Copyright

by

Sebastian Valenzuela

2011
 The Dissertation Committee for Sebastian Valenzuela certifies that this is the
approved version of the following dissertation:

The Affective Citizen Communication Model:

How Emotions Engage Citizens with Politics through Media and Discussion

Committee:

Maxwell McCombs, Co-Supervisor

Homero Gil de Zuniga, Co-Supervisor

Natalie J. Stroud

Sharon Jarvis Hardesty

Renita Coleman
 The Affective Citizen Communication Model:

How Emotions Engage Citizens with Politics through Media and Discussion

BY

Sebastian Valenzuela, B.A.; M.A.

Dissertation

Presented to the Faculty of the Graduate School of

The University of Texas at Austin

in Partial Fulfillment

of the Requirements

for the Degree of

Doctor of Philosophy

The University of Texas at Austin

May 2011
 UMI Number: 3484396

All rights reserved

INFORMATION TO ALL USERS

The quality of this reproduction is dependent on the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.

UMI
Dissertation Publishing

UMI 3484396
Copyright 2011 by ProQuest LLC.
All rights reserved. This edition of the work is protected against
unauthorized copying under Title 17, United States Code.

uest
ProQuest LLC.
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 48106 - 1346
 Dedication

To Tere and Simon, who have motivated me to be a most affective—if not effective!—

husband, father, and friend.

 Acknowledgements

It is said that the most pleasant section to write on a dissertation is the

acknowledgements section. I agree. I wouldn't be writing these lines were it not for the

tremendous assistance provided to me by many faculty, friends and family members.

Here at last I have the formal opportunity of saying "Thank you, I owe you big time!"

Max McCombs has been a supportive advisor to me throughout my graduate

education. Since the day we first met, back in Santiago when I was still a working

journalist for El Mercurio, he has given me a push—both intellectually and financially—

to move forward in my academic career. My first job as a graduate research assistant,

my first book chapter, my first experimental design, my first trip as a guest lecturer, my

first job as conference organizer, and my first reader of the master's thesis—I owe these

formative experiences to Max. Most importantly, he has taught me how to excel in

academia without becoming an arrogant scholar in the process. He has always

respected my work outside agenda setting—including this dissertation—and offered

detailed comments of my writings whenever I asked him. More often than not, I have

found his comments a cause of gratitude (for their detailed consideration of my ideas),

envy ("Why didn't I think of that before?") and amusement (for his ability to weave

humor into the most arcane theoretical discussions). I am honored by his dual role as a

mentor and a friend. I will never forget our joint "content analysis" sessions, especially

when he provided the "raw data" (I'm still savoring that Silver Oak!) and our wives

helped us estimate "inter-coder reliability." Thank you, Max, for helping me these past

six years. I will be forever indebted to you.

I didn't meet Homero Gil de Zuniga until my third year of graduate school, but I

soon realized that we had many things in common. We were both international

scholars, interested in political communication and digital media, whose first language
v
was Spanish, and with a knack for listening to classic rock non-stop, AC/DC included. I

admire his tenacity and passion for work, even when excitement translated into heated

discussions. No one else but Homero could have created and directed the CJCR, a

formidable collective of researchers, and I will be always grateful for his generous

invitation to become part of this "monster"—his words—and operate as student leader.

His generosity spills to my resume in many other ways too. First and foremost, he is co­

author of many of my publications, which speaks in a loud voice about his dedication to

the graduate program and its students. Homero has also helped me immeasurably in my

professional socialization. He has always taken time to introduce me to people within

the discipline, especially during the AEJMC socials, and has been a strong, persistent

advocate for me—the dozen or so letters of recommendation he has written for me are

testament to that. I'm very fortunate of having him not only as co-chair of this

dissertation, but as an enduring friend and colleague that has always gone beyond the

call of duty. jGracias, piramide!

Natalie (Talia) J. Stroud, Sharon Jarvis Hardesty and Renita Coleman graciously

agreed to participate as committee members of this dissertation, and I'm immensely

fortunate for that. Each of them are shining communication scholars, and through their

thoughtful suggestions, sharp critiques and creative connections with existing bodies of

work, they have perfected my research. Talia spent numerous office hours enlightening

me about the benefits of studying emotions in political communication, describing the

advantages and disadvantages of panel data, and helping me sort out obscure

methodological quandaries such as: Should I use weighted or unweighted estimates?

Her first-hand knowledge of the Annenberg surveys was particularly beneficial for this

project. Sharon agreed to be a part of my committee despite the fact that we had never

met before. Nevertheless, she always showed enthusiasm for this project. Her thought-
vi
provoking questions during my comprehensive exams and oral defense of the

dissertation forced me to think about the larger implications of my work and to see

myself as an independent scholar. Renita has influenced me on a number of levels. Her

proseminar class was pivotal in my approach to research papers; as much as I could, I

have tried to follow her mantra of "publish it, publish it, publish it." As I'm about to start

my own path as an assistant professor, I certainly would like to borrow her no-frills

approach to academic life.

Other Journalism faculty also deserve formal acknowledgement for contributing

to my work. Although I took only one of his seminars, Gene Burd has taught me a lot

about positive reinforcement and stimulating the work of graduate students. I thank

him for his good humor, warmth and understanding. Particularly, I appreciate helping

me connect my father's passion for urbanism, city planning and Jane Jacobs with my

own scholarly interests on communication and democratic citizenship. Paula Poindexter

has supported my work since I started my graduate education, and her passion and

commitment towards the graduate program and the School of Journalism are an

example that I would like to follow. I learned a great deal from Maggie Rivas-Rodriguez.

For three years I was her teaching assistant, and I will always take with me her wisdom

about the journalism profession and her courage for defending her ideals. In the short

time I have met Tom Johnson, he has impressed me by being able to combine scholarly

productivity with a laid back attitude that shines best when at the Hole in the Wall.

Thank you, Tom, for your support during the job hunt.

I am extremely grateful for the friends I made in Austin, many of whom I

consider family. Viviana Salinas was our first guide in all things Austin and the University

of Texas. Over time, she made herself indispensable: she was there when our son Simon

was born, she was there to babysit him whenever we asked her to do it, she was there
vii
when we needed advice about the Chilean academia. She is a terrific sociologist, a good

listener, and a generous soul. Viviana also introduced us to Laura Spagnolo and Greg

Landreth, our companions of countless asados, happy hours and trips around Texas.

