Political Efficacy and Campaign News Attention As Catalysts of Discursive Democracy: The Case of the 2004 U.S. Presidential Election


By Hsiang-Ann Liao

WJMCR 18 (December 2009)



This paper argues that models of discursive democracy need to be contextual and domain specific, incorporating cognitive or psychological catalysts or other situational factors that prompt people to talk about politics. Different election campaigns might require different discursive democracy models to encapsulate how and why citizens talk about politics. Based on Kim et al.’s model of deliberative democracy,1 a contextual model of discursive democracy is proposed for the 2004 U.S. presidential election. Political efficacy and campaign news attention were examined as psychological and cognitive catalysts to the contextual model. It was found that both internal political efficacy and attention to campaign news coverage were relevant to a discursive democracy model for the 2004 U.S. presidential campaign. Two structural equation models were formulated to examine the direct and indirect effects between variables.


The role of political conversation in democracy has generated unprecedented attention since scholars started to examine a more discursive aspect of democratic participation. Dryzek noted that theories of democracy have taken a discursive turn in the 1990s.2 Prior to that turn, democratic theories were seen in terms of aggregation of preferences, voting or representation. Discursive democracy, on the other hand, pertains to the aspect of democracy that centers on political discussions among free and equal citizens. In other words, democratic legitimacy is sought in the ability of free, equal, and rational agents to participate in making a collective decision that will affect them; this is the democratic part.3 Moreover, the decision is made by means of arguments and deliberation “offered by and to participants who are committed to the values of rationality and impartiality.”4 This is the discursive part.

In the early decades of the twentieth century, Gabriel Tarde wrote about public opinion by defining public opinion as the result of newspaper reading and political conversation. Tarde’s model of public opinion is: Media » Conversation » Opinion » Action. In fact, Tarde noted that “conversation at all times, and the press, which at present is the principal source of conversation, are the major factors in opinion.”5 In addition to being a key factor for the formation of public opinion, conversation is also the key factor in determining passive or active audience. Tarde noted that the press does not exhibit directive influence; rather it sets the agenda for conversation. According to Tarde, conversation would never have gone above gossip without the press. And without conversation, newspapers would exercise no profound influence on the people. As a result, the audience is active. The third component of Tarde’s model is opinion. For Tarde, opinion is not individual opinion but rather social opinion, which comprises ideas about current issues expressed in public. Individual opinions are transformed into public opinion via conversation.6

Based on Tarde’s model of public opinion, Kim et al. proposed a model of deliberative democracy that comprises news media, talk, opinion, and participation.7 Among the four variables, political talk is the key. The empirical study by Kim et al. found that news media use contributed to the frequency of political conversation, which in turn had positive effect on opinion quality. Both news media use and political conversation contributed to political participation. Cho et al. proposed a model to explore the relationships between campaign advertising exposure, online news use, traditional news use, political messaging, political talk, intrapersonal reflection, political participation, and political knowledge during the 2000 and 2004 presidential election campaigns. It was found that political conversation, political messaging, and intrapersonal reflection mediated the effects of campaign advertising exposure and news use on political participation and knowledge.8


Although models of deliberative democracy were empirically tested and sustained by Kim et al., and later by Cho et al., the psychological or cognitive dimension of the model has not been explored. It is, therefore, unclear what motivates people to use the news media, to hold conversations, to form distinct opinions about issues and events, and eventually to participate in politics. In other words, the psychological and cognitive mechanisms that support the model of discursive democracy have not been examined. To phrase this problem differently, what are the psychological and cognitive catalysts of discursive democracy? In 2009, Southwell and Yzer called for a systematic examination for the social motives of political conversation, such as self-verification and personal relevance of a campaign.9 As a result, this paper intends to examine political efficacy and attention to campaign coverage as psychological and cognitive catalysts that sustain a model of discursive democracy during the 2004 U.S. presidential election.

By investigating the influence of political efficacy and campaign news attention in a discursive democracy model, it is also argued that models of discursive democracy should be contextual and domain specific. In other words, different contexts and situations generate different models of discursive democracy. Although in this paper political efficacy and campaign news attention were examined in the context of the 2004 presidential election, other catalysts, such as issue relevance, attitude importance, the number of people affected, or one’s cultural/ethnic background might be the driving forces in other contexts or situations. Distinct discursive democracy models should be developed to accommodate different circumstances under which political conversations occurred and political actions were taken. Domain specific discursive democracy models have the advantages of providing contextual factors in the framework of discursive democracy and illustrating why and how political conversations are carried out in different contexts, what subjects are at the center of the conversation, and what are the consequences of political talk.

