huckfeldt_alternative contexts of political behavior
TRANSCRIPT
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 1/18
Southern Political Science ssociation
Alternative Contexts of Political Behavior: Churches, Neighborhoods, and IndividualsAuthor(s): Robert Huckfeldt, Eric Plutzer and John SpragueSource: The Journal of Politics, Vol. 55, No. 2 (May, 1993), pp. 365-381Published by: Cambridge University Press on behalf of the Southern Political Science AssociationStable URL: http://www.jstor.org/stable/2132270 .
Accessed: 06/06/2014 14:42
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 from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 2/18
AlternativeContexts fPoliticalBehavior:
Churches, eighborhoods,ndIndividuals
RobertHuckfeldtIndianaUniversity
Eric PlutzerIowaState University
JohnSpragueWashingtonUniversity n St. Louis
This paperexamines he politicalconsequences hat arisedue to multiple andsimultaneousbases of
socialexperience.Two alternativeontexts-those of neighborhoodsndchurches-provide an empiri-
calsetting for the effort;andthe analysis ocuseson twodifferentpoliticalattitudes-policy preferences
regardingabortionand partisanself-identification.Severalquestionsare addressed: n what manner
are the alternative ontextsof politicsdifferentand in what mannerare they similar?To whatextent
are churchesandneighborhoods einforcing n the politicalmessages hey convey,and to whatextent
do they serveas independentbasesof social experience?How do individualdifferencesand individual
discretionmediateand deflectthe impactof these alternative ourcesof political nfluence?
itizens live simultaneouslyn a varietyof socialworlds,any and all of which
might have importantpoliticalconsequences.At one and the sametime they are
rooted socially in neighborhoods,workplaces,churches,clubs, and associations.
Indeed,everycitizen lies at the centerof a social experienceproducedby a series
of intersecting,overlapping, ayeredenvironments.Eachof the environments, n
turn, has potentially mportantconsequencesfor politics becauseeach serves to
modify and deflect the opportunitiesand constraints hat circumscribe ocial in-teraction-social interaction hat servesas a vehicle for the transmission f politi-
cal information ndguidance.
Individualsand individualdifferencesare important oo-they arenot simply
artifactsof largersocialforces.Citizens payheed to someaspectsof socialexperi-
ence and they reinterpretothers.They areattracted o some environmentswhile
theyseekto avoidothers.In theseandotherways, ndividuallymotivated hoicebe-
comes animportant lement of the structuralbasis underlyingdemocraticpolitics.
This researchwas supported n partby NationalScience FoundationGrants:numberSES-8318899
to the Universityof Notre Dame, numberSES-8415572to IndianaUniversity,and SES-8319188to
WashingtonUniversity.Support was alsoprovidedby the SpencerFoundation to Iowa StateUniver-
sity), by the AmericanPolitical Science Association'sSmall Grants Program,and by the Office for
Researchand GraduateDevelopmentandthe Collegeof Arts and Sciencesat IndianaUniversity.
THEJOURNAL OF POLITICS, Vol. 55, No. 2, May 1993, Pp. 365-381
C 1993 by the Universityof Texas Press
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 3/18
366 RobertHuckfeldt,EricPlutzer,andJohnSprague
And thus, politicalpreferencecan be seen as the productof interdependentciti-zens makingchoicesanddecisionswithina rangeof contextsoverwhichthey haveonlypartialcontrol(HuckfeldtandSprague1987,1991).
This paper s motivatedby concernwith the politicalconsequencesof the mul-tiple and simultaneousbases of socialexperience.Two alternativecontexts pro-vide the focalpoint forthe presenteffort-churches andneighborhoods.And twoattitudesareexamined-policy preferencesregardingabortionand partisanself-identification. everalquestionsareaddressed:Whatarethe politicalconsequencesof parishand neighborhoodcontexts?Which context mattersmost for what be-havior,underwhatcircumstances?How do individualdifferencesmediatethe im-pactof bothparishesandcontexts?
DEFINING SOCIALSTRUCTURE
Contextual heories of politics build on the argumentthat individualpoliticalpreference s not a simple functionof individualcharacteristics lone,but ratherthe complex productof an individual'sown characteristics n combinationwiththe characteristicsndpredispositions f othersurroundingndividuals.Thus, thepolitical consequences of being an Irish Catholic might differ dramaticallybetween Massachusettsand New Hampshire, not due to variation n individualcharacteristicsbut rather due to variations in social experience (Ennis 1962;Sprague1981).Correspondingly,he politicalsignificanceof beingan IrishCath-olic could only be understoodrelativeto a particularplaceand time. The centralmotivationfor contextualtheories of politics is the idea that patterns of socialinteraction are influenced by surroundingpopulation distributions (McPhee,Smith, and Ferguson 1963).To push the previousexample somewhatfarther, fthere are proportionallymore Democratsin Massachusetts han in New Hamp-shire,MassachusettsDemocratsmighthavea muchbetterchanceof encounteringotherDemocrats.And thus the likelihoodof receivingsocialsupportfora Demo-craticpreference,andsociallysupplied nformation ompatiblewitha Democraticloyalty,wouldbe higher n Massachusetts.
In practice, he demonstration f suchcontextualeffectson politicalbehavior sa complexundertaking, t least in partdue to the difficultyof identifyingrelevantsocialboundaries. n terms of sustainingDemocratic oyalties,is it more impor-tantto live in a Democraticneighborhood, o worshipat a Democraticchurch,orto drinkat a Democraticbar?Correspondingly,omemightargue hatthe demon-strationof neighborhoodeffects is not plausiblebecausemoderncitizens are nottied to their neighborhoodsany longer;or the study of parisheffects might becalledintoquestionbecausepeopledo not go to churchanylonger;and so on.
