13
What is Important? The Impact of Interpersonal Political Discussion on Public Agendas Bas W. van Doorn, The College of Wooster Objective. This article investigates the role of discussion networks in agenda-setting. More specifically, the focus is on the question whether the partisan composition of people’s discussion networks affects what they judge to be the most important prob- lems facing the country. Method. Using data from the 2000 American National Election Studies (ANES), I employ logistic regression analysis. Results. The findings suggest that discussion networks indeed play a significant role in setting the public agenda, even taking into consideration people’s media use and several other poten- tially relevant variables. Conclusion. While prior research has already linked political discussion to participation and vote choice, this study provides evidence that it also affects what issues people prioritize. The priming literature convincingly demonstrates that vote choice is signif- icantly affected by the issues that are salient to voters on election day. Petrocik’s theory of issue ownership, meanwhile, suggests that the salience of different issues tends to benefit the two major parties differentially: “Democrats have an advantage when problems and issues associated with social welfare and intergroup relations are salient. Republicans have an advantage when issues related to taxes, spending, and the size of government are on the agenda” (Petrocik, Benoit, and Hansen, 2003–2004). Most research on the determinants of issue salience focuses on media con- tent and, in electoral contexts, campaign strategy (e.g., Iyengar and Kinder, 1987; Druckman, 2004; Vavreck, 2009). Lippmann makes the point that the media “is like the beam of a searchlight that moves restlessly about, bringing one episode and then another out of darkness into vision” (1922:229). This changing focus, as Cohen (1963) and McCombs and Shaw (1972) contend, affects what issues the public prioritizes, a process known as agenda-setting. In this article, I argue that in addition to looking at the media as setting the agenda, we also need to look at the role of social networks in this process. Drawing on literatures on discussion networks and issue ownership, I propose Direct correspondence to Bas W. van Doorn, Department of Political Science, The College of Wooster, 1189 Beall Avenue, Wooster, OH 44691 [email protected]. The author will share all data with those wishing to replicate the study. Van Doorn thanks Angie Bos, Logan Dancey, Corrine McConnaughy, Wendy Rahn, Monica Schneider, and Ismail White for helpful advice. SOCIAL SCIENCE QUARTERLY, Volume 95, Number 1, March 2014 C 2012 by the Southwestern Social Science Association DOI: 10.1111/j.1540-6237.2012.00936.x

What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

  • Upload
    bas-w

  • View
    215

  • Download
    1

Embed Size (px)

Citation preview

Page 1: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

What is Important? The Impact ofInterpersonal Political Discussion onPublic Agendas∗

Bas W. van Doorn, The College of Wooster

Objective. This article investigates the role of discussion networks in agenda-setting.More specifically, the focus is on the question whether the partisan composition ofpeople’s discussion networks affects what they judge to be the most important prob-lems facing the country. Method. Using data from the 2000 American NationalElection Studies (ANES), I employ logistic regression analysis. Results. The findingssuggest that discussion networks indeed play a significant role in setting the publicagenda, even taking into consideration people’s media use and several other poten-tially relevant variables. Conclusion. While prior research has already linked politicaldiscussion to participation and vote choice, this study provides evidence that it alsoaffects what issues people prioritize.

The priming literature convincingly demonstrates that vote choice is signif-icantly affected by the issues that are salient to voters on election day. Petrocik’stheory of issue ownership, meanwhile, suggests that the salience of differentissues tends to benefit the two major parties differentially: “Democrats havean advantage when problems and issues associated with social welfare andintergroup relations are salient. Republicans have an advantage when issuesrelated to taxes, spending, and the size of government are on the agenda”(Petrocik, Benoit, and Hansen, 2003–2004).

