The Consequences of Cross-Cutting Networks for Political Participation Author(s): Diana C. Mutz Reviewed work(s): Source: American Journal of Political Science, Vol. 46, No. 4 (Oct., 2002), pp. 838-855 Published by: Midwest Political Science Association Stable URL: http://www.jstor.org/stable/3088437 . Accessed: 01/02/2013 13:44 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].

.

Midwest Political Science Association is collaborating with JSTOR to digitize, preserve and extend access to American Journal of Political Science.

http://www.jstor.org

This content downloaded on Fri, 1 Feb 2013 13:44:28 PM All use subject to JSTOR Terms and Conditions

The

for

of

Consequences Political

Diana

C.

Mutz

Cross-Cutting

Networks

Participation

The

Ohio

State

University

This study advances our understand? a concept ing of "cross-pressures," in the earliest studies of recognized

tradition of research within social psychological studies of voting behavior the "crossemphasizes problems posed by Avenerable pressures" for individuals faced with a voting decision. In early vot?

American voting, but for which em? pirical evidence and theoretical de?

ing research, The People s Choice suggested that conflicts and inconsistencies among the factors influencing an individual's vote decision discourvoters from in involvement the aged early campaign: "Whatever the source

has been sorely lacking. Although the current consensus sug? gests that political cross-pressures velopment

are of little, if any, consequence for I find that political participation, people whose networks involve greater political disagreement are less likely to participate in politics. Two social psychologicai processes

of the conflicting pressures, whether from social status or class identification, from voting traditions or the attitudes of associates, the consistent result was to delay the voter's final decision" (Lazarsfeld, Berelson, and Gaudet 1944,60). The American Voter even more directly acknowledged the problem of conflicting considerations political choices: surrounding The person who experiences some degree of conflict tends to cast his vote for President with substantially less enthusiasm ... and he is somewhat less likely to vote at all than is the person whose partisan feelings are entirely consistent. [... ] If attitude conflict leaves its impress on

are suggested to account for this effect. First, those embedded in cross-cutttng social and political networks are, as a consequence, more likely to hold ambivalent politi?

several aspects of behavior it also influences what we will call the individual's involvement in the election. (Campbell et al. 1960,83, 85)

cal views, which in turn discourage political involvement. Second, social

Likewise, Hovland and colleagues suggested that the effects of conflicting social influences included "vacillation, apathy, and loss of interest in con-

accountability pressures in crosscutting networks discourage political participation; the inherently controver-

flict-laden

sial nature of politics is perceived to pose threats to the harmony of social relationships.

issues" (1953,283). Cross-pressures arising from multiple group affiliations have long been of interest in political sociology as well. Simmel (1955), for example, attributed great significance to the "web of affiliations" and their cross-cutting social relationships

that were contrasted with the highly homogeneous kinof an earlier of era. Studies status simiship-linked groups inconsistency as un? larly conceived of individuals who were experiencing cross-pressures der stress (e.g., Hope 1975). Those exposed to a variety of cues about appropriate

social and political

attitudes

were assumed

to experience

dis-

Diana C. Mutz is Professor of Political Science and Journalism and Communication, The Ohio State University, 2140 Derby Hall, 154 North Oval Mail, Columbus, Ohio 43210-1373 ([email protected]). This study was supported by a grant from the Spencer Foundation and by a National Science Foundation Grant to the Center for Advanced Study in the Behavioral Sciences. American Journal ofPolitical Science, Vol. 46, No. 4, October 2002, Pp. 838-855 ?2002 by the Midwest Political Science Association

838

This content downloaded on Fri, 1 Feb 2013 13:44:28 PM All use subject to JSTOR Terms and Conditions

ISSN 0092-5853

CROSS-CUTTING NETWORKS AND POLITICAL PARTICIPATION

comfort resolved

as a result, though arguments this discomfort varied.

about how people

Interest in testing the cross-pressure hypothesis died out after subsequent analyses repeatedly failed to confirm these early findings. For example, Pool, Abelson, and Popkin (1965) looked for these effects in the 1960 election data, but to no avail. Moreover, in a reof data from the 1948 Elmira study and the 1956 analysis national election study, Horan found that even the earlier evidence had resulted from "interpreting direct effects of

national

social positions more complex 1971, 657).