One way out of these difficulties s to adopt an egocentricdefinitionof socialstructure.That is, a socialstructuremight be definedrelative to each individualthat is uniqueto thatindividual's diosyncraticife space(Eulau1986;Burt 1987).Such a procedureunderliesmanyeffortsat social networkanalysis,wheresocialstructure is often defined according to the individual's own report of social
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 4/18
AlternativeContextsof PoliticalBehavior 367
relationships.Such an approachhasmanyvirtues and advantages,but it includes
shortcomings s well. Most important,t runs theconceptual iskof reducing ocial
structure o a manifestationof individualchoice by creatinga model of individu-
allymanufactured atternsof social interaction.Otheranalysessuggest that suchcontroloversocial nteraction s far fromcomplete: ndividuals xercisechoice dur-
ing processesof social nteraction,but the logic of choice operatesprobabilistically
rather than deterministically(Huckfeldt 1986). Put another way, decisions
to take a job or join a churchare not retakenevery daybut ratherevaluatedonly
periodically.In the meantime and on an everydaybasis, citizens make do with
whatgoes on at work,atchurch, n the neighborhood, nd so on.
Incomplete ndividualcontrolover patternsof socialinteractionpoints toward
the virtueof a multilevelunderstanding f socialstructureandthe politicalbehav-
ior of individualcitizens. The premise of a multilevelcontextualanalysis s thatsocial structure s imposedupon individuals.This does not mean that individuals
fail to exercise discretionover socialinteraction,but only that such control s con-
strainedand boundedby a set of interactionprobabilities hat are contingenton
contextuallydeterminedsupply (McPhee, Smith, and Ferguson 1963;Huckfeldt
1986). Such a perspectiverequiresthat individualpreference, ncludingprefer-
ences regarding ocialinteractionalternatives,be considered n light of surround-
ing populationsand the distributionof variouscharacteristicsand preferences
withinthe population.
The problems hus remain:How should the relevantpopulationsbe identified?Which population boundariesare most important? n short, how should social
structurebe defined relative o politicsand politicalbehavior? n the analysesand
discussionsthat follow we consider two separatecontexts-those of the parish
and the neighborhood.These twoboundary ets do not, of course,exhaustthe list
of politicallyrelevantcontexts,but they providea usefulcomparison orpurposesof the underlyingquestionsof theoryand measurement.
ALTERNATIVECONTEXTS IN SOUTH BEND
Although numerous studies have shown the importanceof a single context(e.g., Wald, Owen, and Hill 1990),very few studies have been designedto mea-
sure the effects of multiplecontexts. As a practicalmatter,this close relationship
betweenresearchdesign and the possibilityof studying variousaspectsof social
structurehastendedto fragment tudiesof social nfluenceandpolitics.Verylittle
seriouscomparisonoccurs between the alternative ontexts of politicsbecausefew
studies are equippedto supportsuch a comparison.Instead we see the develop-ment of a neighborhoodclique, a workplaceclique, a familyclique, and a parish
cliqueamongscholars nterested n contextual heoriesof politics.
The data thatserve as the basis for this paper'sanalysisweretakenfroma 1984electionstudyconducted n SouthBend,Indiana,and its surrounding rea.As partof this study, a three-wavepanelsurveywas conductedwhich includedapproxi-
mately 1,500respondents, upplementedat each wave to offsetpanelattrition,and
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 5/18
368 RobertHuckfeldt,EricPlutzer,andJohn Sprague
most analyses of this paper are based on informationobtainedfrom the third,
postelectioninterview. The study's samplingdesign is crucial to the analysesof
this paper.All 1,500 respondentswere selectedrandomly rom 16 neighborhoods,
with approximately qual numbers chosen from each neighborhood.The neigh-borhoodswerechosen purposefully,however, n an effort to maximizesocial class
variationacross the neighborhoodsand homogeneitywithin neighborhoods.In
this way we are able both to concentrateour samplewithin particularneighbor-
hoods and to insure wide variationacross the neighborhoods.The end product
is a randomsample of respondents iving within a purposefullychosen sample of
neighborhoods-a sample of neighborhoods hosen to reflect the range of living
conditionsamong SouthBendarearesidents.
This samplingdesign providesa doublemeasurementbenefit: t can be used as
a measurementdevice both at the level of individualsand at the level of environ-ments. By aggregating t the level of neighborhoods, ny question ncluded n the
surveycan be used to measureneighborhoodpopulationcomposition.The neigh-
borhoodmeasuresemployedhereare takenfromthe third waveof the survey,and
thus are based on approximately1,500 interviewsequally spread across the 16
neighborhoods. n addition, one of the survey questions asked each respondent,
Whenyou go to services,where do you usually go? More than 20 parisheswere
named by 10 or more respondents, and the respondentschoosing one of these
parishes provide the subset of the survey used for most analyseshere. Such a
practice ntroduces wo biases into the sample.First, Catholicparishes end to belarger han those of other denominations,and thus Catholicsare overrepresented
in the resultingdataset. Second, all the respondents n theseanalysesname a par-
ticular congregation,even if they also report attending infrequently,and thus
a major portionof the population-those who do not name a congregation-is
unrepresented.'