The Discursive Turn in Democratic Theory

The terms, “deliberative democracy” and “discursive democracy,” have been used interchangeably. However, according to Dryzek, discursive democracy is actually a better term for three reasons.10 First, the term “discursive democracy” indicates that decision making is necessarily a collective social process that has to involve communication. On the other hand, deliberation can be a personal decision process upon which communication is not required. Second, deliberation connotes calm and reasoned argument, while a discursive process makes different forms of communication possible, including voices from the margins that do not follow deliberative rules. For example, online discussions do not always follow the high standards set for deliberative democracy. Online political interactions are not always so rational, tolerant, and civil. The rationalist bias of deliberative democracy actually limits its utility.11 Third, discursive democracy is preferred because discourse in the Habermasian sense means freedom in the ability to raise and challenge arguments. As a result, contestations across discourses in the public sphere become the key component of democracy. Based on the above arguments, the term “discursive democracy” will be used throughout this paper. However, the term “deliberative democracy” will be used also to be consistent with the sources cited.

To the extent that the essence of discursive democracy relies on deliberation engaged by equal and free citizens in a non-coercive fashion, discursive democracy is itself the embodiment of communicative rationality. Dryzek specified two types of rationality: instrumental rationality and communicative rationality.12Instrumental rationality is defined in terms of “the ability to devise, select, and effect good means to clarified ends.”13 It is the idea that rational choices concerning theories, beliefs, values and morals should be made through reference to a set of objective standards that are equally applicable � and accessible � to all individuals. On the other hand, communicative rationality is based on reflection, talk and communicative action. It concerns itself with culture, society, and person. It is rationality achieved in the form of uncoerced and undistorted interaction among competent individuals.14

Two types of rationality also generate two possibilities of democracy. Instrumental rationality is associated with both bureaucratic-authoritarian and liberal democratic political forms, while the communicative rationality is associated with more participatory democracy, which Barber called “strong democracy.”15 Politics in its participatory form becomes increasingly “discursive, educational, oriented to truly public interests, and needful of active citizens.”16 In other words, the political form of communicative rationality is discursive democracy, while politics in its instrumental form is “voting, strategy, private interests, bargaining, exchange, spectacle and limited involvement.”17

The Contributions of Everyday Political Conversation

Deliberative democracy as a normative theory centers on the privileges of deliberative democracy to ensure solidarity and a horizontal political will-formation, to generate legitimacy and assure some degree of practical rationality, and to value pluralism, inclusion, and equality.18 Empirical research on deliberation found that deliberation in public forums exhibited a high degree of equality. Moreover, it was political conversation in everyday life with friends, family, and others, not political sophistication, that prepared individuals to participate in deliberation.19 As citizenship becomes more personal and “everyday,”20 the role of political conversation in citizen participation has received unprecedented scholarly attention.

Using ethnographic research, Eliasoph showed that political discussion could engage citizens in politics.21 Empirical research also shows that when conflicting perspectives were presented to participants in an experiment, political conversations actually eliminated the media’s framing effect.22 In examining the effects of television and newspaper campaign coverage on voter turnout, it was found that personal conversation, rather than the media, persuaded the uninterested to vote.23 Political conversation was also found to directly contribute to participating and speaking out in a civic forum,24 and toward a better understanding about information gained from the mass media.25 Deliberative democracy was not only investigated in the context of traditional news media use, but also in the context of infotainment media. Baym argued that The Daily Show with Jon Stewart contributes to critical journalistic inquiry and the advancement of deliberative democracy.26

Studies on the polarizing effects of political conversation showed that political discussion among people who held opposing viewpoints, in some circumstances, motivated political participation,27 and that political discussion in networks of like-minded others leading to the development of extreme attitudes.28 Moreover, a study in Britain showed that talking to a supporter of a particular political party increased a person’s chances of voting for that party and shifting his or her attitudes in the direction associated with that political party.29

While most of the studies on the effects of political talk used data collected in the U.S., the effects of political conversation in democratizing societies have also been investigated. A survey conducted in Hong Kong showed that frequency of interpersonal political discussion related positively to political knowledge, and the relationship was stronger among people who experienced disagreements in discussion.30 A study in Colombia showed that political conversation oriented toward reaching understanding positively related to political involvement and participation.31

Political Efficacy and Campaign News Attention as Cognitive/Psychological Catalysts of Discursive Democracy

Political Efficacy

The functional role of political efficacy was well documented by scholars over the past few decades, and attempts were also made to indentify the contributing factors of political efficacy. In a study on the 1994 New Orleans mayoral election, it was found that political efficacy was positively associated with education and income.32 One recent study in Italy showed that personality traits, such as openness and extraversion, predicted political efficacy.33 Another psychological trait, cynicism toward the political system, was found to reduce political efficacy.34 However, research findings on the link between political efficacy and political actions are mixed. In a study on voter turnout among African-American voters, individual political efficacy did not predict turnout, but group political efficacy did.35 While jury duties were found to have no effects on one’s political efficacy, participation in a political convention contributed to one’s sense of political efficacy.36

With the mixed findings on the link between political efficacy and political involvement, some scholars urged to study political efficacy in a more contextual fashion. Ainsworth proposed that political efficacy should not be studied as an individual level attribute in isolation from macro level events. In other words, one’s sense of political efficacy, a micro level attribute, fluctuated accordingly to macro level events, such as one’s history of successes and failures with collective efforts.37

