Divisive Nominating Mechanisms and Democratic Party Electoral ...

10 downloads 0 Views 1MB Size Report
Democratic party on the party's success in presidential elections. ..... for Alaska and Hawaii could not be determined for their first three election cycles. **Due to ...
Southern Political Science Association

Divisive Nominating Mechanisms and Democratic Party Electoral Prospects Author(s): James I. Lengle, Diana Owen, Molly W. Sonner Source: The Journal of Politics, Vol. 57, No. 2 (May, 1995), pp. 370-383 Published by: Cambridge University Press on behalf of the Southern Political Science Association Stable URL: http://www.jstor.org/stable/2960311 . Accessed: 31/08/2011 17:58 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

and DivisiveNominatingMechanisms DemocraticParty ElectoralProspects

JamesI. Lengle Diana Owen Molly W. Sonner GeorgetownUniversity The purpose of this article is to explore the effect of divisive nominatingcampaignswithin the Democraticpartyon the party'ssuccess in presidentialelections.Divisivenessis definedas a functionof the nominatingmethodand the marginof victory.The individualand combinedeffects of each variable on Democraticpartysuccess is measuredthroughbivariate,multivariate,and probit analyses.The results show that divisivenesshurts the Democraticparty'sprospectsfor winningthe generalelectionafter controllingfor state partyorientationand incumbency.Democratsare more likelyto lose statesthat use a divisivenominatingmechanism(presidentialprimary)and win statesthat use a nondivisivemechanism (caucus).Moreover,among presidentialprimarystates, divisivenesspredicatedupon marginof victoryis stronglyrelatedto outcomes.Democratsaremorelikelyto win primarystatesdecidedby wide marginsand lose those states decided by narrowmargins.The analysiscovers every presidentialprimary,caucus,and generalelection outcomein individualstates from 1932to 1992.

INTRODUCTION

For more than 30 years, political analysts and practitionershave debated the question of whether divisive primarycampaigns-where the margin of victory between the winner and loser is narrow-are detrimentalto a political party's chances of winning in generalelections (Hacker 1965; Lengle 1980, 1981; Stone 1984, 1986; Buell 1986; Southwell 1986; Kenney and Rice 1984, 1987; Stone, Atkenson,and Rapoport1992;Johnson and Gibson 1974;Comer 1976;Bernstein 1977; Born 1981; Abramowitz 1988; Kenney 1988; Piereson and Smith 1975; Miller,Jewell, and Sigelman1988). While inconsistent findings abound in the general researchon primaries,the conclusionthat divisivepresidentialprimarieshurt Democraticcontenders'chances of winningpresidentialelectionshas been consistentlysupportedusing both aggregate (Lengle 1980, 1981; Kenney and Rice 1987) and individual (Lengle 1980; Southwell 1986; Stone 1986) level data. These studies employ a varietyof measures of divisiveness,controlvariables,and statisticalmethodologiesto reachsimiVol. 57, No. 2, May 1995,Pp. 370-83 C 1995by the Universityof Texas Press, P.O. Box 7819, Austin, TX 78713-7819 THE JOURNAL OF POLITICS,

Divisive NominatingMechanisms

371

larconclusions.With this study we join the debateby presentingthe most comprehensiveinvestigationto date of the effects of nominatingmechanismson presidential electionoutcomes. DIVISIVE PRIMARIES REVISITED

Our study differs from, and builds upon, previous researchin severalnotable ways. First, we employ the most expansivedataset ever used in this kind of analysis-the entire populationof state-levelelection datafrom 1932 to 1992. Next, we restrictour analysisto the Democraticpartyonly. The existing literature suggeststhat primarydivisivenessis more consequentialfor Democraticcandidatesthan Republicancandidates.The pluralisticbase, cross-cuttingcleavages, and wide ideologicalrange that have been characteristicof the Democraticparty since its transformationfromminorityto majoritypartystatusin the 1930ssow the seeds of conflict among the party elite as well as between the elite and the mass membershipover the rules, role, priorities,and directionof the party.The minority status,homogeneousmembership,and philosophicalcoherenceof the GOP inoculateit to some degreefrom the effects of divisiveness.We begin our own study with the 1932 presidentialelection to test for the effects of divisivenessduringthis entire periodof Democraticascendancy. Third, we have expandedthe scope of the inquiryby comparingthe effects of two types of presidential nominating mechanisms-presidential primariesand caucuses-on partysuccessratesin generalelections.Previousresearchfocusedon primariesexclusively. To remainconsistentwith the literature,we haveborrowedthe dichotomousindicatorof divisivenessinitiallydevised by Bernstein(1977) to study congressional elections and used by Lengle (1980, 1981) to study presidentialelections. Under this classificationscheme, a primaryis considereddivisive if the marginof victory betweenthe winnerand runner-upis 20% or less. Just as in previousstudies at the aggregatelevel, our workincludes controlsfor the effects of incumbencyand state partyorientation. Finally,we haveextendedearlierresearchin anothermajorrespect.The original studies employed multivariatecontingency table analysis.We follow up this researchby using probitto estimatethe probabilityof a state voting Republicanas a functionof divisiveness,incumbency,and partyorientation. NOMINATING MECHANISMS, DIVISIVENESS, AND ELECTORAL OUTCOMES