Most research on the determinants of issue salience focuses on media con-tent and, in electoral contexts, campaign strategy (e.g., Iyengar and Kinder,1987; Druckman, 2004; Vavreck, 2009). Lippmann makes the point that themedia “is like the beam of a searchlight that moves restlessly about, bringingone episode and then another out of darkness into vision” (1922:229). Thischanging focus, as Cohen (1963) and McCombs and Shaw (1972) contend,affects what issues the public prioritizes, a process known as agenda-setting.In this article, I argue that in addition to looking at the media as setting theagenda, we also need to look at the role of social networks in this process.Drawing on literatures on discussion networks and issue ownership, I propose

∗Direct correspondence to Bas W. van Doorn, Department of Political Science, The Collegeof Wooster, 1189 Beall Avenue, Wooster, OH 44691 〈[email protected]〉. The authorwill share all data with those wishing to replicate the study. Van Doorn thanks Angie Bos,Logan Dancey, Corrine McConnaughy, Wendy Rahn, Monica Schneider, and Ismail Whitefor helpful advice.

SOCIAL SCIENCE QUARTERLY, Volume 95, Number 1, March 2014C© 2012 by the Southwestern Social Science AssociationDOI: 10.1111/j.1540-6237.2012.00936.x

Page 2: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

What is Important? The Impact of Interpersonal Political Discussion 133

a set of hypotheses as to what types of networks will encourage attentionto what types of issues. I then offer an analysis of 2000 American NationalElection Studies (ANES) data, which indeed indicates that the partisan iden-tification of the people with whom one discusses politics affects the issues oneregards as important.

Agenda-Setting

The concept of agenda-setting finds concise expression in Cohen’s claimthat the media “may not be successful much of the time in telling peoplewhat to think, but it is stunningly successful in telling its readers what tothink about” (1963:13). In other words, the agenda-setting hypothesis holdsthat while media content does not have a hypodermic “opinion-changing”effect, it does strongly affect what issues people will focus on and judge to beimportant. This in turn may have consequences for the government’s agenda,translating into public policy (Jamieson, 2000).

Hundreds of agenda-setting studies have been published (e.g., McCombsand Shaw, 1972; Iyengar and Kinder, 1987; for a review see Rogers andDearing, 1988). The majority of these studies focuses on the direct effect ofmedia content on public priorities, neglecting other factors that may affectpeople’s issue importance judgments. In this article, I raise the possibility thatmedia coverage of an issue will not have a uniform effect across the public andthat one reason for this is the discussion networks in which people are located.

Social Context and Discussion Groups

A large body of research demonstrates the effects of interpersonal conver-sation on citizens’ political opinions and behavior (e.g., Beck et al., 2002;Berelson, Lazarsfeld, and McPhee, 1954; Katz and Lazarsfeld, 1955; Gamson,1992; Mondak and Mutz, 2001; Walsh, 2003; Mutz, 2006). The partisanand ideological identification of the people one discusses politics with matterin terms of whether and how dramatically conversation affects people’s opin-ions. It is particularly important whether and to what extent one’s politicaldiscussion group contains individuals with whom one disagrees politically.

Katz (1957) argues that, as a result of self-selection, “[o]pinion leaders andthe people whom they influence are very much alike and typically belong tothe same primary groups of family, friends, and co-workers” (p. 77). This viewis qualified by Huckfeldt and Sprague (1995), who argue that although peoplegenerally want to construct homogeneous (in a partisan and ideological sense)networks, their ability to do this depends to a considerable extent on the socialcontext in which they live. This is consequential: Huckfeldt and Sprague findthat “a politically supportive social context sustains the receiver’s misperceptionof dissonance-producing, socially transmitted, political information” (p. 143).This suggests that social context indeed affects what people learn about politics

Page 3: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

134 Social Science Quarterly

and how they interpret it (or, perhaps more accurately, have it interpreted forthem).

Another important piece in this tradition is Mendelsohn’s (1996) analysis ofrolling cross-section data from the 1988 Canadian election. Mendelsohn findsthat interpersonal communications prime issues, while media consumptionprimes the character of leaders. While this provides evidence that interpersonalconversations matter, it leaves unexplored the differential effects of differentlyconstituted discussion networks.

