3449637

Upload: ionela-broasca

Post on 03-Apr-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/28/2019 3449637

    1/23

    Southern Political Science Association

    An Integrated Model of Women's RepresentationAuthor(s): Leslie A. Schwindt-Bayer and William MishlerReviewed work(s):Source: The Journal of Politics, Vol. 67, No. 2 (May, 2005), pp. 407-428Published by: Cambridge University Press on behalf of the Southern Political Science Association

    Stable URL: http://www.jstor.org/stable/3449637 .Accessed: 11/01/2013 01:25

    Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

    .JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

    content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

    of scholarship. For more information about JSTOR, please contact [email protected].

    .

    Cambridge University Press and Southern Political Science Association are collaborating with JSTOR to

    digitize, preserve and extend access to The Journal of Politics.

    http://www.jstor.org

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject to JSTOR Terms and Conditions

    http://www.jstor.org/action/showPublisher?publisherCode=cuphttp://www.jstor.org/action/showPublisher?publisherCode=spsahttp://www.jstor.org/stable/3449637?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/3449637?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=spsahttp://www.jstor.org/action/showPublisher?publisherCode=cup
  • 7/28/2019 3449637

    2/23

    An IntegratedModel of Women'sRepresentation

    LeslieA. Schwindt-BayerUniversity f MississippiWilliam Mishler

    University f ArizonaThe conceptof representation, s developedin HannaPitkin's seminalwork,is a complex structure,whose multiple dimensions are hypothesizedto be closely interconnected.Most empirical work,however,ignores the integratedcharacterof representationand examines its several dimensions inisolation. The picture of representation hat results is not so much incorrect as incomplete.Thisresearch ests anintegratedmodel of representationinkingformal,descriptive,substantive,andsym-bolic representation.Data on the representation f women in 31 democracies confirmsthe intercon-nections among the several dimensions of representation.The structureof electoral systems exertspowerful nfluenceson bothwomen'sdescriptiverepresentation ndsymbolic representation.Descrip-tive representation,n turn, increases legislatures' responsivenessto women'spolicy concerns andenhancesperceptionsof legitimacy.The effects of substantiverepresentation, owever,are much lessthantheory anticipates.

    The concept of representations a rich brocade whose complex weave is notalwaysappreciated.Hanna Pitkin's(1967) seminal treatment dentifiesfour dis-tinct, but interconnectedmeanings or dimensions of representation ncluding:formal representation, eferring o the institutionalrules andprocedures hroughwhich representativesare chosen; descriptive representation,referring to thecompositional similaritybetween representativesand the represented; ubstan-tive representation or responsiveness, referring to the congruence betweenrepresentatives' ctions and the interestsof the represented;andsymbolicrepre-sentation, referringto the represented's eelings of being fairly and effectivelyrepresented.While there are importantdifferencesamong the four dimensions,Pitkin(1967, 10-11) maintains hatthey areproperlyconceived as integralpartsof a coherentwhole. Yet,despitethe frequencyandapprovalwith which Pitkin'swork is cited,most empiricalworkon representationgnoresher integratedcon-ception. Scholars, typically, choose one or two aspects of representationwhileignoringothersthatare not of interestor for which dataarelacking.Thisnot onlycontributes to a "blindman's understandingof the elephant"but also fails toprovide an adequate empiricaltest of a fundamentalaspect of Pitkin'sconcep-tion: its integratedstructure.THE JOURNALOF POLITICS,Vol. 67, No. 2, May 2005, Pp. 407-428? 2005 Southern Political Science Association

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    3/23

    408 LeslieA. Schwindt-Bayerand WilliamMishlerThis researchredressesthis neglect by developing and testing an integratedmodel of Pitkin'sfourdimensions of representationusing cross-nationaldata onthe representation f women. Womenprovidean ideal focus for testing an inte-

    gratedtheory for threereasons. First,women are a large and easily identifiablegroup whose memberspossess many and variedpolitical interests but also arewidely perceived as sharing some common, identifiable"women's interests"(Sapiro 1981).' Second,althoughwomen'srepresentation as improvedmarkedlyin recentyears, women remainunderrepresentedn most countriesaccordingtomany definitions and measures.Third,while many minority groups also haveidentifiable nterestsand are widely underrepresented,t is much more difficultto comparethem systematicallybecause these groupsare so varied;a groupthatis a minorityin one countrycan be a majorityin anotherand absentaltogetherin a third.Women,however,constituteapproximately 0% of the populationvir-tually everywhere.We begin this analysis by elaboratingPitkin'sconcept of representationanddevelopingan integratedmodel of its several dimensions and their interrelation-ships. We proceed to operationalize he model using data on the representationof women across a broadcross-section of democraticsystems. Structural qua-tion methodsare used to test the validityof the multidimensionalconceptionofrepresentationand to refine its structure.Finally,we use the results to discussboth the structureof representation, enerally,and the dynamicsof women'srep-resentation n particular.

    TheMultidimensionalonceptof RepresentationThe fundamentalsof Pitkin'sconcept of representation rewell known.Pitkinconceives of representationas having four primarydimensions. FormalRepre-sentation ocuses on therulesandprocedures egulating he selection andremovalof representatives.Variants nclude accountability theory,which refers to rulesandproceduresallowingtherepresented o sanctionrepresentatives, x post, whofail to act as the representeddesire,and authorization heory,which refersto theabilityof the represented, x ante, to providemandates o representatives.Whilethe existence of free and fair elections are not a necessary condition for formalrepresentation, n practice elections are considered critical and underlie mostattempts o operationalize his dimension(see Powell 2000).Descriptive representation,or "representativeness,"efers to the extent towhich representatives"standfor" the represented.Typically, his means that thecomposition of representativenstitutions should mirrorthe composition of therepresentedin importantrespects. Varieties includefunctional representation,'The idea that women share distinctivepolitical interests is controversial.Womenare a diverse

    groupwith diverse interests that vary along lines of race, ethnicity,class, etc. Certainlyall womendo not agree on exactly whatwomen's interests are.Yet, in generalterms,women are likely to havesome interests that are distinguishablefrom those of men or those of other identity-based groupshelpingto make them suitable for this analysis.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    4/23

    An IntegratedModel of Women'sRepresentation 409which focuses on the occupationalcorrespondencebetweenrepresentativesandrepresented, nd social representation,whichconcernssocial characteristics uchas gender,race, ethnicity,and class (NorrisandFranklin1997). Descriptiverep-resentationarguablys themost studiedof Pitkin's ourdimensionspartlybecausethe compositionof the legislature s highly visible and easily measured.Substantiverepresentations defined as "actingin the interestsof the repre-sented in a mannerresponsiveto them"(Pitkin 1967, 209). AlthoughEulau andKarps (1977) identifya varietyof ways thatrepresentativesmay act on behalf ofthe represented, he most common interpretations that substantiverepresenta-tion refers to policy responsivenessor the extent to which representatives nactlaws and implementpolicies that areresponsiveto the needs or demands of cit-izens.2 While Pitkin considers substantiverepresentation o be the most impor-tant dimension of representationand the heart of the integratedmodel, othersquestionits priority.Forexample,Wahlke(1971) arguesthatpolicy responsive-ness receivestoo muchemphasisgiven evidence thatcitizens possess few coher-ent policy beliefs and that legislators are poorly informed about the policypreferencesof citizens except in exceptionalcases. Nevertheless,policy respon-sivenesscontinues o be considered hecentralaspectof representation y numer-ous scholars,a varietyof whomhaveattempted o measurepolicy responsivenessboth to overallpublic interests(for example, Miller and Stokes 1963; Miller etal. 1999; Stimson,Mackuen,and Erikson 1995; Thomassenand Schmitt 1997)and to race- and gender-based nterests(for example, Bullock 1995; Hero andTolbert1995; O'Regan2000; Reingold 2000).Finally,symbolicrepresentation efersto the extentthatrepresentatives standfor"the representedwith an emphasison symbols or symbolization.Pitkinpro-vides the example of a flag as a symbol representinga nation. What mattersisnot the symbol itself, but "the symbol's power to evoke feelings or attitudes"(Pitkin 1967, 97). Symbolicrepresentations concernednot with who the repre-sentativesareor whattheydo, but how they areperceived andevaluatedby thosethey represent.Wahlke 1971) embracessymbolic representation s themostreal-istic standardgiven the constraintshe perceives on policy responsiveness (seealso Anderson andGuillory 1997; Hibbingand Theiss-Morse 1999;MishlerandRose 1997, 2001).While Pitkin'smultifacetedconception is well known and widely cited, theconnections among its dimensions are frequently ignored in practice. Evenresearchacknowledging hemultidimensionalnatureof representation nd focus-ing on more than one dimensiontypicallytreats hose dimensions as separateanddistinct (see, for example, Marshand Norris 1997; Mishler and Mughan 1978).Pitkinargues againstseparating he dimensionson two grounds.First,she chal-