One majorlimitationof the currentliteratureis that the relationshipbetweendivisivenessand election outcomes is tested for one type of nominatingmechanism only-the directprimary.The exclusivefocus on primariesin the literatureis understandable.Primariesare the nominatingmethod of choice at the congressional, state, and local level. We believe, however, that divisiveness varies reliably by

372

JamesI. Lengle, Diana Owen, and Molly W. Sonner

nominatingmechanism.Different nominatingsystems, by virtue of their inherent characteristics,are susceptibleto greateror lesser degreesof divisiveness.A nominatingprocessthat is quick,less visible,party-centered,and elite controlled,for instance,is likely to be less polarizing,and hence, less divisive, than a method which is protracted,more visible, candidate-centered,and mass controlled.Direct primariesfit the lattertype. The focus on presidentialelections allowsus to compare the effects of two types of nominatingmechanisms-presidential primaryand caucus-on generalelection outcomes. Caucusesby their very natureare likely to be less divisive than primariesfor a numberof reasons.Campaignsin caucus states are less visible to the partymembership than those in primarystates. With the exception of the Iowa caucuses, fewercandidatesparticipate,less money is spent, fewerads are run, and less media coverageis generated.As a result,the partymembershipin caucusstatescompared with primarystatestends to be less interested,less attentive,less activelyinvolved, and less likely to develop strongemotionaland politicalattachmentsto candidates. In sum, they are less willing to participatein the processat all. On this last point, the historicalrecordis quite clear:turnoutin caucusesis much lower than in primaries.Only 1%-3% of the voting-agepopulationparticipatein caucuses,whereas 20%-25% vote in primaries(Ranney1972, 1977;Lengle 1992). More importantly,however, there is a fundamentaldifferencein the function of primariesand caucuses. For most of the twentieth century, caucuseswere responsible,primarily,for conductingpartybusiness and, secondarily,for selecting delegates. Party members who attended caucuses may have preferreddifferent candidatesfor president,but the caucus was a forum to set partyrules, plot campaign strategy, disseminate information, choose party officeholders,and select party leaders to the next round of caucuses-not to decide upon a presidential nominee. The responsibility for choosing the nominee in a caucus-basednominating processrestedwith the nationalconvention.Thus, if divisivenessaroseduringthe nominatingcampaign,it aroseat the nationalconventionwhere avenuesexisted to resolve the conflict and unite the party. Losing factions were offered patronage, porkbarrel,policy concessions,or the vice presidencyby the winningfactionin return for their support in the generalelection. Divisiveness was reduced through the bargainingand negotiatinginherentin a deliberativenominatingmechanism. The purposeof primaries,however,is to measurepopularsupportfor presidential contenders.By their very nature,therefore,primariesinvite internalpartydissensionif not civil war.They compelcandidatesto criticizeand malignone another before a statewideand nationalaudienceand encourageparty membersto divide themselvesinto opposing camps. Negative and deceptiveadvertisingblanketsthe airwavesand reinforcesvoter loyalty and antipathytowardcandidates.In a presidentialprimary,the price of victoryfor the winneris a tarnishedimage and a split party, and there is no consolationprize for losers. Our argumentis not that divisiveness is absentin caucusesand presentin primaries,but that the magnitudeof

