Theories and Empirical Studies of International Institutions

The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters

Citation Martin, Lisa L., and Beth A. Simmons. 1998. Theories and empirical studies of international institutions. International Organization 52(4): 729-757.

Published Version http://dx.doi.org/10.1162/002081898550734

Citable link http://nrs.harvard.edu/urn-3:HUL.InstRepos:3382862

Terms of Use This article was downloaded from Harvard University’s DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http:// nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of- use#LAA Theoriesand EmpiricalStudies of International Institutions Lisa L.Martin and Beth A.Simmons

Therole of internationalinstitutions has beencentral to thestudy of worldpolitics at leastsince the conclusion of WorldW ar II. Muchof thisresearch was, andcontinues tobe, pioneered in the pages of InternationalOrganization .Inthis article we take stockof pastwork on internationalinstitutions, trace the evolution of majorthemes inscholarship over time, and highlight areas for productivenew research. Our cen- tralargument is that research should increasingly turn to thequestion of howinstitutionsmatter in shaping the behavior of importantactors in world politics. New re- searchefforts shouldemphasize observable implications of alternative theories of institutions.W eadvocateapproaching international institutions as boththe object of strategicchoice and a constrainton actors’ behavior, an ideathat is familiarto scholars ofdomestic institutions but has been neglected in much of the debate between realistand institutionalist scholars of internationalrelations. Thearticle is organized into three major sections. The rst sectionprovides an analyticalreview of the development of studies of international institutions. From thebeginning, the pages of IO havebeen lledwith insightful studies of institutions, insome cases asking questions consistent with the research agenda we proposein thisessay. But the lack of a disciplinaryfoundation in the early years meant that manygood insights were simplylost, not integrated into other scholars’ research. Withthe professionalization of the discipline since the late 1950s, scholarship on internationalinstitutions has become more theoretically informed, and empirical re- searchhas begun more often to conform to social-scienti c standardsof evidence, withresults that provide both caution and inspiration for futureresearch. One of the mostconsequential developments for ourunderstanding of internationalinstitutions camein the early 1970s, when a newgeneration of scholarsdeveloped insights that openedup inquirybeyond that of formalorganizations, providing intellectual bridge- headsto the study of institutionsmore generally.

Ourthanks for comments onpreviousversions go to Marc Busch,Peter Katzenstein,Bob Keohane, SteveKrasner, andparticipants in the IO ftiethanniversary issue conference.

InternationalOrganization 52,4, Autumn 1998, pp. 729– 757 r 1998by The IO Foundationand the Massachusetts Instituteof Technology 730 InternationalOrganization

Thesecond section explicitly addresses a themethat arises from thereview of scholarshipon institutions: whether international politics needs to be treated as sui generis, withits own theories and approaches that are distinct from other eldsof politicalscience, or whether it fruitfullycan draw ontheoriesof domesticpolitics. As ourreview shows, developments in studies of Americanpolitics, such as studies of votingand coalitional behavior, have often in uenced the way that scholars ap- proachedinternational institutions. Most of these efforts did not pay off withmajor insights.The functionalist approach to institutions adopted in the1980sowed little to theoriesof domesticpolitics, drawing more on economicmodels. T oday,we seethe pendulumswinging back, as more scholars turn to modern theories initially devel- opedto studydomestic political phenomena (see HelenMilner’ s articlein this issue). Here, we assesswhether these new attemptsare likely to be anymore successful than previousefforts. Thethird section turns to theproblem of researchagendas. Where does scholarshipon international institutions go next? Our primaryargument in this section is thatattention needs to focus on how, not just whether,internationalinstitutions mat- terfor worldpolitics. T oooften over the last decade and a halfthe focal point of debatehas been crudely dichotomous: institutions matter, or theydo not.This shap- ingof the agenda has obscured more productive and interesting questions about variationin thetypes and degree of institutionaleffects, variations that were infact welldocumented in theless theoretical but well-researched case studies of thejour- nal’s earliestyears. Of course,we donotsuggest a returnto idiographicinstitutional analysis.Rather, we suggesta numberof linesof theoretically informed analysis that maylead to researchthat both asks better questions and is more subject to empirical testing.These paths include more serious analysis of the distributional effects of institutions,the relation between international institutions and domestic politics, the problemof unanticipatedconsequences, and a typologyof institutionaleffects.

The Evolutionof anIdea: Institutions in International Politics EarlyStudies of theInstitutionalization of the Postwar W orld The‘ ‘poles’’ ofrealism and idealism— ofwhich much is madein graduateseminars— hadlittle to dowith the highly practical organizational analysis that dominated the pages of IO inthe rst decadesafter the war. Thefocus of attentionwas onhowwell thesenewly established institutions met the problems that they were designedto solve.On this score, few scholarlyaccounts were overlyoptimistic. Overwhelmed bythe magnitude of the political and economic reconstruction effort, few judged postwarorganizations as upto the task. Central to this debate was ahighlyrealistic understandingthat international politics would shape and limit the effectiveness of postwarinstitutions; virtually no one predicted that these would triumph over poli- InternationalInstitutions 731 tics.The UN, 1 theGeneral Agreement on T ariffs andTrade (GA TT), 2 theInterna- tionalMonetary Fund 3 —allwere thesubject of highlycritical review . Anumberof importantstudies grappled explicitly with the impact of theseinstitu- tionson thepolicies of the major powers and the outcomes for thecentral political andmilitary competition between them. The answers, predictably, were derivedfrom littlemore than informed counterfactual reasoning, but theydisplayed a sensitivityto thebroad range of possibleimpacts that institutions such as the League and the UN couldhave on themajor powers. In their examination of theideal of collective security,Howard C. Johnsonand Gerhart Niemeyer squarely inquired into the role that norms,backed by organizations such as theUN, playin affecting states’ behavior. Theyasked whether states were ‘‘preparedto use force or thethreat of forcefor the sakeof publiclaw and order rather than for thesake of theirnational advantage in relationto that of otherstates. . ..How hasthe behavior of statesbeen affected by thesestandards?’ ’4 Thoughultimately more con dent in thebalanceof powerthan in normsembodied in the rule of law,these scholars were correctto pushfor amecha- nismthat might explain the effects of institutionson behavior:‘ ‘Wecannotclaim to havelearned much about the League experiment until we know how ithasaffected theproblem of harnessingand controlling the factors of forceand their role in the relationsof power.’’5 Aurryof studies in the early 1950s suggested possible answers. Pointing to the U.S. rolein decolonization and military aid for Korea,collective institutions were saidto raise U.S. ‘‘consciousnessof broader issues’ ’ thatmight affect American interestsand thereby make the U.S. moreresponsive to worldopinion. 6 By subject- ingpolicies to globalscrutiny— amechanism not unlike those of transparency and reputationcentral to theliterature in the1980s— the UN was viewedas having had an(admittedly marginal) effect on some of the most central issues of worldpolitics. Thoughlacking the elaborate theoretical apparatus of currentresearch, early studiesof postwarorganizations had many of the sameinsights that have informed ‘ ‘modern’’ institutionalism.Paralleling much contemporary argument on the form ofcooperation,7 onestudy as early as 1949 argued that multilateralism was precludedin caseswhere there were signicant bargaining advantages and discrimination advan- tagesof proceedingbilaterally. 8 Foreshadowingmore theoretically sophisticated treat- mentsof informal versus formal agreements, 9 studiesof GA TTasearly as 1954 recognizedthat some agreements gain strength through their informal nature, and

1.See Goodrich1947, 18; Fox 1951; Hoffmann 1956; Claude 1963; and Malin 1947. But for the optimisticview, see Bloomeld 1960. 2.Gorter 1954. 3.See Knorr1948; and Kindleberger 1951a. 4.Johnson and Niemeyer 1954,27. 5.Niemeyer 1952,558 (italics added). 6.Cohen 1951. For a parallel analysisof institutionaleffects onSoviet behavior, see Rudzinski1951. 7.See Oye 1992;and Martin 1992b. 8.Little 1949. 9.Lipson 1991. 732 InternationalOrganization prescientof theregimes literature viewed the value of GA TTas‘‘afocalpoint on whichmany divergent views on appropriatecommercial policy converge.’ ’10 Lack- inga theoreticalhook on whichto hang these observations, and without a profession- alizedcritical mass ofscholars to develop these insights, many important ndings were onlyrediscovered and advanced more than two decades later . Nowhereis this more true than in the rediscovery of the relationship between internationalinstitutions and domestic politics. The idea that international institu- tionscan in uence state behavior by acting through domestic political channels was recognizedby scholars writing in the mid-1950s. Referring to the example of the InternationalFinance Corporation, B. E.Mateckiwrote that international organiza- tionscould be ‘ ‘ideagenerating centers’ ’ withthe ability to set in motion national forcesthat directly in uence the making of nationalpolicy. 11 Reecting on theefforts ofthe Council of Europe to gain acceptance of its vision for Europein national capitals,an earlystudy by A.GlennMowers pointed out the conscious strategy of directlobbying of national governments through national parliaments. 12 And in a fascinatingstudy of the role of the Security Council in in uencing Dutch colonial policy,Whitney Perkins pointed to the crucialinteraction between authoritative inter- nationaldecisions and democratic politics: ‘ ‘Bydeance of the Security Council the Dutchalerted powerful monitors who allied their strength with domestic forces in requiringthem to liveup toprinciples[of decolonization].’’13 ‘‘Inthis type of inter- actionbetween democratic governments and the UN emergesome of theessential elementsof aworldpolitical process.’ ’14 Anticipatinga mechanismfor institutional effectsthat have recently resurfaced in contemporary studies, he concludedthat ‘ ‘The roleof theUN istoexert pressures designed to enablethe loserin publicsentiment to acceptthe consequences of its loss.’ ’15 Thisresearch approach re ected an effort to eshout the mechanisms by which the policies and perspectives of international institutionscould work through national politics. Inshort, the early postwar literature on international institutions, while highly focusedon formal organizations, was far lessnaive and legalistic, more politically sensitiveand insightful than it isoften given credit for being.Early insights included therecognition that the nature of theinternationalpolitical system provided a context for theeffectiveness of internationalinstitutions, that institutional effectiveness should besubject to empirical investigation, and that elaborate organizational structure is notalways the best approach to achievinginternational cooperation. Moreover, the bestof this early literature was concernednot merely with whether international institutionshad an impact, but how onemight think about a mechanismfor their effects.Transparency, reputation, and legitimacy as wellas domestic political pres- sureswere suggestedin various strands of thought. But there was noconceptual

10.Gorter 1954, 1, 8. 11.Matecki 1956. 12.Mowers 1964. 13.Perkins 1958, 40. 14.Ibid., 26. 15.Ibid., 42 InternationalInstitutions 733 framework thatcould tie these insights together; nor was therea systematiccompara- tiveenterprise to checkfor theirregularity. Rather, another research agenda, replete withfancy methodological tools imported from Americanpolitics, was todemote thesequestions in favor of anonlypartially fruitful examination of the internal poli- ticsof internationalorganizations.