On a methodological level, studies have divided political efficacy into three categories: as an exogenous variable, as an endogenous variable, or as a mediator/moderator. Research has shown that there are clear links between political efficacy and political actions.38 Studies indicated both a single direction relationship and a reciprocal relationship. Political efficacy can function as an exogenous variable and explains campaign involvement, political participation, political outspokenness, and voting.39 Political efficacy can also function as an endogenous variable and can be explained by participation and voting for winning candidates.40 Political efficacy and political participation also reinforce each other in a reciprocal fashion. Finkel noted that campaigning and efficacy had reciprocal causal links: voters who were more efficacious were more likely to campaign, and voters who campaigned were more likely to feel efficacious as a result.41However, findings to the contrary were also reported. In a study on American national elections, neither voting nor campaign activities enhanced efficacy, while voting for winning candidates was associated with increased internal and external efficacy.42 The mediating/moderating role of political efficacy on political involvement was also investigated by scholars over the years. For example, it was found that political efficacy mediated a positive, significant effect of viewing late night TV and local TV news on civic participation among high school students.43It was also found that political efficacy moderated the effect of negative campaign advertising on children’s intention to vote.44 A study in Israel showed that the relationship between socioeconomic status and political participation could be better understood with political efficacy was a mediator.45

Researchers in political science have found that there are two dimensions of political efficacy: internal political efficacy (IPE) and external political efficacy (EPE). IPE pertains to a person’s feelings of political competence or confidences in his/her political abilities. EPE is one’s assessments of the responsiveness of the political system. IPE and EPE also play different roles in the link between political efficacy and political actions. Finkel found that IPE caused voting, but was not influenced by voting, while EPE was both a cause and effect of voting in national elections.46 IPE moderated the influence of anxiety on campaign involvement. It was found that among the highly efficacious voters, anxiety drove involvement. However, among voters with low IPE, anxiety was of little consequence.47 IPE and EPE do not influence each other; a high level of IPE does not necessarily link to a high level of EPE.48 For example, welfare participation contributed to internal efficacy but undermined external efficacy.49 In fact, it was found that a combination of high sense of IPE and low sense of EPE contributed to both unconventional, more elite-challenging political activities and conventional participation.50 Moreover, when a high sense of efficacy is coupled with a low sense of trust, there is optimum mobilization.51

Campaign News Attention

Attention to campaign news coverage, in addition to political efficacy, was also examined during the 2004 presidential election. Researchers have shown that attention, but not exposure, to television news accounted for political learning.52 It was further noted that exposure and attention are separate dimensions.53 Audience self-reports of attention reflected differences in the processing of media messages.54 Attention requires cognitive or psychological involvement, which was found to be a key factor in the political decision making process, while exposure does not.55 Scholars argued that attention is a better indicator of learning and recall of media messages.56 Beaudoin showed that news attention was positively linked to the development of international knowledge.57 News attention, rather than exposure, was the major influence in public perception of nations covered in the news.58 More specifically, campaign news attention contributed to clearer campaign issue opinions and better knowledge of candidate issue positions.59 Attention to television news coverage of the 2000 election campaign had a statistically significant impact on issue learning.60 As a result, news attention is a more robust predictor of issue voting. In a longitude study of various election settings from 1984 to 1992, measures of attention to television news and televised advertisements each significantly accounted for the variance in issue knowledge.61 Campaign news attention also has attitudinal effect on the electorate. It was found that television campaign news attention negatively affected trust in government among less educated but not among the more educated, while the effects of newspaper campaign news attention on trust in government were non-significant across educational level.62

The factor that differentiates attention from exposure is that attention is the prerequisite of learning from the news, while exposure is not. The news media is one of the primary sources where citizens get their political information. As a result, researchers have examined how people learn from the media. For example, researchers have investigated how news audiences differ in the amount of information they learn from various news outlets.

Studies on news learning have focused on learning from the print media and television among either the younger population or the adults. For the younger generation, a study in Taiwan found that television news made a significant contribution to adolescents’ knowledge about the Gulf wars.63 A British study on 10- and 11-year-olds found that children remembered news better from television news that contained semantically redundant pictures than from print or audio versions, regardless of their reading proficiency.64 In the U.S., in an ethnographic study on high school students, researchers found that TV talk and reality shows could be important sources of political conversation and civic engagement.65 For the adult population, viewers of late night comedy programs paid more attention to the campaign in network TV and cable TV than nonviewers.66 Media coverage on political campaign practices, such as televised political ads, candidate appearances, and campaign contacts, promoted learning about politics. It was also found that people who watched more local television news had less political knowledge.67

More recently, scholars started to examine how citizens acquire political information from the Internet. Internet news attention was found to have the most positive effect on international knowledge, in comparison with newspapers, network TV, and cable TV.68 In a digital environment, strategic framing, rather than value framing, was found conducive to learning, especially among people who read political blogs.69 In an experimental study, it was found that online news that contained in-text hyperlinks discouraged learning of the facts that made up news stories, but the same online news structure contributed to users’ more densely interconnected knowledge structures for public affairs topics.70