    2 In addition to public policy, Eulau and Karps(1977) identify three otherways in which repre-sentatives can respondto constituents: ervice responsiveness,which refersto the provisionof par-ticularizedbenefits to individualsorgroups;allocationresponsiveness,which refers to thegenerationof porkbarrelbenefitsfor the constituency;andsymbolic responsiveness,which refersto intangiblegesturesmade in responseto constituent nterests.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    5/23

    410 Leslie A. Schwindt-Bayerand WilliamMishlerlenges the representativequalityof institutions hat manifest one or even severaldimensions of representation ut aresubstantiallyackingthe others.She argues,for example,that a benevolentdictatorship houldnot be consideredrepresenta-tive simply because the dictatoradoptspolicies that provide for citizens' basicneeds (1967, 230-35). Neither should a legislatureto be consideredrepresenta-tive because it "looks like"the public nor because citizens express approval orit. Tobe representative, n institutionmustachieve some minimumon all dimen-sions of representation.Second,Pitkinarguesthat strongcausal connectionsexist amongthe compo-nents of representation.Advocates of formalrepresentation mphasizethatfree,fair, and open elections are importantnot only because they are necessary fordemocracy (Powell 2000; Schumpeter 1947) but also because they facilitatedescriptive representation,encourage policy responsiveness, and enhance thepublic'ssupportfor representativenstitutions.Similarly,descriptiverepresenta-tion is consideredimportant or promotingsymbolic representationand policyresponsiveness,while policy responsiveness is believed to be a principalcon-tributor o symbolic representation Mishler and Rose 1997). It is for both ofthese reasonsthat the concept of representations consideredintegrated.

    An Integrated Model of RepresentationFigure 1 diagramsan integratedmodel showing hypothesized inkages amongthe fourprincipaldimensions of representation.Accordingto this framework,hestructure f the electoralsystem (formalrepresentation)s exogenousanddirectlyinfluencesdescriptiverepresentationlinkA in the model), substantiverepresen-tation (link B), and symbolic representation link C).3The theory is thatpoliti-cal systems with more open and competitiveelections will elect representativeswhose backgroundsmore closely resemble those of the represented.Such po-litical systems also will produce more responsivepolicies which will increasecitizens' confidencein representativenstitutions.Formalrepresentation lso has

    indirect effects on policy responsiveness hrough he mediatingeffect of descrip-tive representation compound link A-D) and on symbolic representationviamultiple paths (links A-E, B-F, andA-D-F). Figure 1 furtherhypothesizes thatdescriptiverepresentationcontributesdirectly to substantiverepresentationbyproducingpolicies moreresponsiveto societal interests linkD). Descriptiverep-

    3 In the long run,the assumption hat formalrepresentations exogenous to descriptiverepresen-tation andpolicy responsivenessprobablycannotbe sustained.As the compositionof the legislaturechanges over time, it is entirelypossible for representatives o enact legislation changing the elec-toral rules of the game. Indeed,the model as diagrammed s a static version of a more general,dynamicmodel in which virtuallyall of the hypothesized inkagescould be conceived as reciprocalover time. Unfortunately,data limitationspreventthe test of a dynamicmodel and necessitate thespecificationof those recursive inkagesthat are most plausiblein the short run.Thus,electoral ruleseffectivelyare fixed andinfluenceboth who is elected and thepolicies they produce.Inthe sameway,the percentageof women in a legislature s more likely, in the shortrun,to affect theresponsivenessof the legislatureto women'sissues, even thoughin the long runthe relationshipmay feed back.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    6/23

    An IntegratedModel of Women'sRepresentation 411FIGURE

    An Integrated Model of Political RepresentationPolicyResponsiveness

    DFormal c SymbolicRepresentation RepresentationDescriptiveRepresentation

    resentationenhances public confidence in representative nstitutions(symbolicrepresentation) othdirectly(linkE) andthrough hemediating nfluenceof moreresponsivepolicies (link D-F). Finally,policy responsiveness s hypothesizedtohave direct effects on symbolic representation link F) consistent with theoriesholdingthatpublictrust n representativenstitutionsvariesin relation o thepro-duction of publicpolicies that are congruentwith public interests.Relatively little empirical research examines representationas a whole, butthere are several literatures, some quite extensive, that examine individualstrands.THE FORMAL-DESCRIPTIVEINK.A substantial literature testifies to the impact ofelectoral rules and procedureson descriptive representation Powell 2000; Rae1967;Taageperaand Shugart1989, amongmanyothers).Among the most criti-cal featuresin this regard s the numberof legislative seats in electoral districts(Duverger1954;Lijphart1994;Taagepera ndShugart1989). Districtmagnitudeis important,n part,because it is a principaldeterminant f the effective numberof partiesin a political system.4This influencesboth the extentof electoral com-petition and the strengthof minority parties,who are more likely to nominatewomen and minority candidates(Jones 1993; Mainwaringand Shugart 1997;Rule 1987; Taageperaand Shugart 1989). District magnitude also facilitatesdiversityin legislatures,since political partiesare more likely to "risk" he nom-inationof nontraditional andidates(i.e., women andminorities)for the nth seatin multimemberdistrictsrather han for the only seat in single memberdistricts(MatlandandBrown 1992;Rule 1987). Empirically, xtensiveresearchconfirmsthat electoralsystemswith greaterproportionality ndhigherdistrictmagnitudeselect largerpercentagesof women to legislatureswhen other factors are con-

    4 Lijphart 1994, 130-31) notesthat the effect of districtmagnitudeon thenumberof partiesvariesby type of electoralsystem. It encourages greaternumbersof partiesin PR systemsbutdiscouragesthem in plurality-majority istricts.Because there arevery few plurality-majority istrictswith a dis-trictmagnitudegreater han 1 (andnone in ourstudy),we can largelyignorethis theoreticalconcernin practiceand treat districtmagnitudeas always havingpositive effects on the numberof parties.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    7/23