TheIn uence of Behavioralism: PoliticsWithin International Institutions If few thoughtinternational organization would liberate the world from politics,it arguablybecame important to understand who has power in these organizations and howthat power was beingexercised. Especially since the use of theveto had appar- entlyrendered the Security Council toothless, concern began to focus on thedevelop- mentof rulesand norms in the General Assembly. The supposed ‘ ‘specter’’ ofbloc votingin thatforum— increasingly of concernto Americanscholars and policymak- ers asthe Cold W arextendedits gelid reach— became a centralconcern. 16 Thisdebate took what appears today to beanodd early direction. Perhaps due to newand exciting work in U.S. legislativebehavior, the research program quickly becamefocused on how to describe patterns of voting in the General Assembly, withouta systematicattempt to sort out the usefulness of the voting behavior approach.Despite warnings that the international system was fundamentallydifferent from domesticpolitical systems, 17 thisresearch program easily accepted that voting inthe UN was aproxyfor powerin that institution. Certainly there were skeptics: RupertEmerson and Inis L. Claude,for example,cautioned that voting in aninterna- tionalbody does not have the same function as in a democraticallyelected parlia- ment;an international conference is a negotiatingrather than a legislativebody. Votingin such a situation,they noted, was unlikelyto playa deliberativerole, since suchvotes were nomorethan propaganda efforts. 18 Few ofthese studies explicitly defendedtheir assumption that General Assembly resolutions somehow mattered to theconduct of world politics. But the fascination with the method for analyzing votingbehavior overcame fairly readily the caution that the domestic– international logicshould be subjectto closescrutiny. Moreover, the hope of providingan explic- itlypolitical (legislative) model inspired by Americanpolitics may have been a reac- tionagainst the overly ‘ ‘anarchic’’ systemsanalysis of thelate 1950s. 19 Muchof this work can be traceddirectly to developmentsin the study of American politics.Hayward Alker and Bruce Russett’ s study InternationalPolitics in the Gen-

16.For one of the earliest studiesof bloc voting, see Ball 1951.For a studyfocusing primarily on the behaviorof theCommonwealth countries, see Carter 1950.Concern with the in uence ofthe Common- wealth grew as former Britishcolonies gained independence and membership in the early 1960s.See Millar1962. 17.Hoffmann 1960, 1– 4. 18.Emerson and Claude 1952. See alsoJebb 1952. 19.Alker and Russett 1965, 145, explicitly refer toLiska1957 and Kaplan 1957. They argue that ‘ ‘[i]t is simplyerroneous to think of international politics as anarchic,chaotic, and utterly unlike national politics.’’ Alkerand Russett 1965, 147. 734 InternationalOrganization eralAssembly ,for example,acknowledged ‘ ‘thatstudies of theAmerican political processby RobertDahl, Duncan Macrae, 20 andDavidTruman were theoreticallyand methodologicallysuggestive of waysin which roll-call data could be usedto testfor theexistence of a pluralisticpolitical process in aquasi-legislativeinternational organization.’’21 Inuenced by James March 22 andRobert Dahl, this study sought to understandvarious in uences on UNvotingbehavior across issue areas in whichthe dimensionsof power and in uence were likelyto differ .Certainly,one factor in u- encingthis research agenda was thepriority given to reproducibleand ‘ ‘objective’’ forms ofsocialscience; the focus on GeneralAssembly voting was acknowledgedto bean artifactof theavailability of fairlycomplete voting records. 23 Largelyrelated to theferment in American voting studies, politics within the UN dominatedthe research agenda for mostof thedecade from themid-1960s. Central was theconcern to explainwhy certain countries had a tendencyto votetogether, to votein blocs, or toform ‘‘legislativecoalitions.’ ’24 Alsoobviously inspired by Ameri- canpolitics, another branch of inquiry focused on thedeterminants of successfully runningfor electiveUN office. 25 Muchof thisliterature was methodologicallyrather thanconceptually driven and highly inductive with respect to its major empirical ndings.26 Littleeffort was madeto explorethe extent to which the concept of repre- sentationor the winning of electionsin thedomestic setting could travel meaning- fullyto aninternationalinstitution. The research program lost steam under heavy re from scholarswho demanded a strongerjusti cation for focusingon the General Assemblyas a microcosmfor worldpolitics. 27 Partiallyin response to the critique that the General Assembly was hardlythe centerof worldpolitics, and partially in uenced by anothertrend in Americanpoli- ticsgrowing out of the study of bureaucraticpolitics and political systems, another researchpath was takenby RobertCox and Harold Jacobson’ s studyof eightspecial- izedagencies within the UN. 28 Intheir edited volume, the focus was onthestructure andprocess of inuence associated with these institutions and their outputs, rather thanon their formal character. Re ecting once again a majorthread in American politics,the underlying assumption was thatinternational organizations could be fruitfullyanalyzed as distinct political systems in which one could trace out patterns ofin uence: ‘ ‘Thelegal and formal character and the content of thedecision is less importantthan the balance of forcesthat it expresses and the inclination that it gives tothe further direction of events.’’29

20.MacRae 1958. 21.Alker and Russett 1965, vii. 22.March 1955. 23.On objectivity, see Alkerand Russett 1965, 2– 3; on availabilityof data see p.19. 24.See Riggs1958; Hovet 1958; Keohane 1967, 1969; Weigert and Riggs 1969; Gareau 1970;Alker 1970;V olgy1973; and Harbert 1976. 25.See Volgyand Quistgard 1974; and Singer and Sensenig 1963. 26.See, forexample, Rieselbach 1960. 27.For two systematic reviews ofthequantitative research ontheUN andinternational organizations, see Riggset al.1970; and Alger 1970. 28.Cox and Jacobson 1973. 29. Ibid. InternationalInstitutions 735

Thework of Coxand Jacobson also encouraged the study of international organi- zationsto consider a moretransgovernmental model of their in uences. Whereas otherresearch inspired by behavioralismtypically assumed a unied model of state interestsand actors, this work focused on transgovernmental coalitions involving partsof governments and parts of internationalorganizations. One of themost impor- tantinsights generated was highlyconsonant with developments in transgovernmen- talrelations that had come on theintellectual scene in the1970s: 30 theobservation thatone channelthrough which international organizations could affect state policies was throughthe potential alliances that could form betweeninternational bureaucra- ciesand domestic pressure groups at the national level. 31 Althoughthis was aninter- estinginsight, and case studies tended to conrm theimportance of such ‘ ‘transnationalcoalitions’ ’ for policyimplementation, their effect on policyformulation remains unclear.32 Meanwhile,the issues facing the international community changed drasti- callyin the early 1970s, giving rise to anewapproach to the study of international institutions,discussed in the following section. Finally,a strandof research stimulated by ErnstHaas’ s ‘‘neofunctionalapproach’ ’ tointegration also left a tellingmark on thestudyof empiricaleffects of international institutionsin the 1970s. Neofunctionalism ascribed a dynamicrole to individuals andinterest groups in the processof integratingpluralist communities. 33 By virtue of theirparticipation in thepolicymaking process of anintegratingcommunity, interest groupsand other participants were hypothesizedto ‘ ‘learn’’ aboutthe rewards of suchinvolvement and undergo attitudinal changes inclining them favorably toward theintegrative system. Accordingto Haas, ‘ ‘politicalintegration is theprocess whereby actorsshift their loyalties, expectations, and political activities toward a newcenter, whoseinstitutions possess or demand jurisdiction over preexisting national states.’ ’34 Theimplications for empiricalresearch on such institutions were readilydrawn: thosewho participate in internationalorganizations should exhibit altered attitudes towardtheir usefulness and effectiveness. Americanpolitics provided yet another methodological instrument that dovetailed nicelywith what was thoughtto be an empirically testable proposition of Haas’ s theory:survey research! From thelate 1950s into the early 1980s, a plethoraof studiestried to establish whether international organizations could contribute to ‘ ‘learning,’’ whethercognitive or affective. 35 Theattitudes of civil servants, 36 politicalap- pointees,and even national legislators 37 were scrutinizedfor evidencethat the length ornature of theirassociation with various kinds of internationalorganizations had inducedattitudinal change. The impact of methodsfrom Americanpolitics was obvi-

30.Keohane and Nye 1974. 31.See Cox1969, 225; and Cox and Jacobson 1973, 214. 32.See, forexample, Russell 1973; and Keohane 1978. 33.See Haas 1958;and Pentland 1973. 34.Haas 1958,10. 35.See Kelman 1962;Alger 1965; and Jacobson 1967. See alsoWolf 1974, 352– 53; and V olgyand Quistgard1975. 36.See Ernst1978; and Peck 1979. 37.See Bonham1970; Kerr 1973;Riggs 1977; and Karns 1977. 736 InternationalOrganization ous:in some cases, indicators were usedthat precisely paralleled the ‘ ‘thermom- eters’’ usedby theNational Opinion Survey Research project. Threeproblems bedeviled this research approach for years.First, it failed to pro- duceconsensus on the effect of international institutions on attitudes. 38 Second,atti- tudeswere neverreconnected with outcomes, policies, or actions. 39 Third,researchers were neverable to overcome the problem of recruitment bias, which itself accountedfor mostof thepositive attitudes held by personnelassociated with internationalinstitutions. As neofunctionalismas atheoreticalorientation lost favor over thecourse of the1970s and integrative international organizations such as theEuro- peanCommunity and the UN seemedto stagnate in the face of growing world prob- lemsbeyond their purview ,thisresearch program declined, though today a versionis pursuedprimarily in studies that attempt to document mass attitudestoward the EuropeanUnion.