Given that the potential of deliberative democracy is more complex than most researchers have acknowledged,71 the model proposed in this study will show that attempts to capture the essence of discursive democracy should pay attention to the contextual dimensions of various political circumstances. In particular, this study examines the catalytic role of political efficacy and campaign news attention in a model of discursive democracy comprises political discussion, opinionation, and political participation (see Figure 1). The preceding arguments and literature suggest the following hypotheses. Kim et al.’s model for deliberative democracy was modified to account for the aims of this study.72

Figure 1. Political Efficacy and Campaign News Attention as Catalysts of Discursive Democracy during an Election

First, Dahlgren argued that the examination of deliberative democracy should include civic cultural factors so as to provide deliberative democracy an anchoring point where citizens’ lived experiences, personal resources, and subjective dispositions can be taken into consideration. Second, political efficacy was found to contribute to political participation and mass media use.74 Third, research has shown that IPE is generally more stable than EPE, and is less sensitive to the fluctuation of macro level events.75 It is thus hypothesized that IPE will be a stronger catalyst than EPE in the model of discursive democracy.

H1: Internal political efficacy, instead of external political efficacy, predicts each of the four components of the model � attention to campaign news, political conversation, opinionation and participation.

Past research on the relationship between media exposure and political conversation found that hard news media use strongly related to political talk.76Specifically, political conversation functioned as a moderator between Internet campaign exposure and political efficacy, political knowledge, and campaign participation.77 Kim et al. found that both general media use and issue-specific media use related to political conversation, with issue-specific media use having a stronger effect. Kim et al. also found that both issue specific news media use and issue specific political conversation contributed to willingness to argue.78Research to date, however, has not shown the effect of attention to campaign coverage on political conversation. Given the effects of campaign news attention on issue knowledge and issue opinions,79 it is thus hypothesized that attention to campaign news coverage will contribute to political conversation and opinionation.

H2: Attention to campaign news predicts political conversation.

H3: Both political conversation and attention to campaign news predict opinionation.

Research show that media’s influence on people’s information levels and political participation is mostly a function of people talking to each other first and learning more and becoming more likely to participate in politics.80 The same can be said about general social interaction. McClurg found that the effect of social interaction on political participation was contingent on the amount of political discussion that occurred in social networks.81 Moreover, McClurg found that political discussion had a substantively strong influence on the participatory behavior of lower status individuals.82 Political conversation also functioned as a mediator between hard news media use and political participation. In other words, the impact of hard news media use on political participation was more pronounced among those who talked about politics more frequently.83 People who talked about politics more often also displayed a higher level of political knowledge and political participation.84 It is thus hypothesized that political participation is predicted by attention campaign news coverage, political conversation, and opinionation.

H4: Political participation is predicted by attention to campaign news, political conversation, and opinionation, in addition to being predicted by internal political efficacy.

Although attention to campaign news is hypothesized to positively relate to political conversation, opinionation, and political participation, different modes of campaign news attention can have different effects. Research on news attention show that attention to newspaper campaign news contributed to voting, and that attention to newspaper local news also promoted political participation.84 Simply having access to the Internet and to online election news predicted voting in the 1996 and 2000 election.85 Previous studies also show that newspaper and the Internet make more significant contribution to people’s political knowledge than television.86 There is no systematic study yet on the effects of different types of campaign attention on political behavior. It is thus further hypothesized that the reading of newspaper about the campaign and getting news online about the campaign will be stronger predictors of conversation, opinionation and participation than other types of campaign attention.

H5: Attention to newspaper and online campaign news will be stronger predictors than attention to television or radio campaign news to account for political conversation, opinionation, and participation.



A survey was conducted in Monroe County, New York, from Oct. 7 to Oct. 21, 2004. The sample consists of both a random telephone sample and a convenient sample from local shopping malls. A total of 187 residents was interviewed. Students registered in an undergraduate communication course conducted the interviews. The sample contains 40% females, 39% males, and 22% were missing. 40% of the interviewees were between 18 to 25 years old, 17% were between 26 to 35 years old, 15% were between 36 to 45 years old, 16% were between 46 to 60 years old, 11% were over 60 years old, and 2% were missing. 39% of the respondents had a 12th grade education, 32% finished 2-year college, 21% finished 4-year college, 8% had master’s degree, .5% had degree higher than a master’s, and .5% was missing. In terms of party identification, 27% of the respondents were Republican, 31% were Democrat, 22% were Independent, and 20% were missing.


There are two limitations in this study. First, the sample is small. Students who conducted the interviews stated that people were reluctant to answer questions on the phone. A lot of them simply hung up. As a result, the alternative to conduct interviews at local malls became an option for students to complete the project. The number of interviews each student needed to do was also reduced. Second, the sample does not represent the population across all demographic variables. According to data from the U.S. Census Bureau, in 2007, 63.6% of the Monroe County residents were between the ages of 18 and 65. Data collected for this study has 88% of the respondents between 18 and 65 in 2006. Consequently, sample in this study contains more respondents between 18 and 65 than the general population in Monroe County. In terms of education attainment, sample data represent the population. The U.S. Census Bureau data show that in 2000, 53.7% of the Monroe County residents 25-year-old or older finished high school or a 2-year college, and 31.2% have a bachelor’s degree. Sample data show that 29.5% of the respondents finished 4-year college.


