators are seekingnovel packages that might improve both sides' positions,mis- representationofone's win-setcan be counterproductive.Creative solutions thatexpand the scope forjoint gain and improvethe odds ofratification are likelyto requirefairly accurate information about constituents' preferences andpoints of special neuralgia. The analysisof two-level games offers many illustrationsof Zartman's observation that all negotiationinvolves "the con- trolledexchange of partialinformation. 73

71. Synder and Diesing, ConflictAmong Nations, p. 517. 72. I am gratefulto Robert0. Keohanefor pointing out theimpact of uncertaintyon the expectedvalue of proposals. 73. I. WilliamZartman, The 50% Solution(Garden City, N.J.: Anchor Books, 1976),p. 14. The presentanalysis assumes that constituents are myopicabout the other side's Level II, an assumptionthat is not unrealisticempirically. However, a fullyinformed constituent would considerthe preferences of keyplayers on theother side, for if the current proposal lies well withinthe other side's win-set,then it wouldbe rationalfor the constituent to voteagainst it, hopingfor a second-roundproposal that was morefavorable to himand stillratifiable abroad; thismight be a reasonableinterpretation of Senator Lodge's positionin 1919(Bailey, Wilson and theGreat Betraya[). Considerationof such strategic voting at Level II is beyondthe scope of thisarticle.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 454 InternationalOrganization

Restructuringand reverberation

Formallyspeaking, game-theoretic analysis requires that the structureof issues and payoffsbe specifiedin advance. In reality,however, much of whathappens in any bargainingsituation involves attempts by theplayers to restructurethe game and to alterone another'sperceptions of thecosts of no-agreementand thebenefits of proposedagreements. Such tacticsare moredifficult in two-levelgames than in conventionalnegotiations, because itis harderto reachconstituents on theother side with persuasive messages. Nevertheless,governments do seekto expandone another'swin-sets. Much ambassadorialactivity-wooing opinion leaders, establishing contact with oppositionparties, offering foreign aid to a friendly,but unstablegovern- ment,and so on-has preciselythis function. When Japanese officials visit CapitolHill, or Britishdiplomats lobby Irish-American leaders, they are seekingto relax domestic constraints that might otherwise prevent the admin- istrationfrom cooperating with their governments. Anotherilluminating example of actionsby a negotiatorat theopposing Level II to improvethe odds of ratificationoccurred during the 1977ne- gotiationsbetween the International Monetary Fund and theItalian government.Initial IMF demandsfor austerity triggered strong opposition from theunions and left-wingparties. Although the IMF's bargainingposition at Level I appearedstrong, the Fund's negotiatorsought to achievea broader consensuswithin Italy in supportof an agreement,in orderto forestall involuntarydefection. Accordingly, after direct consultations with the unions and leftistleaders, the IMF restructuredits proposal to focuson long-term investmentand economicrecovery (incidentally, an interestingexample of targeting),without backing off from its short-termdemands. Ironically, the initialCommunist support for this revised agreement subsequently collapsed because of conflictsbetween moderate and doctrinairefactions within the party,illustrating the importance of multilevelanalysis.74 In some instances,perhaps even unintentionally,international pressures "reverberate"within domestic politics, tipping the domesticbalance and thusinfluencing the internationalnegotiations. Exactly this kind of rever- berationcharacterized the 1978summit negotiations. Dieter Hiss, theGer- mansherpa and one of thosewho believedthat a stimulusprogram was in Germany'sown interest,later wrote that summits change national policy onlyinsofar as theymobilize and/or change public opinion and theatti- tude of political groups.... Oftenthat is enough, if the balance of