PoliticsBeyond Formal Organizations: TheRise of InternationalRegimes As thestudy of international institutions progressed over the post– W orldW ar II years,the gulf between international politics and formal organization arrangements beganto openin ways that were noteasy to reconcile. The major international con- ictfor arisinggeneration of scholars—the V ietnamW ar—raged beyond the formal declarationsof the UN. Twodecades of predictable monetary relations under the BrettonW oodsinstitutions were shatteredby a unilateraldecision by the United Statesin 1971 to close the gold window and later to oatthe dollar. The rise of the Organizationof Petroleum Exporting Countries and their apparent power to upset previouslyunderstood arrangements with respect to oilpricing and availability took placeoutside the structure of traditional international organizations, as didconsum- ers’response later in the decade. For some,the proper normative response seemed to beto strengthen international organizations to deal with rising problems of interdependence.40 Othersmore familiar with the public choice literature argued that a proper extensionof propertyrights, largely underway in areassuch as environmentalprotec- tion,rather than a formalextension of supranationalauthority per se, was theanswer tosolving problems of collectiveaction. 41 Overall,few doubtedthat international life was ‘‘organized,’’ but,increasingly, it became apparent that much of theearlierfocus

38.Studies that failed to con rm expectationsof attitudinal change include Siverson 1973; and Bonham 1970.A few studieseven found negative impacts onattitudes due toassociation with international organizations:Smith 1973; and Pendergast 1976. 39.To the extent that such associations affected outcomes,the results were generallyinnocuous. See, forexample, Mathiason 1972. 40.Brown and Fabian, for example, modestly call for‘ ‘acomprehensiveocean authority,an outer space projectsagency, a globalweather andclimate organization,and an international scienti c commis- sionon globalresources andtechnologies.’ ’ See Brownand Fabian 1975. See alsoRuggie 1972, 890, 891; andGosovic and Ruggie 1976. 41.Conybeare 1980. InternationalInstitutions 737 onformal structures and multilateral treaty-based agreements, especially the UN, hadbeen overdrawn. 42 Theevents of theearly 1970s gave rise to the study of ‘‘internationalregimes,’ ’ dened as rules, norms, principles, and procedures that focus expectations regarding internationalbehavior. Clearly, the regimes movement represented an effortto substi- tutean understandingof internationalorganization with an understandingof interna- tionalgovernance more broadly. 43 Italso demoted the study of internationalorgani- zationsas actors: prior to the study of internationalregimes an inquiry into the effects ofinternational institutions meant inquiring into how effectively a particularagency performedits job, for example,the efficiencywith which the W orldHealth Organiza- tionvaccinated the world’ s needychildren. 44 Whenregimes analysts looked for effects,these were understoodto be outcomes in uenced by a constellationof rules ratherthan tasks performed by acollectiveinternational agency. Butjust what effects regimes analysis sought to uncover has changed as the re- searchprogram has unfolded. 45 Arst collectiveeffort by thescholarly community to addressregime effects was primarilyinterested in the distributive consequences of thenorms of theinternational food regime, arguing that it isimportantto consider the ‘‘waysin whichthe globalfood regime affects . ..wealth,power, autonomy, community,nutritional well-being, . ..andsometimes physical survival.’ ’46 Inthis view , regime‘ ‘effects’’ were tobereckonedin termsof the distributive consequences of the behaviorof amyriadof producers, distributors, and consumers, and, in aminorway, byinternational organizations and state bureaucracies. Certainly, there was inthis earlyvolume little thought that regimes were somehowefficient or efficiency- improvingoutcomes, as later theorizing would imply; rather, the food regime was characterizedby ‘ ‘broadand endemic inadequacies,’ ’ whichare the result of national policiesthat are ‘ ‘internationallybargained and coordinated . ..bymultilateralagree- mentor unilateraldictate.’ ’47 Furtherresearch on internationalregimes moved thinking in three important directions.First, distributive consequences soon fell from thecenter of considerationas researchbegan to focuson howinternational regimes are created and transformed in the rst placeas wellas thebehavioral consequences of normsor rules, 48 rather than thedistributive consequences of behavior itself. (W earguelater that attention to distributiveissues ought to berestored.)Second, in one(thoughnot dominant) strand

42.On skepticism regarding the centrality of the GA TTregime, see Strange1988. On the declining importanceof ‘‘publicinternational agencies’ ’ ingeneraland the FAO in particular,see McLin1979. 43.See, forexample, Hopkins and Puchala 1978, especially 598. 44.Hoole 1977. The focus on internationalorganizations as actors providingcollective or redistribu- tivegoods has alonghistory. See Kindleberger1951a; Ascher 1952;Wood 1952; Loveday 1953; Sharp 1953;and Gregg 1966. 45.We focushere oneffects ofinternational regimes because, as arguedlater, we thinkthis is the questionon which future research shouldconcentrate. For a review oftheoriesthat purport to explain internationalregimes, see Haggardand Simmons 1987. 46.Hopkins and Puchala 1978, 598. 47.Ibid., 615– 16. 48.Krasner 1983b,introduction and conclusion. 738 InternationalOrganization ofresearch, attention to the normative aspects of internationalregimes led naturally toconsideration of the subjective meaning of such norms and to aresearchparadigm thatwas insympathy with developments in constructivistschools of thought. 49 (See theessay by MarthaFinnemore and Kathryn Sikkink in thisissue of IO.) Third,by the mid-1980s explanations of internationalregimes became intertwined withexplanations of internationalcooperation more generally. The work of Robert Keohaneespecially drew from functionalistapproaches that emphasized the effi- ciencyreasons for rulesand agreements among regime participants. 50 Based on ra- tionalityassumptions shared by a growingliterature in political economy, this re- searchsought to show that international institutions provided a wayfor statesto overcomeproblems of collectiveaction, high transaction costs, and informationde - citsor asymmetries.This approach has produced a numberof insights,which we will discussand extend later. But its analytical bite— derived from itsfocus on statesas unied rational actors— was purchasedat the expense of earlierinsights relating to transnationalcoalitions and, especially, domestic politics. Furthermore, the strength ofthis approach has largely been its ability to explain the creation and maintenance ofinternational institutions. It has been weaker in delineatingtheir effects on state behaviorand other signi cant outcomes, an issue to which we willreturn. Thisweakness opened the way for animportant realist counterthrust in the late 1980s:the challenge to show that international institutions affect state behavior in anysigni cant way. Some realists, particularly neorealists, raised logical and empiri- calobjections to the institutionalist research agenda. On the logical side, Joseph Grieco51 andJohn Mearsheimer argued that relative-gains concerns prevent states from intensivecooperation. The essence of theirargument was thatsince the benets ofcooperation could be translated into military advantages, states would be fearful thatsuch bene ts would disproportionately owto potential adversaries and therefore wouldbe reluctant to cooperate in substantial, sustained ways. Responses by DuncanSnidal and Robert Powell showed that, even if states did put substantial weighton suchrelative-gains concerns, the circumstances under which they would greatlyinhibit cooperation were quitelimited. Mearsheimer, in his extensive chal- lengeto institutionalism,also argued that the empirical evidence showing that insti- tutionschanged patterns of statebehavior was weak,especially in thearea of security affairs. Whilewe mightdispute the extreme conclusions drawn by Mearsheimer,we takeseriously his challenge to providestronger empirical evidence. In the third sec- tionof thisarticle we suggestlines of institutionalist analysis that should lend them- selvesto rigorous empirical testing, avoiding some of theinferential traps and falla- ciesthat Mearsheimer and other realists have identi ed. 52

49.See Haas 1983;and Ruggie 1972. 50.Keohane 1984. 51.See Grieco 1988;and Mearsheimer 1994. 52.See Snidal1991; and Powell 1991. See alsoBaldwin 1993. InternationalInstitutions 739

Institutions Across the Level-of-Analysis Divide: Insights fromDomestic Politics

Earlystudies of international institutions were oftenmotivated by the attempt to applynew methods used in thestudy of domesticpolitics. As justreviewed, studies ofvoting behavior in theGeneral Assembly, electoral success in the UN governing structure,and surveysregarding attitudinal change as aresultof international organi- zationexperience are all prime examples. Similar studies continue today, for example,in calculationsof powerindexes for memberstates of theEuropean Union. 53 Theseapproaches have not, however, been widely in uential recently and have been subjectto trenchant criticisms. 54 Inspite of thisless-than-promising experience, scholars todayare turning once again to modelsof domestic politics to suggestnew ques- tionsand approaches to the study of international institutions. In this section, we briey considerwhether these new approaches are more likely to bear fruit. Wendreasons to be relatively optimistic about today’ s attemptsto transport modelsacross levels of analysis,as longas suchattempts are undertaken with some caution.In particular, we seesubstantial potential in lookingat theoriesof domestic institutionsthat are rooted in noncooperative game theory. Rationalist theories of institutionsthat fall into the category of the‘ ‘newinstitutionalism’ ’ haveapplicabil- ityat both the domestic and international levels. V irtuallyall the early attempts to applytechniques and research strategies from domesticpolitics to theinternational levelwere implicitlybased on the assumption that agreements among actors are enforceable.Indeed, this was the only assumptionunder which it madesense to look atthe politics that underlay voting and decision making in international institutions atall. Models that assume that agreements will be enforced by aneutralthird party areespecially inappropriate for theinternational setting; calculating voting power in theGeneral Assembly in aworldof unenforceableagreements may have more than a passingresemblance to arranging deck chairs on theTitanic.Thus, it isnot surprising thatthese models have not had great in uence when transported to theinternational level. However,recent models of domestic institutions as a ruledraw, often explicitly, on noncooperativegame theory. The basic assumptions of noncooperativegame theory arethat actors are rational, strategic, and opportunistic, and that no outside actor will stepin to enforce agreements. Therefore, agreements that will make a difference mustbe self-enforcing. These conditions are remarkably similar to the usual charac- terizationof internationalpolitics as asituationof anarchyand self-help. 55 As long as modelsuse the same basic assumptions about the nature of actorsand their environment,the potential for learningacross the level-of-analysis divide could be enor- mous.