The survey asked questions based on a revised version of the model proposed by Kim et al. and explored the following variables: internal political efficacy, external political efficacy, attention to news about the campaign, political conversation, opinionated, and political participation.87 All of the questions were answered on 5-point Likert scales. A list of questions pertaining to each variable examined in this study is in Table 1. Mean scores and standard deviations are included.

Table 1: Measures
Internal Political Efficacy
UNDRSTND:“I feel that I have a pretty good understanding of the important political issues facing our country.” (M = 3.76, SD = 1.08, N = 184)
SELFQUAL:“I consider myself well qualified to participate in politics.” (M = 3.26, SD = 1.33, N = 182)
PUBOFF:“I feel I could do as good a job in public office as most other people.” (M = 2.71, SD = 1.30, N = 180)
INFORMED:“I think that I am better informed about politics and government than most people.” (M = 2.90, SD = 1.26, N = 184)
External Political Efficacy
NOSAY:“People like me don’t have any say about what the government does.” (M = 3.11, SD = 1.33, N = 180)
NOCARE:“I don’t think public officials care much about what people like me think.” (M = 2.76, SD = 1.26, N = 180)
COMPLEX:“Sometimes politics and government seem so complicated that a person like me can’t really understand what’s going on.” (M = 2.99, SD = 1.30, N = 184)
Attention to Campaign Coverage
LOCALTV:Attention to campaign coverage in local television newscast. (M = 3.30, SD = 1.17, N = 185)
PAPER:Attention to campaign coverage in newspapers. (M = 3.06, SD = 1.25, N = 186)
NATLTV:Attention to campaign coverage in national television newscast. (M = 3.24, SD = 1.19, N = 187)
RADIO:Attention to campaign coverage in radio newscast. (M = 2.59, SD = 1.24, N = 186)
ONLINE:Attention to campaign coverage in news online. (M = 2.66, SD = 1.39, N = 180)
Campaign Talk
T/GOV:Talk about what the city, state or national government is doing. (M = 2.77, SD = .94, N = 186)
T/ECONMY:Talk about how the economy is doing. (M = 3.06, SD = .84, N = 187)
T/FOREIGN:Talk about what is happening in foreign countries. (M = 2.91, SD = .92, N = 185)
T/RELIGION:Talk about religion or religious beliefs. (M = 2.50, SD = 1, N = 185)
Noncampaign Talk
T/CRIME:Talk about crime and violence in society. (M = 2.99. SD = .84, N = 187)
T/EDU:Talk about what is going on in schools or education in general. (M = 2.97, SD = .88, N = 185)
PRESIDNT:How George W. Bush is handling his job as president. (M = 2.86, SD = 1.43, N = 182)
ECONOMY:How George W. Bush is handling the economy. (M = 2.76, SD = 1.34, N = 182)
FOREIGN:How George W. Bush is handling our foreign relations. (M = 2.76, SD = 1.41, N = 181)
TERROR:How George W. Bush is handling the war on terrorism. (M = 2.86, SD = 1.56, N = 180)
IRAQ:How George W. Bush is handling the war in Iraq. (M = 2.68, SD = 1.53, N = 181)
Political participation
LETTER:Have written letters to the media. (M = 1.28, SD = .45, N = 183)
OFFICIAL:Have contacted a candidate for public office or an elected official. (M = 1.26, SD = .44, N = 182)
ATTEND:Have attended a meeting regarding public affairs. (M = 1.43, SD = .50, N = 181)
SPOKEN:Have spoken out at a meeting regarding public affairs. (M = 1.15, SD = .36, N = 182)
MARCH:Have taken part in a public demonstration or march. (M = 1.21, SD = .41, N = 184)
CAMPN:Have worked in a political campaign. (M = 1.16, SD = .37, N = 183)
MONEY:Have contributed money to a political campaign. (M = 1.22, SD = .42, N = 184)
Note: All questions were coded on 5-point Likert scales with “Agree strongly” coded as 5 and “Disagree strongly” coded as 1.

Internal Political Efficacy

Using the 1988 National Election Study (NES) data, Niemi, Craig, and Mattei found that IPE and EPE are two different constructs and standardized the measurement of IPE and EPE.88 Morrell used the 1992 and 2000 NES data and found Niemi et al.’s measurement of IPE to be robust.89 Morrell further conducted experiments to test the measurement for IPE, and again, confirmed Niemi et al.’s measures.90 As a result, IPE and EPE were measured using the scales proposed by Niemi et al.91 Respondents were asked whether they “agree strongly,” “agree somewhat,” “neither agree nor disagree,” “disagree somewhat” or “disagree strongly” with the following statements: “I feel that I have a pretty good understanding of the important political issues facing our country,” “I consider myself well qualified to participate in politics,” “I feel I could do as good a job in public office as most other people” and “I think that I am better informed about politics and government than most people.”