on nonvoting

(and interest)

as due to a

(Horan cross-pressures phenomenon" In other words, the investigators had uninthe direct effects of membership confounded

tentionally in social categories

with the effects of being linked to conflicting social categories. Subsequent studies have differed that were evaluin terms of the kinds of cross-pressures ated (e.g., primary group, class-based, and so forth), and whether bivariate or multivariate approaches were used (Jackson

and Curtis

promising

1972; Davis 1982), but despite a by the late 1970s studies of cross-

beginning, due to an accumulapressures had largely disappeared tion of negative evidence (see Knoke 1990, for a review).

the theory of political cross-pres? As Horan summarized, sures became part of "that category of plausible theories whose empirical support has been cut out from under them" (1971,659). In the early studies,

of whether a permeasurement was typically acson was experiencing cross-pressures such as complished using social category memberships

the fact that a person was both white collar and Catholic, for example. Conflicts were defined purely at the level of social categories deemed potentially conflictual by the rethat might exert pressure on not even were documented people though interaction was to be responmicro-level assumed the process generally searchers. Actual interactions

sible for producing cross-pressures. Today several data sets that include batteries of items on individuals' political networks make it possible to test this hypothesis in a manof actual (as opposed to into cross-pressures and in a manner that

ner that allows measurement ferred) exposure allows insight into potential

processes of influence.

839

(Horan 1971, 659). Most versions are in agreement with the assumption that "social interaction is the primary mechanism social and indi? group membership linking vidual

how people sort out their opinions in the face of conflict? ing social pressures rather than on how such exposure alters their political participation. As Horan (1971) exbecause this evolved from plains, theory gradually and of it has often lacked data, interpretation analysis In this formulation. clarity as an abstract theoretical study I attempt to remedy this problem by focusing specifically on developing theory and evidence relevant to of conflicting

the effects

Context, and

Networks,

Participation

of this concept in early voting Despite the prominence is in research, the "'theory of political cross-pressures' and assumptions" fact a rather mixed bag of propositions

social

influences

on political

participation. In one sense, the link between the composition of social environments and people's political participation For example, has already been widely acknowledged. studies of social context

and social networks already have on a consensus that political activity is strong converged rooted in social structure. But for the most part this conclusion refers to the idea that highly participative social and active social networks

contexts

further

enhance

for an individual's

the Mo-

political participation. has been recognized as one ofthe major factors underlying turnout (e.g., Rosenstone and Hansen 1993; Verba, Schlozman, and Brady 1995). In

prospects bilization

via social networks

within the immediaddition, the extent of participation has been found to have significant ate social environment consequences pation

(e.g.,

for the likelihood

of individual partici? 1979, 1986; Giles and Dantico it remains unclear whether social envi?

Huckfeldt

1982), although ronment affects all or only some particular kinds of participatory acts (see Kenny 1992; Leighley 1990). Whether these studies are based on aggregated contextual

measures

of social

environment

or measures

of

an individual's

immediate social network, the general is that a participatory social environment still more participation, and the mechanism as-

conclusion

begets sumed to account

for this effect is the same in both cases; that is, the more people interact with one another within a social context, the more norms of participation will be

transmitted, Social

behavior," (1971, 650) but beyond this, of the term vary. For example, the emstudies of cross-pressures has been on

political

understandings phasis in many

and the more people

will be recruited

into

political activity. To be sure, social context appears to make a difference in the extent to which individuals become politically ac? of political beliefs within tive, but does the homogeneity the social environment also have consequences for politi? cal participation? Some scholars have theorized that people

may be more likely to participate

This content downloaded on Fri, 1 Feb 2013 13:44:28 PM All use subject to JSTOR Terms and Conditions

if their social

DIANA C. MUTZ

84O is consistent

environment

with their political beliefs (e.g., Leighley 1990; cf. Oliver 1999),1 but the kind of data most appropriate to testing this hypothesis have been in short

amount

of vacillation

and uncertainty in people's views, most of which appeared to be driven by competing val? ues and considerations as applied to political questions rather than from a lack of political

expertise. Likewise, ef? responses to survey questions that citizens' opinions are comprised of

supply.

forts to better understand

have suggested competing ideas and considerations Potential

In order to avoid confusion mulations

of

Mechanisms

Influence

with the many different

ofthe

forI use

(Zaller and Feldman is 1992; Zaller 1992), and that, as a result, ambivalence often difficult to distinguish from nonattitudes as typically measured (see also, Alvarez and Brehm 1995,1997).