The benefit of the resultingdata set is that it can be used to obtainaggregate
measuresof parishesas well as neighborhoods.Thus, the data set providesobser-
vation at three distinct levels-the neighborhood, he parish,and the individual.
The bulk of the analysis s conducted n terms of individuals,but foreachindivid-ualwe are ableto estimatethe attitudinal ompositionof the parishand the neigh-
borhoodcontexts withrespectto bothpartisanship nd abortionattitudes.
AGGREGATES,INDIVIDUALS,AND METHODOLOGICAL MPERATIVES
The theoreticalpremise being considered s that the behaviorof individuals s
contingenton the behaviorof otherswithin an individual's ife space (Huckfeldt
and Sprague1991).But this theoreticalpremise generatesseveralmethodological
'The first question asked: When you go to services,where do you usuallygo? The next question
was: Would you say you go to church/synagogueevery week, almostevery week, once or twice a
month, a few times a year, or never? A relatively mall numberof respondentsprovidean answerto
the firstquestionbut thenreport hat they never attend.
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 6/18
AlternativeContextsof PoliticalBehavior 369
complications. First, it is potentially tautologicalto argue that individuals are
more likely to be Democratsor pro-life in environmentswhere more people are
Democratic or pro-life. This is an especially severe problemwhen the sample
within contexts s small, andwhen such a sample s used to infer both the behaviorof individualsand the aggregatepatternsof behaviorwithin the contexts.In such
an instance the behaviorof the individual-the phenomenonbeing explained-
has a pronounced ffecton the constructionof the compositionalproperty.Part of
the problem s resolvedif we can be assured that, as a matter of construction,the
behaviorof the individual s independentof the behaviorof the aggregate. n the
analysesconductedhere, the aggregatemeasuresare constructedto insure such
independence.In calculatingthe mean abortion attitude within a respondent's
parish,for example,the respondent'sown attitudeis excluded from the calcula-
tion, and the resultingmeasure s preciselyand only an estimateof the opinionssurroundinghe respondentwithin the parish.
Moreover,we takethe additionalprecaution n the analysesbelow of including
controlsfora varietyof individual evel factors Hauser1974). Thus, forexample,
we do not simply examinethe relationshipbetween holding a pro-life opinion and
attendinga church where others hold pro-life opinions.Rather,we ask whether
people who attendpro-lifechurches are more likely to be pro-lifewhen we take
accountof theirreligiousaffiliations,heirfrequenciesof churchattendance, heir
incomes, and their educations.As we shall see, includingcontrols such as these
makes t problematicwhethersuch a relationshipholds.A second problem n the analysis s less easily resolved.The insights of the Co-
lumbia school of political sociologists ead us to expect that the effects of various
dimensionsof socialexperienceupon politicalbehaviorshould be interdependent
(Lazarsfeld,Berelson,andGaudlet 1948;Berelson,Lazarsfeld, nd McPhee 1954).
Thus, the effect of the parishshould dependon the neighborhood,and the effect
of the neighborhoodshould depend on the parish. Unfortunately,such inter-
dependenteffectsaresometimesdifficultto sort out as anempiricalmatterdue to
excessivecollinearityamongthe variousmeasures.The standardprocedureused
to test for such a contingenteffect is to calculatea simple multiplicative nter-action term for the contextualmeasuresand enter it as a regressor n the estima-
tion procedure.Unfortunately or these purposes,the interactionvariable s often
highly correlatedwith one or the other of the component contextual variables,
thereby makingthe analysisproblematic.A major culprit in the productionof
such collinearity s a lackof variation n one or the other of the componentvari-
ables. These statisticalproblemsarenot without substantive orollaries.The vari-
ation in partisanshipand especially abortion attitudes2is particularly imited
across the Catholicparishesof our sample,and thus it is sometimes difficult to
2Partisan rientationsare measuredon the familiar even-point party dentification cale, where 0 isstrong Democrat and 6 is strong Republican.Abortionattitudes are measuredon a seven-pointscale
formedby summing the responses o six questionssoliciting whethera respondentbelieves that abor-
tion should be legal or illegal in particular ircumstances.A score of 0 indicates that the respondent
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 7/18
370 RobertHuckfeldt,EricPlutzer,andJohnSprague
sort out the importanceof being Catholicas opposedto the importanceof being
amongother Catholics-a problemof trulyancientlineagein the socialsciences
(Durkheim1897).Ourgeneralpracticehereis to includesuch an interaction erm
in modelswherethe evidencesupports t.Finally,to whatextent,andon what basis,do individualscontroltheirlocations
in churchesandneighborhoods?We choosea neighborhood orits schools,its ser-
vices, the house or apartmentwe find within its boundaries.We choosea church
because t is convenientto the neighborhood,becausewe like the pastoror priest
or rabbi,becauseour parentsraisedus in its tradition,becauseit offersa shared
communityof faith. This paperis less concernedwith these overt functionsof
parishesand neighborhoods hanit is with an important atentconsequence-the
roleeachcontextplays n structuring pportunities ndconstraintshatacton social
interaction,and thus the roleeachplaysin structuring ocialinfluenceon politics(Wald,Owen, andHill 1990).Of course,overt functionsand latentconsequences
inevitablybecome ntertwined, ndthus it is importanto consider he role of indi-
vidualchoiceas it affectsthe locationof individualswithinvariouscontexts.