53.Hosli 1993. 54.Garrett andTsebelis 1996. 55.Waltz 1979. 740 InternationalOrganization

As oneexample, consider what international relations scholars might learn from lookingat currentdebates on the nature of legislative institutions. 56 Analogouslyto howrealist theory portrays states with a mixtureof commonand con icting interests butwithout supranati onalenforcemen t,these models treat legislato rs asself- interested,individualistic actors in a situationwhere they must cooperate with one anotherto achieve mutual bene ts. 57 Theyask how legislatorsunder these conditions mightconstruct institutions— such as committees or parties—that will allow them to reachgoals such as reelection. 58 Similarly,international relations scholars are inter- estedin how states or otherentities design institutional forms (organizations,procedures,informal cooperative arrangements, treaty arrangements) that assist in there- alizationof theirobjectives. The point is not,asmuchof the earlier literature assumed, that‘ ‘legislativeactivity’ ’ atthe international level is interesting per se. The powerof theanalogy rests solely on howactors choose strategies to cope with similar strategic environments.In general, we suggestthat more progress can be madeby drawingout theaspects of domestic politics that are characterized by attempts to cooperate by actorswith mixed motives, who cannot turn easily to external enforcement, and applyingthem selectively to thestudy of internationalrelations. Thedebate about legislative organization, which we arguemay provide insights intointernational institutions more generally, has been roughly organized into a con- trastbetween informational and distributionalmodels. Informational models concentrateon theways in whichlegislative structures allow legislators to learn about the policiesthey are adopting, thus avoiding inefficient outcomes. 59 Researchershave arguedthat properly structured legislative committees can efficiently signal informa- tionabout the effects of proposedpolicies to the oor,and that informational con- cernscan explain both the pattern of appointmentof legislators to committees and thedecision making rules under which committees operate. All of these claims have stimulatedintense empirical investigation, which has been challenged by the distri- butionalperspective discussed later. Informational models can be used to extendand clarifyarguments in the international literature that stress the role of institutionsin theprovision of information, as Keohanehas argued, and in the learning process, as Ernstand PeterHaas haveemphasized. They can lead to predictions about the condi- tionsunder which international institutions can effectively provide policy-relevant informationto states,about the kinds of institutionsthat can provide credible information,and about the effects of suchinformation provision on patternsof statebehavior.Anexample of an issuearea where these effects might be prominentis environ-

56.The work on legislative institutions is justone example oftheapplication of noncooperativegame theoryto domestic institutions. But since itis a particularlywell-developed literature, we concentrateon it here,without wishing to imply that this is the only branch of research ondomestic institutions that may haveinteresting analogies to international institutions. 57.Shepsle and Weingast 1995. 58.Although much of the work on legislativeorganization concentrates onthe American context,in recent years creative effortshave been made todevelop such models in non-U.S. settings. See Huber 1996b;Tsebelis andMoney 1995; Ramseyer andRosenbluth 1993; G. Cox1987; and Shugart and Carey 1992. 59.See Gilliganand Krehbiel 1990; and Krehbiel 1991. InternationalInstitutions 741 mentalinstitutions, where it is highly likely that the ability of organizations to provide reliable,credible information about the effects of human activities on the environ- mentis akeyfactor in explainingthe success or failure of negotiationson environ- mentaltreaties. Another possible application might be thecreation of international nancialinstitutions, such as the Bank for InternationalSettlements, an original func- tionof whichwas toprovide credible information to marketson Germancreditwor- thiness.60 Withinthe EuropeanUnion, the Commission’s roleas arelativelyindepen- dentcollector of policy-relevantinformation is a plausibleexplanation for itsability toexercise considerable in uence over policy outcomes. 61 Distributionalmodels, on the other hand, assume that information is notall that problematic.Instead, they concentrate on thefact that legislators are heterogeneous intheir tastes, caring differentially about various issues. 62 Achievingmutual gains, in thisframework, means cutting deals that will stick across different issues. Since exchangesof votescannot always be simultaneous,legislators have developed struc- turessuch as committees and agenda-setting rules that allow them to put together majoritieson theissues of mostintense particularistic interest to them. This structure providespredictions about the distribution of benets to individuallegislators. Distri- butionalbene ts owthroughappointment to powerful legislative committees. Like researchersin the informational tradition, those in the distributional tradition have usedsuch models to explainand predict various aspects of legislativeorganization. For example,they argue that committees will be composedof preferenceoutliers— thoselegislators who care most intensely about particular issues— and that such com- mitteeswill be grantedagenda-setting power, which is necessaryto keep cross-issue dealsfrom unravelingon the oor.Distributional models may be especiallyuseful in exploringin arigorousfashion the role of internationalinstitutions in facilitatingor hamperingmutually bene cial issue linkages that have been an importantresearch agendain internationalrelations. 63 Thedebate between informational and distributional models of legislative organi- zationhas been highly productive, in both theoretical and empirical terms. It has providednew insights into the types of problems confronted by legislators,the types ofsolutions available to them, and the role of institutions in democracies. On the empiricalside, it has generated a plethoraof alternativeobservable implications, for example,about the composition of congressionalcommittees or the conditions under whichactors gain gatekeeping or amendment power .Empiricalresearch on both sideshas led to deep insights about how the structure of institutions,such as legisla- tivecommittees, in uences their ability to help individuals overcome collective- actionproblems, and the conditions under which individuals will be willingto del- egatesubstantial decision-ma kingauthority to such institutio ns.Both types of questionsare highly relevant and essential to an understandingof therole of institutionsin international politics as well.For example,the informational model suggests

60.Simmons 1993. 61.See Haas 1989;and Bernauer 1995. 62.Weingast and Marshall 1988. 63.On issue linkage,see Stein1980; and Martin 1992c. 742 InternationalOrganization thatinstitutions should be most in uential in promoting cooperation when they are relativelyindependent, ‘ ‘expert’’ sourcesof information and when such information isscarce and valuable to states. W eshouldexpect this model to be most useful in internationalissue areas characterized by informationasymmetries or in the developmentof expertknowledge (such as nancialand banking regulation). The distribu- tionalmodel predicts that institutions will be most successful in allowingfor credible cross-issuedeals between states when those with the most intense interest in any particularissue dominate policymaking on that dimension and when institutional mechanismsinhibit states from renegingon cross-issue deals, even if performance ondifferent dimensions is notsimultaneous.Institutions that try to copewith environ- mentalprotection and development needs in the same package (such as UNCED and theAgenda 21 program) provide a plausibleexample. For ourinterests, another strikinganalogy between the international arena and the legislative literature is the degreeto which the termsof thedebate— information versus distribution— re ect the emergingdebate about the signi cance of international institutions. Inmany essential respects the problems faced by individual legislators mirror thosefaced by individual states in the international system. Individual actors face situationsin which they must cooperate in order to achieve bene ts but also face temptationsto defectfrom cooperativearrangements. No external authority exists to enforcecooperative agreements; they must be self-enforcing.Self-enforcement takes theform ofexclusion from thebene ts ofcooperation,a coercivemeasure. Given theseanalogies, there is every reason to expectthat some of themethods, insights, andresults of these new studies of legislatorscould usefully inform new studies of internationalinstitutions, inspite of the fact that legislators (usually) operate in a moredensely institutionalized environment .64 Moregenerally, rationalist models of institutionsthat have been developed in domestic settings have the potential to be translatedto the international level. As longas we areconsidering mixed-motive situationsin which actors must cooperate in order to pursue their objectives, the incentivesto construct instituti onsto structure and encourage cooperatio nare similar.