    412 Leslie A. Schwindt-Bayerand WilliamMishlertrolled (Darcy, Welch, and Clark 1994; Duverger 1955; Matlandand Studlar1996; Norris 1985; Rule 1987; see, however,Welchand Studlar1990).THEFORMAL-SUBSTANTIVEINK.Thehypothesisthatformalelectoralrules influ-ence policy responsivenessalso is widely supported see, for example,Miller etal. 1999; Stokes2001). Powell (2000) demonstrates, or example, that morepro-portional systems generally experience higher levels of policy congruencebetween voters and governmentsthan do more majoritarian lectoral systems,andMilleret al. (1999) concurs. Less researchconnectselectoralrulesto the sub-stantiverepresentation pecificallyof women. Still, given thatthe mediancitizenin most democracies s female and that the interestsof women traditionallyhavebeen underrepresented,t is reasonableto expect that women's interestsshouldbenefit as electoralrules andprocedurespush governmentpolicy in the directionof the median citizen.THE FORMAL-SYMBOLICINK.The impact of electoral rules and procedures onsymbolic representationalso has been relatively neglected. Nevertheless, theavailableevidence sustains the hypothesisthatsupportfor representativenstitu-tions varies with the extent of political competitionandchoice providedby elec-toralsystems (AndersonandGuillory 1997).Norris (1999) reports,for example,thatpublic supportfor legislaturesis significantlyhigher in multipartyandpar-liamentarysystems, andAnderson andGuillory(1997) demonstrate hatsupportfor democratic nstitutions s higherin proportionalas comparedto majoritarianpolitical systems.The usual hypothesisregardingwomen is that they should express less trustthan men given that legislatures traditionallyare male bastions. Hibbing andTheiss-Morse(1995) report,however, hatAmericanwomenexpressgreater rustin Congressthando men, andNorris (1985) obtainssimilar resultscross-nation-ally. This raises questions about the utility of egocentric models of politicalbehavior and recalls the continuingdebate about egocentric versus sociotropicmodels of voting (see, e.g., KinderandKiewiet 1979; Lewis-Beck 1988). Whileegocentricmodels reflectthe traditionalbelief that individualsoperate"selfishly"and are motivated argely by individualself-interest,sociotropicmodels assumea more "enlightened" elf-interestin which individualsrecognize that theirper-sonal fortunesdependon the fortunes of the group.From a sociotropicperspec-tive, individual interests are more likely to be evaluatedin positive-sumterms.Thus, electoral rules increasingthe representationof women in the legislaturemay be embracedby women egocentricallybecause of the expected impact ofthose rules on the descriptiverepresentation f women. However,they also maybe valued sociotropicallyby men and women because those rules are conduciveto more equal representation f all groups in the political system, includingbutnot limitedto women.

    Separating he potentialegocentricandsociotropiceffects of formalrepresen-tation on symbolic representation n this model involves distinguishingdirectfrom indirect effects. To the extent that formalrepresentations importantego-

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    8/23

    An IntegratedModel of Women'sRepresentation 413centrically, hen its effects on symbolic representation hould vanish or at leastbe significantlyattenuatedwhen descriptiverepresentationand policy respon-siveness to women's issues are controlled.Conversely, he sociotropiceffect offormalrepresentation an be measuredsimply as the direct effect of formalonsymbolic representationhatpersistswhen descriptiveandpolicy representationare controlled.The differencebetween the egocentricand sociotropiceffects offormalrepresentation n legislativelegitimacyalso shouldbe manifestin genderdifferences.If women's confidencein the legislatureis influencedby egocentricassessments of the representativeness r responsivenessof the system to womenand theirinterests, henwomen's confidencein the system shouldbe stronglyandpositivelyaffectedby variations n formalrepresentationwhile men'sconfidencein the system should be either unaffected or negativelyaffected.THEDESCRIPTIVE-SUBSTANTIVEINK.One of the most widely studied hypothe-ses regardingrepresentationholds that variationsin descriptive representationhave substantialeffects on policy responsiveness.John StuartMill arguedmorethan a century ago that, "in the absence of its naturaldefenders,the interest ofthe omitted is always in danger of being overlooked;and when looked at, isseen with very differenteyes from those of the persons whom it directly con-cerns"(1967, 22; chapter3). Pitkinmakesthe samepointreasoning hatit is onlylogical "to expect the composition [of a legislature]to determinethe activities"(1967, 63). The literatureon women'srepresentation onfirmsthe importanceofdescriptiverepresentation or women's policy responsiveness. Several studieshave found stronglinks between increasingthe percentageof female legislatorsandwomen'spolicy outputs(Bratton2002; Reingold2000; Swers2002; Thomas1991), although he effects appear o be limited to women'sissue areas andoftenare observableonlywithinpoliticalparties Swers (2001) providesa good review).While much of the literatureassumes a linear relationshipbetween the pro-portionof femalerepresentatives ndtheproductionof women-orientedpolicies,Kanter(1977) hypothesizesa threshold effect. She arguesthat there is likely tobe only a modest relationshipbetween the numberof women in the legislatureandpolicy responsivenessuntil women'srepresentation eaches a critical mass.Beyondthatthreshold,Kanterspeculates,women'sinterestswill begin to diffusebroadlyamongall members of the assemblyresultingin acceleratedresponsive-ness to women'spolicy concerns.A recent study supportsthis hypothesiswithregardto the impactof women's descriptiverepresentation n child care policyin Norway (Brattonand Ray 2002). To date, however,little empiricalwork hasbeen producedin supportfor the thresholdhypothesis,perhapsbecause of thesmall numberof legislatureswhere the proportionof female legislators is largeenough to constitutea criticalmass.THEDESCRIPTIVE-SYMBOLICINK.Even where female legislators do not advo-cate a distinctly"femaleagenda"orrespond o women'spolicy concerns,a visiblepresence of women in the legislaturemay still enhancewomen's confidence inthe legislativeprocess.Thehypothesisis that constituentsaremorelikelyto iden-

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    9/23

    414 LeslieA. Schwindt-Bayerand WilliamMishlertify with the legislatureand to defer to its decisions to the extent thatthey per-ceive a significant percentage of "people like themselves" in the legislature(Mansbridge1999; Phillips 1995). Forwomen, this means that increases in theproportionof female legislatorsshould increasethe legitimacyof the legislatureas perceivedby femaleconstituents.The evidenceon thispointis limitedbutgen-erally supportiveof the expectation(Norrisand Franklin1997).Again, differentegocentric and sociotropic effects of descriptive representa-tion should be manifest in gender differences in supportfor the legislature.Tothe extent thatmen and women assess the legitimacy of the legislatureegocen-trically, hen thereoughtto be significantdifferencesin the effects of descriptiverepresentationon male and female confidence in the legislature.Conversely, othe extentthatthe effects of descriptiverepresentation resociotropic,we wouldexpect higher percentagesof women in the legislatureto have similareffects onboth male and female confidencein the legislature.THESUBSTANTIVE-SYMBOLICINK.For Pitkin and many others, substantive rep-resentationorpolicy responsiveness s the vital core of whatrepresentationmeansand is the heartof the integratedmodel. Formalanddescriptiverepresentationsconsidered mportantn largepartbecauseof theirhypothesizedeffects on policyresponsiveness.Policy responsivenessalso is consideredkey to the legitimacyorsymbolic representation f the legislature.This is consistentwith bothneo-insti-tutionaland rationalactortheories which hold thatcitizens' supportfor politicalinstitutionsdependslargelyon citizen evaluationof an institution'sperformance(Jackmanand Miller 1996; Mishler and Rose 2001; Powell 2000).As this discussiondemonstrates,hereare numerousandcomplex causalcon-nectionshypothesized n an extensive literatureo exist amongthe severaldimen-sions of representation.As a result,attempts o extractany one linkage fromthenetwork and examine it in isolation raise serious, though usually unrecognizedproblemsof model misspecification.Forexample,muchof the workdemonstrat-ing a causal connectionbetweendescriptiveandsubstantive epresentations mis-specified because it fails to control for the hypothesized,antecedenteffects offormal representationon both descriptiveand substantiverepresentation.Thismeans that the apparent ffects of descriptiverepresentation n substantive epre-sentationmaybe spurious,eitherwhollyor in part.Similarly, tudiesshowingthatsubstantive epresentationontributes o symbolicrepresentation remisspecifiedunless they controlforbothdescriptiveandformalrepresentation. roperestima-tion of these relationships equiresa fully specified,integratedmodel.