Most peopledo not chooseneighborhoodsand churchesbasedon the abortion
attitudesand partisanorientationsof parishionersand neighborhoodresidents,
but even thoughthese locationalchoicesare predicatedon other bases,theybring
implicationsfor a whole range of surroundingpreferences-including abortion
attitudesand partisanorientations.At the same time, and perhapsmore impor-
tantly, individualcontrol does not end once an environment s chosen. Rather,individualsareableto controltheirexposureto manyof the environmentswithin
which they are located,and thus social influenceis rightfullyseen as an ongoing
giveandtakebetweenthe sourceand the recipientof a particular oliticalmessage
(Finifter1974;HuckfeldtandSprague1987, 1991).This interplaywill providean
important ocusforthis paper.
PARTISANSHIP N PARISHAND NEIGHBORHOOD
In the analyses hat follow, we considerthe partisanorientationsof individuals
within the contextsof parishesand neighborhoods. n view of the attitudinalho-
mogeneityacrossthe Catholicparishesof our sample,we havechosento conduct
feels abortion houldnever be legal. A scoreof 6 indicates hatthe respondent eelsabortion houldal-
waysbe legal.The completebatteryof questionsfollows:
There has been a great dealof talk aboutabortion n this electioncampaign.Under whichof
the followingconditionsdo you think abortion hould be legal?First, if a woman s marriedand
doesn'twant anymore children.Shouldabortionbe legalor illegalin this situation? f a woman
is not marriedanddoesn't wantto marry he man. If a familyhas a very low income andcannot
affordanymore children.If there is a strongchance thereis a seriousdefect in the baby.If thewomanbecamepregnantas a resultof rape. If the woman'sown health is endangeredby the
pregnancy.
A slightlydifferentversionof this questionwasused in a follow-up studythat askedclergyto provide
the positionsof theirchurchesregardingwhetherabortion houldbe legal.
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 8/18
AlternativeContextsof PoliticalBehavior 371
TABLE 1
PARTISANSHIP N PARISHESAND NEIGHBORHOODS
Catholics Non-Catholics Both
Constant -1.09 6.36 -.94
(1.94) (1.77) (1.50)
Education .08 .28 .12
(1.72) (3.16) (2.78)
Income .15 -.06 .11
(2.38) ( .55) (1.89)
MeanNeighborhood .55 -2.90 .21
Partisanship (2.79) (2.62) (1.28)
Mean Parish .005 -1.76 .47
Partisanship (.03) (1.78) (3.48)
ParishMean X .80
NeighborhoodMean (2.61)
Catholic -.51
(1.99)R .12 .20 .18
N 430 142 572
MinimumTolerance .41 .02 .41
Source:1984 SouthBendStudy.
Note:Ordinary east squares; oefficient -valuesare in parentheses.
analysesseparately or Catholicandnon-Catholicrespondents,as well as for both
groupscombined.3Table 1 considersthe effects of neighborhoodand parishcon-
texts on individualpartisanship, ontrolling or individual ncomeandeducation,4and the results of the Catholicregression are shown in the first column of thetable. Two things areespeciallynoteworthyregarding he regressionamong Cath-olics. First, the model does not explain a high proportionof the variation n par-
tisanshipamong Catholicswith an R2 of only .12. Second, the only contextualattribute hatproducesaneffectamongCatholics s neighborhoodpartisanship.
Does this meanthatparishpartisanships unimportant o the individualparti-san attitudesof Catholics?Perhaps,but analternativenterpretationmight be that
3Byconductingthese analysesseparately or Catholicsand non-Catholics,andby restrictingatten-
tion to respondentswithin parisheswhere 10 or more respondentsattend, the samplesize for non-
Catholics n particular rows quite small.Thus, the resultsof this paperhave beenconfirmedusingall
respondentswho attendparisheswith morethanthree respondentsattending.This reanalysiswascon-
ducted usingordinary east squaresas well as weighted leastsquares o reflectthe small parishn-sizes
(Hanushekand Jackson1977), and these reanalysesproducedestimatesgenerally n keeping with the
resultsreported n this paper.
4Education s measuredaccording o the years of schooling to a maximumof 17 years.Family in-
come is measuredaccordingto a scale that rangesfrom 1 to 8. In view of the previouslydiscussed
collinearityproblems, he minimumtolerance s shownforeachregressionmodelin this paper,where
the tolerance oreachexplanatory ariable s definedas (1-R2) forthe R2that is obtainedwhen one ex-
planatory ariable sregressedon all otherexplanatory ariables.
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 9/18
372 RobertHuckfeldt,Eric Plutzer,andJohnSprague
the level of variation n the strongly Democratic partisansignal coming from
Catholicparishes s not sufficientto explainvariation n individualpartisanship.
Viewedfrom one perspective his wouldmean thatthe levelof partisanship mong
parishionershas little consequencefor the partisanshipof Catholicswhen onlyCatholicsareconsidered.Viewedfromanotherperspective, hehighlevelof Demo-
craticidentificationamongthe Catholics n our samplemaybe due to the gener-
allyhighlevelsof Democratic upportwithinCatholicparishes.
These resultsfor Catholicscontrastsharplywith the resultsobtained n the sec-
ond columnof table 1 whereonly non-Catholicsare considered.Indeed,all three
of the contextualmeasures-neighborhood and parishpartisanship s well as their
interaction-yield at least marginal -values.Furthermore, he R2 for this model
is increased,explaining20%of the variationn partisanship mongnon-Catholics.