373

Divisive NominatingMechanisms

the problemis less and the numberof partymembersaffectedis fewer in caucuses than in primaries. The effect of differentnominatingmechanismson party(dis)harmonyand electoraloutcomesis an especiallyinterestingquestion given the recent changesin the presidentialnominatingprocessover the last 20 years.Before 1972, an overwhelming majorityof states used caucusesto select delegatesto the DemocraticNational Convention,and a smallminorityused primaries.Since 1972, the ratioof caucuses to primarieshas been reversed.In 1992, 38 statesused primariesand only 12 states used caucuses. The proliferationof primariesand demise of caucuses was one of the unintended consequencesof the McGovern-FraserCommission reforms adoptedby the Democraticpartyin 1971 (Shafer 1983, 1988;Polsby 1983;Lengle 1987). See figure 1. One simpletest of our hypothesisthat nominatingmechanismsmatteris to compare Democratic party success during the prereformand postreformeras. From 1932to 1968,when delegateswere selectedby caucusesand nomineeswerechosen by nationalconventions,the Democraticparty won seven of 10 presidentialelections. Since the inception of the primary-basedsystem, Democratshave lost four

FIGURE 1 NUMBER OF DEMOCRATIC PRIMARIES AND CAUCUSES,

1932-1992

40

30.

Pre-ReformEra 0

|

PostreformEra

1932 1936 1940 1944 1948 1952 1956 1960 1964 1968 1972 1976 1980 1984 1988 1992

Primaries + Caucuses

JamesI. Lengle, Diana Owen, and Molly W. Sonner

374

TABLE 1 CANDIDATE CHOICE BY DIVISIVENESS OF DEMOCRATIC DELEGATE SELECTION MECHANISM

1932- 1992*

Democrat Republican

Divisive Democratic Primaries

Nondivisive Democratic Primaries

Caucuses

23% 77%

49% 51%

51% 49%

100% (105)

100% (220)

100% (451)

Chi-squarep< .00; Phi = .19. *The total numberof caucusesin these tables is 451 due to inclusionof Alaskaand Hawaiifor their firstthree election cycles.

of six elections.Obviously,otherfactorscontributedto Democraticpartysuccesses and failuresfrom 1932-1992. We believe, however,that it is more thancoincidental that victoriesand defeatsare so stronglytied to structuralchangesin leadership recruitment. Although a direct test of the hypothesis that primariesare more divisive than caucuses is impossible using aggregatedata, an indirect test is available.If divisivenesshurts, and if primariesaremore divisivethan caucuses,then the Democratic party should lose more primary states than caucus states in presidential elections. Moreover,if divisivenessmatters,then states with divisive Democratic primariesshould be more likely to vote Republican,and states with nondivisive primariesshouldbe more likelyto vote Democratic.Consequently,if our inference about the less divisive natureof caucusesis correct,then caucus states should be more supportiveof Democraticnominees than either divisive or nondivisiveprimarystates. Table 1 crosstabulatespartychoice of statesin generalelectionsby type of delegate selectionmechanismemployedby statesand, if statesused a presidentialprimary,by whetherthe primarywas divisiveor not.' 1The dependent variablethroughoutthe study is candidatechoice in generalelection campaigns. States carriedby the Democraticpartynomineeare coded "Democratic."States carriedby the Republicannomineearecoded "Republican."From 1932to 1992, 10 stateswerecarriedby a thirdpartyor independent candidatefor president. They were eliminatedfrom the analysisbecause they voted for a nonmajorpartycandidate. Our dataset consistsof 776 Democraticprimariesand caucuses.The nominatingmechanismused by the Democratsdifferedfromthat used by the Republicansonly 31 times. Out of a totalof 325 primaries, there were 22 cases where the Democratsheld a primaryand the Republicansheld a caucus. Of these 22 cases, 12 were carriedby the Democratsin the generalelection, 9 voted for the Republicannominee, and in one case a third-partycandidateprevailed.Fourteenof the 22 primarieswere nondivisive;8 were