They are a smart, affective, loyal and fun couple to be around with. I can say with

confidence that the best conversations about world affairs, news and politics have been

with Laura and Greg, especially when in front of a bottle of Malbec and Laura's delicate

Middle Eastern recipes. I am sure that, wherever you guys decide to raise Noah, we will

remain in close contact. I also want to thank our good friend Alvaro Quezada, one of the

founding members of Austin's Chilean mafia. We have been blessed with his loyalty and

generosity; always willing to give us a hand when we needed one. He is an intelligent,

trustworthy and happy person, and we will be forever grateful for the many evenings he

has spent playing with Simon. Thank you, also, for inviting me to play with La Roja; we

may have not won a tournament, but we sure had a great time playing together in

Intramural Fields!

My dear friends Dean Graber, Lou Rutigliano, Paul Alonso, Alejandra Ramirez and

Vijay Parthasarathy also deserve special mention, particularly because they have given

me the opportunity to let off steam about the graduate program when I needed it. It has

been great to share my graduate education with Ingrid Bachmann, whom I know since

my undergrad years in Chile; her "concise" literature review on emotions helped me

clarify my ideas in the early stages of this dissertation. My former and current

colleagues at the CJCR—Amber Willard Hinsley, Seth Lewis, Yonghwan Kim, Kelly

Kaufhold, Jae Kook Lee, Sun Ho Jeong, Alex Avila, Brian Baresch, Monica Chadha, Sandra

Hsu, Sung Woo Yoo, Ting Chen, Ingrid Bachmann and, of course, Teresa Correa—made

staying on campus Friday afternoons a much more pleasant experience than I could ever

viii
imagine. Thank you for your willingness to listen to my ideas and helping me put them

into paper.

I am extremely grateful for the support that my family in Chile has given me for

so long. I appreciate the patience and unrelenting support over the years of my parents,

Jaime Valenzuela and Maria Loreto Leighton; my siblings, Jaime, Maria Loreto and Juan

Cristobal; and my in-laws, Jorge Correa and Tiki Reymond. I'm also fortunate of having

the best cunados/as one could wish for: Rossana Faieta, Cuca Mena, Mana Correa and

her husband Cristobal "Tatol" Avendano, Jorge Correa, Elvira Correa, Maida Correa and

Ale Correa. I can't wait to resume our regular sobremesa de domingo and catch up with

you all.

Last, but without a doubt not least, I most want to thank my wife, best friend

and talented colleague Teresa Correa. She has been a constant source of love,

motivation and dedication ever since we decided to build a life together seven years

ago. I am lucky to be even more indebted to Tere because of her knowledge of

communication and media research. Her name should be written all over this

dissertation and, to be honest, in all my other papers. Tere has spent hours upon hours

discussing with me ways to move forward this project and her insight was critical to this

dissertation. She has the ability to throw out brilliant ideas with disturbing frequency; I

admire her for that and for much, much more. You are fun, you are smart, you are

generous, you are loyal, you are the most amazing woman I ever met, you rock. So Tere,

thanks for always being there. I really mean it. Thanks a million. Te amo.

And Simon, our beautiful son, when you finally grow up and get to read this, I

want you to know: you have been inspiring and amazing me every day since you were

born. You truly make me want to be a better person. Eres lo mejor.

ix
 The Affective Citizen Communication Model:

How Emotions Engage Citizens with Politics through Media and Discussion

Publication No.

Sebastian Valenzuela, Ph.D.

The University of Texas at Austin, 2011

Co-Supervisors: Maxwell McCombs and Homero Gil de Zuniga

This dissertation seeks to improve our understanding of the process by which

emotions enable citizens to learn about public affairs and engage in political activities

during electoral campaigns. It advances a theoretical model that incorporates the

dynamics of emotions, various forms of media use, interpersonal communication and

political involvement. This affective citizen communication model integrates into a

single framework the insights of affective intelligence theory (Marcus, Neuman, &

MacKuen, 2000) and the work on communication mediation (McLeod et al., 1999, 2001)

and its two iterations, cognitive mediation (Eveland, 2001) and citizen communication

mediation (Cho et al., 2009; Shah et al., 2005, 2007). More specifically, it suggests that

the effects of emotions triggered by political candidates (e.g., enthusiasm, anxiety,

anger) on knowledge of the candidates' stands on issues and on political participation

are largely mediated by communication variables, including news media use, political

discussion and debate viewing. By positing emotions as an antecedent of both mediated

x
and interpersonal communication, the study extends current research based on

affective intelligence theory. At the same time, the study adds emotions to

communication mediation processes, which to date have been studied from a mostly

cognitive perspective.

To test the relationships between the variables identified in the affective citizen

communication model, I rely on panel survey data collected for the 2008 and 2004 U.S.

presidential elections by the American National Election Studies (ANES) and the National

Annenberg Election Surveys (NAES), respectively. Two types of structural equation

models are tested, cross sectional (to relate individual differences) and auto-regressive

(to relate aggregate change across waves). Results suggest that positive emotions spark

media use, whereas negative emotions spark political discussions, and both types of

communication behavior influence issue knowledge and participation in campaign

activities. Furthermore, the theorized structure is found to perform better than an

alternative structure where communication variables cause positive and negative

emotions. Thus, results provide strong support for the proposed affective citizen

communication model. Refinements to the proposed model, connections with existing

theories of political communication, such as agenda setting and partisan selective

exposure, and directions for future research are also discussed.

xi
 Table of Contents

List of Tables xiv

List of Figures xvii

Chapter 1: Introduction 1
Research Objectives 3
Relevance of the Study 6
Organization 8

Chapter 2: From Emotions to Communication to Political Involvement 10

Introduction 10
A Brief Summary of Affective Intelligence 13
A Brief Summary of Communication Mediation 16
The Affective Citizen Communication Model 18
Refining the Affective Citizen Communication Model 24
Positive vs. Negative Emotions 24
News Use, Debate Viewing and Personal Communications 28
Hypotheses 30
Direct Effects 30
Indirect Effects 31