The best way to understand he patternof effects suggestedby the non-Catholicmodelof table1is to portray hemin the three dimensional raphsof figure1. The
two graphsprovidealternativeperspectiveson the samepatternof effects, where
the individual evel factorsareheld constantat intermediate alues,andmeanpar-
tisanship evelsin the neighborhood ndthe parishare allowed o varyacrosstheir
observedranges.We emphasizethat the two partsof figure1 providealternative
perspectivesof the same three-dimensional pace.The only difference s that the
orderingof the neighborhoodpartisanship xis is reversed, herebyproviding wo
differentvantagepointsfrom whichto viewthe three-dimensionalurface.
As figure 1 shows, the-highest level of Republican upportis found amongre-spondentswho are located n Republicanparishesand Republicanneighborhoods.
Perhaps ess intuitively,the highest level of Democraticsupportis foundamong
respondentswho are located in Democraticparishesand Republicanneighbor-
hoods. In such a situationit would appearthat the parishshields the individual
from the neighborhoodand even sustainsa reaction o it (Huckfeldt1986).If we
can assumethat people exercisemore controlover the choice of a parishthan a
neighborhood, hese dataare consonantwith the idea that people takerefuge in
protectiveenvironments churches)whichserve to shieldthem from politicalsig-
nalsover which theyhaveless control(Finifter1974;Coleman1964).How do these resultsrelate o the literatureon cross pressures?Whilethe cross-
pressureargumentwas put forwardby the Columbiaschool to explain various
aspectsof the vote decision, it can be appliedreadilyto partisan oyaltyas well.
Individualswhoreceiveconsistentcues should tendto be resolute n their political
loyaltiesas Democratsor Republicans,while individualswho receive conflicting
cues should be ambivalentn theirloyaltiesandmore likelyto adoptan indepen-
dentorientation.
The resultsof table 1andfigure1 do not fully supportsuch a direct translation
of the cross-pressuresargument.ConsistentlyRepublicancontextsdo encourage
strong Republicanloyalties, but the other results present a more complicated
pattern.ConsistentlyDemocraticcontexts do not producethe highest levels of
Democratic dentification. ndeed, the strongest evels of Democraticsupportare
producedwhen individualsattend Democratic parishesand live in Republican
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 10/18
FIGURE 1
INDIVIDUALPARTY IDENTIFICATIONBY MEAN PARTY IDENTIFICATION N THE
NEIGHBORHOODAND IN THE PARISH, FORNON-CATHOLICS.
Strong
Republican
Strong
Democrat
Strong
Republican
Strong
Democrat
Source:Estimated ornon-Catholics rom regression stimatesof table 1.
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 11/18
374 RobertHuckfeldt,EricPlutzer,andJohn Sprague
neighborhoods.Thus, the cross-pressures rgumentmust be modified by impor-
tant asymmetriesn the influenceof varioussocial structuralpropertieson various
politicalbehaviors-asymmetries producedthrough the complexintersectionsof
multiple and overlapping nvironmentswith individualchoice.The problem still remains:why does the same patternof effects fail to appear
among Catholics?One might arguethat it is a mistake to separateCatholic and
non-Catholicanalyses n the first place, but the problem s not easily or neatly re-
solved by combiningCatholic and non-Catholicrespondents n a single regres-
sion. In an alternativemodel, shown in the third column of table 1, the effect of
the parishcontext and the neighborhoodcontext are considered for the whole
samplewith a dummy variablecontrol for being Catholic.In this instance, only
the regressioncoefficientfor the parish context produces a statisticallydiscern-
ableeffect, perhapssupporting he earlierargumentregarding he importanceofDemocraticparishesamongCatholics.
Why do we see differentpatternsof effects?A clue lies in the rangeof variation
andinterrelationships resentamongthe socialworldsinhabitedby Catholicsand
non-Catholics. n particular,he relationshipbetween the partisan ompositionof
the two contexts is very strong amongCatholics R2= .50) but nonexistentamong
non-Catholics R2= .02). In short, the Catholics n our sample are less likely to
experienceneighborhoodsand parishesthat emit divergentpolitical signals,and
hence it is more difficultto uncover the patternof interdependent ffectspresent
amongnon-Catholics.The level of diversity n the social worldsof Catholicsandnon-Catholics s not
simply a methodological ssue. Being Catholic (or not being Catholic)is much
more than an intrinsic, idiosyncratic haracteristic f individuals.Being Catholic
(ornot being Catholic) s a determinant f socialexperience,and the consequences
extend farbeyond systemsof beliefs to include the realitiesof dailylife. The roots
of these differentpatternsof existenceare historicaland institutional.First, as a
matterof history, manyof South Bend's Catholicsare the descendantsof Eastern
European mmigrantswho came to Indiana ate in the nineteenthandearlyin the
twentiethcenturies.That commonhistoricalexperienceproduceda Catholic,eth-nic communityin South Bend with strongDemocratic roots (Stabrowski1984),
and even thoughthe communityhasbeentransformed n a continualbasis,a his-
toricalresidue s still evident.Second, as a matterof institutionaldesign,Catholic
parishesaregeographically ased-individuals areassigned o nearbyparishes.As
a result these two elements of social structure-parish and neighborhood-tend
to traveltogether.The substantive andtheoreticalandmethodological)esson we
takefromthis analysis s that the partisan ffects of parishandneighborhood on-
texts are betterunderstoodrelative o the religiousgroup n question.