Divisive NominatingMechanisms

375

Outcomesarerelatedto both nominatingmechanismsand to divisiveness.From 1932 to 1992, Democratslost 60% of those states that held a presidentialprimary but won 51% of those states that held a caucus. However, when divisiveness is taken into account, the dynamics of this relationshipchange substantially.As table 1 indicates,Democratslost 77% of those states that experienceddivisiveprimaries.The partyis almostequallysuccessfulamongthose stateswith nondivisive primaries,losing only 51% of the time, and caucuses,losing 49%. Thus, the evidence points to divisivenessas being a more importantdeterminantof Democratic successthan nominatingmechanismalone. The relationshipuncoveredin table 1 may be a function of state partyorientation ratherthan divisiveness.If traditionallyRepublicanstates use primariesand traditionallyDemocraticstatesuse caucuses,then the relationshipfound in table 1 betweennominatingmechanismand success is spurious,a productof the relationship betweennominatingmethod and statepartyorientation. There is some reasonto believe that nominatingmechanismsmight be tied to state partyorientation.States with strongstate and localDemocraticpartiesmight opt for caucusesbecausecaucusesgive state and local partyleadersgreatercontrol over delegate selection and greaterinfluenceat nationalconventions. States with weakDemocraticparties,on the other hand, would be more likely to succumbto populistor progressivepressurefor more mass democracyin the formof presidential primaries.In fact, directprimariessproutedfirstin those stateswith weakparty organizations.If strengthof Democraticpartyorganizationis relatedto nominating method, we would find a tendency for caucus states to vote Democratic,not because caucuses are less divisive, but because the Democratic party in caucus statestends to be stronger.We also would find a tendencyfor primarystatesto vote Republican,not becauseprimariesare more divisive, but becausethe Democratic partyin primarystatestends to be weaker. In addition,we also must control for the effects of state partyorientationwhen examiningthe relationshipbetween divisivenessand success. Since most caucus/ convention systems of delegate selection in the past were tightly controlled by state and local partyorganizations,candidatesneeded strong party ties to win the nomination.Candidateswithout these ties had only one strategy-to compete for Democraticdelegatesin states where their own party'sorganizationwas weakest. Such strategicconsiderationsby candidateswould produce divisive Democratic primariesin traditionallyRepublicanstates.Thus, statepartyorientation,not divisiveness, would explainthe poor performanceof Democraticnomineesin divisive primarystates. divisive. The Republicancandidatewon the generalelection in 6 out of the 8 states that experienced divisive primaries. Of the 451 caucusesin the dataset, there were only nine cases in which the GOP held a primaryand the Democratsheld a caucus. All of the primarieswere nondivisive.The Democratscapturedthe four statesbefore 1972;the Republicanswon the five statessince 1972.

376

JamesI. Lengle, Diana Owen, and Molly W. Sonner TABLE 2

CANDIDATE CHOICE BY DIVISIVENESS OF DEMOCRATIC DELEGATE SELECTION MECHANISM CONTROLLING FOR STATE PARTY ORIENTATION,

1932-1992* TraditionallyDemocratic

Democrat Republican

Divisive Primary

Nondivisive Primary

Caucus

47% 53%

78% 22%

78% 22%

100% (17)

100% (32)

100% (120)

Chi-square:p ? .01; Phi = .22. TraditionallyCompetitive

Democrat Republican

Divisive Primary

Nondivisive Primary

Caucus

21% 79%

51% 49%

52% 48%

100% (66)

100% (113)

100% (193)

Chi-square:p < .00; Phi = .23. TraditionallyRepublican

Democrat Republican

Divisive Primary

Nondivisive Primary

Caucus

9% 91%

32% 68%

24% 77%

100% (22)

100% (75)

101%** (132)

Chi-square:p < .07; Phi = .15. *The totalnumberof caucusesin these tablesis 445 insteadof 451 becausethe state partyorientation for Alaskaand Hawaiicould not be determinedfor their firstthreeelection cycles. **Due to rounding

To test for these possible sourcesof spuriousness,table 2 presentsthe relationship between nominatingmechanism,divisiveness,and party success controlling for traditionalstate partyorientation.2 2Stateswhose Democraticpercentagein each of three presidentialelectionsimmediatelypriorto the year of the primaryexceeded the nationalDemocraticaveragewere categorized"traditionallyDemocratic."States whose Republicanpercentagein those threeelectionsexceededthe nationalaveragewere categorizedas "traditionallyRepublican."States which split their partisanloyaltiesor failed to exceed