74. JohnR. Hillman,"The MutualInfluence of Italian Domestic Politics and the International MonetaryFund," The Fletcher Forum 4 (Winter1980), pp. 1-22.Luigi Spaventa, "Two Letters of Intent:External Crises and StabilizationPolicy, Italy, 1973-77," in JohnWilliAmson, ed., IMF Conditionality(Washington, D.C.: Institutefor International Economics, 1983), pp. 441-73, arguesthat the unions and theCommunists actually favored the austerity measures, but found theIMF demandshelpful in dealingwith their own internal Level II constituents.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 455

opinionis shifted,providing a bare majorityfor the previously stymied actions of a strongminority.... No countryviolates its own interests, butcertainly the definition of its interestscan changethrough a summit withits possibletradeoffs and give-and-take.75 Fromthe point of view of orthodox social-choice theory, reverberation is problematic,for it impliesa certaininterconnectedness among the utility functionsof independentactors, albeit across differentlevels of thegame. Two rationalesmay be offeredto explain reverberationamong utility- maximizingegoists. First, in a complex,interdependent, but often unfriendly world,offending foreigners may be costlyin thelong run. "To get along, go along" may be a rationalmaxim. This rationaleis likelyto be more commonthe moredependent (or interdependent)a nation, and it is likely to be morepersuasive to Level II actorswho are moreexposed internationally,such as multinationalcorporations and internationalbanks. A secondrationale takes into account cognitive factors and uncertainty. It would be a mistakefor political scientists to mimicmost economists' disregardfor the suasive elementin negotiations.76Given the pervasive uncertaintythat surrounds many international issues, messages from abroad can changeminds, move the undecided, and heartenthose in thedomestic minority.As one reluctantGerman latecomer to the"locomotive" cause in 1978explained his conversion,"In theend, even the Bank for International Settlements[the cautious Basle organizationof centralbankers] supported theidea ofcoordinated relation." Similarly, an enthusiasticadvocate of the programwelcomed the internationalpressure as providinga useful"tail- wind" in Germandomestic politics. Suasivereverberation is more likely among countries with close relations and is probablymore frequent in economicthan in political-militarynego- tiations.Communiques from the Westernsummits are oftencited by par- ticipantsto domesticaudiences as a wayof legitimizing their policies. After one such statementby ChancellorSchmidt, one ofhis aides privatelychar- acterizedthe argument as "not intellectuallyvalid, but politically useful." Conversely,it is widelybelieved by summitparticipants that a declaration contraryto a government'scurrent policy could be used profitablyby its opponents.Recent congressional proposals to ensuregreater domestic pub- licityfor international commentary on nationaleconomic policies (including hithertoconfidential IMF recommendations)turn on theidea thatreverber- ationmight increase international cooperation.77 75. DieterHiss, "Weltwirtschaftsgipfel: Betrachtungen eines Insiders [World Economic Sum- mit:Observations of an Insider],"in JoachimFrohn and ReinerStaeglin, eds., Empirische Wirtschaftsforschung(Berlin: Duncker and Humblot,1980), pp. 286-87. 76. On cognitiveand communicationsexplanations of internationalcooperation, see, for example,Ernst B. Haas, "WhyCollaborate? Issue-Linkage and International Regimes," World Politics32 (April1980), pp 357-405;Richard N. Cooper,"International Cooperation in Public Healthas a Prologueto MacroeconomicCooperation," Brookings Discussion Papers in In- ternationalEconomics 44 (Washington,D.C.: BrookingsInstitution, 1986); and Zartman,50% Solution,especially Part 4. 77. Henning,Macroeconomic Diplomacy in the 1980s, pp. 62-63.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 456 InternationalOrganization

Reverberationas discussedthus far implies that international pressure expandsthe domesticwin-set and facilitatesagreement. However, rever- berationcan also be negative,in thesense that foreign pressure may create a domesticbacklash. Negative reverberation is probablyless commonem- piricallythan positive reverberation, simply because foreigners are likelyto forgopublic pressure if it is recognizedto be counterproductive.Cognitive balancetheory suggests that international pressure is morelikely to rever- beratenegatively if its sourceis generallyviewed by domesticaudiences as an adversaryrather than an ally.Nevertheless, predicing the precise effect offoreign pressure is admittedlydifficult, although empirically, reverberation seemsto occurfrequently in two-levelgames. The phenomenonof reverberation (along with synergistic issue linkage of the sortdescribed earlier) precludes one attractiveshort-cut to modeling two-levelgames. If nationalpreferences were exogenous from the point of view of internationalrelations, then the domesticpolitical game could be moldedseparately, and the"outputs" from that game could be used as the "inputs" to the internationalgame.78 The divisionof laborbetween comparativepolitics and internationalrelations could continue,though a few curiousobservers might wish to keep trackof theplay on bothtables. But if internationalpressures reverberate within domestic politics, or if issues can be linkedsynergistically, then domestic outcomes are not exogenous, and thetwo levels cannot be modeledindependently.