ABORTIONATTITUDES IN PARISH AND NEIGHBORHOOD
The politicalreachof social influenceextends farbeyond partisanship. n the
following analyseswe considerthe consequencesof neighborhoodsand parishes
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 12/18
AlternativeContextsof PoliticalBehavior 375
for the abortionattitudes of individual respondents.The initial question thus
becomes: Does the structure of social influence vary as a function of the be-
havior being influenced?If citizens live simultaneously n a variety of social
worlds, do all these social worldshave similar politicalconsequencesacrossdif-ferentbehaviors?
Using table 1 as a point of reference,a differentpatternof contextualeffects is
demonstratedwhen the neighborhoodand parisheffectson abortionattitudesare
consideredin table 2(a). In this instance no contextual effect is present among
Catholics.The neighborhood ffect amongnon-Catholicsdisappears,but a parish
effect is sustained-non-Catholic respondentswho attendpro-choiceparishesare
more likely to hold pro-choice attitudes.Both models demonstratean effect for
church attendance hat is especiallypronouncedamongCatholics:peoplewho at-
tend more frequentlyappear o be morepro-life in theirattitudes.In summary, he evidence suggeststhat the neighborhood s relevant or some
behaviorsbut not forothers, amongsome respondentsbut not amongothers. In a
similarvein, parisheffects are present for both partisanshipand abortionatti-
tudes among some people but not among others. At least among non-Catholics,
it would appearthat social influenceis specializedand structuredby the multi-
dimensionalnatureof socialexperience.Peoplelook to differentplacesfor differ-
ent cues regardingdifferent behaviors. Coreligionistsmay be more influential
regardingpolitical ssues that involve faith and morality.Secular settingsmay be
more importantas sources of influenceregardingsecularmatters-politics andpartisanship.
When the model is estimated for a combined sample of Catholicsand non-
Catholics with a dummy variablefor Catholics (third column of table 2a), the
non-Catholic patternof effects is maintainedbut the magnitudesof the parish
effect and its corresponding -value are substantiallyreduced. And thus adding
Catholicsto the analysistends to attenuatethe parisheffect. Why do we fail to
find a parisheffect among Catholics?Or do we fail to find an effect? Recallthat
the variation n abortionattitudes is restrictedacrossCatholicparishes.In other
words,to attenda Catholicparish s to be exposedto a context wherepro-lifeatti-tudes prevail.The differenceamong Catholicparishesmay not be sufficientto
providestatisticalpurchase,but this is not the same as saying thatthe parishdoes
not matter.
An alternativeway to addressthe importanceof the parish s in terms of indi-
vidualexposure.Recallthe strongeffect of church attendanceamongCatholics.
At least two interpretationsmight accountfor this effect. First, it may simplybe
the case that regularattendersare more religious,and religious people are more
pro-life.Alternatively,t maybe the case that regularattendersexposethemselves
morefrequently o the normative tanceof theircongregations egardingabortion(Sprague 1981).We cannoteasilychoosebetween these two alternatives orCath-
olics because here is relatively ittle parishvariationn abortionattitudes.That is,
we cannot observethe consequenceof Catholicsattendinga pro-choiceCatholic
parish.
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 13/18
TABLE 2
ABORTIONATTITUDES IN NEIGHBORHOODSAND PARISHES
A. Effects of neighborhood ndparishcontexts
Catholics Non-Catholics Both
Constant 4.12 5.32 3.97(3.17) (1.76) (3.23)
PartyIdentification -.07 -.07 -.07(1.31) ( .76) (1.43)
ChurchAttendance -.66 -.29 -.58(7.85) (2.02) (7.96)
Education .07 -.01 .06(1.34) ( .01) (1.30)
Income -.01 .08 .004(.18) (.69) ( .07)
MeanNeighborhood .31 -.59 .16Attitude (1.03) (.82) (.58)
MeanParish -.07 .53 .23Attitude (.30) (2.42) (1.43)
Catholic -.53
(2.01)R2 .16 .14 .18N 348 108 456MinimumTolerance .71 .71 .64
B. Contingenteffectsof parishcontexts,dependingon attendance.
Catholics Non-Catholics Both
Constant 4.58 13.72 9.18(1.56) (4.42) (5.08)
PartyIdentification -.06 -.02 -.07(1.17) ( .25) (1.54)
ChurchAttendance -.57 -2.99 -1.69
.86) (4.12) (4.43)Education .08 .003 .06
(1.46) (.04) (1.21)Income -.002 .12 .02
(.04) (1.13) (.36)MeanParish .09 -2.22 -1.08
Attitude ( .10) (2.94) (2.28)AttendanceX -.03 .67 .33
ParishAttitude ( .14) (3.77) (2.97)Catholic -.76
(2.80)R2 .16 .24 .20N 348 108
456MinimumTolerance .01 .03 .03
Source:1984SouthBendStudy.Note:Ordinary eastsquares; oefficient -valuesare in parentheses.
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 14/18
AlternativeContextsof PoliticalBehavior 377
A far differentsituationexists among non-Catholics,however.The composi-
tion of their parisheswith respectto abortionattitudesshows much more varia-
tion, andthus a directtest is available orthe two previouslystatedalternatives. f
attendance s importantbecause it exposes church members to the climate ofopinionwithinthe parish, hen we shouldexpectthat:(1) the effectof the climate
of opinionshoulddependon attendance,and (2) the effect of attendance hould
dependon the climateof opinion.