Chapter 3: Methods 32
Data Sets 32
Overview of the ANES 2008-2009 Panel Study 33
Overview of the NAES 2004 Debates Panel Study 35
Definition of Key Variables 36
Political Knowledge 37
Political Participation 41
Emotions 45
News Media Use 47
Political Discussion 50
Measurement of Variables 51
Variables from the ANES 2008-2009 Panel Study 52
Variables from the NAES 2004 Debates Panel Study 57
Statistical Analysis 62
Structural Equation Modeling 62
Assessing Model Fit in SEM 64
Analytical Strategy 65

xii
 Indirect Effects 72

Chapter 4: Results 74
Introduction 74
Analyses of the ANES 2008-2009 Panel Study 75
Cross-Sectional Model 75
Auto-Regressive Model 80
Comparison with Alternative Model 85
Analyses of the NAES 2004 Debates Panel Study 87
Cross-Sectional Model 87
Auto-Regressive Model 90
Comparison with Alternative Model 95
Integrating Results across Data Sets 96

Chapter 5: Discussion and Conclusions 101

Introduction 101
Discussion of Findings 102
Strengths of the Study 112
Methodological Limitations 115
Expanding the Affective Citizen Communication Model 119
Bringing in the Moderators 119
Beyond Political Involvement 122
Convergence with Other Communication Theories 123
Agenda Setting and Priming 124
Selective Exposure 127
Summary and Conclusions 128

Appendix 131

References 153

Vita 180

xiii
 List of Tables

Table 3.1 Descriptive Statistics for Emotion Items in the ANES 2008-2009 Panel 54

Table 3.2 Descriptive Statistics for Emotion Items in the NAES 2004 Debates Panel 59

Table 4.1 Comparison of Theorized Cross-Sectional Models using the ANES 2008-2009

Panel 76

Table 4.2 Direct, Indirect and Total Effects in the Final Cross-Sectional Model using the

ANES 2008-2009 Panel 79

Table 4.3 Comparison of Theorized Auto-Regressive Models using the ANES 2008-2009

Panel 81

Table 4.4 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using the

ANES 2008-2009 Panel 84

Table 4.5 Comparison of Theorized Model with Alternative Model using the ANES 2008-

2009 Panel 86

Table 4.6 Comparison of Theorized Cross-Sectional Models using the NAES 2004 Debates

Panel 87

Table 4.7 Direct, Indirect and Total Effects in the Final Cross-Sectional Model using the

NAES 2004 Debates Panel 90

Table 4.8 Comparison of Theorized Auto-Regressive Models using the NAES 2004

Debates Panel 91

Table 4.9 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using the

NAES 2004 Debates Panel 94

Table 4.10 Comparison of Theorized Models with Alternative Models using the NAES

2004 Debates Panel 96

xiv
Table 4.11 Proportion of Hypotheses Supported using the ANES 2008-2009 Panel

Sample and the NAES 2004 Debates Panel 98

Table 4.12 Average Direct, Indirect and Total Effects Weighted by Sample

Size 105

Table A.l Demographic Profile of ANES 2008-2009 Panel Sample and Census Data 132

Table A.2 Demographic Profile of NAES 2004 Debates Panel Sample and U.S. Census

Data 133

Table A.3 Exploratory Factor Analysis of Emotional Reactions to Presidential Candidates

in the ANES 2008-2009 Panel (Unrotated Solution) 134

Table A.4 Exploratory Factor Analysis of Emotional Reactions to Presidential Candidates

in the ANES 2008-2009 Panel (Rotated Solution) 135

Table A.5 Exploratory Factor Analysis of Emotional Reactions to Presidential Candidates

in the NAES 2004 Debate Panel (Unrotated Solution) 136

Table A.6 Partial Correlation among Key Variables in the ANES 2008-2009 Panel 137

Table A.7 Partial Correlation among Key Variables in the NAES 2004 Debates Panel ... 139

Table A.8 Regression Analysis of Key Variables on Controlling Variables in the ANES

2008-2008 Panel 141

Table A.9 Regression Analysis of Key Variables on Controlling Variables in the NAES 2004

Debates Panel 143

Table A.10 Comparison of Theorized Auto-regressive Models using the ANES 2008-2009

Panel 145

Table A.11 Comparison of Theorized Auto-regressive Models using the NAES 2004

Debates Panel 146

Table A.12 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using the

ANES 2008-2009 Panel 147

xv
Table A.13 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using the

NAES 2004 Debates Panel 150

xvi
 List of Figures

Figure 2.1 Structural Paths Predicted by Affective Intelligence 19

Figure 2.2 Structural Paths Predicted by Communication Mediation 20

Figure 2.3 Structural Paths Predicted by Affective Citizen Communication 23

Figure 2.4 Affective Citizen Communication with Positive and Negative Emotions 25

Figure 2.5 Affective Citizen Communication with Habitual and Particular Media Use 29

Figure 3.1 Theorized Model Using Cross-Sectional Data 66

Figure 3.2 Theorized Model using Panel Data 69

Figure 3.3 Alternative Model using Panel Data 71

Figure 4.1 Final Cross-Sectional Model using the ANES 2008-2009 Panel 77

Figure 4.2 Final Auto-Regressive Model using the ANES 2008-2009 Panel 82

Figure 4.3 Final Cross-Sectional Model using the NAES 2004 Debates Panel 88

Figure 4.4 Final Auto-Regressive Model using the NAES 2004 Debates Panel 92

Figure 4.5 Significant Paths of the Affective Citizen Communication Model 99

xvii
 Chapter 1: Introduction

White students appalled by the violence in the South in the 1960s went on buses
into a world they had not known to seek justice for others. Similarly, the
environmental movement, the AIDS movement, the pro-choice and pro-life
movements, and many others recognize not only that claims of justice must be
advanced but also that people get angry, that people get attentive, that people
get hopeful, and that they can be moved to action by emotions evoked by well-
crafted campaigns (Marcus, 2002, p. 45).

Political communication is built on emotions. Fear ads are a mainstay of

American electoral campaigns (Geer, 2006; Mark, 2009). Candidates tailor their

messages to spark enthusiasm and hope in their voters—and anxiety and anger toward

their opponents (Brader, 2006b). Journalists are accused of a negativity bias towards

politicians and the political process in general (Cappella & Jamieson, 1997; Patterson,

1993). Anger and resentment are the defining traits of many grassroots movements and

their messages, from the New Left of the 1960s to the tea party of the Obama age

(Bunch, 2010; Goodwin, Jasper, & Polletta, 2001).