HowInstitutions Matter

Sincethe 1980s,work on internationalinstitutions has been de ned for themost part bythe demand that scholars respond to a realistagenda: to prove that institutions havea signicant effect on statebehavior. While structuring the debate in thisman- nermay have stimulated direct theoretical confrontation, it has also obscured some importantand tractable research paths. Allowing realism to set the research agenda hasmeant that models of international institutions have rarely taken domestic poli-

64.One couldmake asimilar argumentabout domestic theories of delegation. See Epsteinand O’Halloran1997; Lohmann and O’ Halloran1994; and Lupia and McCubbins 1994. The analogy between politiciansdeciding to delegate authorityto bureaucrats or committees andstates delegatingauthority to internationalinstitutions is strong. InternationalInstitutions 743 ticsseriously, treating the state as a unit.The debate has also been reduced to a dichotomy:either institutions matter or theydo not. Insufficient attention has been givento the mechanismsthrough which we mightexpect institutional effects to work. Institutionalists,in responseto realism,have treated institutions largely as indepen- dentvariables, while playing down earlier insights that international institutions are themselvesthe objects of strategic state choice. Treating institutions as dependent variableshas mistakenly been understood as an implicit admission that they are epiphenomenal,with no independenteffect on patternsof behavior .65 Althoughit has been important to gobeyond merely explaining the existence of internationalinstitutions, productive new lines of researchemerge if we acceptthat institutionsare simultaneouslycauses and effects ;thatis, institutions are both the objectsof state choice and consequential. In a rationalist,equilibrium framework, thisstatement is obviousand unexceptionable: states choose and design institutions. Statesdo sobecause they face certain problems that can be resolvedthrough institu- tionalmechanisms. They choose institutions because of theirintended effects. Once constructed,institutions will constrain and shape behavior, even as they are con- stantlychallenged and reformed by theirmember states. In this section, we outlinea numberof linesof research that show promise to take us beyondthe ‘ ‘dotheymatter ordon’ t they’’ structureof research on internationalinstitutions. Thefollowing research agenda is rmlyin the rationalist tradition. Although this approachallows for substantialvariation in patterns of preferencesover outcomes, andindeed provides predictions about outcomes based on exogenouschange in such preferences,it provides relatively little explanatory leverage with respect to the sources ofchange in such preferences. A few wordson how this agenda is related to the constructivistresearch program may be in order .Tothedegree that constructivist approachesprove powerful at makingchanges in actors’ fundamental goals endog- enous,providing refutable hypotheses about the conditions for suchchange, the con- structivistand rationalist approaches will be complementary. Although rationalist approachesare generally powerful for explaininghow policy preferences change whenexternal constraints or information conditions change, alternative approaches, suchas constructivism, are necessary for explainingmore fundamental, internal changesin actors’ goals. However, the rationalist research program has much to contributeeven without strong theories about the reasons for changein actors’goals. Oneof the core insights of theories of strategic interaction is that, regardless of actors’speci c preferences,they will tend to facegeneric types of cooperationprob- lemsover and over again. Many situations give rise to incentives to renegeon deals orto behavein time-inconsistent ways that make actors happy in theshort run but regretfulin the long run. Likewise, many situations of strategicinteraction give rise tobene ts from cooperation,and con icts over how to divide up this surplus will plaguecooperative efforts. Thus, considerations of howto prevent cheating and how toresolve distributional con ict, to give two prominent examples, are central to theoriesof cooperation regardless of the speci c goalsof actors. Rationalist ap-

65.Mearsheimer 1994. 744 InternationalOrganization proachesare powerful because they suggest observable implications about patterns ofcooperation in thefaceof suchdilemmas, even absent the kind of preciseinforma- tionabout preferences that scholars desire. It is to suchdilemmas that we nowturn ourattention.

CollaborationV ersusCoordination Problems Themost productive institutionalist research agenda thus far ininternational rela- tionshas been the rationalist– functionalist agenda, originating with Keohane’ s After Hegemony andSteve Krasner’ s editedvolume on internationalregimes. 66 This work was informedby a fundamentallyimportant insight, inspired by themetaphor of the Prisoners’Dilemma (PD). Individuallyrational action by statescould impede mutu- allybene cial cooperation. Institutions would be effective to the degree that they allowedstates to avoid short-term temptations to renege, thus realizing available mutualbene ts. Someauthors, recognizing that PD was onlyone type of collective-action problem,drew adistinctionbetween collaboration and coordination problems. 67 Collabo- rationproblems, like PD, arecharacterized by individualincentives to defectand the existenceof equilibria that are not Pareto optimal. Thus, the problem states face in thissituation is ndingways to bind themselves and others in order to reach the Paretofrontier .Incontrast, coordination games are characterized by the existence of multiplePareto-optimal equilibria. The problem states face in thissituation is notto avoidtemptations to defect,but to chooseamong these equilibria. Such choice may berelatively simple and resolved by identication of afocalpoint, if theequilibria arenot sharply differentiated from oneanother in terms of thedistribution of ben- e ts.68 Butsome coordination games, like the paradigmatic Battle of theSexes, in- volvemultiple equilibria over which the actors have strongly divergent preferences. Initially,most authors argued that institutions would have little effect on patterns of statebehavior in coordination games, predicting substantial institutional effects only incollaboration situations. Interestingly, these arguments led both to expectations aboutinstitutional effects on state behavior and tostateincentives to delegateauthor- ityto institutions, consistent with the kind of equilibrium analysis we ndmost promisingfor futureresearch. As thelogic of modern game theory has become more deeply integrated into internationalrelations theory, and as authors have recognized the limitations of the collaboration–coordination distinction, we havebegun to see work that integrates the efficiencyconcerns associated with collaboration and the distributional concerns as- sociatedwith coordination. Krasner madea seminalcontribution to this line of analysis.69 He arguedthat when states are attempting to cooperatewith one another, achiev- ingefficiency gains— reaching the Pareto frontier— isonly one ofthe challenges they

66.See Keohane1984; and Krasner 1983b. 67.See Snidal1985a; Stein 1983; and Martin 1992b. 68.Garrett andWeingast 1993. 69.Krasner 1991. InternationalInstitutions 745 faceand often not the most difficult one. Many equilibria may exist along the Pareto frontier,and specifying one of theseas the locus of cooperation,through bargaining andthe exercise of statepower, dominates empirical examples of international cooperation.Krasner’ s insightis perfectlycompatible with the folk theorems of noncooperativegame theory that show that repeated play of a PD-typegame gives rise to many—infact, in nite— equilibria. Thus, repetition transforms collaboration prob- lemsinto coordination problems. In most circumstances, states have simultaneously toworry aboutreaching efficient outcomes and resolving distributional con ict. Oncewe recognizethis fact, our approach to international institutions becomes bothmore complex and more closely related to traditional international relations concernsabout power and bargaining. T obeeffective, institutions cannot merely resolvecollaboration problems through monitoring and other informational functions.They must also provide a mechanismfor resolvingdistributional con ict. For example,institutions may construct focal points, identifying one possible equilib- riumas the default or ‘‘obvious’’ one,thus reducing state-to-state bargaining about thechoice of a particularpattern of outcomes. The role of the European Court of Justice(ECJ), discussedelsewhere in this article, is captured in part by this type of constructedfocal-point analysis. The Basle Banking Committee’ s rolein devising internationalstandards for prudentialbanking practices similarly helped to coordi- natenational regulations where a numberof plausible solutions were available. 70 Wherestates fear thatthe bene ts ofcooperation are disproportionately owingto others,institutions can provide reliable information about state behavior and the realizedbene ts of cooperationto allay such fears. Tradeinstitutions perform many functions;one function that could stand more analytical scrutiny is the provision of suchinformation about the distribution of bene ts among members. Another way institutionscould mitigate distributional con ict is to‘‘keepaccount’ ’ ofdealsstruck, compromisesmade, and gains achieved, particularly in complexmulti-issue institutions.The networks created within the supranational institutions of the European Union,for example,provide the necessary scope for issue-linkageand institutional memoryto performthe function of assuringthat all members, over time, achieve a reasonablyfair shareof the bene ts of cooperation. 71 Unlessthe problem of equilib- riumselection is resolved,all the third-party monitoring in theworld will not allow for stableinternational cooperation. Thus,a promisingline of researchwill involve bringing distributional issues back intothe study of international institutions, issues that were infact the focus of some ofthe early regimes literature discussed earlier. Institutions may interact with distributionalcon ict in a numberof ways.Most simply, they re ect and solidify settle- mentsof distributional con ict that have been established through more traditional means.These means include the exercise of statepower, which Krasner emphasizes, marketdominance, and alternative methods of bargaining such as making trades acrossissues. 72 Inthis perspective, institutions can make a differenceif theylock in a

70.Simmons 1998. 71.Pollack 1997. 72.Fearon 1994a. 746 InternationalOrganization particularequilibrium, providing stability. But rather than merely re ecting power in anepiphenomenal fashion, as realists would have it, institutions in thisformulation preventpotential challengers from underminingexisting patterns of cooperation,ex- plainingwhy powerful states may choose to institutionalizethese patterns rather than relyingsolely on adhoccooperation. Institutionsmay also serve a lesscontroversial signaling function, therefore mini- mizingbargaining costs. This would be thecase if institutions construct focal points orif theyprimarily keep account of thepattern of benets over time, as discussed earlier.Ineither case, they effectively increase path dependence. Once a particular equilibriumis chosen, institutions lock it in.Researching the ways in which institu- tionsdo this—how do theyenhance path dependence, and under what conditions?— wouldbe intriguing.Normative questions also rise to the top of theagenda once we recognizethe lock-in role of institutions. If theydo in fact solidify a patternof cooperationpreferred by the most powerful, we shouldquestion the ethical status of institutions,turning our attentionto equity, as well as efficiency,questions. Inthe most traditional, state-centric terms, institutions re ect and enhance state power;in T onyEvans and Peter Wilson’ s words,they are ‘ ‘arenasfor actingout powerrelations.’ ’73 Onthe other end of the spectrum, we maywant to ask about situationsin which institutions play a moreactive role in resolving distributional conict. Perhaps institutions sometimes do more than lock in equilibria chosen through theexercise of state power, having an independent part in the selection of equilibria. Suchan argumenthas beenmade most clearly in the case of the ECJ. Here, Geoffrey Garrettand Barry R. Weingast ndthat there are a numberof ways in which the EuropeanCommunity could have realized its goal of completingthe internal market.74 TheECJ madea bigdifference in thecourseof Europeanintegration because it was ableto construct a focalpoint by choosing one of these mechanisms, that of mutualrecognition. This choice had clear distributional implications but was ac- ceptedby member states because it was aParetoimprovement over the reversion pointof failing to completethe internal market. A distinctresearch tradition empha- sizesthe legitimizing role that international institutions can play in focal-pointselec- tion.Some scholars point out that institutionally and legally enshrined focal points cangain a highdegree of legitimacy both internationally and domestically. 75 This legitimacy,in turn,has important political consequences. 76 Todevelop a researchagenda on how institutions resolve problems of multiple equilibriaand distribution, we wouldhave to build on these insights to ask condi- tionalquestions. When are states, particularly the powerful, willing to turn the problemof equilibrium selection over to an institution? What kinds of institutions are mostlikely to perform this function effectively— those that are strategic or those that

73.Evans and Wilson 1992. 74.Garrett andWeingast 1993. They also argue that the multiple equilibria were notsharply distin- guishedfrom one another in terms ofefficiency anddo not concentrate on distributionalcon ict among equilibria.They have been criticized onthese points.See Burleyand Mattli 1993. 75.See Franck1990; and Peck 1996, 237. 76.Claude 1966, 367. InternationalInstitutions 747 arenaive; those that rely on politicaldecision making or those that rely heavily on relativelyindependent experts and/ orjudicialprocesses; those that broadly re ect the membershipof the institution or thosethat are dominated by the powerful? Under whatconditions are constructed focal points likely to gain international recognition andacceptance? Overall, bringing the traditional international relations focus on distributionalcon ict back into the study of internationalinstitutions holds the poten- tialfor generatingresearchable questions that are both positive and normative in nature.