The role of the chiefnegotiator

In the stylizedmodel of two-levelnegotiations outlined here, the chief negotiatoris the onlyformal link between Level I and Level II. Thus far,I haveassumed that the chief negotiator has no independentpolicy views, but acts merelyas an honestbroker, or ratheras an agenton behalfof his constituents.That assumption powerfully simplifies the analysis of two-level games.However, as principal-agenttheory reminds us, thisassumption is unrealistic.79Empirically, the preferences of the chiefnegotiator may well divergefrom those of his constituents. Two-level negotiations are costlyand

78. This is theapproach used to analyzethe Anglo-Chinese negotiations over Hong Kong in Bruce Bueno de Mesquita,David Newman,and AlvinRabushka, Forecasting Political Events: The Future of Hong Kong (New Haven: Yale UniversityPress, 1985). 79. For overviewsof thisliterature, see TerryM. Moe, "The New Economicsof Organi- zation," AmericanJournal of Political Science 28 (November 1984), pp. 739-77; JohnW. Pratt and Richard J. Zeckhauser, eds., Principals and Agents: The Structureof Business (Boston, Mass.: HarvardBusiness School Press, 1985); and BarryM. Mitnick,"The Theoryof Agency andOrganizational Analysis," prepared for delivery at the1986 annual meeting of thte American PoliticalScience Association. This literatureis onlyindirectly relevant to ourconcerns here, forit has notyet adequately addressed the problems posed by multipleprincipals (or constituents,in ourterms). For one highlyformal approach to theproblem of multiple principals, see R. DouglasBernheim and MichaelD. Whinston,"Common Agency," Econometrica 54 (July 1986),pp. 923-42.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 457

riskyfor the chief negotiator, and theyoften interfere with his otherprior- ities,so it is reasonableto ask whatis in it forhim. The motivesof thechief negotiator include: 1. Enhancinghis standingin theLevel II gameby increasinghis political resourcesor by minimizingpotential losses. For example,a head of governmentmay seek thepopularity that he expectsto accrueto himif he concludesa successfulinternational agreement, or he may anticipatethat the results of theagreement (for example, faster growthor lowerdefense spending) will be politicallyrewarding. 2. Shiftingthe balance of powerat Level II in favorof domesticpolicies thathe prefersfor exogenous reasons. International negotiations sometimesenable government leaders to do whatthey privately wish to do, butare powerlessto do domestically.Beyond the now-familiar 1978case, thispattern characterizes many stabilization programs that are (misleadingly)said to be "imposed"by theIMF. For example,in the 1974and 1977negotiations between Italy and theIMF, domestic conservativeforces exploited IMF pressureto facilitatepolicy moves thatwere otherwise infeasible internally.80 3. To pursuehis own conceptionof thenational interest in theinterna- tionalcontext. This seemsthe best explanation of JimmyCarter's prodigiousefforts on behalfof thePanama Canal Treaty,as well as of WoodrowWilson's ultimately fatal commitment to theVersailles Treaty. It is reasonableto presume,at leastin theinternational case oftwo-level bargaining,that the chiefnegotiator will normally give primacyto his do- mesticcalculus, if a choicemust be made,not least because his own incum- bencyoften depends on his standingat Level II. Hence, he is morelikely to presentan internationalagreement for ratification, the less of his own politicalcapital he expectsto haveto investto winapproval, and the greater thelikely political returns from a ratifiedagreement. This expandedconception of therole of thechief negotiator implies that he has, in effect,a veto overpossible agreements. Even ifa proposeddeal lieswithin his Level II win-set,that deal is unlikelyto be struckif he opposes it.81Since thisproviso applies on bothsides of theLevel I table,the actual internationalbargaining set may be narrower-perhapsmuch narrower- thanthe overlap between the Level II win-sets.Empirically, this additional constraintis oftencrucial to theoutcome of two-levelgames. One momen- tousexample is thefate of the Versailles Treaty. The best evidence suggests, first,that perhaps 80 percentof theAmerican public and of the Senate in 1919favored ratification of thetreaty, if certain reservations were attached, and second,that those reservations were acceptableto the otherkey sig- 80. Hillman,"Mutual Influence," and Spaventa,"Two Lettersof Intent." 81. Thispower of thechief negotiator is analogousto whatShepsle and Weingastterm the "penultimate"or "ex postveto" powerof themembers of a Senate-Houseconference committee.(Shepsle and Weingast,"Institutional Foundations of CommitteePower.")