External Political Efficacy

Questions to assess external political efficacy were also drawn from Niemi et al.92External political efficacy measures how well the respondents thought the government responded to their concerns. Respondents were asked whether they have any say regarding what the government does, whether they think public officials care about what they think, and whether politics is too complicated for people like them to understand.

Attention to Campaign Coverage

Campaign news attention was assessed using 2000 NES survey questions. Respondents were asked how much attention they paid to the news about the election campaign via the following media channels: local television, national television, newspaper, radio, and the Internet.

Political Conversation

Political conversation was examined in two dimensions: campaign talk and non-campaign talk. Given that the purpose of the study is to formulate a situational model of discursive democracy, the measurement of political conversation will have to be sensitive to the content of respondents’ political conversation during the 2004 campaign season when certain topics caught people’s attention more than others. Adapting a scale developed by Wyatt et al.,93 which contains nine content specific questions, items used to assess campaign talk included “what the city, state or national government is doing,” “how the economy is doing” and “what is happening in foreign countries” because government, the economy and foreign relations were centers of people and the media’s attention during the 2004 campaign season. Items used to assess non-campaign talk included “crime and violence in society,” “what’s going on in schools and education in general,” and “your religion and religious beliefs.” This categorization is sensitive to the focus of each campaign. A different election will generate different content items for campaign talk and for noncampaign talk.


Opinionation was also investigated in two dimensions: hard opinionation and soft opinionation. Hard opinionation pertains to answers such as “agree strongly” or “disagree strongly” to a 5-item measure created to assess the extent to which respondents having opinions about the way George W. Bush was handling the following: his job as president, the economy, foreign relations, the war on terrorism, and the war in Iraq. Soft opinionation pertains to answers like “agree somewhat” or “disagree somewhat.” Each respondent will have a score for hard opinionation and a score for soft opinionation ranging from 5 to 0. “Don’t know” and “neither agree nor disagree” were coded as 0. For example, a respondent with 3 answers of “disagree somewhat” and 2 answers of “disagree strongly” will have a score of 3 for soft opinionation and a score of 2 for hard opinionation.

Political Participation

Adapted from Kim et al. and the 2000 NES survey, a set of 7 “yes-no” questions were administered.94 Respondents’ levels of political participation were assessed based on whether they have written letters to the media, contacted a candidate or a public official, attended a public affairs meeting, spoken out at a public affairs meeting, taken part in a demonstration or march, worked in a political campaign or contributed money to a political campaign.

Explanatory Factor Analysis

One explanatory factor analysis was conducted to assess the reliability and validity of this investigation. All variables specified in Table 1 were entered with Direct Oblimin rotation, and results are presented in Table 2. The only discrepancy between Table 1 and Table 2 is the “talk about religion” variable, which was clustered with the campaign talk variables and is thus treated as one of the campaign talk variables.

Table 2: Factor Analysis of Predictors
VariableFactor Loading
Attention to Campaign Coverage
External Efficacy
Political Participation
Noncampaign Talk
Internal Efficacy
Campaign Talk
Total Variance Explained: 68.2%


H1: Internal political efficacy is a stronger catalyst than external political efficacy. To examine whether internal political efficacy is a stronger catalyst than external political efficacy in the model of discursive democracy, a correlation analysis was performed. Results are presented in Table 3, which shows that IPE has significant, positive associations with all variables in the model, except soft opinionation. EPE predicts attention to campaign coverage in the media and political participation, but none of the other variables. Hypothesis 1 is thus sustained.

Table 3: Linear multivariate regression models for attention to campaign coverage
 Equation 1Equation 2
Party ID.096.103
Internal efficacy.531***
Model fit
Note: Cell entries are standardized coefficients (ß). *=.05, **=.01, ***=.001

H2: Attention to campaign news predicts political conversation. As previously mentioned, political conversation was examined at two dimensions: campaign talk and noncampaign talk. Table 3 shows that attention to campaign news in the media predicts both campaign talk and noncampaign talk, with campaign talk exhibiting a stronger relationship. To further assess whether political conversation can be explained by campaign news attention, three separate multiple regression analyses were performed for each dimension of political conversation. With each dimension of political conversation as the dependent variable, four demographic variables were entered as the independent variables in the first equation. For the second equation, IPE was entered as the fifth independent variable. Attention to campaign news was entered as the sixth independent variable in the third equation. Results are in Table 4. For campaign talk, three demographic variables and party identification were entered in the first equation, and no statistical significant relationships were generated. IPE was entered in the second equation and is significant in predicting campaign talk. The regression model explained 23% of the variance in campaign talk. Campaign news attention was added in the third equation, which explains a very significant degree of variance in campaign talk, R2= 44, p < .001. However, the effect of IPE in the third equation is insignificant. This is primarily because IPE and attention to campaign news are highly correlated, r = .645, p < .001, and only one of the two will contribute significantly to the regression when they are used in combination. The same procedure was performed for non-campaign talk, and only the model that contains attention to campaign news is significant, R2 = 18, p < .01. For both dimensions of political conversation, campaign news attention appears to be a robust indicator. Hypothesis 2 is sustained. The difference in variance explained in the third equation of both dimensions shows that campaign news attention has a much stronger association with campaign talk than with non-campaign talk.