These expectationsareset againstevidencein table2(b), wherea simplemulti-
plicative nteractionvariables included n the regressionmodels.This interaction
variable s calculatedas the productof individualchurch attendancefrequency
and the mean parish abortionattitude. Not unexpectedly,given the restricted
rangeof mean abortionattitudesamongCatholicparishes, his additionalvariable
fails to provideexplanatorypurchaseamongCatholics.Amongnon-Catholics, ncontrast,all three variables-church attendance,parishabortionattitudes,and
the interaction-produce crispt-valuesandsubstantial oefficients.A similarpat-
tern is presentfor the sampleas a whole in the thirdcolumnof the table,but the
magnitudeof the coefficientsandtheir t-valuesarediminished.
Once again, the interactioneffect among non-Catholics is best considered
throughthe use of three-dimensionalprojections(see figure 2). As before, the
figureis displayedfrom two differentperspectives n this instance accordingto
alternativeorderingsof the attendancefrequencyaxis. Carefulexaminationof
figure2 showsthat:1. Increasedattendancen pro-lifeparishes s related o pro-lifeattitudes.
2. Increasedattendance n pro-choiceparishes s relatedto pro-choiceattitudes.
3. Peoplewho attendregularlyaremorelikelyto havepro-lifeattitudes f theyat-
tendpro-lifechurches.
4. Amongthosepeoplewho attend nfrequently,an inverserelationship xistsbe-
tween individualattitudesand the parish-individuals aremore likelyto have
pro-choiceattitudes f theybelongto pro-lifeparishes.
The first three effects aredirectlyinterpretable,but the fourtheffect requiresmore carefulconsideration.We maybe witnessinga situation n whichnonatten-
danceserves as a shield againsta parishcontext which the individualfinds to be
disagreeable.Thus, attendance ervesto exposeindividuals o the climateof opin-
ion of the parish,but parishionerscontrol their own attendancepatterns.And
thus theymaychoosenonattendance san articulated r unarticulated esponse o
an environment hat is at variancewith their own viewpoint(compare o Finifter
1974).
In summary,parishesmattermore than neighborhoodswith respect to influ-
encing abortionattitudes.The effect of parishes s, however,mediatedby expo-sure. Regularattendersare more likely to be affectedthan infrequentattenders.
Thus, we see a patternof effects in which the structureof socialinfluencevaries
acrossattitudesandindividuals n systematicways.
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 15/18
FIGURE 2
INDIVIDUALABORTIONATTITUDE BY MEAN ABORTIONATTITUDE IN PARISH AND
INDIVIDUAL ATTENDANCEFREQUENCY,FORNON-CATHOLICS.
Pro-choice
Pro-life
CC *eCv'
Pro-choice
-5
Pro-life
Source: Estimated for non-Catholics from regression estimates of table 2b.
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 16/18
AlternativeContextsof PoliticalBehavior 379
THE SOURCE OF PARISH INFLUENCE
We may,of course, be witnessingthe waggingtail rather han the barkingdog.
What is the source of the parisheffect on parishioners?Do they respond to com-
monly held beliefs among theircoreligionists,or do their own opinions-and the
opinions of other parishioners-respond to the theological positions of their
churches?Four yearsafter the original South Bend study, a team of interviewers
conducted a follow-up study in which they interviewed the pastors of many
parishesattendedby the respondents.In some instances,of course, the inter-
viewed parish pastor (or priestor rabbi) s not the samepastorwho was in place
fouryearsearlier.For present purposes,however,this does not pose a significant
problem.
Ourgoalis to assess the theologicalpositionof the respondent's hurchregard-
ing abortion.Each pastorwas askedto providehis or her own interpretation f the
church'steachingon the abortion ssue relativeto the same set of circumstances
regardingwhich the respondentshad been askedto give their positions.Thus, we
have the judgmentof a well-informedobserverregarding he theologicalposition
of the church which the respondentreports attending-a judgmentthat is likely
to be quitestableacross ime and acrossclergywithin the same church.
The initial question becomes: how well do the theological positions of the
churchescorrespond o the attitudes of the parishioners?They correspondquite
well for the sampleas a whole (n = 514), generatinga correlationof .44. The cor-
relation s dramatically educed,however,when Catholicsand non-Catholicsare
considered separately:.04 among Catholics (n = 371) and .14 among non-
Catholics(n = 125). This means that the theologicalposition of the church ex-
plains,atmaximum,2% of the variationn parishioner ttitudes.
The weak correlationamongCatholics s due at least in partto the extremely
restrictedrangeof variation n the interpretedpositionsof church teaching pro-
vided by Catholic clergy. A different situation prevails among non-Catholics,
however, where church teachings vary from consistently pro-choice to consis-
tently pro-life.The prevailingattitudesof parishionersend to be more moderate
than the positions of their churches. Parishioners n pro-life churches tend to
be more pro-choicethan the clergy interpretedposition of their churches, and
parishioners n pro-choicechurches tend to be more pro-life. Thus, while the
non-Catholic parishionersgenerallytake the theologicalcue of the church, the
slopeof the relationships fairlyflat,and there are substantial ggregatedeviations
even from this modestrelationship.
Finally,which factordoes a betterjobof explaining ndividualattitudesregard-
ing abortion: he theologicalpositionof the individual'schurch or the positions
held amongotherparishioners?n analysesnot shownhere, the modelsof table 1
and table2 were rerunwith the clergy-supplied heologicalposition substitutedfor the meanparishattitude,and the results were much less impressive.The cli-
mate of opinionmeasureprovidesmuchmorestatisticalpurchase han the clergy-
reported heologicalposition.