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 458 InternationalOrganization natories,especially Britain and France.In effect,it was Wilsonhimself who vetoedthis otherwise ratifiable package, telling the dismayed French Am- bassador,"I shallconsent to nothing."82 Yet anotherconstraint on successfultwo-level negotiation derives from the leader's existingdomestic coalition. Any politicalentrepreneur has a fixedinvestment in a particularpattern of policypositions and a particular supportingcoalition. If a proposedinternational deal threatensthat invest- ment,or ifratification would require him to constructa differentcoalition, thechief negotiator will be reluctantto endorseit, even if (judged abstractly) it couldbe ratified.Politicians may be willingto riska fewof theirnormal supportersin thecause ofratifying an internationalagreement, but the greater thepotential loss, thegreater their reluctance. In effect,the fixedcosts of coalition-buildingthus imply this constraint on thewin-set: How greata realignmentof prevailing coalitions at Level II wouldbe requiredto ratifya particularproposal? For example,a tradedeal may expandexport opportunities for Silicon Valley, but harmAliquippa. Thisis finefor a chiefnegotiator (for example, Reagan?) who can easilyadd NorthernCalifornia yuppies to his supportcoalition and who has no hope of winningAliquippa steelworkers anyhow. But a differentchief negotiator witha differentsupport coalition (for example, Mondale?) might find it costly or even impossibleto convertthe gains from the same agreementinto po- liticallyusable form.Similarly, in the 1978 "neutronbomb" negotiations betweenBonn and Washington,"asking the United States to deploy[these weapons]in WestGermany might have been possiblefor a ChristianDem- ocraticGovernment; for a Social Democraticgovernment, it was nearly impossible.'"83 Undersuch circumstances,simple "median-voter" models of domesticinfluences on foreignpolicy may be quitemisleading. Relaxingthe assumptionthat the chiefnegotiator is merelyan honest broker,negotiating on behalfof his constituents,opens the possibility that theconstituents may be moreeager for an agreement(or more worried about "no-agreement")than he is. Empiricalinstances are not hardto find:in early1987, European publics were readier to acceptGorbachev's "double- zero" armscontrol proposal than European leaders, just as inthe early 1970s theAmerican public (or at leastthe politically active public) was moreeager fora negotiatedend to theVietnam War than was theNixon administration. As a rule,the negotiator retains a vetoover any proposed agreement in such cases. However,if the negotiator's own domesticstanding (or indeed,his incumbency)would be threatenedif he wereto rejectan agreementthat falls withinhis Level II win-set,and ifthis is knownto all parties,then the other sideat Level I gainsconsiderable leverage. Domestic U.S. discontentabout

82. Bailey, Wilson and the Great Betrayal, quotationat p. 15. 83. RobertA. Strongand Marshal Zeringue, "The NeutronBomb and the Atlantic Alliance," presentedat the 1986annual meeting of theAmerican Political Science Association, p. 9.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions Diplomacyand domesticpolitics 459 theVietnam War clearly affected the agreement reached at theParis talks.84 Conversely,if the constituents are (believedto be) hard-line,then a leader's domesticweakness becomes a diplomaticasset. In 1977,for example, the Americanscalculated that "a delayin negotiatinga treaty . .. endangered [PanamanianPresident Omar] Torrijos' position; and Panamawithout Tor- rijosmost likely would have been an impossiblenegotiating partner."85 Sim- ilarly,in the1954 Trieste negotiations, the weak Italian government claimed that"'Unless somethingis done in our favorin Trieste,we can lose the election.'That card was playedtwo or threetimes [reported the British negotiatorlater], and it almostalways took a trick."86 My emphasison thespecial responsibility of centralexecutives is a point of affinitybetween the two-level game model and the"state-centric" literature,even thoughthe underlying logic is different.In this"Janus" model of domestic-internationalinteractions, transnational politics are less prom- inentthan in some theoriesof interdependence.87However, to disregard "cross-table"alliances at Level II is a considerablesimplification, and it is moremisleading, the lower the political visibility of the issue, and themore frequentthe negotiations between the governments involved.88 Empirically, forexample, two-level games in the EuropeanCommunity are influenced bymany direct ties among Level II participants,such as nationalagricultural spokesmen.In somecases, thesame multinational actor may actually appear at morethan one Level II table.In negotiationsover mining concessions in some less-developedcountries, for example, the same multinationalcor- porationmay be consultedprivately by both the home and host governments. In subsequentwork on the two-levelmodel, the strategicimplications of directcommunication between Level II playersshould be explored.