Theory and Expectations

Given the oft-demonstrated importance of partisanship in U.S. politics, Iargue that the partisan leaning of people’s networks has significant impacton their importance judgments. For directional hypotheses, I rely on thetheory of issue ownership. As Petrocik and his colleagues (1996; Petrocik etal., 2003–2004) have demonstrated, the two major parties “own” differentissues. Republicans, for example, are seen as particularly strong on defense,while Democrats are considered strong on social welfare issues. In addition,Petrocik finds, there are “performance issues”—the economy, foreign relations,and the functioning of government—the ownership of which depends on whatparty is considered to be in charge and whether that party’s performance onthe issue is rated as good or bad.

Presumably, networks that overwhelmingly consist of supporters of oneparty will engender attention to and perceived relevance of issues that themajority party (in that particular network) is strong on. That is, people inmore Republican networks will be more likely to name a Republican-ownedissue as most important than those in more Democratic networks, who willbe more likely to name Democrat-owned issues. The reason for this is thatparty elites tend to highlight issues their party is strong on (see Petrocik et al.,2003–2004, for evidence from presidential campaigns from 1952–2000) andattentive citizens tend to take their cues from the elites associated with the partythey identify with (Zaller, 1992; Lupia and McCubbins, 1998; Druckman,2001), heightening the probability that people in networks skewing towardsone party will focus on party-owned issues.

The general expectation is that an increase in the proportion of Republi-cans in a network will heighten the probability that a respondent will namea GOP-owned issue as most important. However, respondent party identifi-cation matters here such that this probability will be lower for Democrats infully Republican networks than for Republicans in fully Republican networks.Democratic respondents in mixed networks, the expectation is, will displayprobabilities of mentioning a GOP-owned issue that are higher than those forDemocrats in fully Democratic networks, but lower than those for Democratsin fully Republican networks. Similarly, Republicans in mixed networks willbe more likely to name a GOP-owned issue than those in fully Democratic

Page 4: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

What is Important? The Impact of Interpersonal Political Discussion 135

networks and less likely to do so than Republicans in fully Republican net-works.

Data and Measures

Data

For this analysis, I use data from the 2000 ANES. This data-set has theadvantage that it contains measures of who respondents report they discusspolitics with and for whom respondents think their discussion partners voted.This combined with respondents’ own reported partisanship allows me toconstruct variables that get at the discrepancy or match between discussionnetwork and one’s own partisan preference.1 Moreover, these data also includemeasures of what respondents judge to be the “most important problems fac-ing the nation”.2 In addition, and as is traditional for the ANES, the datainclude a host of control variables. Despite some significant shortcomings,this data-set is unique in combining measures of discussion partners and theirpolitical preferences, most important problem measures, and other poten-tial determinants of the public agenda. For my model, five independent andone dependent variable are of interest. The dependent variable of interestis the problem mentioned by the respondent as most important, while theindependent variables of interest are education, trust in media, media atten-tion, political knowledge, and—most important for this analysis—networkcomposition.3

Most Important Problem

This variable is measured through an open-ended question that asks therespondent: “What do you think are the most important problems facing thiscountry?” If the respondent provides an answer, she is asked whether thereare any other problems she judges to be important. Subsequently, she is askedto indicate which of the issues she has listed is the most important in heropinion. The responses to these open-ended questions are coded accordingto subcategories and listed under different, more inclusive categories, such as“social welfare problems,” “foreign affairs,” and “public order.” For my analysis,I adhered to the ANES coding scheme, with one exception. Where the ANESlisted abortion under social welfare problems and gay rights, euthanasia, andother moral issues under public order, I combined these items into one “moralissues” category.

1Unfortunately, the data-set does not have information on the party identification of thediscussants, thus leaving me no other option than to use presumed vote choice as a proxy fordiscussant partisanship.