    Measures and MethodsTo test the integratedmodel we use aggregate data on the representationofwomen in the mid-1990s in 31 countries.5Countrieswere chosen based on dataSThecountries include:Argentina,Australia,Austria,Belgium, Britain,Bulgaria,Canada,Chile,Denmark,Estonia, Finland,France,Germany, celand,Ireland, taly,Japan,Latvia,Lithuania,Nether-

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    10/23

    An IntegratedModel of Women'sRepresentation 415availabilityand democraticstatus. Thejustificationfor the first of these is self-evident.While data on formalanddescriptiverepresentation rewidely available,data on policy responsivenessandon public confidencein representativenstitu-tions are in short supply.Thus, we have constructedwhat is, in essence, an"opportunity ample"of countriesfor which we were able to assemblerelativelycomplete data on all four dimensions of political representation.We also limitour analyses to countriesthat are considered"free"accordingto the FreedomHouse indices of civil andpolitical freedoms.6While Pitkin(1967, 2-3) acknowl-edges that democracyand representationare not synonymous, she argues thatpolitical representation nddemocracyare closely linked. Formalrepresentativestructuresmayexist in nondemocracies,buttheyareusuallysubsumedby author-itarian leaders and do not operate as functioning representative nstitutions.Therefore, there is little reason to examine representation n nondemocraticstates.7Data used in the analysis consist of three measuresof formalrepresentationincluding district magnitude, parliamentaryversus presidential systems andsingle-memberdistrictversusproportional epresentation ystems (the appendixprovidesdescriptions,means, standarddeviations,and sources for all variables).Descriptiverepresentations measuredsimplyas thepercentageof womenin thelowerhouse of the nationallegislature.Measuringpolicy representationproved more challenging both because theavailablemeasuresof this concept are highly subjectiveand because there is achecker-board atternof missing datafor the differentmeasures.To compensatefor the potentialbias in any single measure,we employ four indicatorsmeasur-ing genderequality in political rights, gender equality in social rights,nationalmaternity eave policy, andgenderequalityin marriageanddivorcelaws.8Whilethese measures in no way exhaust the range of issues that might be consideredlands,Norway,Poland,Portugal,SouthAfrica, SouthKorea, Slovenia, Spain, Sweden, Switzerland,Uruguay,andUnited States.

    6 Specifically,we limit the analysisto countrieswith averagescores between 1 and 2.5 on the sep-arate,7-point Freedom House scales of civil liberties and political rights;where lower scores indi-cate greaterfreedom(http://www.freedomhouse.org/ratings/index.htm).7We experimentedwith the inclusion of a nearly equal number of "unfree"countries in prelimi-narytests of the model. Unlike the clearpatternreportedbelow for democraticregimes,virtuallynosignificantrelationshipswere found to exist among the different dimensions of representation nundemocraticsystems. Moreover,the inclusion of undemocraticregimes severely confoundedtheresults and caused the model for democraticregimes effectively to collapse. The decision to focus

    only on democraticregimes is supportedboth by theory and evidence.8 deallya measure of policy responsivenessshould assess congruencebetweenlegislativeoutputsand the interests(needs or demands)of the represented.Unfortunately, ross-nationalmeasures of

    publicopinionin specific policy areasarelargelyunavailable, s arereliablemeasuresof publicneeds.Lacking cross-nationaldataon women'spolicy needs or demands,we make the heroic assumptionthat women share a numberof policy interestscross-nationallyandthat we can measureresponsive-ness by focusing solely on legislativeoutputson issues we assumeto be especially salientto women.To compensatefor the lack of dataon women'spolicy demands,we includea controlvariablemeas-uring publicattitudes owardwomen'sroles in society as discussed in detailbelow.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    11/23

    416 Leslie A. Schwindt-Bayerand WilliamMishlerwomen'sissues, they do providean indication of policy responsivenesson somekey issues of particularimportance to many women. To measure symbolicresponsiveness,we use aggregatedWorldValues Survey (1995-97) data froma question about citizen confidence in the legislature. Separatemeasures ofwomen's andmen's confidencein the legislatureare calculatedas the percentageof each group in each country respondingthat they had "a great deal of confi-dence" or "quitea lot of confidence" n thatcountry's egislature.9In addition to modeling the four dimensionsof representation,t is importantto control for exogenous factors that threaten he validity of observedlinkages.Because of the small numberof cases in oursample,however,we had to be eco-nomical in our choice of controls.Therefore,we experimentedwith severaldif-ferent candidates ncludingthe percentageof women in the workforce,per capitaGross Domestic Product,and a series of variablesaggregatedfrom the WorldValuesSurvey measuringpublicattitudes oward he roles of women in society.10The several controls were highly correlated,however,andthe choice of controlshad very little effect on the overall structureof the model. Thus, we rely on asingle, compositevariablereflectingculturalattitudes oward he role of womenin society (see Table1 and the appendixfor details).Wetreated his eminist atti-tudes variable as exogenous, hypothesizingthat it is causally priorto the otherdimensions of representation ncludingthe percentageof women in parliament,policy responsivenessto women's concerns, and women's confidence in parlia-ment. Inclusion of this control is important for ensuring that any linkagesobserved among the componentsof representationare causal and not spuriousartifactsof broaderculturalattitudesand values.

    ModelEstimationStructural quationmodeling (SEM) proceduresareused to estimatethe inte-grated model. These permit the simultaneousestimation of a complex causalmodel and of a series of measurementmodels fortheprincipalconceptsor"latentvariables"(in this case, feminist attitudes, formal representation,and policy

    responsiveness) that are measured with multiple indicators. The statisticalpackage used for the analysis,AMOS (Arbuckleand Wothke 1995), calculatesFull InformationMaximumLikelihood(FIML)estimates.Among otherbenefits,

    9Because the legislature s not the only representativenstitution n some systems and is not theonly policymaking nstitution n others,we experimentedwith othermeasures of symbolicrepresen-tation includingmeasures of publicconfidence in "thegovernment"and average public confidenceacross a range of various institutions. In fact, public confidence in different institutionsis highlycorrelatedand the model results are the same regardlessof the variableused to measuresymbolicrepresentation.

    l?Thenumber of possible influences on women'srepresentations virtuallylimitless. We wouldhave liked, for example, to have measuresof the strengthof women'smovements in each country,women's educationlevels, and women'smembership n higher level areas of the workforce.Unfor-tunately,suchdata arein very shortsupplyand could not be found in consistentcross-national orm.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    12/23

    An IntegratedModel of Women'sRepresentation 417TABLE

    MeasurementModelsof FeministAttitudes, ormalRepresentation,ndSubstantiveRepresentationLatentVariables/Measures Loading StandardError StandardizedLoadingFeminist Attitudes

    UniversityMoreImportant or Men -1.00 -.75WomenNeed Children -3.23* .90 -.97WomenEarningMore a Problem -.77* .31 -.58

    Formal RepresentationDistrictMagnitude 1.00 .67PR vs. SMD .04* .01 .76Presidentialvs. Parliamentary -.02 -.02 .23Substantive RepresentationWeeks of MaternityLeave 1.00 .62Women'sPoliticalEqualityIndex .04* .02 .79Women'sSocial EqualityIndex .03* .01 .48MaritalEqualityin Law .04* .02 .49Notes: Analyses areconfirmatory actoranalyses.*p < .05.N=31.See AppendixA for variablecoding and sources.