Table 4: Linear multivariate regression models for political conversation
 Campaign TalkNoncampaign Talk2
 Equation 1Equation 2Equation 3Equation 1Equation 2Equation 3
Party ID.009-.024-.078-.138-.173-.197*
Internal efficacy.432***.126.176-.035
Attention to campaign news.557***.369**
Model fit
Note: Cell entries are standardized coefficients (ß). *=.05, **=.01, ***=.001

H3: Both political conversation and attention to campaign news predict opinionation. Very different results were generated for the two dimension of opinionation. Correlation analysis results in Table 3 show that hard opinionation has statistically significant associations with IPE, campaign attention, and campaign talk. Soft opinionation, however, fails to show significant associations with any variables investigated in this study. As a result, only hard opinionation was further assessed. 70% of the respondents in this study gave at least one hard opinionation answer. Following the same procedure previously used, four equations were generated. The results are in Table 5. Campaign talk has a positive, significant association with hard opinionation, ß = .308, p < .05. IPE has a consistent, positive and strong effect on hard opinionation across various equations. This further shows that people who demonstrated hard opinionation had a higher level of self-confidence in their ability to function well in politics.

Attention to campaign news and noncampaign talk, however, were negatively and insignificantly associated with hard opinionation. The negative signs indicate that people who demonstrated hard opinionation were less likely to turn to the media for information and were less likely to engage in noncampaign talk during the election season. Hypothesis 3 is partially supported.

Table 5: Linear multivariate regression models for hard opinionation
 Equation 1Equation 2Equation 3Equation 4
Party ID .075 .076 .078 .081
Internal efficacy .351*** .404*** .372**
Attention to campaign-.062-.224
Campaign talk .308*
Noncampaign talk-.060
Model fit
Note: Cell entries are standardized coefficients (ß). *=.05, **=.01, ***=.001

H4: Political participation is predicted by attention to campaign news, political conversation, and opinionation, in addition to being affected by internal political efficacy. Results for political participation in Table 6 indicate that IPE, attention to campaign news, and campaign talk have positive associations with political participation. However, consistent with Kim et al.’s results, opinionation has no significant effect on political participation.95 Hypothesis 4 is partially supported. The effect of IPE was reduced from the second equation to the third equation when attention to campaign news was added. As previously mentioned, this is because campaign news attention and IPE are highly correlated with each other. The same phenomenon occurred again when campaign talk was added to the fourth equation, rIPE.campaign talk = .557, p < .001, rattention.campaign talk = .607, p < .001. However, the increases in R2 across the five equations show that the model for discursive democracy is robust. The final equation explains 41% of the variance of political participation.

Table 6: Linear multivariate regression models for political participation
 Equation 1Equation 2Equation 3Equation 4Equation 5
     Age .214* .187* .167 .113 .125
     Gender .102 .096 .119 .113 .121
     Education .394*** .311** .299** .284** .298**
Party ID .025 .043 .022 .044 .029
Internal efficacy .261** .049 .011
Attention to campaign .316** .134 .151
Campaign talk .364** .336**
Noncampaign talk-.099-.099
Model fit
Hard opinionation .135
Soft opinionation .053
Model fit
Note: Cell entries are standardized coefficients (ß). *=.05, **=.01, ***=.001

H5: The effects of newspaper and online campaign news attention

Four multiple regressions were carried out to assess how different types of campaign news attention affect people’s political behavior. Results in Table 7 indicate that both local and national television news on the election got people to talk about the government, the economy, the U.S. foreign relations and their religious beliefs, while other media channels failed to contribute to campaign talk. For noncampaign talk, radio is the only significant predictor. People of hard opinionation tended to use the Internet for news about the election, ß = .261, p < .01, and much less likely to use the newspaper, ß = -.287, p < .01. As for the effect of various types of campaign news attention on political participation, congruent with previous literature, people who turned to newspaper and the Internet for campaign information were more likely to participate in politics. Hypothesis 5 is partially sustained.

Table 7: Effects of types of campaign news attention on political conversation, opinionation, and political participation
Local TV.310***.163.130.014
National TV.209*.188.183.161
Model fit
Note: Cell entries are standardized coefficients (ß). *=.05, **=.01, ***=.001

A Model of Discursive Democracy during an Election

Based on both theory and the above proposed hypotheses, a structural equation model (SEM) was constructed to assess the direct and indirect effects among variables. IPE was entered as the exogenous variable. Attention to campaign news, campaign talk, and hard opinionation served first as endogenous and later as exogenous variables in the model. Political participation is the final, endogenous variable that the model is intended to explain. The model is constructed using LISREL 8.7 and is depicted in Figure 2.