2http://www.electionstudies.org/studypages/2000prepost/2000prepost_qnaire_pre.pdf.3No other data-set I am familiar with contains all of the variables of interest.

Page 5: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

136 Social Science Quarterly

TABLE 1

Issue Ownership (2000 Ownership of Performance Issues in Parentheses)

Republican Democratic Performance

Defense Social welfare Economy (Democrat)Crime Agriculture Foreign relations (Democrat)Morality Racial issues Government functioning (Republican)Public order Natural resources

Labor

SOURCE: Petrocik (1996), Petrocik et al. (2003–2004).

In addition to being more reasonable, this way of coding the data alsomakes it consistent with Petrocik’s issue-ownership categories, thus providinga tighter fit between theory and test (see Table 1). After collapsing the subcat-egorized data into the revised ANES category scheme, I created a number ofvariables consistent with the issue ownership scheme. That is, I created dummyvariables that distinguished Democrat-owned issues from Republican-ownedissues.

Issue-ownership theory identifies the economy, foreign affairs, and gov-ernment functioning as performance issues, i.e., issues whose ownershipis subject to change according to current or recent party performance onthem. Given the strength of the economy and the lack of U.S. involve-ment in any large-scale armed conflict at this time—when the President wasa Democrat—I assign the economy and foreign affairs to the Democrats.ANES and other data provide support for this: a higher proportion of 2000ANES respondents (29.1 percent) believed that the Democrats were betterable to handle the economy than the Republicans (22.5 percent). A GallupPoll from October 8-9, 2000, asked “which political party will do a betterjob of keeping the country prosperous?” Again, more respondents (47 per-cent) named the Democratic than the Republican party (40 percent). The2000 ANES’ question as to which party people believe will do a better job“keeping us out of war” sheds some light on ownership of foreign affairs:more respondents (20.6 percent) name the Democrats than the Republicans(17.9 percent). Because of Clinton’s reputation of lacking personal ethics,especially post-Lewinsky scandal, I assign government functioning to theRepublicans.4

4Some readers may be skeptical about the assignment of the performance issues. However,although assigning the performance issues to the parties along these lines does strengthen therelationship between networks and agendas, a robust relationship remains when excludingthem from the analysis altogether.

Page 6: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

What is Important? The Impact of Interpersonal Political Discussion 137

Network Composition

Network composition is the independent variable of greatest interest inthis study.5 Using respondents’ answers to the question who they think theirdiscussion partner(s) voted for, I constructed a variable that captures the rel-ative proportion of Democrats and Republicans in respondents’ discussionnetworks.6 This is not a particularly precise measure of network composition,since it does not directly ask discussion partners for their partisan identifica-tion, nor does it allow for inclusion of people perceived to not have votedor for a consideration of the strength of partisanship of discussion partners.However, as discussed above, this data-set is the only one that includes in-formation about networks as well as a most important problem variable andother potentially relevant variables.

The variable is coded 0–1, with a value of 0 signifying a 0 percent Republican(and thus a 100 percent Democratic) network, and a value of 1 signifying a100 percent Republican (and thus a 0 percent Democratic) network. Usingthese variables, we can ascertain whether those in more Republican networksjudge different problems to be important than those in more Democraticnetworks.7

Party Identification

Party identification strongly influences people’s political attitudes and be-havior (e.g., Campbell et al., 1960; Lewis-Beck et al., 2008). Applied toagenda-setting, it is plausible that partisans will prioritize issues that theirpreferred party is perceived as strong on, especially because candidates tend tohighlight these issues (Petrocik, 1996; Vavreck, 2009). Party identification ismeasured using the traditional 7-point ANES scale, which I rescaled so it runsfrom 0–1, with 0 being strong Democrat and 1 being strong Republican.

Education

Most prior research on agenda-setting is either experimental or uses aggre-gate time-series data. Thus, little attention has been given to what demographic

5The use of the term network may confuse slightly, since in my conception it also includesrespondents who list only one discussant.