Divisive NominatingMechanisms

377

As table 2 shows, the effects of divisivenessremainregardlessof state political culture, while the effect of nominatingmechanismis somewhatmuted. In traditionallyDemocraticstates, the Democraticpartyloses 53% of the divisive primary statesbut wins 78% of the nondivisiveprimaryand caucusstates. The effect in traditionallytwo-party competitive states is equally dramatic. Democratslose 79% of the two-partycompetitivestates that underwenta divisive primarybut capture 51% of the nondivisiveprimarystates and 52% of the caucus states. Among traditionallyRepublicanstates, the relationshipstill exists althoughin a weakenedform.Democratslose traditionallyRepublicanstatesgenerallybut do far worse among divisive primarystates, losing 91%, and significantlybetter among nondivisiveprimaryor caucusstates,losing 68% and 77% respectively. Our findingsin table 1 could be spuriousfor anotherreason.The currentera of the presidentialprimaryparallelsthe recent string of Republicanpresidents.We argue that party success and failureis related to the type of nominatingmethod employed and the degree of divisiveness experiencedby the party during these periods.Others, however,could arguejust as easily that incumbencyaccountsfor the relationshipbetween nominatingmechanismand success. Caucusstates voted Democratic because caucuses coincided with Democratic incumbents. Primary statesvoted Republicanfor the same reason-primaries proliferatedduringan extended periodof Republicanpresidents. Incumbencyalso might explain the relationshipbetween divisivenessand electoral success. The partyout of office is likely to experiencedivisive primariesbecause its nomination is open and also more likely to lose the general election becauseit does not controlthe resourcesof the Oval Office. The incumbentparty, on the otherhand,would have fewercontestednominationsand more generalelection success. Over time the result would be the same for the Democraticparty: more nondivisiveprimariesand more victorieswhen competingas the incumbent party, and more divisive primariesand losses when competing as the out-party. The explanationfor failureand success would be incumbency,not divisiveness. the nationalaveragefor either party (e.g., in presidentialelections with majorthird partyor independent candidates)werecategorizedas "two-partycompetitive."Alaskaand Hawaiicould not be classified until each had experiencedthree election cycles in 1972. This measureof traditionalpartyorientationis identicalto the one used by Lengle (1980). The purpose of the measureis to captureparty orientationat the presidentiallevel only. Including state and local voting history in the measureis inappropriatesince states have differentpoliticalorientationsat differentelectorallevels. The South, for instance, is traditionallyRepublicanat the presidentiallevel but Democraticat the state and local levels. Also, since "tradition"does not preclude the possibility of change, and since state party traditions have changedover time (e.g., South), we kept the numberof pastelectionslow to capturechangeacross time withoutstackingthe deck to our advantage. The validityof the measureis supportedby the strengthof the relationshipbetween state partyorientationand candidatechoice. Tau,= .38.

JamesI. Lengle, Diana Owen, and Molly W. Sonner

378

TABLE

3

CANDIDATE CHOICE BY DIVISIVENESS OF DELEGATE SELECTION MECHANISM CONTROLLING FOR INCUMBENCY, 1932-1992* DemocraticIncumbent

Democrat Republican

Divisive Primary

Nondivisive Primary

Caucus

12% 88%

69% 31%

74% 26%

100% (17)

100% (85)

100% (186)

Chi-square:p < .00; Phi = .31. No Incumbent

Democrat Republican

Divisive Primary

Nondivisive Primary

Caucus

15% 85%

25% 75%

34% 66%

100% (26)

100% (52)

100% (114)

Chi-square:p < .25; Phi = .10. RepublicanIncumbent

Democrat Republican

Divisive Primary

Nondivisive Primary

Caucus

29% 71%

42% 58%

35% 65%

100% (62)

100% (83)

100% (151)

Chi-square:p < .12; Phi - .12. 'See note for table 1.

To test for this possibility,table 3 controlsour findingsfor the effects of incumbency.3As table3 shows, divisivenesstakesits toll regardlessof incumbency,while caucuses benefit the party in two of three instances. In election years with Democraticincumbents,divisivenessseverelyhandicapsthe party.Democratslose 88% of the divisive primarystates, but win 69% of the nondivisiveprimaryand 3If the Democratic incumbent sought renomination,the election year was coded "Democraticincumbent."If the Republicanincumbentsought renomination,the election yearwas coded "Republican incumbent."If the incumbentpresidenteither could not run or chose not to run, the election yearwas coded "no incumbent."