The commonality of affect in politics notwithstanding, most theories of political

communication tend to deemphasize the role played by emotions. Agenda setting

(McCombs, 2004), priming (Iyengar & Kinder, 1987) and framing (D'Angelo & Kuypers,

2009; Reese, Gandy, & Grant, 2001)—the three most salient paradigms of media effects

in the field (Graber & Smith, 2005)—as well as longstanding models of citizen

communications, including the two-step flow (Katz & Lazarsfeld, 1955) and democratic

deliberation (Gastil, 2008), are all rooted in models that stress cognitive aspects, such as

attention, elaboration and knowledge activation (cf., Chaffee & Schleuder, 1986;

Eveland, 2004; Price & Tewksbury, 1997).

If included at all, affective aspects such as arousal, valence and discrete positive

and negative emotions are brought in as auxiliary variables, perhaps as moderators of a

1
more important relationship or as additional consequences of a causally prior cognitive

process. With some notable exceptions (e.g., Brader, 2005; Coleman & Wu, 2010;

Druckman & McDermott, 2008; Holbert & Hansen, 2008; Miller, 2007; Parsons, 2010b;

Valentino, Hutchings, Banks, & Davis, 2008; Way & Masters, 1996a), political

communication research has not fully realized the fact that emotions are central to

human experience. This is neither surprising nor unique to the field of political

communication. There is a long tradition in Western thought—think Plato, Descartes

and Kant—that downplays the contribution of affect (i.e., emotion) while highlighting

the benefits of reason (i.e., cognition). Furthermore, the cognitive revolution that swept

social psychology—one of the mother fields of political communication—in the 1950s

and 1960s, with its human-mind-as-computer-processor metaphor, reinforced the

prevalence of emotionally-devoid models of political communication (Marcus, 2003).

In the last two decades, however, the era of "cognitive imperialism" (Lau &

Sears, 1986, p. 8) has receded, prompting observers to talk of an "affective revolution"

within the social sciences (Holbert & Geidner, 2009, p. 353). While the idea of revolution

may sound hyperbolic, certainly more scholars in political communication are studying

the determinant role played by emotions in human behavior and decision making (a

good overview is the edited volume by Neuman, Marcus, MacKuen, & Crigler, 2007). In

the last few years, various specialized journals in the field of politics and communication

have devoted special issues to the theme of emotions. This growing interest

notwithstanding, current scholarship has only scratched the surface on the affective

determinants of political communication processes.

If political communication is to become a more relevant field, researchers need

to be aware of existing models in related fields of inquiry and update their

understanding of the processes of communication about politics accordingly. Perhaps

2
the best example of this productive route is the work on selective exposure, which

started firmly entrenched in social psychology (e.g., Erlich, Guttman, Schonbach, & Mills,

1957; Freedman & Sears, 1965) and now is a central body of work within political

communication (e.g., Stroud, 2011). Fortunately, this exercise is more attainable than

before because we can borrow from existing theoretical approaches to emotion from

psychology and political science—fields that have developed in the last two decades a

rich literature on the links between emotion, cognition and behavior (for brief literature

reviews on the study of affect in each of these fields, see Cacioppo & Gardner, 1999;

Marcus, 2000, respectively).

RESEARCH OBJECTIVES

This dissertation seeks to improve our understanding of how citizens learn about

public affairs and engage in political activities by advancing a model that incorporates

the dynamics of emotions, various forms of media use, and interpersonal

communication to explain citizens' political involvement during electoral campaigns.

This model, which I have labeled the affective citizen communication model, integrates

into a single framework the insights of affective intelligence theory (Marcus, Neuman, &

MacKuen, 2000) and the work on communication mediation (McLeod et al., 1999;

McLeod et al., 2001) and its two iterations, cognitive mediation (Eveland, 2001; Eveland,

Shah, & Kwak, 2003) and citizen communication mediation (Cho, Shah, McLeod,

McLeod, & Scholl, 2009; Shah, Cho, Eveland, & Kwak, 2005; Shah et al., 2007). More

specifically, the affective citizen communication model proposes that the effects of

positive (e.g., hope) and negative (e.g., anger) emotions triggered by political candidates

on citizens' knowledge of the candidates' stands on the issues and on participation in

3
campaign activities are largely mediated by information-seeking behaviors, such as news

media use, debate viewing and political discussions with peers.

By positing emotions as an antecedent of both media use and interpersonal

communication, the study extends current research based on affective intelligence

theory. In the realm of political communication, this theory has focused mostly on

campaign news attention as an outcome of political emotions. Conversely, studies using

affective intelligence theory to study the impact of emotions on political discussion tend

to so in isolation of media variables. At the same time, the dissertation adds emotions to

communication mediation processes, which to date have been studied from a purely

cognitive perspective only. Here lies the greatest contribution of the current study.

Importantly, the affective citizen communication model also builds a theoretical

bridge between affective intelligence and communication mediation, two strands of the

political communication literature that have developed separately but, nonetheless, are

compatible. Affective intelligence has studied the direct link between feeling emotions

towards political objects and individuals' learning of, and engagement with, political

affairs, without specific attention to the intervening processes by which this influence

occurs (MacKuen, Wolak, Keele, & Marcus, 2010; Marcus & MacKuen, 1993; Marcus,

Wood, & Theiss-Morse, 1998). Communication mediation models, on the other hand,

usually elaborate on the intervening psychological and communication processes

between media use (e.g., attention, elaboration and deliberation) and outcomes such as

civic participation, paying far less attention to the psychological underpinnings of media

use beyond political interest and surveillance motivations (Cho, Shah, McLeod, McLeod,

& Scholl, 2009; Eveland, 2004; Holbert, 2005; Shah et al., 2007). Showing the

compatibility between these two theoretical approaches is the main purpose of the

study.
4
 To examine the relationships between the variables identified in the affective

citizen communication model, I rely on panel data from the American National Election

Study (ANES) of the 2008 U.S. presidential election and the National Annenberg Election

Survey (NAES) of the 2004 U.S. presidential election. There are several advantages to

employing separate data sets: (1) it provides a check on the generalizability of the

theorized model across different election cycles; (2) it allows for variation in the time-

span to measure effects; and (3) it permits to test the robustness of the substantive

findings to alternative measurement of key variables. The use of two-wave panel data,

on the other hand, allows stronger causal inference than cross-sectional data because it

explicitly builds in the time dimension of a dynamic process that is thought to be of a

causal nature (Finkel, 1995). That is, change in a set of variables can be directly

measured. Furthermore, by having measured each key variable at two different points

in time, alternative specifications to the proposed model can be tested and compared to

find the best-fitting model. Prior measurement of control variables, on the other hand,

makes it more plausible to regard them as exogenous to the processes connecting the

variables of interest.