6There are two potential sources of slippage here. First, respondents may misperceive whotheir discussion partner voted for. Second, even if respondents correctly report who theirdiscussion partner voted for, voting for a presidential nominee of a certain party does notnecessarily mean you identify with that party. Despite these two potential sources of error, Ibelieve this variable satisfactorily captures network composition.

7Among partisans who have at least one discussion partner, 34.9 percent are in matchingand homogenous networks, over 30 percent are the sole Democrat/Republican in an otherwiseRepublican/Democrat network, and the rest are part of mixed networks. In other words, thereis nice variance in network type.

Page 7: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

138 Social Science Quarterly

factors may be related to agenda-setting. It is possible that education—togetherwith its covariates income and social class—has an effect on what issues peoplejudge to be important, independent of media focus and media attention. Anissue that may be particularly important to highly educated people—who aremore likely to have postmaterial values (Inglehart, 1981)—is the environment.People with less education may be more interested in economic issues and is-sues concerning social welfare. The variable is coded 1–7 with 1 signifyingthe lowest level of education (8 grades or less) and 7 signifying the highest(advanced degree).

Media Exposure

Exposure to media content is a prerequisite for agenda-setting in the originalformulation of the hypothesis, the expectation being that people who followthe news regularly will prioritize the issues salient in media coverage. Therefore,I include a variable that measures media exposure. I constructed a scale out ofthree measures8: number of days the respondent watched national televisionnews, number of days the respondent watched the local television news, andthe number of days the respondent read a daily newspaper.

Trust in Media

As Miller and Krosnick (2000) have argued, “both priming and agenda-setting may be most likely to occur among people who trust the competenceand motives of media personnel the most” (p. 303). For this reason, I includeda measure of trust in media in the analysis. I reverse-coded the ANES data sothat 1 indicates that the respondent “almost never” trusts the media “to reportthe news fairly” and 4 indicates that the respondents “just about always” trustthe media to do this.9

Knowledge

Scholars have debated what role knowledge plays in agenda-setting andpriming. One line of reasoning portrays people that are less politically knowl-edgeable as victims that are “swept away” by media content (Iyengar et al.,

8I added up the responses for the three questions and divided the total by three, producinga scale that runs from 0 (no media exposure) to 7 (maximum media exposure). The measureof local news exposure was constructed by adding up the reported number of days respondentswatched the local morning TV news and the local night TV news and dividing this numberby two.

9For complete original question-wording, see: ftp://ftp.nes.isr.umich.edu/ftp/nes/studypages/2000prepost/2000prepost_qnaire_post.pdf.

Page 8: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

What is Important? The Impact of Interpersonal Political Discussion 139

TABLE 2

Logistic Regression Coefficients

I. Full Sample II. Republicans III. Democrats(GOP Issues) (GOP Issues) (Democrat

IVs N = 567 N = 221 Issues) N = 282

Party identification 0.090 (0.268) N/A N/APolitical knowledge −0.006 (0.092) 0.185 (0.147) 0.124 (0.129)Media exposure −0.073

+(0.049) −0.129 (0.080) 0.024 (0.068)

Trust in media −0.350∗∗(0.135) −0.413∗∗ (0.218) 0.347∗ (0.192)Education −0.089 (0.066) −0.005 (0.112) 0.089 (0.091)Network composition 0.958∗∗∗ (0.229) 1.389∗∗∗ (0.381) −0.804∗∗ (0.321)Constant 0.316 (0.477) −0.096 (0.745) −0.440 (0.657)

+: p < 0.15; ∗: p < 0.10; ∗∗: p < 0.05; ∗∗∗: p < 0.01. Standard errors in parentheses.

1984:780). However, Miller and Krosnick (2000) provide evidence that peo-ple with high political knowledge and high media trust are most likely totake cues from the media as to what the most important problems facingthe country are. Either way, there is enough reason to include a measure ofknowledge in the analysis.