Divisive NominatingMechanisms

379

74% of the caucus states. In years with no incumbents,Democrats lose 85% of the divisive primarystates, but only 75% of nondivisivestates and 66% of caucus states. During election years with Republicanincumbent, Democrats lose 71% of the divisiveprimarystatesbut farefarbetterin the nondivisiveprimaryand caucus states. PROBIT ANALYSIS OF STATE VOTE CHOICE

We now takean additionalstep in our analysisby estimatinga probitmodel4to determinethe influenceof type of nominatingmechanism/divisiveness,stateparty orientation,and incumbencyon presidentialvote choice.' Table 4 presentsthe resultsof the analysis.The probitcoefficientsdemonstrate that all of the variableswe employ in the analysis, with the exception of a state having a Republicanincumbent, are statisticallysignificant,and the signs are in the expecteddirection.Democraticstatepartyorientationand partyof the incumbent are negativelyrelatedto voting Republican.6The goodnessof fit statisticsindicatethat the model fits the datareasonablywell. Seventy-onepercentof the cases were correctly classified, the pseudo R2 = .26, and the asymptotic F-statistic is significant. The results of the probit analysisfurther support our contention that divisive Democraticprimarieshelp elect Republicans,and caucuseshelp elect Democrats even after controlling for state party orientation and incumbency. As table 4 demonstrates,the probabilityof voting Republicanis highest for divisive primary statesand lowest for caucusstates for every combinationof state partyorientation and incumbency.Again, we find that divisivenessis the factor that renders primariesdamagingto the Democrats,as the differencesin the probabilityof losing nondivisiveprimariesand caucusesare small. Although Democratsdo best in "friendly"states in elections with Democratic incumbents,the effect of nominatingmechanismon performanceis nonetheless dramatic.A Democraticincumbenthas a 29% chance of losinga "friendly"state after a divisive primary,and only an 8% chance of losinga "friendlystate"which employs a caucussystem. The worst scenario for Democrats occurs in "hostile"states with a predominantlyRepublicanculture.Democratslose"hostile"states97% of the time when a divisive primaryhas occurred,comparedwith 87% when a caucus has been held. 4LIMDEP 6.0 was used for the analysis. 'The dependentvariable,votechoice,was coded 0 for the Democraticcandidateand 1 for the Republicancandidate.The predictorvariableswereenteredinto the equationas sets of dichotomousvariables. Type of nominatingmechanismwas representedby two indicators-divisive primaries/allothercontests and caucuses/allothercontests.State partyorientationand partyof the incumbentwere coded into categories of Republican /all othersand Democrat/allothers. 6The probit coefficientsthemselvescannot be compareddirectlv (see Hanushekand Jackson1977, 206).

JamesI. Lengle, Diana Owen, and Molly W. Sonner

380

TABLE 4 PROBIT ANALYSIS Probabilityof Voting Republican State Party Orientation

Party Incumbency

Divisive Primary

Nondivisive Primary

Caucus

Democratic Democratic Democratic Republican Republican Republican Competitive Competitive Competitive

Democratic Republican Open Democratic Republican Open Democratic Republican Open

.29 .63 .58 .83 .97 .95 .51 .95 .78

.11 .40 .32 .59 .89 .83 .26 .62 .52

.08 .38 .29 .55 .87 .81 .24 .60 .49

MODEL STATISTICS

Div. primary Nondiv. prim. Caucus Dem. SPO Rep. SPO Competitive Dem. incumb. Rep. incumb. Open seat Constant

coeff

se

coeff/se

sign t

.88 .31 .20 -.72 .79 .55 -.87 .29 .17 .51

.15 .10 .09 .13 .12 .12 .11 .09 .12

5.84 3.00 2.13 5.50 6.58 4.61 7.42 3.19 1.30

.00 .00 .03 .00 .00 .00 .00 .00 .19

N = 780; correctlyclassified= 71%; cases in modal category- 53% (n = 414); chi-square= 224.33; df= 5; pseudoR 2 = .26; asymptoticF = 44.87; significance= .00;PRE = .45.