In studies

of issues ranging from race to abortion, am? has been found to play an important part in the formation of citizens' attitudes. The consequences of

original cross-pressure hypothesis, the term cross-cutting networks and refer to the extent of cross-cutting exposure taking place within them. In this

bivalence

a theory to explain the pro? study I focus on developing cess by which social interactions that cross lines of politi? cal difference might affect political participation. Assum-

have been less widely explored, al? to include more moderate political though they appear less positions, certainty in political judgments (Guge and

ing for the moment that cross-cutting exposure does, in there are at least two po? fact, discourage participation, tential social psychological mechanisms that might ex?

Meffert

plain such an effect.

anced or even-handed

Political First, political nal ambivalence

Ambivalence

inaction

could be induced by the attitudithat cross-cutting exposure is likely to an individual. If citizens are embedded

engender within in networks that do not reinforce

their viewpoints, but them with political information

instead tend to supply that challenges their views, then such cross-cutting ex? could make uncertain of their own posure people posi? tions with respect to issues or candidates, and make them less likely to take political action as a result. In this case it is intrapersonal conflict that drives the effect, and the chain of events

which,

of voting intentions, 1998), delayed formation and instability in candidate evaluations (Lavine 2001). Ambivalence also has been tied to having more bal-

about political issues judgments 1981; Guge and Meffert 1998). For ex? (e.g., Sniderman awareness of conflicting consider? ample, simultaneous ations bearing on a given issue can lead to higher levels of (see Green, Visser, and Tetlock integrational complexity 2000), which is similar to what others call "balanced judg? ment"; that is, an awareness that many political questions are not black and white, and a recognition that there is to be said for "the other side" (Sniderman something 1981). This condition is distinct from having a middle-ofthe-road position or no position at all, although the typical approach to the measurement of political attitudes makes such distinctions

cross-cutting expo? in turn, reduces politi?

cal participation. A relatively recent resurgence of interest in ambiva? lence?that is, the simultaneous presence of both posi? tive and negative considerations directed toward the been noted in both qualitative approaches to understanding political

object?has

and quantitative attitudes. For example, Americans,

ambivalence

is one in which

sure leads to ambivalence

same attitude

political

Hochschild

in her in-depth interviews with noted a tremendous (1981,1993)

1 Leighley (1990), for example, operationalizes exposure to conflict in one's personal network as respondents' reports of whether a friend has tried to convince him/her to vote for a candidate of the opposite party, and finds, contrary to her hypothesis, that conflict enhances participation. In contrast to Leighley's hypothesis, Oliver (1999) suggests that economic diversity in cities should produce competition and greater conflict over resources and that macrolevel conflict should encourage participation. His analyses suggest that this is only true for participation in local politics, and the re? sults do not speak directly to the question of cross-pressures.

Social

difficult to observe.

Accountability

The second

possible reason that cross-cutting political would discourage political participation is be? cause cross-cutting networks create the need to be acnetworks

countable

to conflicting constituencies. According to this social psychological the problem is not that mechanism, one is internally conflicted over which side to support, but rather that one feels uncomfortable

taking sides in the face of multiple competing constituencies. The need for social accountability creates anxiety because interper?

sonal disagreement threatens social relationships, and there is no way to please all members of one's network and thus assure social harmony. As Green, Visser, and Tetlock suggest, "The decision maker is caught in the middle, pushed one way by part ofthe group, and pulled the other way by an opposing faction. The individual is forced to defend a position in what may be perceived as a 'no win' situation, in which one side will inevitably be

This content downloaded on Fri, 1 Feb 2013 13:44:28 PM All use subject to JSTOR Terms and Conditions

CROSS-CUTTING NETWORKS AND POLITICAL PARTICIPATION

alone were at (2000, 4). If this mechanism work, we would expect mainly public forms of political to be affected; in private situations such as participation alienated"

networks should pose few the voting booth, cross-cutting problems due to social accountability. evidence in support of the idea that Qualitative

84I

Steiner 1966). By devaluing

politics and avoiding political controversy, people effectively resolve the problem. In a recent experiment exposure to arguments manipulating on opposite sides of an issue, as well as whether subjects were accountable to conflicting or unified constituencies,

interper? people avoid politics as a means of maintaining sonal social harmony has been around for some time. In