    theseprovidea superiormethod forhandlingmissing datathanis typicallyavail-able when OrdinaryLeast Squaresestimatorsare employed(Kline 1998).The first concernin the analysisis the validityof the measurementmodels forthe latentconcepts.Consistentwith expectations, he model for feministattitudesin Table1 suggeststhat the severalmeasures oad stronglyon a single latentvari-able. Countriesscoring high on one indicatorof feministattitudesscore high onall of the others as well. All three indicatorsare statistically significant,and allarenegatively signed so thatpositive scores indicategreatersocietal supportforfeminist values."1The measurementmodel for women's policy responsiveness includes fourvariablesall of which load substantiallyon a single dimension.All of the load-ings are statisticallysignificantand their standardizedcoefficients range from.48, for a variablemeasuringwomen's social equality, to .79 for the variablemeasuringthe political equalityof women. The mean loading for the fourvari-ables is .60, which is especially impressive given the small numberof cases, the

    "While we label this dimensionfeminist values, we might also have labeled it modernism-tradi-tionalism or something similar.Additional analyses, not shown, demonstratethat this variable ishighly correlatedwith the percentageof women in the workforce,GDP per capita,and (negatively)infantmortality. ndeed,all three of these additionalvariableshave loadings greaterthan .40 on thislatent variablewhen includedtogetherin the analysis.We includeonly feministvalues here becausethey have the strongest oadings. They also have the strongestcausal connectionsto other variablesin the model andthusprovidethe strongestcontrols on the representationalinkages.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    13/23

    418 Leslie A. Schwindt-Bayerand WilliamMishlerTABLE

    Structural Equations Estimates of an Integrated Model of the PoliticalRepresentation of WomenStandardized

    Variables Coefficient StandardError CoefficientDescriptive Representation R2= .52

    ProportionalElectoralSystems .72* .31 .48Presidentialvs. Parliamentary ystems 1.06 3.86 .05FeministAttitudes -82.7* 29.70 -.54Substantive Representation R2 = .83

    ProportionalElectoralSystems -.82 .52 -.86Presidentialvs. Parliamentary ystems -1.79 3.15 -.12Percentageof FemaleLegislators .56* .28 .88FeministAttitudes -.09 28.10 -.01

    Symbolic Representation R2= .27ProportionalElectoralSystems .07* .01 .38Presidentialvs. Parliamentary ystems -.02 .07 -.07PercentageFemaleLegislators .002 .002 .16Women'sPolicy Responsiveness -.002 .02 -.09FeministAttitudes -.36 .56 -.19Notes: Estimates are Full InformationMaximumLikelihoodEstimates.*p < .05.N= 31.See AppendixA for variablecoding and sources.

    subjectivenatureof these indicators,and the missing data in several of thesemeasures.In contrast o feministattitudesand women'spolicy concerns,the threemeas-ures of formalrepresentationdo not fit a single dimension.While the two indi-catorsof electoral system proportionality district magnitudeand PR vs. SMD)are closely related, he standardizedoading of the variabledistinguishingpresi-dential and parliamentary ystems is very weak (only .23). Based on this, werevised the model to include two measuresof formalrepresentation: latentvari-able measuringelectoralproportionalityand the observed variabledistinguish-ing presidentialandparliamentary ystems.Table 2 reports nitial estimatesof the fully specified,integratedmodel of thepolitical representation f women across 31 democraticcountries.The diagnos-tics atthe bottomof the tabledemonstratehatthe model fits the datafairlywell.12'2The relative chi square statistic (cmin/df) is considerablyless than 2.0, which, accordingtoCarminesandMcIver,demonstrates n"acceptable it betweenthehypotheticalmodel andthesampledata"(1981, 80). The RMSEA is .13, which is higherthan the .05 level normallyrecommendedbutreasonablyclose given a small sample (Bollen 1989).The IncrementalFit Index (IFI) andCompar-ative Fit Index (CFI) are .79 and .70, respectively, ndicatingthat the model representsa 70-79%improvementover the null model.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    14/23

    An IntegratedModel of Women'sRepresentation 419The estimated model representsa significant improvementover the null modelin which all variables are assumed to be unrelated,and it closely approximatesthe fit of a "saturatedmodel" in which all possible linkages are specified.Moreimportantly,he estimatedmodel accounts forone-quartero one-half of the vari-ance in each of the threeendogenousdimensionsof representation.Despite the model'soverall"goodnessof fit,"a numberof individual inkagesare weak or insignificant.Feminist attitudes have strongand significanteffectson the percentage of women serving in national legislatures,but they do nothave significanteffects eitheron women'spolicy responsivenessor on women'sperceptions of legislative legitimacy when other variables in the model arecontrolled. Similarly, formal representationhas importanteffects, consistentwith theory, on the percentage of female legislators but not on substantiverepresentation.13 iven the large number of variables in the analysis and thesmall sample size, statistical degrees of freedom are precious; even smallchanges in the model can significantlyalter model estimates. Therefore, it isimportantto correct the model specification based on the initial results andto reestimate a properly specified, reduced form of the model using theextra degrees of freedom. The results, reportedin Figure 2, confirmboth thatthe overallfitof the model is improvedandmulticollinearityamongthe estimatesis eliminated.

    Importantly,he structureof the revised model largely confirmsexpectationsregardingthe integratednature of political representation,albeit with severalimportant xceptionsto Pitkin's heory.Consistentwith thehypothesis hat formalrepresentation ontributes o descriptiverepresentation,he revised model clearlydemonstrates hatPR electoralsystems and those withhigherdistrictmagnitudeselect significantly largerpercentagesof female legislators.The effect is strong(.56), positive and direct.This is the case, moreover,even after controllingforthe strong, positive effects (.51) that feminist attitudeshave on the election offemale legislators.Also consistentwith the integratedmodel, the percentageof female legislatorshas a substantialandstatisticallysignificanteffect (.42) on the responsivenessoflegislatureson women'sissues. This effect persistseven when controlsare intro-duced for public attitudes owardwomen (and,alternatively, or GDP per capitaand the percentagewomen in the work force). Even in countrieswith compara-ble economies and similarattitudesaboutwomen, legislaturesare morerespon-sive to women's issues when there are more women in legislature.

    13Anotherproblemwith the model is multicollinearity r = .91) in the parameter stimates for link-ages A and B (i.e., those connecting proportionalelectoral systems with the percentwomen in thelegislatureand women'spolicy responsiveness).The effect of this multicollinearity s to inflate con-siderablythe two parameters,makingboth appear largerthan either one does when enteredsepa-rately.Given thatthe linkagebetween electoralproportionality nd women'spolicy responsiveness snot significantat traditional evels (p < .10 for a small sample)and remainsnonsignificantwhen thelinkagebetweenpercentwomen andpolicy responsiveness s removed,we have deleted the electoral-* policy link from the revised model.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    15/23

    420 LeslieA. Schwindt-Bayerand WilliamMishlerFIGURE

    ReducedFormStructuralquationModelof Women'sRepresentationmaternity socialpolicy equalitypolitcal 4maritaldistrict

    quaelit P "ui

    magnitude 94 POTC R2= 0.7PRsy ~ ~m 0 Electoral 04setem 0.42System 0.45 R2 = 0.28

    Women's0.03 Confidenceducate 0.56men 0.13,o by Feminist 0.51 Percentpay Attitudes *Womenproblem a,,i