The Democraticpartydoes not farewell when a divisiveprimaryhas takenplacein a competitivestate holdingan election for an open seat. Democratslose 78% of the time under these conditions.However,they face a 51% chanceof winning if a caucus is held. CONCLUSION

The relationshipsuncoveredin this study should not be surprisingto students of politicalparties, generally,and of party reform,specifically.Historically,party successhas been intricatelytied to formalinstitutionalmechanismsand procedures for selectingpartynominees(Ranney 1975;Ceasar1979;Heale 1982;McCormick 1982; Cavala1974; Lengle and Shafer 1976; Polsby 1983; Lengle 1987). Consequently,any changesin these rules and mechanismsaffectthe legitimacyof the de-

Divisive NominatingMechanisms

381

cision, the qualityof the nominees,the attitudesand loyaltyof partymembers,the powerof groupsand stateswithin the partycoalition,and, ultimately,the competitivenessof the partyin the electoralarena. From a more technical perspective, our findings lay a foundation for future theoreticaland empiricalinvestigationsusing both aggregateand individual-level data.One way to test the "divisiveprimaryhypothesis"furtheris to refinethe core conceptof "divisiveness."The use of a 20% marginleavesmanyunansweredquestions which can be addressedby using the "actualmarginof victory"in the analySiS.7 For example, is the relationshipbetween candidate choice and margin of victory strictly linear?Do Democratic party prospects improve as the marginof victory increases?Are primariesdecided by 40% one-half as divisive and, therefore, one-half as detrimentalto Democratic party success as primariesdecided by 20%? Aside from specificmeasurementissues, the concept of divisivenessalso can be more carefullyspecifiedby takinginto account the dynamicsof nominatingcampaigns.Undoubtedly,some primariesdecidedby less than 20% canbe less divisive than other primariesdecidedby more than 20%. Contextualfactors,includingthe size of the field of candidatesrunning in a state and the stage of the campaignat which a primarytakesplace, can influencelevels of divisivenessin a state. While our study answersconclusivelyone importantquestion using aggregatelevel data,it also raisesmany other relatedones at the individuallevel. Obviously, our findings suggest that politicalloyalties,attitudes,and perceptionsof voters in nominatingcampaignsareinfluencedenormouslyby the structureand intensityof the competition fostered by nominatingmechanisms.The voters' psychological and politicalreactionto bitterly fought nominatingcampaignsdoes not disappear afterthe nationalconvention.Instead,it remainswith the voters and becomespart of the mind set that influencestheir behaviorin generalelections. Issues concerning the relationshipof divisive primariesto the developmentof voters' short- and long-term attitudestowardcandidates,politicalparties, and the politicalprocess, their willingnessto workfor the presidentialnominee,and their propensityto turn out in the generalelection need to be explored.Only when such studies have been conductedcanwe hope to understandmorefully the divisiveprimaryphenomenon. Manuscript submitted18January 1994 Final manuscriptreceived 9 September1994

REFERENCES Abramowitz,Alan I. 1988. "ExplainingSenate Election Outcomes."AmericanPoliticalScienceReview 82:385-403. Atkenson, Lonna Rae. 1992. "Moving Toward Unity: An ExperimentalLook at Attitude Change 7Kenneyand Rice (1987) employeda measurebasedmore preciselyon marginof victory.