InternationalInstitutions and Domestic Politics Inallowing their agenda to be dened by respondingto therealist challenge, institu- tionalistshave generally neglected the role of domestic politics. States have been treatedas rational unitary actors and assigned preferences and beliefs. This frame- workhas been productive in allowing us tooutline the broad ways in which institu- tionscan change patterns of behavior.Butinprivileging the stateas anactor,we have neglectedthe ways in whichother actors in international politics might use institutions(a centralinsight of earlier studies of transgovernmentalorganization) and the waysin which the nature or interestsof thestate itself are potentially changed by the actionsof institutions(an implication of theearly neofunctionalist literature). Here we outlinea few linesof analysis that should be fruitful for integratingdomestic politicsand international institutions in a systematicmanner, rather than treating domesticpolitics as aresidualcategory of explanation. Because the lines of analysis herehave foundations in speci c analyticalframeworks withexplicit assumptions, applyingthem to theproblem of internationalinstitutions should result in productive researchpaths, rather than merely the proliferation of possible ‘ ‘explanatoryvari- ables’’ thathas characterized many attempts to integratedomestic politics and internationalrelations. W eshouldnote that bringing domestic politics back into the study ofinternational institutions is an agenda that should be understood as analytically distinctfrom thatof applyinginstitutionalist models developed in thedomestic set- tingto the international level, an agendaaddressed elsewhere in thisarticle. As we willargue, one of themore fundamental ways in which international insti- tutionscan change state behavior is by substitutingfor domesticpractices. If policies formerlymade by domesticinstitutions are now made on theinternational level, it is reasonableto expectsubstantial changes in thepatterns of worldpolitics. Three re- latedquestions are central to understandingthe relationsbetween domestic and internationalinstitutions. First, under what conditions might domestic actors be willing to substituteinternational for domesticinstitutions? Second, are particular domestic actorsregularly advantaged by theability to transferpolicymaking authority to the internationallevel? Third, to what extent can international institutional decisions and rulesbe enforcedby domesticinstitutions, and what are the implications for compli- ance?These questions are tied together by the assumption that domestic actors inten- tionallydelegate policymaking authority to the international level when this action furtherspursuit of theirinterests. 748 InternationalOrganization

Domesticinstitutions can at times be a barrierto the realization of bene ts for societyas a whole.Failures of domesticinstitutions can arise through a numberof mechanisms.Perhaps most obviously, domestic institutions can be capturedby preferenceoutliers who hold policy hostage to their demands. Recent research suggests thatthis may be thecase with respect to the settlementof territorialdisputes between borderingstates in someregions: repeated failure to ratifyborder agreements in the legislatureis oneof themost important domestic political conditions associated with thewillingness of statesto submit their disputes to internationalarbitration. 77 More generally,this situation is likelyto arisewhen some actors, such as thoselooking for particularisticbene ts, nditeasierto organizethan do actorsmore concerned with thewelfare ofthe average citizen. Such is the story often told about trade policy. Import-competingproducers and others with an interest in protectionist policies may ndit easierto organize than those who favor free trade,a coalitionof exporters and consumers.This differential ability to organize will bias policy in favor of protection, decreasingoverall welfare. Transferring the policymaking process to the internationallevel, where exporters can see that they have a stakein organizationin order to gainthe opening of foreignmarkets, can facilitate a moreevenhanded representation ofinterests. Those actors who have the most to gain from pursuitof generalwelfare— suchas executives elected by anationalconstituency— will show the most interest in turningto international institutions under such circumstances. Judith Goldstein pro- videsan analysisalong these lines when she explains the paradox of theU.S. presi- dentagreeing to bilateral dispute-resolution panels in the U.S.– Canada Free Trade Act(FT A), inspiteof thefact that these panels predictably decide cases in a waythat tendsto denyprotection to U.S. producers. 78 Wecanidentify other incentives for domesticactors to transfer policymaking to theinternational level. One common problem with institutions that are under the controlof politicalactors is thatof time-inconsistent preferences. Although running anunexpectedly high level of ination today may bring immediate bene ts topoliti- ciansup for reelection,for example,allowing monetary policy to be made by politi- cianswill introduce a welfare-decreasingin ationary bias to the economy. Putting additionalconstraints on policy, for example,by joining a systemof xedexchange ratesor acommoncurrency area, can provide a mechanismto overcome this time- inconsistencyproblem, as argued by proponents of a singleEuropean currency. In general,if pursuitof gainsover time involves short-term sacri ces, turning to internationalinstitutions can be anattractive option for domesticpolicymakers. Asecondand related question about domestic politics is whetherparticular kinds ofactors will regularly see an advantage in turningto the international level. At the simplestlevel, it seems likely that ‘ ‘internationalist’’ actors—those heavily engaged ininternational transactions, 79 thosewho share the norms of internationalsociety, 80

77.Simmons 1998. 78.See Goldstein1996; and Gilligan 1997. 79.Frieden 1991. 80.Sikkink 1993a. InternationalInstitutions 749 orthose who have a stakein a transnationalor global resource 81 —willhave an interestin turning to the international level. This may especially be the case when suchgroups or parties are consistently in a minorityposition in domestic politics. Drawingon these ideas, we couldbegin to develop hypotheses about the kinds of domesticinterest groups that will most favor transferring some authority to the internationallevel. Certaindomestic institutional actors may also have a tendencyto bene t from international-levelpolicymaking. One such actor, which is just beginning to enter politicalscientists’ analysis of internationalinstitutions, is thejudiciary. Increasingly, internationalagreements are legal in form. Thismeans that they often are interpreted bydomestic courts, and that judges can use international law as a basison which to makejudgments .82 Becauseinternational law provides this particular actor with an additionalresource by which to pursue agendas, whether bureaucratic or ideological, we mightexpect that the judiciary in generaltends to besympatheticto international institutions. Overall,as we worktoward more sophisticated speci cation of thecausal mechanismsthrough which institutions can in uence behavior, we willhave to pay much moreattention to domesticpolitics than studies of internationalinstitutions have thus far.Thedevelopment of generaltheories of domesticpolitics provides an opening for systematicdevelopment of propositionsabout domestic actors. W enolongerneed to treatthe domestic level as merelythe source of statepreferences, nor as aresidual categoryto explain anomalies or patterns of variation that cannot be explained by internationalfactors. Instead, we canmove toward genuinely interactive theories of domesticpolitics and international institutions, specifying the conditions under which certainactors are likely to prefer thatpolicy be madeon theinternational level. This focusallows us to specify conditions likely to leadto thedelegation of policymaking authorityto theinternational level, some of whichwe haveoutlined here.

UnanticipatedConsequences Ina rationalistframework, institutions are both the object of statechoice and consequential.The link that ties these two aspects of institutionstogether, and allows the analystto develop refutable propositions about institutions within an equilibrium framework,is theability of actors to anticipatethe consequences of particulartypes ofinstitutions. For example,in the preceding discussion of domestic politics, we assumedconsistently that domestic actors were ableto gauge with some degree of accuracythe ways in which working within international institutions would affect theirability to pursuetheir material or ideationalgoals. Therationalist approach stands in distinction to a historicalor sociological ap- proachto institutions. 83 Theseapproaches see institutions as more deeply rooted and