The data sets used in this dissertation, however, cannot fully address the

problem of measurement error; three-wave panel surveys would be better equipped to

separate instability from unreliability (Bartels, 2006). Furthermore, while panel surveys

are more consistent with causal inference than cross-sectional surveys, it is important to

note that they are not a cure all. The issue of omitting alternative causal factors looms

all research that is not based on random assignment of participants to treatment

conditions. Perhaps lab experiments could better address the causality quandary

(although the variety of causal relationships implied by the affective citizen

communication model are not easily captured in one single experiment). These
5
limitations notwithstanding, use of the ANES and NAES panels was warranted because

both are based on nationally representative samples, which are well equipped to

address concerns over the external validity of the proposed model.

RELEVANCE OF THE STUDY

A focus on political knowledge and participation as outcomes of interest is

justified on both normative and empirical grounds. Democratic theory assumes a

knowledgeable citizenry that is able to reach political decisions after careful

consideration and evaluation of available information on current and future courses of

action (Dahl, 1989, 1999). When people are informed about the affairs of the polity,

they can better safeguard their preferences and check that the government acts in the

public good (Delli-Carpini & Keeter, 1996; Luskin, 2003). A basic assumption of

elections—the most visible manifestation of democracy—is that individuals use their

knowledge to understand the issues being discussed and to vote in line with their

interests (Althaus, 2001; Bartels, 1996; Wlezien & Soroka, 2007). On the other hand, it

has long been recognized that the average level of political knowledge of the public, at

least in the U.S., is low (Berelson, Lazarsfeld, & McPhee, 1954; Campbell, Converse,

Miller, & Stokes, i960; Converse, 1964; Lippmann, 1922) and skewed towards the better

educated, higher income and dominant groups of society (Holbrook, 2002; McDevitt &

Chaffee, 2000). In fact, some scholars have argued that the current media environment,

particularly with the diffusion of the Internet and cable television, has increased the

gaps in political knowledge and involvement (e.g., Prior, 2007). This is a cause of

concern because previous empirical research (Junn, 1991; Leigh ley, 1991) has

demonstrated that political knowledge is positively correlated with active participation

in politics (although the strength of this relationship is still a matter of dispute; see

6
Levendusky, 2011). And when people participate, they have a voice in public affairs,

they can hold authorities accountable, and they are empowered to act on their own

behalf (Verba, Schlozman, & Brady, 1995). As Marcus (2002) argued, "participation is

the irreducible requirement to ensure that the citizens retain, individually and

collectively, the status of autonomous citizens" (pp. 42-43).

Furthermore, there is an empirical advantage for studying political knowledge

and participation. Both concepts have been widely examined in the literature. By

studying how emotions and communication relate to knowledge and engagement, the

dissertation gains theoretical leverage: it can use and expand extant scholarship on the

subject.

Examining the emotional foundation of communication, knowledge and

participation is also important. From a scientific perspective, we can develop a more

nuanced and accurate description of how citizens think, decide and behave in public

affairs. Our models may be cognitive-dominated, but this does not mean that

individuals' learning and engagement involves the cognitive system only; in fact,

neuroscience tells us that it may be more appropriate to say that individuals are rational

and cognitive as a consequence of being affective and emotional (Carver & White, 1994;

Gray, 1985, 1987).

From a normative perspective, studying the important role played by emotions

in driving information processing and stimulating democratically desirable outcomes

could lead to new ways of reinvigorating citizen engagement. For government and

policymakers, a better understanding of the role played by emotions may lead to

developing more effective campaign messages, while politicians and individuals seeking

public office would benefit from more successful mobilizing and recruitment efforts. For

journalists wishing to spark greater attention to public affairs, understanding the role of
7
emotions could lead them to develop storytelling techniques that foster affective

responses positively related with news attention and learning information. Most

importantly, a better understanding of the role of affect can help citizens make better

decisions. For instance, it has been shown that in an electoral campaign, affect towards

the candidates can provide a useful—and, perhaps, more accurate—heuristic for one's

evaluations of the merits of each candidate (Brady & Sniderman, 1985; Bucy &

Newhagen, 1999).

ORGANIZATION

This dissertation is organized into four chapters in addition to this introduction.

Chapter 2 summarizes the two main bodies of work informing the study, affective

intelligence and communication mediation. Based on these two paradigms, a new

model for linking emotions, communication, political information and behavior is

introduced, the affective citizen communication model. This model posits that positive

and negative emotions people feel towards political candidates trigger exposure to

public affairs content in the media (i.e., news use and debate viewing) and political

discussions (i.e., conversations with family and friends about politics and elections),

which in turn influence levels of political knowledge (i.e., candidates' issue stances) and

campaign behavior (i.e., political participation).

Chapter 3 describes the methods used in the analysis. First, I provide a general

overview of the data sets, the ANES 2008-2009 Panel and the NAES 2004 Debates Panel.

After addressing some of the problems in the conceptualization and operationalization

of the main variables of the affective citizen communication model, I detail the

measures used in the subsequent statistical analyses. Because I employ structural

equation modeling, this chapter includes with a discussion of the advantages of this

8
technique over more traditional forms of multivariate analysis. The chapters ends with a

detailed account of the analytical strategy that will be used to estimate the proposed

model.

Chapter 4 presents the results of the structural equations testing the fit of the

affective citizen communication model to the data employed. The findings are

presented separately for the ANES and NAES surveys, as each deal with a different

election cycle. The chapter also presents the results of an alternative causal ordering of

the variables, so as to assess the robustness of the theorized model. At the end of the

chapter, a summary of the hypotheses that were supported is presented.