Green, Visser, and Tetlock (2000) found that cross-pressured subjects engaged in many decision-evasion tactics and (including buckpassing, procrastination, exiting the

noted in his in1950s, Rosenberg (1954-55), har? interviews that the threat to interpersonal was a significant deterrent to political activity.

to contradicsituation) in order to avoid accountability constituencies. If we these tory generalize findings outside the laboratory, we would expect those with high levels of

the mid depth

mony More recent case studies

have provided further support for this thesis. In her study of New England town meet(1980) workplace, Mansbridge ings and an alternative an avoidance was that conflict observed impor? similarly to political participation (see also Eliasoph Conover and In their focus group discussions, evidence that considerable also found Searing (1998) of of and ofthe risks aware wary politi? people were both

tant deterrent 1998).

As one of for interpersonal cal discussion relationships. their focus group participants put it, "It's not worth it... to try and have an open discussion if it gets them [other citizens] upset" (1998,25). Verba and Nie (1972) applied a similar logic to their in which activities were analysis of political participation

cross-cutting off political thus

making

exposure decisions their

in their personal networks to put as long as possible or altogether,

political

unlikely. In the analyses extent cross-cutting

participation

that follow,

particularly

I first examine

to what

within social networks exposure for political par? does, in fact, have adverse implications of I various kinds. evaluate the extent Second, ticipation to which these two proposed processes of influence? intrapersonal

ambivalence

and interpersonal for the overall impact

countability?account diversity on political

not only on the basis ofthe extent of initiative required, and the scope of the outcome, but also on the extent to which conflict with others was involved.

social

ac?

of network

participation.

differentiated

Moreover, in a recent analysis of national survey data, Ulbig and Funk (1999) found that individual differences in conflict avoidance were negatively related to political participation

more public kinds, particularly acts such as protesting, working on a cam? of some

participatory paign, and political

discussion.2

avoidance discourages partici? research with social psychological consistent is also pation dis? on how people handle nonpolitical interpersonal agreements. When a person confronted with a difference The idea that conflict

does not shift to the other person's views or them to adopt his or her own views, the most persuade reaction is to devalue the issue itself (e.g., alternative likely of opinion

2Research on the "spiral of silence" similarly contends that per? ceived minority status will affect political preferences by discouraging the expression of political viewpoints that are perceived to be unpopular (Noelle-Neumann 1974). However, in this case po? litical discussion is the dependent, rather than the independent variable. But if one considers discussion as a form of political par? ticipation, the spiral of silence can be interpreted as suggesting that fear of interpersonal conflict inhibits participation, as also shown by Ulbig and Funk (1999). Nonetheless, the hypothesis that per? ceived support for one's opinions in the broader political environ? ment relates to willingness to speak out publicly has received very limited support to date (see Glynn, Hayes, and Shanahan 1997).

Data

and

Methods

these questions, I drew on two representasurveys, both including large batteries of tapping characteristics of respondents' political

To investigate tive national measures

The first survey was supported by the Spencer and executed by the University of Wisconsin Survey Center in the fail of 1996, immediately preceding the presidential election in November (see Appendix A). networks.

Foundation

This telephone interview included a battery of items talked addressing the frequency with which respondents about politics with up to three political discussants, plus five separate items assessing the extent to which respon? dents agreed or disagreed with the views of each of the that were named. These five mea? political discussants sures per discussant were combined into an additive scale the extent to which people's networks exrepresenting them to views unlike their own.3 In total the posed sample included 780 respondents providing information on over 1700 of their discussion partners, thus providing tremendous about the extent of ex? depth of information posure

to political

disagreement

in personal

networks.

3Cronbach,s alpha indicated that these five items scaled relatively well, with alphas of .78, .81, and .81 for the first, second and third named discussants, respectively

This content downloaded on Fri, 1 Feb 2013 13:44:28 PM All use subject to JSTOR Terms and Conditions

DIANA C. MUTZ

842 Moreover, because this survey included multiple indica? it variable for each discussant, tors of the independent was possible to create a reliable measure of the extent to a person's political network included exposure to views, a measure that assessed the extent to oppositional which a source provided dissonant contact independent ofthe frequency of that contact.

which

be noted that this kind of measure is very from what has typically been used in studies of are used as the where group memberships cross-pressures It should

different

basis for inferring that cross-cutting contact has occurred. It is a huge operational leap from knowing that a person is both Catholic and a businessman, for example, to infer that they are subject to political cross-pressures from proCatholic acquaintances and pro-Republican Democratic business people. It is far less of a leap when that same per? son names the members mediate

network.