Conclusion

The mostportentous development in thefields of comparativepolitics and internationalrelations in recentyears is the dawningrecognition among practitionersin each fieldof the need to take intoaccount entanglements betweenthe two. Empiricalillustrations of reciprocalinfluence between domesticand internationalaffairs abound. What we neednow are concepts

84. I. WilliamZartman, "Reality, Image, and Detail:The ParisNegotiations, 1969-1973," in Zartman,50% Solution,pp. 372-98. 85. ZbigniewBrzezinski, Power and Principle(New York:Farrar, Straus and Giroux, 1983), p. 136,as quotedin Habeeb and Zartman,Panama Canal Negotiations,pp. 39-40. 86. Harrisonin Campbell,Trieste 1954, p. 67. 87. SamuelP. Huntington,"Transnational Organizations in WorldPolitics," World Politics 25 (April 1973), pp. 333-68; Keohane and Nye, Power and Interdependence;Neustadt, Alliance Politics. 88. BarbaraCrane, "Policy Coordination by MajorWestern Powers in Bargainingwith the ThirdWorld: Debt Reliefand theCommon Fund," InternationalOrganization 38 (Summer 1984),pp. 399-428.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions 460 InternationalOrganization andtheories that will help us organizeand extend our empirical observations. Analysisin termsof two-levelgames offers a promisingresponse to this challenge.Unlike state-centric theories, the two-level approach recognizes the inevitabilityof domesticconflict about what the "nationalinterest" requires.Unlike the "Second Image" or the "Second Image Reversed," the two-levelapproach recognizes that central decision-makers strive to reconciledomestic and international imperatives simultaneously. As we have seen, statesmenin thispredicament face distinctivestrategic opportunities and strategicdilemmas. Thistheoretical approach highlights several significant features of the links betweendiplomacy and domesticpolitics, including: * theimportant distinction between voluntary and involuntarydefection frominternational agreements; * thecontrast between issues on whichdomestic interests are homogeneous, simplypitting hawks against doves, and issues on whichdomestic interestsare moreheterogeneous, so thatdomestic cleavage may ac- tuallyfoster international cooperation; * thepossibility of synergisticissue linkage,in whichstrategic moves at one game-tablefacilitate unexpected coalitions at thesecond table; * theparadoxical fact that institutional arrangements which strengthen decision-makersat homemay weaken their international bargaining position,and vice versa; * theimportance of targetinginternational threats, offers, and side- paymentswith an eye towardstheir domestic incidence at homeand abroad; * thestrategic uses of uncertaintyabout domestic politics, and thespecial utilityof "kinkywin-sets"; * thepotential reverberation of internationalpressures within the domestic arena; * thedivergences of interestbetween a nationalleader and thoseon whosebehalf he is negotiating,and in particular,the international implicationsof his fixedinvestments in domesticpolitics. Two-levelgames seem a ubiquitousfeature of social life,from Western economicsummitry to diplomacyin theBalkans and fromcoalition politics inSri Lanka to legislative maneuvering on CapitolHill. Far-ranging empirical researchis needednow to testand deepenour understandingof how such gamesare played.

This content downloaded from 130.64.11.153 on Wed, 10 Jun 2015 17:31:18 UTC All use subject to JSTOR Terms and Conditions