    R2 = 0.56need Diagnostics:child CMIN/DF= 1.58

    RMSEA= 0.051IFI= 0.960CFI = 0.918

    Theproportionality f the electoralsystem also has substantial,albeit indirect,effects on women'spolicy responsiveness.14 Rsystemstendto elect morefemalelegislators whose greater numbers facilitate the enactment of more women'slegislation,otherthings being equal.Electoralproportionality, owever,does nothave independenteffects on substantive epresentation;he effects of formalrep-resentationarefully mediatedthroughdescriptiverepresentation.Althoughelec-toral proportionality ncouragesthe election of a more diverse legislature,thisdiversitymay impede the formation of a governingmajority.Indeed,a substan-tial literaturedebatesthe characteristic trengthsand weaknessesof the two typesof systems in this regard Lijphart1994; Powell2000). Consistentwith this sug-gestion, the simple correlation between electoral proportionalityand women'spolicy responsiveness is negative, although the correlation is weak and notsignificant.Additionalsupportfor the integratedconceptionof representations providedby the observation that formal representationhas a statistically significantandsubstantialeffect on women'sconfidencein the legislature(symbolic representa-tion). Importantly, however, neither descriptive representation nor policy

    14Themagnitudeof the indirecteffect is calculatedby multiplying he standardized oefficientsinthe compound path. Thus, the indirect effect of the electoral system on policy responsivenessviadescriptiverepresentations .57 x .42 = .24.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    16/23

    An IntegratedModel of Women'sRepresentation 421responsiveness has appreciableeffects in this regard.'"The failure of policyresponsiveness to influence symbolic representation s especially problematicgiven Pitkin'semphasison policy responsivenessas the centerpieceof the inte-gratedmodel.The absenceof directlinks betweendescriptiveand substantiverepresentationon the one hand and symbolic representationon the other admitsseveralpossi-ble explanations.First,it may result fromthe choice of policy areasincluded inthe measure of women'spolicy responsiveness.The influenceof policy respon-siveness on women's confidence in the legislature may be greater for otherwomen's issues such as abortionrightsor domesticviolence legislation. Second,while policy responsiveness may be the centralaspect of representation rom anormativeperspective, t maybe less important mpirically.Giventhe low levelsof political knowledge characteristicof citizens in most democratic polities,manycitizens simply maynot have the knowledgeto accurately udge the policyresponsivenessof the system. Third,even when citizens can assess the qualityof policy responsiveness, t may be difficultfor themto apportionresponsibilityfor policy responsiveness among the executive, legislature, bureaucracy,andcourts.Lackingdata on most of these matters,we can only speculateaboutthelack of relationshipbetween policy responsivenessand women's confidence inlegislatures.NONLINEARFFECTS.Theabsence of a substantial elationshipbetweendescrip-tive representation nd symbolic representation .13 in Figure2) also may indi-cate model misspecification.As noted,the literatureanticipates he possibilityofa nonlinearrelationshipbetween the percentageof women in a legislatureandpolicy responsiveness (Kanter 1977). One possibility is that a critical mass ofwomen must be elected to the legislaturebefore the powerof women reaches athreshold sufficient to affect legislation;anotherpossibility is the existence of amultiplier effect whereby the power of women increases exponentially withwomen'sincreasingpresencein the legislature.The logic of this argumentmightwell be extended to symbolic representation s well. Womenmaybe more likelyto accept the legislatureas legitimateonly afterthe percentageof female legis-latorscrosses some criticalthreshold.

    To test these possibilitieswe estimated several alternative pecificationsof therelationshipbetween thepercentageof female legislatorsandbothpolicy respon-siveness and symbolic representation Table 3). First,we created a crudetest ofthe thresholdhypothesisusing a dummyvariablecoded "1"for those legislatureswith more than the average(i.e., 15%)percentageof women in the legislaturel61 The parameterbetween substantiveand symbolic representationalso is highly correlatedwith

    the parameterbetweendescriptiveandsymbolic representation uch that the two cannot be includedin the model simultaneously.16There is no basis in theory or in previous researchto predictwhat the specific percentageofwomen must be before a critical threshold s reached.Thus we arbitrarily ividedthe 31 legislaturesat the mean. Severalother,higherthresholds the upper quartileand 85thpercentile)werealso testedwith equally negligible results.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    17/23

  • 7/28/2019 3449637

    18/23

    An IntegratedModel of Women'sRepresentation 423for the normativeprimacy of substantiverepresentationare compelling. Theanalysis does suggest, however,that descriptive representation s the glue thatbinds the several dimensions of representation ogetherinto somethingresem-bling a coherentwhole.EGOCENTRICOR SOCIOTROPIC SSESSMENTSOF REPRESENTATION. finalquestion concerns the extent to which perceptions of legislatures' legitimacyhinge on egocentricversussociotropicevaluationsof representation.One way ofassessingthis is by comparingsurvey responsesof men and women. Specifically,if assessments of representationare made on the basis of narrowself-interest,thenmen and women'sconfidencein the legislatureshouldbe affecteddifferen-tially by variations n both descriptiverepresentation ndpolicy responsiveness.Specifically,women shouldbe morestronglyandpositivelyaffectedthanmen byhigher levels of women'srepresentation.Conversely, f assessmentsof represen-tation are based more on enlightened self-interest,then both men and womenshouldrecognizethat theirinterestsaresignificantly ntertwined,andthey shouldbe relatively equally affected by women's descriptiverepresentationand policyresponsiveness.To test this, we replicated the analysis in Figure 2 substitutingmen's forwomen's confidence in the legislature.The results in Table4 are unambiguous.The effects of women's descriptiverepresentation nd policy responsivenessonmen's confidence in the legislature are virtually identical to their effects onwomen. Men do not expressmore confidence in legislatureswith fewerwomenorin contexts wherewomen'spolicy interestsarerelativelyneglected.Tothe con-trary,men and women respond quite similarlyto variations n the quantityandquality of women's representation,Men clearly respond sociotropically,inter-pretingwhat is good forthe representation f women to be good, moregenerally,for society as a whole. While the absence of data on men's representationpre-vents a test of women's sociotropicevaluations,there is no reason to think thatwomen areanymoreorless enlightened hanmen. Thefact thatwomen'sabsolutelevel of trust in the legislatureis almost identical to men's17despite decades, ifnot centuries,of gender inequality s strong,albeitcircumstantial vidence in thisregard.

    Women'sRepresentations UniversalRepresentationPitkin conceives of representationas a complex structurewhose multipledimensions areclosely integrated.Ouranalysisof womenin 31 democraticcoun-tries confirmsPitkin'sinsights regarding he stronginterconnectionsamong thedimensions of representation. t also supports many of the specific hypotheses

    17Men on averageare2.5%more likely to trust their legislaturethan are women. The largest gapacross these countries s in Australiawheremen are 11%moretrustingof the legislature.Sweden issecond with a 9% maleedge. Womenaremoretrustingof the legislature henmen in nine countries;the largest gap is in Switzerlandwhere women are 6% more likely to trust the legislature.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    19/23

    424 LeslieA. Schwindt-Bayerand WilliamMishlerTABLE

    Direct, Indirect, and Total Effects (Standardized Maximum LikelihoodEstimates) f Formal,Descriptive,nd SubstantiveRepresentationnMen'sandWomen'sSymbolicRepresentation

    Women'sConfidencein Men's ConfidenceinLegislature Legislature

    Direct Indirect Total Direct Indirect TotalFormalRepresentation .45 .07 .50 .34 .16 .50Descriptive Representation linear) .13 .00 .13 .28 .00 .28Descriptive Representation squared) .28 .00 .28 .32 .00 .32Policy Responsiveness -.10 .00 -.10 -.02 .00 -.02

    Notes:Total= Direct + IndirectEffects.See AppendixA for variablecoding and sources.