teristics

of those

of those groups as part of his imBut even knowing the political charac? in one's network does not ensure that

contact has occurred. For this reason, the cross-cutting measures used in this study also take into account the fre? quency of political discussion with each discussant. Even if one's network

people with differing political are viewpoints, it is difficult to argue that cross-pressures at work if politics is hardly ever discussed. These American

includes

data were supplemented with data from the of the Cross-National Election component

Project (CNEP), a telephone survey executed during the 1992 elections (see Appendix B).4 The CNEP data pro? vided an item measuring the extent of exposure to disagreement (based on choice of presidential candidate) for each of up to five discussants.5 The other major difference between the Spencer and CNEP surveys was that the CNEP questionnaire used a discussant generator which asked respondents to volunteer the names of people with whom

whom

they talked most "about the events of the recent presidential election campaign," thus generating a more explicitly political discussion partner. Previous comparisons of name generators suggest that the explicitly politi? cal frame will produce more nonrelatives and discussants

who are weak ties (Huckfeldt and Sprague 1995b), thus making the Spencer survey more likely to generate dis? cussants who will be politically dissimilar to the main respondents. The two surveys complemented one another well for these purposes. The Spencer survey provided extensive information on exposure to oppositional political views and some variables

useful for pinning down mechanisms of influence, while providing more limited information on political participation. The CNEP study, in contrast, included more participation measures plus a question time of presidential vote decision, but it inless information on corporated exposure to political dif? ference within the respondent's network. Unquestionaddressing

ably, both

surveys

operationalization ditional approach

an improvement in the represent of cross-pressures relative to the tra? that simply assumes cross-cutting ex?

in combinations of par? posure based on membership ticular religious, economic, occupational, age, or racial that may (or may not) be central to an categories individual's social network, that may (or may not) repre? sent oppositional political perspectives, and that may (or on respondents may not) exert cross-pressures through communication. political In general, the extent of accuracy in respondents' on the political leanings of political discus? self-reports sants is relatively high (see, e.g., Huckfeldt and Sprague 1995a).7 Moreover, dependent themselves

because

the CNEP data included

in?

reports of candidate choice by the discussants which have been analyzed in other studies,

matters" for the first they discussed "important four discussants, while the Spencer survey asked explicitly about people with whom respondents talked about "gov?

known.

dyads showed

of the re?

ernment, elections and politics."6 For the fifth discussant in the CNEP questionnaire, respondents were asked with

spondents' 78 percent

and a full

4See Beck, Dalton, and Huckfeldt (1992) for details. 5Although the CNEP data also included an item tapping the frequency of disagreement with the discussant f/they had talked about politics, because the discussant generator asked for "impor? tant matters" discussants, respondents were not asked this ques? tion about a large proportion of the discussants who did not claim to talk politics. In order to avoid losing a large proportion of re? spondents due to missing data, I did not include this measure in the operationalization of cross-cutting exposure for the CNEP sample. 6If a respondent in the Spencer survey was unable to name a politi? cal discussant, they were then asked about an "important matters" discussant.

the extent

of projection in respondents' perceptions 12 ofthe Only percent respondent-discussant potential evidence of projection political views onto the discussant,

is

of respondents' perceptions were accurate re? of the discussants' views; the remaining 9 percent ports were situations in which perceptual errors were made in the direction

of a candidate other than the respondent's favored one (Mutz and Martin 2001). Only 8 percent of dyads involved perceptual errors in which the respondent claimed that preferred one candidate and erroneously 7Although respondents are likely to perceive somewhat greater agreement in their networks than actually exists, it is their percep? tions of their discussants that should shape their tendency to engage or withdraw politically. For this reason perceptual measures are preferable to assessments drawn from the perspectives of dis? cussants, which may be subject to inaccuracies as well.