Lastly, Chapter 5 discusses the findings, poses the methodological strengths and

limitations of the study and elaborates on directions for future research. More

specifically, I discuss possible ways of refining and expanding the model examined in this

study, and provide possible applications of the model to existing theories of

communication (namely, agenda setting, priming and selective exposure). The chapter

ends with a discussion on the implications of the dissertation for democratic citizenship.

9
 Chapter 2: From Emotions to Communication to Political Involvement

Reason is, and ought only to be the slave of the passions, and can never pretend
to any other office than to serve and obey them (Hume, 1739/1888, p. 415).

There can be no knowledge without emotion. We may be aware of a truth, yet

until we have felt its force, it is not ours. To the cognition of the brain must be
added the experience of the soul (Bennett, 1932).

INTRODUCTION

The resurgence of scholarly interest on the effects of emotions about politics and

political figures on American voters' gained full traction in the 1980s (e.g., Abelson,

Kinder, Peters, & Fiske, 1982; Conover & Feldman, 1986; Marcus, 1988; Roseman,

Abelson, & Ewing, 1986; Sears & Citrin, 1982; Sullivan & Masters, 1988). The seminal

work laid out then translated later into several lines of research relating political affect

to citizens' cognitions and judgments. Chief among these new theoretical accounts is

affective intelligence, which has garnered substantial scholarly attention in the last few

years. According to Google Scholar, as of April 2011, the first incarnation of this

theory—a journal article by Marcus and MacKuen (1993)—has been cited in 301 works.

The more developed version of the theory—distilled in book form (Marcus, Neuman, &

MacKuen, 2000)—has received 480 citations, which translates into 40 citations per year.

This is comparable to the amount of scholarly attention garnered by Iyengar and

Kinder's (1987) News that Matters, a bedrock of the political communication literature.

With 1,141 citations since its first year of publication, this classic work on media priming

effects has had an average of 46 annual citations.

The use of affective intelligence theory is justified for several reasons. Unlike

other theories of emotions and politics, it is explicitly based on the work of

neuroscientists—most notably, Gray's dual affect system (1985; , 1987; , 1990)—thus

10
providing direct links between political behavior, evolutionary theory and current

knowledge on the way the human brain works. Contrary to appraisal theories of

emotion (Scherer, 1999; Smith & Kirby, 2001), affective intelligence is specific enough to

predict relationships between feelings, media attention, political learning and

participation—central concepts of the affective citizen communication model proposed

in this study. Furthermore, affective intelligence posits dynamic (i.e., changing over

time) associations between emotions and concrete behaviors, which fits with the

dynamic nature of political campaigns and the longitudinal nature of the data used in

this study.

Another reason for employing affective intelligence relates to the cumulative

nature of social science. There is substantive evidence collected using a variety of

methods, including cross-sectional surveys, panel designs, controlled lab experiments

and field experiments, supporting the basic tenets of the theory as it applies to the

realm of political campaigns (see, e.g., Brader, 2006b; Crigler, Just, & Belt, 2006;

MacKuen, Wolak, Keele, & Marcus, 2010; Redlawsk, Civettini, & Lau, 2007). Lastly, and

from a more practical perspective, this theory can be operationalized using self-reports,

rather than more costly and cumbersome physiological measures adopted by other

work relating emotions and communication, such as skin conductance, heart rate, facial

EMG data, and secondary task reaction times (e.g., Lang & Newhagen, 1996; Lang, Park,

Sanders-Jackson, Wilson, & Zheng, 2007). For all these reasons, affective intelligence

provides a solid framework on which to build research on emotions, communication,

and political outcomes.

On a similar vein, I rely on a particular theoretical approach to examine the

effects of mediated and interpersonal communication on political knowledge and

participation—the second component of the affective citizen communication model.

11
This approach is the communication mediation model (McLeod et al., 1999; McLeod et

al., 2001) and its two iterations, the cognitive mediation model (Eveland, 2001; Eveland,

Shah, & Kwak, 2003) and the citizen communication mediation model (Cho, Shah,

McLeod, McLeod, & Scholl, 2009; Jung, Kim, & Gil de Zuniga, forthcoming; Shah, Cho,

Eveland, & Kwak, 2005). While the field of political communication is ripe with media

effects theories, few bodies of work explicitly integrate the effects of political media use

and talk on citizens' knowledge and participation. The two-step flow of communications

(Katz & Lazarsfeld, 1955; Robinson, 1976), diffusion of innovations (Rogers, 2003),

communicatory utility (Atkin, 1972) and the differential gains hypothesis (Scheufele,

2002) come readily to mind, but not much else. Perhaps this is a natural division of

labor, similar to the divide between mass media and interpersonal communication

research. In any case, this is somewhat surprising, considering that political discussions

do not occur in a vacuum of media content (Chaffee, 1986; Southwell & Yzer, 2007). On

the contrary, news coverage often sparks informal political conversations, particularly

during election campaigns (Gamson, 1992; Just et al., 1996; Walsh, 2004). Talk, on the

other hand, influences what people get from the media, such as when individuals

monitor the news more closely because they anticipate having discussions with others

(Eveland, 2004). These examples should make it apparent that the effects of political

conversation and news consumption on political outcomes are related. In fact, it may

well be that the effects of news are mediated by discussion.

Against this context, the model of communication mediation stands out as a

useful framework for studying the direct and indirect relationships between media use,

discussion, political knowledge and participation. Particularly, it provides a framework of

structural paths between various forms of media use and communication within social

networks. There are other reasons for choosing this model, too. It forces one to consider
12
the process and intervening processes of communication effects, rather than a focus on

simple, direct effects. Among the outcomes for which this model has been found to be

applicable are political knowledge and campaign participation, the same outcomes

studied in this dissertation (see, e.g., Cho, Shah, McLeod, McLeod, & Scholl, 2009;

Eveland, Hayes, Shah, & Kwak, 2005b; Gil de Zuniga & Valenzuela, Forthcoming).

Furthermore, its premises have been found to be robust to tests conducted using both

cross-sectional and longitudinal data, that is, there is cumulative evidence of the causal

links between the variables of the model.