    about these interconnectionsgleanedboth fromPitkin and from diverseempiri-cal work.Althoughourfindingsarelimitedto representative emocracies, he integratedmodel provides strongevidence, consistentwith theory,that formalrepresenta-

    tive structuresandprocesses exertpowerful nfluenceson the extentof women'sdescriptiverepresentation,policy responsiveness,and symbolic representation.The integratedmodel corroboratespreviousresearchdemonstrating hathigherlevels of descriptive representation increase legislatures' responsiveness towomen'spolicy concernsand enhanceperceptionsof legitimacy.It also providessome of the strongestevidence, to date, supporting he idea that the effects ofdescriptiverepresentation n policy responsivenessandsymbolic representationare nonlinear and accelerate as the percentage of women in the legislatureincreases. While even a few women in a legislaturecan generate importantben-efits for women, real gains in policy responsiveness and political legitimacyappear o dependupon the achievementof a criticalmass.Althoughthe results of ouranalysessubstantiallyvalidate both Pitkin's heoryandprevious empiricalresearch, wo principalanomaliesemerge.The firstis theunexpected findingthatwomen'spolicy responsivenesshas little or no influenceon women'sperceptionsof the legitimacyof the legislature.This is troublingnotonly because it contradicts he centralityof policy responsiveness n Pitkin's he-oretical frameworkbut also because it appearsantagonistic o rationalactorthe-ories of political behavior.While we advancedseveral alternativehypothesestoaccount for the unexpectedempiricalresults, includinglimited informationandproblems of blame attribution,systematic assessments of these possibilitiesrequiredatacurrentlyunavailable.Nevertheless,while citizens' lack of direct concernwithpolicy responsivenesscontradicts heory, t maypose fewerproblems hanit seems. Bothpolicy respon-

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    20/23

    An IntegratedModel of Women'sRepresentation 425siveness and legislative legitimacy are otherwise securely tied into this causalweb, especially through their connections to descriptive representation.Thisensures that the two will be closely correlatedeven if they are not directly orcausallyconnected. Pitkin'sconceptionof representationmayrequireelaborationand refinementin specific empirical contexts, but its fundamentalstructure sconfirmed.

    A second apparentanomalyin the results is the observation hat men respondto the representation or misrepresentation) f women almost identically as dowomen. Men's confidence in the legislature s just as sensitiveto the proportion-ality of the electoralsystem and to the proportionof women holding seats in thelegislature.Farfrom undermining he integratednatureof women'srepresenta-tion, this unexpectedfindingcorroborates he feministargument hat the promo-tion of women'srights inevitablyadvancesthe rights of men as well. In effect,women (andminorities,too, we presume)can be seen as the "miners'canaries"of political representation.Political systems that nurtureand protect the repre-sentationof less-advantagedgroupscan be trustedby membersof more advan-taged groups to protect their interests as well. The quality of women'srepresentation,n this sense, is universal.

    AppendixA:VariableDefinitions,Means, andStandard DeviationsMean Std

    Presidential vs.Parliamentary Coded"1"if presidentialsystem and"0"if .26 .44System parliamentary ystem.Electoral SystemDistrictMagnitude A country'smean numberof legislative seats 7.60 8.30

    per electoraldistrict.PR System Coded"l" if system employs some form of .71 .46proportional epresentation nd"0"otherwise.a

    Percent Women Percentageof legislators(lower house) who 15.00 10.40are women.bWomen's Policies

    Maternitypolicy Lengthof nationalmaternity eave in weeks.d 21.60 16.70Politicalequality Index of GenderEquityin PoliticalRights: 2.00 .351 = substantial nequality;2 = someinequality;3 = substantialequalityofrights.eSocial equality Index of GenderEquityin Social Rights: 1 = 1.90 .33substantial nequality;2 = some inequality;3 = substantialequalityof rights.eMaritalequality Equityof the sexes in marriageand divorce 2.60 .50proceedings:1 = substantial nequality;2 = some inequality;3 = substantialequalityof rights.e

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    21/23

    426 LeslieA. Schwindt-Bayerand WilliamMishlerAppendixA:continued

    Mean StdWomen's "... could you tell me how much confidence .35 .13Confidence you have in (Parliament): greatdeal ofconfidence,quitea lot of confidence,not

    very much confidence,or none at all?"Percentageof womenresponding"agreatdeal" or "quitea lot.'fFeminist Attitudes

    Need Child "doyou thinkthat a woman has to have .50 .23children... to be fulfilled..." 1 = yes;0 = no.'PayProblem "If a woman earnsmoremoney than a man, .44 .09

    it's almost certain to causeproblems"1 = StronglyAgree/Agree;0 = Disagree/StronglyDisagree.'Men Education "Auniversityeducation is more important or .20 .09a boy than for a girl."1 = StronglyAgree/Agree; 0 = Disagree/StronglyDisagree.f

    Sources: a. Beck,Thorsen,GeorgeClarke,AlbertoGroff,PhilipKeefer,and PatrickWalsh.2001."NewToolsin ComparativePoliticalEconomy:The Databaseof Political Institutions."WorldBank Economic Review 15 (1):165-176.b. IPU. 1995. Women n Parliaments,1945-1995: A WorldStatistical Survey, ReportsandDocuments,No. 23. Geneva: IPU.c. UN DevelopmentProgram.1997. TheHumanDevelopmentReport.New York:UnitedNations.d. UN Statistics Division. 2000. Table 5.C-Maternity leave benefits.http://unstats.un.org/unsd/demographic/ww2000/table5c.htm.e. Humana,Charles. 1992. WorldHumanRightsGuide. 3rd ed. New York:OxfordUni-

    versity Press.f. WorldValuesSurvey,1995-1997.

    AcknowledgmentThis is a revisedversionof apaperpresentedat the AnnualMeetingof the Amer-ican PoliticalScienceAssociation,August2002. BrianCrisp,SuzanneDovi, andBarbaraNorranderprovidedconstructivecriticism of earlierdrafts.ManuscriptsubmittedDecember 31, 2003Final manuscriptreceived October25, 2004

    ReferencesAnderson,ChristopherJ., and ChristineGuillory.1997. "Political Institutionsand Satisfaction withDemocracy:A Cross-NationalAnalysisof ConsensusandMajoritarian ystems."AmericanPolit-ical Science Review91(1): 1-16.Arbuckle,JamesL., andWernerWothke.1995. Amos 4.0 User's Guide.Chicago:SPSS Inc.Bollen, KennethA. 1989. "ANew IncrementalFit Index for General StructuralEquationModels."Sociological Methods and Research 17(3): 303-16.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    22/23