382

JamesI. Lengle, Diana Owen, and Molly W. Sonner

During a PresidentialElection."Presentedat the annualmeeting of the AmericanPoliticalScience Association,Chicago. Bartels, Larry M. 1988. PresidentialPrimariesand the Dynamicsof Public Choice.Princeton, NJ: PrincetonUniversityPress. Bernstein, Robert A. 1977. "Divisive PrimariesDo Hurt: U.S. Senate Races, 1956-1972." American PoliticalScienceReview71: 540- 55. Born, Richard. 1981. "The Influence of House PrimaryElection Divisiveness on General Election Margins, 1962-76." Journalof Politics43:640-61. Boyd, Richard. 1989. "The Effects of Primariesand StatewideRaces on Voter Turnout."Journalof Politics51:730-39. Buell, Jr., Emmett H. 1986. "Divisive Primariesand Participationin Fall PresidentialCampaigns:A Study of 1984 New HampshirePrimaryActivists."AmericanPoliticsQuarterly14:376-90. Cavala,William. 1974. "Changingthe Rules, Changes the Game."AmericanPoliticalScienceReview 68: 27-42. Ceasar,James. 1979. PresidentialSelection:Theoryand Development.Princeton,NJ: PrincetonUniversity Press. Comer,John. 1976. "AnotherLook at the Effects of the Divisive Primary:A ResearchNote." American PoliticsQuarterly4:121-28. Crotty,William. 1983.Party Reform.New York:Longman. Epstein, Leon. 1986. PoliticalPartiesin the AmericanMold. Madison:University of WisconsinPress. Hacker,Andrew. 1965. "Does a 'Divisive' PrimaryHarm a Candidate'sElection Chances?"American PoliticalScienceReview59:105-10. Hanushek, Eric A., and John E. Jackson. 1977. StatisticalMethodsfor Social Scientists.New York: Academic. Heale, M. J. 1982. ThePresidentialQuest.New York:Longman. Johnson, Donald Bruce, and JamesR. Gibson. 1974. "Divisive PrimaryRevisited:Activists in Iowa." AmericanPoliticalScienceReview68: 67-77. Kenney, PatrickJ. 1988. "Sorting Out the Effects of Primary Divisiveness in Congressionaland PoliticalQuarterly41:765-77. SenatorialElections." Western Kenney, PatrickJ., and Tom W. Rice. 1984. "The Effect of PrimaryDivisiveness in Gubernatorial and SenatorialElections."Journalof Politics46:904-15. Kenney, PatrickJ., and Tom W. Rice. 1987. "The Relationshipbetween Divisive Primariesand GeneralElectionOutcomes."AmericanJournalof PoliticalScience31: 31-44. Key, V. 0. 1964.Politics,Parties,L PressureGroups.(5th ed.). New York:Crowell. Lengle, James I. 1980. "Divisive PresidentialPrimariesand Party ElectoralProspects, 1932-1976." 8: 261-77. AmericanPoliticsQuvarterly andPresidentialPrimaries:TheDemocratic Partyin thePost-Reform Lengle,JamesI. 1981.Representation Era. Westport,CT: Greenwood. Lengle, James I. 1987. "DemocraticParty Reforms:The Past as Prologue to the 1988 Campaign." Journalof Law & Politics4:223-73. Lengle, James I. 1992. "Reformingthe PresidentialNominating Process."In TheQuestfor National Office,ed. Stephen Wayneand Clyde Wilcox. New York:St. Martin'sPress. Lengle, James I., and Byron E. Shafer. 1976. "PrimaryRules, Political Power, and Social Change." AmericanPoliticalScienceReview70: 25-40. McCormick,Richard.1982. ThePresidentialGame.New York:OxfordUniversityPress. Miller, Penny M., MalcolmE. Jewell,and Lee Sigelman.1988. "DivisivePrimariesand PartyActivists: Kentucky, 1979and 1983."Journalof Politics50:459-70. Piereson,JamesE., and Terry B. Smith. 1975. "PrimaryDivisivenessand GeneralElection Success:A Re-examination."Journalof Politics37: 555-62. Pomper,Gerald. 1981.Party Renewalin America.New York:Praeger. of Party Reform.New York:OxfordUniversityPress. Polsby, Nelson. 1983. TheConsequences

Divisive NominatingMechanisms

383

Ranney, Austin. 1972. "Turnout and Representationin PresidentialPrimaryElections." American PoliticalScienceReview66: 21-37. Ranney,Austin. 1975. CuringtheMischiefsof Factions:Party Reformin America.Berkeley:Universityof CaliforniaPress. Ranney,Austin. 1977. Participationin AmericanPresidentialNominations.Washington,DC: American EnterpriseInstitute. Southwell, Priscilla L. 1986. "The Politics of Disgruntlement: Nonvoting and Defection among Supportersof NominationLosers, 1968-1984." PoliticalBehavior8:81-95. Shafer,Byron. 1983.QuietRevolution.New York:Sage Foundation. Shafer,Byron. 1988. BifurcatedPolitics.Cambridge:HarvardUniversityPress. Stone, Walter J. 1984. "PrenominationCandidate Choice and General Election Behavior: Iowa PresidentialActivistsin 1980."AmericanJournalofPoliticalScience28:361-78. Stone, WalterJ. 1986. "The CarryoverEffect in PresidentialElections."AmericanPolitical Science Review80:271-79. Stone, WalterJ., Lonna Rae Atkenson,and RonaldB. Rapoport.1992. "TurningOn or Turning Off? Mobilizationand DemobilizationEffects of Participationin PresidentialNominationCampaigns." AmericanJournalof PoliticalScience36:665-91. Ware,Alan. 1979. "Divisive Primaries:The ImportantQuestions."BritishJournalof PoliticalScience 9:381-84.

James I. Lengle is associateprofessorof government,GeorgetownUniversity, Washington,DC 20057. Diana Owen is assistant professor of government, Georgetown University, Washington,DC 20057. Molly W. Sonner is a doctoralcandidateof government,GeorgetownUniversity, Washington,DC 20057.