In what follows, I will provide brief summaries of affective intelligence theory

and the communication mediation model, and subsequently integrate the insights of

these approaches into a novel theoretical model, labeled affective citizen

communication. After examining existing research about the various hypothesized links,

the last section presents the formal research hypotheses that will be tested in the

dissertation.

A BRIEF SUMMARY OF AFFECTIVE INTELLIGENCE

Affective intelligence posits that people have a dual emotional system that

produces specific emotional appraisals, which in turn determine both thought (e.g.,

information-processing and cognitive activities) and behavior (e.g., media use,

discussion, political participation) (Marcus & MacKuen, 1993; Marcus, Neuman, &

MacKuen, 2000). While the disposition system triggers emotions that fall along the

continuous ranges of happiness or satisfaction, the surveillance systems gives rise to

emotions of anxiety and unease (Brader, 2006b, p. 60). Which emotional system is

activated depends on incoming information. When received information suggests that

the execution of one's plans is consistent with expectations, the disposition system kicks

13
in by giving rise to emotions of enthusiasm, such as hope and pride. Think, for instance,

when your candidate of choice is doing well in the polls and seems likely to win the

election. Conversely, if there's inconsistency between the executions of one's plans and

expectations, sadness and depression, even aversion, may arise. Think, for instance, if

your candidate is not doing well in the polls and, furthermore, you can attribute this

failure to a specific something (e.g., the opposition party's attack ads) or someone (e.g.,

the candidate's own gaffes). Thus, enthusiasm, and lack thereof, reflect the typical

affective components of liking and disliking, or what psychologists term approach and

avoidance (Wolak, MacKuen, Keele, Marcus, & Neuman, 2003, p. 2). The enthusiasm

and aversion dispositions act as guiding cues; they indicate that things are going as

planned, whether we like it or not.

Nevertheless, as Wolak and her colleagues (2003) noted, "single-minded reliance

on routines is efficient only so far as one can be confident they produce outcomes in

line with their best interests" (p. 3). Here lies the importance of the surveillance system,

the second component of the emotional system. According to the theory of affective

intelligence, the surveillance system is activated by threatening stimuli, that is, when

things are unexpectedly novel and merit careful attention. Think, for instance, when

your political party nominates a candidate you believe is incompetent or will not be able

to attract independents. The surveillance system signals that continued reliance on

routines may not produce the best course of action—in fact, careful attention to the

threatening stimuli may lead to a new course of action (i.e., voting for the candidate of

another political party).

In the course of a political campaign, affective intelligence suggests two

alternative scenarios. If a political candidate activates in a voter his/her disposition

system, as suggested by the emotions of enthusiasm and hope, the individual will rely
14
on existing habits and established preferences when deciding what to think, how to

think and what to do about that particular candidate. Simply put, the election for that

voter will be business as usual: voting according to established predispositions (e.g.,

party ID and ideological matching). If, however, the candidate activates in a voter

his/her surveillance system, as manifested by the emotions of anxiety, uneasiness and

worry, it is more likely that the election will be anything but business as usual. These

emotions will stimulate increasing attention on the source of the threat in order to learn

and have a more informed understanding of it. Therefore, the surveillance system

motivates the abandonment of established predispositions and promotes seeking out

novel information. Thus, emotions determine one of two courses of action: maintaining

current political habits or engaging in "stop-and-think" moments in which existing

beliefs are less useful.

The key assumption of affective intelligence, then, is that emotions precede, and

trigger, cognitive processes. In the now classic academic debate between the primacy of

affect and the primacy of cognition (Lazarus, 1984; Zajonc, 1984), affective intelligence

sides clearly with the former. Another important assumption of affective intelligence is

that myriad emotions towards political figures and issues can be reduced to a few

dimensions. So, unlike discrete models of emotions (e.g., Roseman, 1991; Roseman,

Wiest, & Swartz, 1994), individuals can experience emotions on two or three different

dimensions that may well be orthogonal to each other (e.g., voters can be worried and

hopeful about the candidates at the same time).

Evidence supporting the theory of affective intelligence comes in many forms:

neuroscientific research (Gray, 1990), survey analysis—particularly work using ANES

data (Marcus, 2000; Marcus & MacKuen, 1993)—and experimental research (Brader,

15
2006b; Marcus, Wood, & Theiss-Morse, 1998; Valentino, Banks, Hutchings, & Davis,

2009). Therefore, it provides a solid ground for the current research endeavor.

A BRIEF SUMMARY OF COMMUNICATION MEDIATION

When researchers analyze the direct effects of media use on political outcomes

such as issue knowledge and campaign behavior, their analyses tend to follow a simple

stimulus-response framework (for a good example of this line of work, see Drew &

Weaver, 1991, 1993, 1998, 2006; Weaver & Drew, 1995, 2001). Nevertheless, for

decades researchers in communication have rejected the notion of media use as a

"magic bullet" or "hypodermic needle." Oftentimes, media effects on citizenship are

indirect and mediated by psychological and communication processes (McGuire, 1972).

The communication mediation model developed initially by McLeod and his colleagues

(McLeod et al., 1999; McLeod et al., 2001) takes this notion by heart by concluding that

"informational media use and political discussion largely channel the effects of

background dispositions and orientations on citizen learning and participation" (Shah,

Rojas, & Cho, 2009, p. 216). More specifically, as an outgrow of Markus and Zajonc's

(1985) Orientation-Stimulus-Orientation-Response (O-S-O-R) model, the

communication mediation model integrates and extends past research on the effects of

audience members' predispositions (0%) on the reception of media messages (S) and the

ways in which they interact with media content (0 2 ) and respond to it (R).