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 17/18
380 RobertHuckfeldt,EricPlutzer,andJohn Sprague
CONCLUSION
People locate themselves in neighborhoods,churches, workplaces,clubs, and
associations. They make these locationalchoices for good reasons on rational
grounds, but in the processthey also define-even if indirectlyand unintention-
ally-the dimensionsof their social experience.This social experiencehas rele-
vance far beyond the basis of the original choice. In particular, t defines the
compositionof politicalpreferences o which the individual s exposed.
As a matterof convenienceand necessity, contextualstudies of politics have
typicallysummarized ocial experience n terms of geographical nits-precincts,
neighborhoods, ounties,states. All too frequently hesemeasurementprocedures
have obscuredthe multidimensionalnature of social experiencethat is produced
by the separate lementsof social structure.Each structural lement carrieswith it
political implications hat are potentially mportant.In some instances,for some
behaviors,one structuralelement is importantwhile another s not. In other in-
stances, multiplestructural lements must be seen as being interdependentn the
constructionof a politicallyrelevantsocialexperience.In still other instancesthe
importanceof particular tructural lements is best seen as the result of individual
exposure,and individualexposure s best seen as a furthermanifestationof indi-
vidual choice.
In summary,social influencein politics is structurallybased, but it frequently
dependson multipleelements of socialstructure.It imposesitself on individuals,
but individualchoice playsa majorrole in determiningwhich individualsare ex-
posedto whichpoliticalcues andviewpoints.The emergingview is one in which
social nfluence s structurally eterminedbut in which the social nfluenceprocess
cannot be understoodaccording o any simpledeterministic ogic (Boudon 1986).
Ratherthe end resultof social influence n politics is createdby the points of in-
tersection between individualchoice and the multiplebases of social experience.
Manuscriptubmitted3 April 1991
Finalmanuscripteceived 6July 1992
REFERENCES
Berelson, BernardR., PaulF. Lazarsfeld,and WilliamN. McPhee. 1954. Voting.Chicago:University
of ChicagoPress.
Boudon, Raymond.1986.Theories f Social Change.Berkeley:Universityof CaliforniaPress.
Burt, Ronald S. 1987. SocialContagionand Innovation:Cohesionversus StructuralEquivalence.
Americanournalof Sociology 2:1287- 1335.
Coleman,James S. 1964.Introductiono Mathematical ociology.New York:FreePress.
Durkheim,Emile. 1897. Suicide.Translationby John Spaulding publishedin 1951. New York: Free
Press.
Ennis, Philip H. 1962. The ContextualDimension in Voting. In Public Opinionand Congressional
Elections, d. WilliamMcPheeandWilliamA. Glaser.New York:FreePress.
Eulau, Heinz. 1986. Politics,Self, andSociety.Cambridge,MA:HarvardUniversity Press.
This content downloaded from 189.125.96.10 on Fri, 6 Jun 2014 14:42:29 PMAll use subject to JSTOR Terms and Conditions
8/12/2019 HUCKFELDT_Alternative Contexts of Political Behavior
http://slidepdf.com/reader/full/huckfeldtalternative-contexts-of-political-behavior 18/18
AlternativeContextsof PoliticalBehavior 381
Finifter,AdaW. 1974. The FriendshipGroup as a ProtectiveEnvironment or PoliticalDeviants.
American oliticalScienceReview68:607-25.
Fuchs, LawrenceH. 1955. American ews and the PresidentialVote. AmericanPoliticalScienceRe-
view49:385-401.
Hanushek,EricA., andJohnE. Jackson.1977. StatisticalMethodsor Social Scientists.New York:Aca-demicPress.
Hauser, Robert M. 1974. ContextualAnalysis Revisited. SociologicalMethodsand Research
2:365-75.
Huckfeldt,Robert. 1986.Politics n Context.New York:Agathon.
Huckfeldt,Robert,andJohnSprague.1987. Networks n Context:The Social Flow of PoliticalInfor-
mation. American oliticalScienceReview81:1197-1216.
Huckfeldt,Robert,and John Sprague.1991. DiscussantEffectson VoteChoice: Intimacy,Structure,
and Interdependence. ThejournalofPolitics53:122-58.
Lazarsfeld,Paul F., BernardBerelson,and Hazel Gaudet. 1948. The People'sChoice.New York:Co-
lumbiaUniversityPress.
Leege, David C., andMichaelR. Welch. 1989. ReligiousRootsof PoliticalOrientations. TheJournal
ofPolitics52:137-62.
McPhee,WilliamN., RobertB. Smith, andJackFerguson.1963. ATheoryof InformalSocialInflu-
ence. In FormalTheories fMass Behavior, d. WillamN. McPhee.New York:FreePress.
Sprague,John. 1981. IsThere a MicroTheory ConsistentwithContextualAnalysis? n Strategies f
Political nquiry, d. ElinorOstrom.BeverlyHills: Sage.
Stabrowski,Donald. 1984. APoliticalMachine, anEthnicCommunity,andSouth Bend'sWest Side:
1900-1980. Ph.D. diss. Universityof NotreDame.
Wald, KennethD., Dennis E. Owen, andSamuelS. Hill, Jr. 1990. PoliticalCohesionin Churches.
TheJournal fPolitics52:195-215.
RobertHuckfeldtis professorof politicalscience, IndianaUniversity,Bloom-
ington,IN 47405.
Eric Plutzer is assistantprofessorof political science, Iowa State University,
Ames,IA 50011.
JohnSprague s professorof politicalscience,WashingtonUniversity,St. Louis,
MO 63130.