81.Y oung1979. 82.See Alter 1996;and Conforti 1993. 83.See Steinmo,Thelen, and Longstreth 1992; and Pierson 1996b. Historical institutionalism stresses thepath-dependent nature of institutions,explaining why apparently inefficient institutions persist. Socio- 750 InternationalOrganization draw attentionto their unanticipated consequences. Although we mayquestion whether manyinternational institutions reach the same degree of ‘ ‘taken-for-grantedness’’ thatwe seein domestic politics or smaller-scalesocial relations, it seems undeniable thatthey sometimes have effects that surprise their member states. It is important to differentiatebetween unintended and unanticipated effects. Effects may be antici- patedbut unintended. For example,it is generally expected that arrangements to lowerthe rate of in ation will lead to somewhat higher levels of unemployment. Thus,higher unemployment is ananticipated, although unintended, consequence of stringentmonetary policies. It is best understood as a priceactors are sometimes willingto bearto gain the bene ts oflow in ation. Such unintended but anticipated consequencesof institutionspresent little challenge to arationalistapproach, since they tneatlyinto a typicalcost-bene t analysis.Genuinely unanticipated effects, however,present a largerchallenge. Specic examplesof apparentlyunanticipated consequences of internationalinsti- tutionsare not difficult to nd.States that believed that human-rights accords were nothingbut meaningless scraps of paperfound themselves surprised by theability of transnationalactors to usethese commitments to forcegovernments to change their policies.84 Inthe European Community, few anticipatedthat the ECJ wouldhave the widespreadin uence on policy that it has. 85 Prime MinisterMargaret Thatcher was apparentlyquite surprised at theresults of agreeingto changevoting rules within the EuropeanCommunity, such as theadoption of qualied-majority voting, which she acceptedin the Single European Act. 86 How mighta rationalistapproach deal with these events? One productive ap- proachmight be to attempt to specifythe conditions under which unanticipated con- sequencesare most likely. This speci cation would at leastallow us tosuggestwhen asimplerationalist model will provide substantial explanatory leverage and when it mightbecome necessary to integrate the insights of other schools of thought. If unanticipatedconsequences dominate political outcomes, we wouldhave to draw on alternativesto rationalistmodels in a waythat goes far beyondusing them as away tospecify preferences and goals. Here, we beginspecifying when unanticipated con- sequencesare most likely to confoundpatterns of internationalcooperation. Inductively,it appears that changes in secondary rules— that is, rules about rules— arethe changes most likely to work in unexpected ways. Changes in voting rules withinan institution, for example,can give rise to new coalitions and previously suppressedexpressions of interest,leading to unpredictedpolicy outcomes. Changes indecision-making procedures can have even more widespread and unexpected ef- fectsif theyopen the policy process to input from newactors. Many examples of unanticipatedconsequences arise from decision-makingprocedures that provide ac- cessto nongovernmental and transnational actors, as, for example,Kathryn Sik- logicalinstitutionalism emphasizes thesocial natureof institutions, stressing their role in de ning individuals’identities and the fact thatmany important institutions come tobe takenfor granted and therefore notseen as susceptibleto reform. 84.Sikkink 1993a. 85.Burley and Mattli 1993. 86.Moravcsik 1991. InternationalInstitutions 751 kink’s workhas shown. 87 Bothas sourcesof newinformation and as strategicactors intheir own right, such groups are often able to use new points of access to gain unexpectedleverage over policy. Changes in decision-makingrules will have wide- spreadeffects on a varietyof substantive rules and are thus more likely to have unanticipatedeffects on outcomes than changes in substantive rules themselves. If thisobservation is correct,we shouldsee more unanticipated consequences in situa- tionsthat have relatively complex and permutable secondary rules, such as legalized institutions.Traditional state-to-state bargaining with a unitveto, which has little secondaryrule structure, should provide less opportunity for nonstateactors or coali- tionsof theweak to inuence outcomes unexpectedly. Onequestion that often arises, especially in theinternationalarena, is why govern- mentsare willing to live with unanticipated outcomes. After all,participation in internationalinstitutions is voluntary. If unpleasantand unexpected outcomes fre- quentlyoccur, states as sovereign actors retain the right to pull out of institutions. Whymight they choose to remainin? The trivial answer is that the bene ts ofremain- ingin are greater than the costs. But we canturn this answer into something non- trivialby thinking about the conditions when institutional membership is likely to providethe greatest bene ts. Some of thesehave been spelled out in functionalist theory.Keohane argues that the demand for internationalinstitutions will be greatest underconditions of interdependence, when states face a densenetwork of relations withone another and where information is somewhat scarce. 88 Wecouldgeneralize thatstates are least likely to be willing to withdrawfrom aninstitution in theface of unanticipatedconsequences when they are dealing with issues that exhibit increasing returnsto scale,which, in turn,create conditions of path dependence. Consider the creationof regionaltrading arrangements in the1990s. These arrangements provide theirmembers with economic bene ts, and those on theoutside of the arrangements ndthemselves losing investment and trading opportunities. W ethereforesee east- ernEuropean, Caribbean, and other states clamoring to become members of the relevantregional trading arrangements. This is a goodexample of how increasing returnsto scale create a highdemand for institutionalmembership. Under these conditions,it seemslikely that these states will be willingto putup witha highlevel of unexpectedoutcomes before they would seriously consider withdrawing from an institution.However, this example begs the question of whethertrade agreements are likelyto have substantial unanticipated effects. They are only likely to do soin the caseof rapidtechnological change or largeinternational economic shocks, such as theoil shocks of the1970s.

Typologyof Institutional Effects As we turnour attention to the problem of how,notjust whether, international insti- tutionsmatter, it becomes essential to understand alternative mechanisms through whichinstitutions might exert their effects. To prod our thinking in this direction, we

87.Sikkink 1993a. 88.Keohane 1983a. 752 InternationalOrganization introducea preliminarytypology of institutionaleffects. The reasoning behind this typologyis that different institutions, or perhapssimilar institutions in differentset- tings,will have different types of effects.Specifying these effects will not only allow usto developbetter insights into the causal mechanisms underlying the interaction betweeninstitutions and states or societies. It will also provide for moretestable propositionsabout how and when we shouldexpect institutions to exertsubstantial effectson behavior. Thetypology we suggestis analyticallyinformed but aims rst toprovide a lan- guagefor describingpatterns of changein statebehavior after creation of aninternationalinstitution. Here we spellout the typology and present some illustrative examples.The nextstep will be tolinkthe typologyto causal processes, and we suggest somepreliminary ideas along these lines. W ebeginby suggestingtwo types of institutionaleffects: convergence and divergence effects.Of course,the null hypothesis is thatinstitutions have no effect.Development of acleareranalytical framework may forceus to consider situations in which we combineeffects: for example,perhaps sometypes of states are subject to convergence effects and others to divergence effects. Webeginwith convergence effects, since the logic of most rationalist, economis- tic,and functionalist theories of international institutions leads us to expect such effects.These models posit goals that states nditdifficult to achieveon theirown, whetherfor reasonsof time-inconsistent preferences, collective-action problems, old- fashioneddomestic political stalemate, or otherfailures of unilateral state action. In thisfunctionalist logic, states turn to international institutions to resolvesuch problems;institutions allow them to achieve bene ts unavailablethrough unilateral ac- tionof existing state structures. Functionalist analysis sees international institutions asimportant because they help states to solve problems. Many of these problems havetheir roots in thefailures of domesticinstitutions, and their resolution involves turningsome types of authorityover to the international level. Once policy is del- egatedto an international institution, state behavior will converge: members will tendto adopt similar monetary, trade, or defense policies. Whathas been missing from functionalistaccounts of institutionalizatio nisthe systematicconnection between domestic political conditions and incentives to con- structand comply with international institutions. But once we recognizethat internationalinstitutions may make a differencebecause they effectively substitute for domesticpractices (making policy decisions, setting policy goals, or undertaking monitoringactivities), our attention turns to the domestic political conditions that makesuch substitution a reasonablepolicy alternative. If domesticinstitutions are thesource of persistent policy failure, if they somehow prevent the realization of societalpreferences, or iftheyinterfere with the pursuit of mutualbene ts withother states,turning functions over to the international level can enhance national welfare.89 Monetarypolicy is a primeexample of this logic. Other examples might

89.Some would argue that this process isantidemocratic. See Vaubel1986. However, such an argument rests onweak foundations.First, it assumes thatdomestic institutions are necessarily responsiveto InternationalInstitutions 753 includetrade policy, if domestictrade policy institutions are captured by protection- ists;or environmentalpolicy, if domestic institutions encourage a short-termrather thana long-termperspective on theproblem.Thinking about the logic of substitution requiresmuch more attention to inefficient domestic politics than most functional theorieshave provided to date. Aclassicexample of international institutions acting as substitutesfor domestic institutionsand therefore having convergence effects lies in arguments about why high-ination states such as Italy might choose to enterthe European Monetary Union (EMU).90 Highin ation is a publicbad, leading to lower overall welfare thanlow ination. However, the short-term bene ts to politicians from allowingspurts of unanticipatedin ation make it difficultto achievelow rates of ination unless institutionsthat set monetary policy are independent of political in uence. 91 Thus,trans- ferringauthority to aninstitution that is relatively insulated from politicalin uence, andthat itself has a preferencefor lowin ation, can provide overall welfare benets for thecountry. This is thelogic that leads a statelike Italy to takethe unusual step (for arelativelyrich, developed country) of transferringa coreaspect of sovereignty— controlover the currency— toa EuropeanCentral Bank. Giventhis logic of delegation,states that become members of theEMU shouldsee aconvergencein theirrates of ination. 92 Althoughthe debate rages among econo- mistsabout whether the European Monetary System has in factworked in this man- ner,93 thereis littledoubt that one of themajor motivations for monetaryunion is for high-ination states to ‘ ‘import’’ lowGerman rates of in ation, leading to similar ination rates in all member states. If we lookedat the variation in in ation rates priorto entryinto monetary union (or intoa monetarysystem more generally), and comparedit to in ation rates after entry, we shouldsee a declinein the level of variation. Althoughmonetary union is aprominentand intriguing example of convergence effects,we canimagine a similardynamic in otherissue areas as well.Environmental institutionsshould lead to convergence of environmentalindicators, such as carbon dioxideemissions. 94 Human-rightsinstitutions acting as substitutesshould lead mem- bersto adoptincreasingly similar human-rights practices. Even if full convergence doesnot occur,the majoreffect of aninstitution that is actingas asubstitutewill be to bringstate practices more closely in line with one another. Aconvergenceeffect could be measuredand identi ed bydecreasedvariation in relevantindicators of state practices, whether in ation rates, pollution, or human- nationalpreferences. Forthe kinds of reasons justdiscussed, such as time-inconsistentpreferences, or institutionalcapture, this assumption is oftenfalse. Second,the argument assumes thatinternational institutionsare necessarily more difficultto monitor, constrain, and in uence thandomestic institutions. Al- thoughthis may beareasonableassumption for some kindsof societal actors andsome states, itis not universallytrue. 90.For a contrastingargument on the logic of EMU,see Gruber1996. 91.Rogoff 1985. 92.Fratianni and von Hagen 1992. 93.See Giavazzi andGiovannini 1989; and W eber 1991. 94.Levy 1993. 754 InternationalOrganization