Figure 2: A Model of Discursive Democracy with Estimated Coefficients

The structural equation model of discursive democracy for the 2004 election presented here fits the data well, with IFI = .99, CFI = .99, and RMSEA = .024. The insignificant χ2 demonstrates that the observed and implied matrices are similar, indicating that the implied theoretical model significantly reproduces the observed relationships in the model.96 In addition to model fit, it is also important to examine the statistical significance of individual estimates for the paths in the model. According to Figure 2, IPE significantly predicts campaign news attention, which in turn predicts campaign talk, which further predicts political participation. IPE, however, fails to predict campaign talk and political participation in this model. Moreover, IPE does significantly predict hard opinionation, but hard opinionation does not predict participation. In fact, quite different patterns emerged from the model concerning hard opinionation: congruent with the results of regression analysis in Table 5, IPE successfully predicts hard opinionation in the model, and in contrast with the results in the same table, campaign talk does not predict hard opinionation. Because of the insignificant role hard opinionation plays in the model, and because previous research also failed to show a significant relationship between opinionation and participation,97 hard opinionation was removed from the model to construct a reduced model to better account for the discursive aspect of democracy during an election. The path between campaign news attention and participation was also removed. The reduced SEM is presented in Figure 3.

Figure 3: First Reduced Model of Discursive Democracy: Figure 2 Deleting Hard Opinionation and Path from Campaign Attention to Participation

The reduced model fits the data well, with IFI = .99, CFI = .99, and RMSEA = .026. The insignificant χ2 indicates congruence between the theoretical model and the observed relationships. All paths in the model show at least marginal statistical significance, with strong coefficients from IPE to campaign news attention, from campaign news attention to campaign talk, and from campaign talk to political participation. The path between IPE and participation is also significant, and the path from IPE to campaign talk is marginally significant. There are high r-squares for campaign news attention, campaign talk, and political participation in all models.


This study sought to demonstrate that models of discursive democracy should be domain specific. Different models are necessary to account for various forms of political conversation that took place in different contexts. This study showed that internal political efficacy and attention to campaign news were two catalysts that stimulated political conversation and sustained discursive democracy during the 2004 presidential election.

One of the advantages of a domain specific discursive democracy model is that specific variables can be added to the model to better account for the context of people’s political conversation. Because of the nature of a presidential election, especially the first one that took place after the September 11th terrorist attacks and one when the nation was at war, the investigation and inclusion of motivational variables such as internal political efficacy and campaign news attention especially make sense.

Another advantage of domain specific discursive democracy models is that the model will reflect predominate topics of people’s political conversation. Contextual models of discursive democracy will also show that the content and amount of people’s political conversation fluctuate as a result of macro level events. This study demonstrated that people were talking about politics during the election season, and precisely talking about the government, the economy, the U.S. foreign relations and their religious beliefs. This shows that the media and politicians do set the agenda regarding what people will talk about and be concerned about. More studies are needed to examine the role of political conversation in media’s agenda setting effect.

The strongest predictor of campaign talk is attention to campaign news in the media. This relationship shows the important role the media play in discursive democracy, and also presents some challenges to democracy overall. The literature criticizing the concentration of media ownership and documenting the effect of conglomerization on media content is abundant.98 Moreover, Dahlgren argued that current research in deliberative democracy downplayed relations of power that are built into the communication systems.99 It is thus imperative to take a critical stand in the study of discursive democracy. In other words, do we cheer simply for a significant relationship between media attention and political conversation? Or do we pause and ask ourselves what does this relationship mean for democracy?

Opinionation was investigated in two dimensions: hard opinionation and soft opinionation. It is quite interesting to find that soft opinionation was not associated with any of the variables examined in this study. This result shows that opinion crystallization is consequential to discursive democracy, and that respondents who hold neutral, or close to neutral, positions on issues might do so because they did not have the adequate amount of information or knowledge for a stronger position on issues. The portrayal of hard opinionators comprises a higher level of internal efficacy, the preference over online campaign news, and campaign talk. However, hard opinionation did not an impact on political participation. The missing link between opinionation and participation shows that there is room for improvement of models of discursive democracy.

One particularly interesting finding of this study is that television news contributed to campaign talk over other types of medium. Given that most Americans get their news from television, and that most people watch television at home, this finding shows that home is an important locale for everyday, discursive political conversation. Another interesting finding is that people of hard opinionation went to the Internet for news instead of any other medium. This finding shows that hard opinionators were more inclined to distrust the mainstream media and go online for alternative views.

As political scientists have found decades ago, education was a strong predictor for political participation, which was also predicted by internal efficacy, campaign news attention, and campaign talk, according to the regression analysis results.

Two structural equation models were constructed to find the model that can best describe the catalytic role of IPE and campaign news attention in discursive democracy. The reduced model emerged as providing a better account of how political conversation functions during an election campaign. The deletion of hard opinionation in the simplified model does not mean that opinionation is insignificant in discursive democracy. It actually means that opinionation is at least as important as political participation and requires its own investigation and models, instead of being included in a model aiming to explain political participation.

Hsiang-Ann Liao is an assistant professor of communication at Buffalo State College. This article is based on the 2007 Faculty Top Paper, Civic and Citizen Journalism Division, AEJMC.

Leave a comment

Your email address will not be published. Required fields are marked *