    An IntegratedModel of Women'sRepresentation 427Bratton,KathleenA. 2002. "TheEffect of Legislative Diversityon Agenda Setting:Evidence fromSix StateLegislatures."AmericanPolitics Research30(2): 115-42.Bratton, KathleenA., and LeonardP. Ray. 2002. "DescriptiveRepresentation,Policy Outcomes,and Municipal Day-Care Coverage in Norway."American Journal of Political Science 46(2):428-37.Bullock, Charles S. III. 1995. "TheImpactof Changingthe Racial Compositionof CongressionalDistricts on Legislators'Roll-CallBehavior."AmericanPolitics Quarterly23(2): 141-58.Carmines,EdwardG., and John P.McIver. 1981. "AnalyzingModelswith UnobservedVariables."nSocial Movements: Current ssues, eds. G.W Bohrnstedtand E.E Borgatta.Beverly Hills: SagePublicationsInc., pp. 65-115.Darcy,Robert,SusanWelch,andJanet Clark. 1994. Women,Elections, and Representation.2nd ed.Lincoln:Universityof NebraskaPress.Duverger,Maurice. 1954. Political Parties. New York:Wiley.Duverger,Maurice. 1955. ThePolitical Role of Women.Paris:UNESCO.Eulau,Heinz, and Paul D. Karps. 1977. "The Puzzle of Representation:Specifying ComponentsofResponsiveness."LegislativeStudiesQuarterly2(3): 233-54.Hero,Rodney,and CarolineTolbert.1995."Latinosand SubstantiveRepresentationn the U.S. Houseof Representatives:Direct,Indirect,or Nonexistent?"AmericanJournalof Political Science 93(3):640-52.Hibbing,John R., and ElizabethTheiss-Morse. 1995. Congressas Public Enemy:Public AttitudesTowardsAmericanPolitical Institutions.Cambridge:CambridgeUniversityPress.Jackman,RobertW., and Ross A. Miller. 1996. "A Renaissance of Political Culture?"AmericanJournalof Political Science 40(3): 632-59.Jones, Mark P 1993. "The Political Consequencesof Electoral Laws in Latin America and theCaribbean." lectoral Studies12(1): 59-75.Kanter,Rosabeth M. 1977. "Some Effects of Proportionon GroupLife: Skewed Sex Ratios andResponseto TokenWomen."AmericanJournalof Sociology 82(5): 965-90.Kinder,Donald,and D. RoderickKiewiet. 1979. "EconomicDiscontent and Political Behavior:The

    Role of PersonalGrievancesand Collective EconomicJudgmentsn CongressionalVoting."Amer-ican Journalof Political Science 23(3): 495-527.Kline, Rex. 1998. Principles and Practices of StructuralEquationModeling. New York: GuilfordPress.Lewis-Beck,Michael. 1988.Economics and Elections: TheMajorWesternDemocracies. Ann Arbor:

    Universityof MichiganPress.Lijphart,Arend.1994. ElectoralSystemsandPartySystems:A Studyof27 Democracies,1945-1990.New York:OxfordUniversityPress.Mainwaring,Scott, and MatthewSoberg Shugart.1997. "Conclusion:Presidentialismandthe PartySystem."In Presidentialismand Democracy in LatinAmerica, eds. S. Mainwaringand M.S.Shugart.New York:CambridgeUniversityPress,pp. 394-439.Mansbridge,Jane. 1999. "ShouldBlacks RepresentBlacks and WomenRepresentWomen?A Con-

    tingent 'Yes.'"Journalof Politics 61(3): 628-57.Marsh,Michael,andPippaNorris. 1997."PoliticalRepresentationn theEuropeanParliament." uro-pean Journalof Political Research32(2): 153-64.Matland,RichardE., and DeborahDwightBrown. 1992. "DistrictMagnitude'sEffect on FemaleRep-resentation n U.S. StateLegislatures."LegislativeStudiesQuarterly 17(4):469-92.Matland,RichardE., and Donley T. Studlar. 1996. "TheContagionof Women Candidates n SMDandPR ElectoralSystems:CanadaandNorway." ournalof Politics 58(3): 707-33.Mill, John Stuart.1967[1861]. Considerationson RepresentativeGovernment.London:Longmans,Green and Co.Miller,WarrenE., Roy Pierce, JacquesThomassen,RichardHerrera,Soeren Holmberg,Peter Esa-iasson, and BernhardWessels. 1999. Policy Representationin WesternDemocracies. Oxford:OxfordUniversityPress.Miller,WarrenE., and Donald Stokes. 1963. "Constituency nfluence n Congress."AmericanPolit-ical Science Review 57 (March):45-56.

    This content downloaded on Fri, 11 Jan 2013 01:25:16 AMAll use subject toJSTOR Terms and Conditions

    http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
  • 7/28/2019 3449637

    23/23

    428 LeslieA. Schwindt-Bayerand WilliamMishlerMishler,William,andAnthonyMughan.1978. "Representinghe Celtic Fringe:Devolution andLeg-islativeBehaviorin Scotland and Wales."LegislativeStudiesQuarterly3(3): 377-408.Mishler, William, and Richard Rose. 1997. "Trust,Distrust and Skepticism:PopularEvaluationsof Civil and Political Institutions n Post-CommunistSocieties."Journal of Politics 59(2): 418-51.Mishler,William,andRichardRose. 2001. "Whatare the Originsof PoliticalTrust?TestingInstitu-tional and CulturalTheoriesin Post-CommunistSocieties."ComparativePolitical Studies34(1):30-63.Norris,Pippa.1985. "Women'sLegislativeParticipationn WesternEurope."WestEuropeanPolitics

    8(4): 90-101.Norris, Pippa. 1999. "InstitutionalExplanations or Political Support." n Critical Citizens: Global

    Supportor DemocraticGovernance,ed. P.Norris. Oxford: OxfordUniversityPress,pp. 217-35.Norris, Pippa, and Mark Franklin.1997. "Social Representation."EuropeanJournal of PoliticalResearch32(2): 185-210.O'Regan,ValerieR. 2000. GenderMatters:FemalePolicymakersInfluencen IndustrializedNations.Westport,CT:Praeger.Phillips,Anne. 1995. ThePolitics of Presence. Oxford: ClarendonPress.Pitkin,Hanna. 1967. TheConcept of Representation.Berkeley: Universityof CaliforniaPress.Powell, G. Bingham. 2000. Elections as Instrumentsof Democracy. New Haven: Yale UniversityPress.Rae, Douglas W. 1967. ThePolitical Consequencesof ElectoralRules. New Haven:YaleUniversityPress.Reingold,Beth. 2000. RepresentingWomen:Sex, Gender,and LegislativeBehavior in Arizona and

    California.ChapelHill: Universityof NorthCarolinaPress.Rule, Wilma. 1987. "ElectoralSystems, ContextualFactors and Women'sOpportunity or Electionto Parliament n Twenty-ThreeDemocracies." WesternPolitical Quarterly40(3): 477-98.Sapiro,Virginia.1981. "WhenAre InterestsInteresting?The Problemof PoliticalRepresentation fWomen."AmericanPolitical Science Review75(3): 701-16.

    Schumpeter,JosephA. 1947. Capitalism,Socialism,and Democracy.2nd ed. New York:Harpers.Stimson,JamesA., Michael B. Mackuen,and RobertS. Erikson. 1995. "DynamicRepresentation."AmericanPolitical Science Review89(2): 543-65.Stokes, Susan C. 2001. Mandates and Democracy.Cambridge:CambridgeUniversityPress.Swers,Michele L. 2001. "Researchon Women in Legislatures:WhatHave We Learned,Where areWe Going?"Women& Politics 23(1-2): 167-85.Swers, Michele L. 2002. TheDiference WomenMake: the Policy Impact of Women n Congress.

    Chicago: Universityof ChicagoPress.Taagepera,Rein, and MatthewSobergShugart.1989. Seats and Votes:TheEffectsand Determinantsof ElectoralSystems.New Haven:YaleUniversityPress.Thomas,Sue. 1991. "TheImpactof Womenon StateLegislativePolicies."Journalof Politics 53(4):958-76.Thomassen,Jacques,and Hermann Schmitt. 1997. "Policy Representation."EuropeanJournal ofPolitical Research32(2): 165-84.Wahlke,John C. 1971. "PolicyDemandsandSystemSupport:The Role of the Represented."BritishJournalof Political Science 1 (July):271-90.Welch, Susan, and Donley T. Studlar. 1990. "MultimemberDistricts and the RepresentationofWomen." ournalof Politics 52(2): 391-412.

    Leslie A. Schwindt-Bayer [email protected])is assistantprofessor ofpolitical science, Universityof Mississippi,University,MS 38677-1848. WilliamMishler([email protected])s professorof political science, Universityof Arizona,Tucson,AZ 85721-0027.