The knowledge scale is based on 4 questions that ask the respondent toidentify Janet Reno, Tony Blair, Trent Lott, and William Rehnquist. Thevariable is scaled from 0 to 4, with the score indicating the number of questionsthe respondent answered correctly.

Results

I first present results for all respondents who were asked the Most ImportantProblem question and who reported discussing politics with at least one person(Table 2, column I).10

As evidenced by the sign of the coefficient (positive), an increase in the pro-portion of Republicans in a network increases the likelihood of the respondentmentioning a Republican-owned issue as most important. Conversely, a higherproportion of Democrats in a network is associated with a higher probabilityof selecting a Democrat-owned issue as most important.11

10After all, in this scheme the chance of mentioning an issue owned by one party is 1 minusthe chance of mentioning an issue owned by the other party, leading the results to be mirrorimages of each other.

11Omitting Network Composition from the model leads to little change in coefficient andstandard error for the Party Identification variable (0.079 vs. 0.090 and 0.231 vs. 0.268,respectively). Leaving out Party Identification similarly barely affects coefficient and standarderror for the Network Composition variable (1.03 vs. 0.958 and 0.228 vs. 0.229, respectively).This indicates that including both variables in the model produces no significant bias.

Page 9: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

140 Social Science Quarterly

Thus, there appears to be a substantially and statistically significant re-lationship between partisan composition of discussion networks and whatissue respondents consider the most important. These results are all the moreconvincing because the respondents’ own party identification is controlled for.

These results allow us to draw conclusions about the effects of networkcomposition in general, but not about the effect of matching versus discrepantnetworks. Following previous work on polarization (e.g., Isenberg, 1986;Huckfeldt and Sprague, 1995; Mendelberg, 2002), I expect an increase inthe proportion of fellow partisans in a network to lead to an increase inprobability of choosing party-owned issues greater than the increase observedwhen disregarding the subject’s partisanship. To test for the occurrence ofpolarization in matching networks, I ran logistic regressions for subsamples ofpartisans.

The results (Table 2, column II) show that for Republicans an increasein the proportion of fellow partisans in their networks leads to a higherprobability of naming Republican-owned issues as most important. Politicalknowledge, trust in media, and media exposure are all also significant. AmongDemocrats (column III), the negative coefficient for network compositionindicates that the higher the proportion of Republicans in their networks, theless likely Democrats are to name Democrat-owned issues as most important.Conversely, it means that the higher the proportion of Democrats in thenetwork, the higher the likelihood they will pick a Democrat-owned issue asmost important.

The predicted probabilities depicted in Figure 1 indicate that for Republi-cans the probability of mentioning a Republican-owned issue as most impor-tant increases by 30.7 percentage points as we move from a fully Democraticto a fully Republican network. Among Democrats, we see an 18 percentagepoint increase in the probability of mentioning a Democrat-owned issue asmost important as we move from a fully Republican to a fully Democraticnetwork, thus suggesting that a change in network composition has less of animpact among Democrats than among Republicans.12

A possible reason for this smaller effect amongst Democrats is that 8 yearsunder a Democratic president may have led to a policy agenda dominatedby Democrat-owned issues. This may have made it harder for Republicans topersuade their Democratic discussion partners to perceive Republican-ownedissues as most important. Given a Democrat-dominated policy agenda, how-ever, it may have been easier for Democrats to persuade Republican discussionpartners that Democrat-owned issues are most important. This would leadto large differences between Republicans in matching versus Republicans indiscrepant networks.

12This is even clearer when taking into account 95 percent confidence intervals for the prob-ability estimates: among Democrats they overlap for all networks except the fully Republicanand fully Democratic ones. For Republicans, all majority Republican networks are differentfrom both the 0 and the 25 percent Republican networks.