This content downloaded on Fri, 1 Feb 2013 13:44:28 PM All use subject to JSTOR Terms and Conditions

CROSS-CUTTING NETWORKS AND POLITICAL PARTICIPATION

the discussant

preferred the same one. The remaining 4 12 percent mentioned above) were cases percent (ofthe in which a neutral discussant was erroneously perceived to favor the respondent's own candidate. Despite these high levels of accuracy in respondent some might consider the discussants' reperceptions, to those

provided by the respondents. of operationalizing social influ?

ports superior However, for purposes it makes ences on the respondent, that discussants' views will influence

little sense to argue the respondent even

when these views have not been clearly communicated. Although the choice of measure should make little differ? ence in these particular data, it makes more theoretical will experience crosssense to argue that respondents that their netto the extent that they recognize pressures hold differing political views. of results proceeds by first analyzing to the general question of whether pertaining

work members

The discussion findings

cross-cutting

networks

have implications for political of net? the characteristics

I break down

participation. works into three

variables

their

representing separate size, frequency of political discussion, and degree of heterogeneity. Next, I evaluate the two potential social psyfor this relationship. Toward that chological explanations items in the the six participation end I disaggregate or not do involve direct CNEP survey into those that do

843 Zanna, and Griffin 1995; also Breckler (see Thompson, 1994).9 When applied to choice of presidential candidate,

this formula

takes into account the overall intensity of toward the candidates, corrected by the extent to feelings which the valence of respondents' reactions to the candi? dates differ. Using this measure, ambivalence decreases as a function of increasing differentiation between the can? didates and increases as a function of the average inten? sity of feelings toward them. The advantage of Griffin's formula over measures that simply take the absolute value of the difference in evaluations is that it assigns those

who are highly ambivalent (such as people who rate both candidates as 90s on a candidate-feeling ther? a different score who from those are mometer) simply indifferent as to their choice of candidate (such as when two candidates

are both rated a neutral 50). With this ophigh-intensity feelings with highly similar ratings quite logically produce the greatest am? bivalence scores. erational

measure,

of social accountability, To examine the importance the Spencer survey included a battery of items tapping people's reluctance to involve themselves in face-to-face conflict

(see Appendix A).10 This index allowed for more of the extent to which conflict avoid? direct assessment ance facilitates

the effects of cross-cutting

of opposing views.8 As Verba activities that involve have suggested, are more likely to engender conflict, public expression but it is not the fact that the act is performed with others confrontation

exposure.

with those

and Nie (1972)

present that is crucial so much as whether one must confront those of oppositional views, with all the potential social awkwardness of such encounters. I created mea? To examine the role of ambivalence, of Griffin's formula, a prefersures using a modification because it simultaneously able measure of ambivalence considers

both the dissimilarity

and intensity of attitudes

8Although these items have been disaggregated a number of ways in the past?particularly based on individual versus socially based participation?here the criterion was whether performing the act generally necessitates face-to-face contact with those of differing views, rather than whether it requires getting together with others. For example, attending a fundraiser or rally for a candidate is clearly social, but it involves contact almost exclusively with likeminded individuals attending the same function and thus does not require a willingness to confront people with differing views. Giving money to candidates can also easily be accomplished without confrontation, even though such records are technically public. In contrast, trying to persuade someone else to one's own viewpoint and actively working to support a particular candidate require the individual to make it known to potentially unsupportive others that this is one's preference. Thus these two items?persuasion and working for a candidate?were classified as potentially confrontational, while putting up a sign, giving money, and attending a meeting were considered largely nonconfrontational.

Findings Using both the CNEP and Spencer surveys, Table 1 summarizes the relationship between cross-cutting exposure and the likelihood of participation, for after controlling and a host of political interest, strength of partisanship, other

variables.

As shown

Table 1, the likelihood gressional

elections

in the first two columns

of

of voting in presidential and con? is a function of the usual predictors

9The formula used to tap both the intensity and dissimilarity of views was adapted to the case of a three-candidate race as follows: A = I-D Where A = ambivalence; I = absolute value of average intensity of feeling for the two most preferred candidates; D = absolute value of differentiation among the two most preferred candidates. In the case of feeling thermometers in which 5 represents the low? est intensity of feeling, this translates to: A = ((abs(5 - candidate A rating) + abs(5 - candidate B rating))/2) - abs(candidate A - candidate B). 10The alpha for these for four items was .60, and the scale was then dichotomized to increase reliability and facilitate tests for interactive effects.

This content downloaded on Fri, 1 Feb 2013 13:44:28 PM All use subject to JSTOR Terms and Conditions

DIANA C. MUTZ

844 Table

1

Effects

of Network

Composition

on Political

Participation

Note: Entriesare coefficients fromfive ordered probitanalyses with z-values in parentheses. ***p