FIGURE 1. Internationalinstitutions with convergence or divergence effects rightsabuses. The existence of aconvergenceeffect could also be identied through graphicalmeans. Figure 1 givesan example.On the x-axis,states are arrayed in order oftheir performance on theoutcome dimension, say in ation rates. These rates are indicatedon the y-axis.The solid line indicates in ation rates prior to entry into monetaryunion. Its steepslope indicates that the states exhibit substantial variation inin ation rates. The dotted line represents the outcome of monetaryunion acting as substitute,causing convergence in in ation rates. The more shallow slope indicates lessvariation than observed before entry into monetary union. Thenotion that international institutions might substitute for domesticones under- liesfunctionalist theories of institutions.However, some empirical work on the ef- fectsof institutionshas found a patternquite different from theconvergence of out- comespredicted by such a mechanism.Instead, some authors have found that the primaryeffect of institutions is to exaggerate preexisting patterns of behavior.For example,Andrew Moravcsik has found, in a regionalcomparison of human-rights institutions,that these institutions only led to an improvement in practices in those statesthat already exhibited a highlevel of respectfor humanrights. 95 Thus, West Europeanstates, through participation in institutions, have improved their already verygood human-rights records, whereas Latin American states, according to his evidence,show little impact of institutional participation. Thispattern suggests that international institutions sometimes lead to divergence ofstate practices, in effect complementing and magnifying preexisting tendencies ratherthan overriding them. In this case, institutions will have a divergenceeffect. Thiseffect results when states whose initial practice falls far from institutionalguide- lineswill show little change from behavior,whereas those near the guidelines move

95.Moravcsik 1995. InternationalInstitutions 755 evencloser to them. In contrast, a convergenceeffect appears when institutions exert theirgreatest in uence on precisely those states whose behavior deviates substan- tiallyfrom institutionalnorms. Divergence is likelyto emergewhen institutions ex- aggeratedomestically generated tendencies of state behavior or whenthey primarily mimicdomestic institutions. Anne-Marie Slaughter has argued something along these linesin pointing out that liberal states are the ones most likely to createand abide by relativelyliberal international institutions. 96 Accordingto this logic, liberal institu- tionswill change the behaviorof liberalstates but notilliberal ones, leading to diver- genceof statebehavior. Adivergenceeffect means that those states that already come close to institutional normswill move further toward them, whereas the behavior of those that deviate from suchnorms will remain unchanged. If we were todevelop a measureof state behavior,we wouldsee a divergenceeffect in increased variation of state behavior afterinstitutional creation. W ecanalso illustrate divergence effects graphically, as in Figure1. Here, institutionaleffects result in asteeperline, indicating greater diver- gencein the relevant outcome variable. For easeof comparison,we continueto use theEMU-in ation example. Although such an outcome seems unlikely in practice, for thesake of argument we couldimagine that monetary union that allowed for decentralized,unconstrainedscalpolicymakin gwhileproviding additional re- sourcesto cover national debts could lead to such a perverseoutcome. Another, perhapsmore plausible, example of a divergenceeffect is in the area of overseas developmentaid. In the 1970s, OECD countriesagreed to devotea setpercentage of theirGDP ,0.7percent, to development assistance. Although some countries have comeclose to providingthis level of aidand use the target gureas a toolin domestic debates,others have wholly neglected this target and instead decreased the percent- ageof theirnational income that they devote to foreign aid. If thistypology provides a usefulway to describe alternative institutional effects, thenext challenge is to begin to link up these patterns of behavior to alternative causalmechanisms. This project appears promising, and we outlinepreliminary ideas here.As suggestedearlier, institutions that lead to convergence of statebehavior link upnicely to the functionalist approach that has dominated studies of international institutions,regimes, and organizations over the last fteenyears. In this situation, thefailure of domesticinstitutions or ofunilateralstate action creates incentives to relyon internationalmechanisms. The kinds of problemsthat would prompt states to useinternational institutions that lead to convergenceof behavior are relatively well understood.They include time-inconsistency problems that create incentives for states tobind themselves and collective-action problems among states or withinpolities. Whenstates turn to international institutions as the result of such problems, and whenthese institutions are operating as intended, we wouldexpect to see convergenceof statebehavior. Theconditions that would prompt states to useinstitutions that lead to divergence ofbehavior are not as well understood. W ecanbegin by noting that states facing

96.Slaughter 1995. 756 InternationalOrganization collective-actionproblems, such as aPDoracoordinationgame, would be unlikely torely on aninstitution that exaggerated differences in state behavior. The fundamen- talproblem in suchcases is to create incentives for statesto adopt similar policies: free trade,stringent scalpolicies, arms control,and so on. 97 Insuch a situation,an institutionthat led to increased divergence of statepractice would quickly become irrelevantas states ignored its constraints. Thus, one initialexpectation is thatinstitu- tionsshould not lead to divergence in situations where incentives exist to adopt similarpolicies, as whenstrong externalities to divergent or unilateralstate behavior exist.Perhaps this helps us understandwhy we appearto see some divergence effects inthe human-rightsissue area. Although human rights are a matterof concernaround theglobe, human-rights practices usually do not involve the kinds of externalities andincentives for strategicinteraction that exist in issue areas such as theenviron- mentor monetarypolicy. 98 However,lack of externalities does not provide a directanswer to why divergence wouldoccur. To understand this effect, it islikely that we needto consider domestic politics,returning us toan argument made earlier in thisarticle. International agreements,even those without enforcement mechanisms such as the OECD aidtarget, canprovide ‘ ‘hooks’’ bywhichinterest groups that favor the international agreement canincrease their in uence on the domestic agenda. For example,in Scandinavian countriesthe OECD targethas become a potentarguing point in parliamentary debates.In states without a well-organizedgroup to grab onto this hook, or in those witha moreclosed political process, agreements without enforcement mechanisms orsubstantial pressure from otherstates to comply are unlikely to have any effect. Thesecontrasting domestic political dynamics are likely to giverise to divergence of statebehavior among members of the institution. Arationalistresearch agenda for thestudy of internationalinstitutions is rich and promising.This agenda begins by recognizing that, in equilibrium, institutions are bothcauses and effects, and that empirical researchers must begin to consider the questionof howinstitutionsmatter, not just whether they do. Thinking in theseterms turnsour attention to theproblem of howinstitutions might resolve bargaining and distributionalcon ict as well as themore recognized problems of cheating.It forces usto differentiate anticipated from unanticipatedeffects of institutions and to ask aboutthe conditions under which unanticipated effects are most likely. Rationalist theoriesprovide a mechanismfor bringingdomestic politics more systematically intothe study of internationalinstitutions, an area of researchthat has been slighted bythe development of the eldthus far .Finally,a rationalistapproach allows us to

97.There may besome coordinationsituations, for example, some discussedby Simmons,in which thesolution to the coordination problem does not involveadoption of similar policiesby all states butclear divisionof responsibilitiesamong states. See Simmons1994. The Bretton Woods systems, forexample, coordinatedstate behaviorby creatingexpectations that the United States wouldbehave differently from othermembers ofthesystem. 98.Donnelly surveys the landscape of human rights regimes. See Donnelly1986. There may beexcep- tionsto the generalization that international strategic interactionon humanrights is minimal,for example, whensevere human-rightsabuses lead tomassive refugee ows.This kind of logic could lead totestable propositionswithin the issue area, forexample, that institutions should function differently when such externalitiesexist than under ‘ ‘normal’’ circumstances. InternationalInstitutions 757 begindifferentiating between different types of institutionaleffects and developing refutablepropositions about the conditions under which we aremost likely to ob- servesuch effects.

Conclusions

Studiesof internationalinstitutions have varied in their theoretical sophistication and frequencyover time but have remained a stapleof international relations research andthe pages of IO overthe last ftyyears. In this article we haveexamined the developmentof thesestudies and outlined some promising directions for futurere- searchon international institutions. Early studies of institutions were verymuch problem-driven,focusing on the problems of the postwar world that some hoped internationalorganizations could solve. Although on balancerealistic and insightful, theresults of these studies failed to cumulate,likely due to the lack of adisciplinary ortheoretical framework inwhichto situatethe studies. A morescienti c approach showeditself in a newerwave of work on institutions, drawing on methods and modelsof American politics. But because these models were ingeneralpoorly suited tothe realities of international politics, they failed to generate substantial new insights.It was notuntil the 1980s, with the development of work on international regimesand functionalisttheories, that a moreprogressive research program on insti- tutionsarose. Onefailing of thecurrent research program, however, has been its intensefocus on provingthat institutions matter, without sufficient attention to constructing well- delineatedcausal mechanisms or explaining variation in institutional effects. W e considertwo approaches that might move research beyond this impasse. First, we askwhether applying recent models of domesticpolitics might be moresuccessful thanhave past attempts. W endscopefor optimismhere, since modern theories of domesticinstitutions typically draw onsimilarassumptions of unenforceableagree- mentsand opportunistic behavior by individualsthat characterize most work in internationalrelations. Finally, we turnto somemore speci c researchdirections that are likelyto giverise to importantand testable propositions. These include more careful considerationof distributional issues, the role of domestic politics, unanticipated consequences,and a typologyof institutional effects. As we considerinternational institutions as both objects of strategic choice and consequential,allowing them to serveas both dependent and independent variables inour models, the potential for increasingour understanding of institutionsand of internationalpolitics in general is substantial, as preliminary empirical work has begunto show. The earliest work on internationalinstitutions produced insights that failedto addup tomuchbecause of thelack of ananalytical framework inwhich to situatethese insights; the next generation of work had the bene t ofsucha framework,but one that was poorlysuited to thetask at hand. In this article, we hopeto haveidenti ed linesof researchthat will combine the best of bothworlds: theoreti- callygrounded research on institutionsthat draws onassumptionsthat are appropri- atefor thepersistent problems of internationalrelations.