Page 10: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

What is Important? The Impact of Interpersonal Political Discussion 141

FIGURE 1

Probability of Prioritizing Party-Owned Issue

On the whole then, these analyses provide evidence that having a match-ing network (i.e., one that is homogeneous and consistent with a person’spartisanship) increases the likelihood of making issue importance judgmentsconsistent with the partisan make-up of the network. Furthermore, if myargument holds water, polarization only occurs amongst people who identifywith the party that does not dominate the policy agenda.

Discussion

These analyses suggest that the composition of one’s discussion networkaffects what problems one finds most important to the country. Deliberativetheory argues that for democracy to function properly, citizens need to be ex-posed to different perspectives on issues (e.g., Benhabib, 1996; see also Mill,1989). In keeping with this, much of the framing literature provides evidencethat exclusive exposure to one viewpoint has potentially great effects on publicopinion. This study, however, suggests that exclusive exposure to one perspec-tive may only be part of the problem; people may also just focus on a limitednumber of issues and these issues differ in part according to the composition of

Page 11: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

142 Social Science Quarterly

people’s discussion networks. In other words, American democracy may notonly suffer because people are exposed to one-sided information, it could alsobe weakened because people only consider a limited amount of (and different)issues, thus prohibiting consideration of and deliberation about the remainderof the universe of issues.

As the grasp of traditional broadcast media on the public continues to di-minish (Prior, 2007), the fragmentation of public agendas may increase andthe beneficial effects of a focused public agenda—transcendence of parochialand self-interested concerns, broadening horizons—may disappear (e.g., Sun-stein, 2001). Also, to the extent one believes the levels of partisan segregationin the United States will continue to increase, these findings spell troublebecause people will increasingly be able to self-select into matching networks.

There are some limitations to this study. First, it is cross-sectional and it isthus hard to firmly establish a causal connection between independent anddependent variables. Second, the data do not contain information on whatexactly people discuss with the people they talk to about politics. A studythat would include the content of discussions and follow respondents overtime may provide more insight into the processes at work in the setting ofpublic agendas. As it is, the results of the analyses presented in this articleoffer sufficient promise to address these limitations and to pursue this topicfurther.

REFERENCES

Beck, Paul Allen, Russell J. Dalton, Steven Greene, and Robert Huckfeldt. 2002. “The SocialCalculus of Voting.” American Political Science Review 96:57–73.

Benhabib, Seyla. 1996. “Toward a Deliberative Model of Democratic Legitimacy.” Pp. 67–92in Seyla Benhabib (ed.), Democracy and Difference: Contesting the Boundaries of the Political.Princeton: Princeton University Press.

Berelson, Bernard, Paul Lazarsfeld, and William McPhee. 1954. Voting: A Study of OpinionFormation in a Presidential Campaign. Chicago: University of Chicago Press.

Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960. TheAmerican Voter. Chicago: University of Chicago Press.

Cohen, Bernard C. 1963. The Press and Foreign Policy. Princeton: Princeton University Press.

Druckman, James N. 2001. “Using Credible Advice to Overcome Framing Effects.” Journal ofLaw, Economics, and Organization 17:62–82.

———. 2004. “Priming the Vote: Campaign Effects in a U.S. Senate Election.” PoliticalPsychology 25:577–94.

Gamson, William A. 1992. Talking Politics. New York: Cambridge University Press

Huckfeldt, Robert, and John Sprague. 1995. Citizens, Politics, and Social Communication. NewYork: Cambridge University Press.

Inglehart, Ronald. 1981. “The Silent Revolution in Europe: Intergenerational Change inPost-Industrial Societies.” American Political Science Review 65:991–1017.

Page 12: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

What is Important? The Impact of Interpersonal Political Discussion 143

Isenberg, Daniel J. 1986. “Group Polarization.” Journal of Personality and Social Psychology50:1141–51.

Iyengar, Shanto, and Donald R. Kinder. 1987. News That Matters:Television and AmericanOpinion. Chicago: University of Chicago Press.

Iyengar, Shanto, Donald R. Kinder, Mark D. Peters, and Jon A. Krosnick. 1984. “The EveningNews and Presidential Evaluations.” Journal of Personality and Social Psychology 46:778–787.

Jamieson, Kathleen H. 2000. Everything You Think You Know About Politics . . . and WhyYou’re Wrong. New York: Basic Books.

Katz, Elihu. 1957. “The Two-Step Flow of Communication: An Up-To-Date Report on anHypothesis.” Public Opinion Quarterly 21:61–78.

Katz, Elihu, and Paul Lazarsfeld. 1955. Personal Influence: The Part Played by People in the Flowof Mass Communications. New York: Free Press.

Lewis-Beck, Michael S., Helmut Norpoth, William G. Jacoby, and Herbert Weisberg. 2008.The American Voter Revisited. Ann Arbor: University of Michigan Press.

Lippmann, Walter. 1922. Public Opinion. New York: Free Press.

Lupia, Arthur, and Matthew D. McCubbins. 1998. The Democratic Dilemma: Can CitizensLearn What They Need to Know? Cambridge: Cambridge University Press.

McCombs, Maxwell E., and Donald L. Shaw. 1972. “The Agenda-Setting Role of MassMedia.” Public Opinion Quarterly 36:176–187.

Mendelberg, Tali. 2002. “The Deliberative Citizen.” Pp. 151–193 in Michael X. Delli Carpini,Leonie Huddy, and Robert Y. Shapiro, eds., Research in Micropolitics, Vol. 6. New York: Elsevier.

Mendelsohn, Matthew. 1996. “The Media and Interpersonal Communications: The Primingof Issues, Leaders, and Party Identification.” Journal of Politics 58:112–125.

Mill, John Stuart. 1989 [1859]. On Liberty and Other Writings. In Stefan Collini (ed.), NewYork: Cambridge University Press.

Miller, Joanne M., and Jon A. Krosnick. 2000. “News Media Impact on the Ingredients ofPresidential Evaluations: Politically Knowledgeable Citizens are Guided by a Trusted Source.”American Journal of Political Science 44:301–315.

Mondak, Jeffery, and Diana C. Mutz. 2001. “Involuntary Association: How the WorkplaceContributes to American Civic Life.” Presented at the annual meeting of the Midwest PoliticalScience Association.

Mutz, Diana C. 2006. Hearing the Other Side: Deliberative versus Participatory Democracy. NewYork: Cambridge University Press.

Petrocik, John R. 1996. “Issue Ownership in Presidential Elections, with a 1980 Case Study.”American Journal of Political Science 40:825–850.

Petrocik, John R., William L. Benoit, and Glenn J. Hansen. 2003–2004. “Issue Ownershipand Presidential Campaigning, 1952–2000.” Political Science Quarterly 118:599–626.

Prior, Markus. 2007. Post-Broadcast Democracy: How Media Choice Increases Inequality inPolitical Involvement and Polarizes Elections. New York: Cambridge University Press.

Rogers, Everett M., and James W. Dearing. 1988. “Agenda-Setting Research: Where Has ItBeen, Where Is It Going?” Pp. 555–594 in James A. Anderson (ed.), Communication Yearbook11. Newbury Park, CA: Sage.

Sunstein, Cass. 2001. Republic.com. Princeton: Princeton University Press.

Page 13: What is Important? The Impact of Interpersonal Political Discussion on Public Agendas

144 Social Science Quarterly

Vavreck, Lynn. 2009. The Message Matters: The Economy and Presidential Campaigns. Princeton:Princeton University Press.

Walsh, Katherine Cramer. 2003. Talking About Politics: Informal Groups and Social Identity inAmerican Life. Chicago: University of Chicago Press.

Zaller, John R. 1992. The Nature and Origins of Mass Opinion. Cambridge: Cambridge Uni-versity Press.