Originally, this model was interested in studying how individual-level

orientations (e.g., values) had an effect on civic participation via news media use and

interpersonal communication. However, subsequent research led to refinements in the

original model. Eveland and colleagues (Eveland, 2001, 2004; Eveland, Shah, & Kwak,

2003) used cross-sectional and panel survey data to examine more closely the 0 2 part of

16
the model and asked: What cognitive mechanisms mediate the effects of news attention

on political knowledge? Their findings showed that mental elaboration of the messages

partially mediated the effects of attention to media messages on learning, while both

elaboration and attention fully mediated the motivations to use informational media on

political knowledge. Thus, a cognitive mediation model was advanced. Subsequently, a

team of researchers led by Shah (Cho, Shah, McLeod, McLeod, & Scholl, 2009; Shah,

Cho, Eveland, & Kwak, 2005; Shah et al., 2007) proposed a citizen communication

mediation model by theorizing that the effects of media use (i.e., exposure and

attention combined) on participatory behaviors were largely mediated through face-to-

face and online discussions about news. Using panel data and a series of alternative

causal orderings of the communication and outcome variables, they found that both

interpersonal political conversation and interactive political messaging channeled the

effects of traditional and online media effects on civic engagement.

The contribution of the cognitive mediation model and the citizen

communication mediation model to the original communication mediation model has

led researchers in the area to revise the O-S-O-R framework and propose an O-S-R-O-R

model (Cho, Shah, McLeod, McLeod, & Scholl, 2009; Jung, Kim, & Gil de Zuniga,

forthcoming). In this revised framework, activities such as mental elaboration and

political discussion are treated as reasoning (R%) about media stimuli (S). As Shah and

colleagues (2009) explained,

Currently, the S-0 portion of the model is a jumble of factors, including news
consumption, thinking and talking about issues, and cognitions and attitudes
that arise from this process. Mental elaboration and interpersonal discussion are
particularly difficult to situate in this framework. They are not stimuli in the
formal sense, since they have been found to be causally antecedent of exposure
to mass media (Eveland et al., 2003; Shah et al., 2005). However, they are also
not conventional outcome orientations in the sense of altered attitudes or

17
 developed cognitions. Instead, they are between stimuli and outcome
orientations, indicative of efforts to form an understanding and reason through
ideas (p. 218).

Recently, researchers have started to map out the second set of orientations

(0 2 ), which have been neglected in previous research. For instance, it has been found

that internal political efficacy partially mediates the effects of discussion on political

participation, both online and offline (Jung, Kim, & Gil de Zuniga, forthcoming). This

makes sense. A perceived sense of ability to influence politics can be stimulated by

political news use (Semetko & Valkenburg, 1998) and through political discussions (Min,

2007).

The refined and detailed account of media effects posited by communication

mediation provides a robust foundation for studying the interactions between

emotions, information-seeking behaviors and political outcomes.

THE AFFECTIVE CITIZEN COMMUNICATION MODEL

So far, the reviews of affective intelligence theory and communication mediation

and its various iterations have led to two alternative models explaining political

knowledge and participation. As Figure 2.1 shows, affective intelligence predicts a direct

link between emotions and attention to campaign news. In particular, surveillance-

related emotions such as fear and anxiety should motivate increased attention to the

media environment so as to extract useful information to decide how to act (Marcus,

Neuman, & MacKuen, 2000; Valentino, Banks, Hutchings, & Davis, 2009). It also suggests

a direct link between emotions and involvement in the activities of a campaign. In

particular, emotions of enthusiasm towards the preferred candidate, such as hope and

pride, and emotions of aversion towards the other candidate, such as anger, can

motivate voters to go to the polls and get involved in different campaign activities of

18
their preferred candidate (Marcus & MacKuen, 1993; Valentino, Gregorowicz, &

Groenendyk, 2009; Valentino, Hutchings, Banks, & Davis, 2008). Thus, while both

feelings of enthusiasm, anxiety and anger can be conducive to media use and political

participation, it is anxiety that most directly leads to knowledge acquisition.

Figure 2.1 Structural Paths Predicted by Affective Intelligence

Positive Media Use

Emotions (News, Debate
(Disposition) Viewing)

Negative Involvement
Emotions (Knowledge,
(Surveillance) Participation)

Processes of communication mediation, in turn, focus our attention on the

indirect effects of informational uses of media, such as news use and debate viewing, on

political knowledge and participation through citizen-to-citizen political discussion. In

graphic form, the expectation is that political media use motivates conversations about

politics, which in turn shape citizens' learning and behavior in electoral campaigns (see

Figure 2.2). The rationale for the mediating role of conversations on news media effects

is that by talking, people can make better sense of political information. Relatedly, there

is a strong expectation that certain forms of communication, such as media use and

debate viewing, lead to other forms of communication, such as interpersonal political

discussion—in other words, "the most likely 'effect' of communication (...) is further

communication" (Chaffee, 1986, p. 76).

19
Figure 2.2 Structural Paths Predicted by Communication Mediation

Media Use Political Involvement

(News, Debate Discussion (Knowledge,
Viewing) (Frequency) Participation)

Having laid out the essential components and relationships predicted by each of

the theoretical models that inform the current study, the next step is to synthesize the

predicted effects of emotions, media use and political discussion on both political

knowledge and behavior. This synthesis will take the form of the affective citizen

communication model, which combines into a single framework the insights of affective

intelligence theory and the communication mediation model. As explained in Chapter 1,

the purpose of doing such a synthesis is twofold. On the one hand, the new model

connects the burgeoning area of communication mediation with the affective route to

political information and behavior. On the other hand, it forces affective intelligence

theory to look at media use and particularly political discussion as a central element in

the process of emotions leading to cognition and behavior.

At a theoretical level, emotions triggered by candidates, issues and other

political stimuli should lead individuals to be more motivated to engage in

communication behaviors and markers of political involvement, such as knowledge and

participation. That is, individuals who feel emotions (whether positive or negative) are

more likely to pay attention to the campaign and exchange information and opinions

with others compared to individuals who experience apathy—or lack of emotion.

Likewise, individuals with stronger emotions about the candidates and the campaign in

general are more likely to learn about the policy stands of the candidates (in part

20
because they are consuming more information) and also to engage in campaign activity.

In contrast, individuals for whom the campaign does not trigger any emotion are more

likely to stay like that. Their emotional system will not cause them to reconsider paying

attention to the campaign or engaging in political acts. They will remain in their habit of

not following political news, not discussing about the campaign, not learning about the

issue policies of the candidates and not volunteering for the candidates. In the words of

Brown-Kramer (2009),

They are not enthusiastic, so they feel no need to volunteer their time to support
a candidate. They are not angry, so they are unmotivated to act to prevent a
candidate from being elected. They are not anxious, so they need not reconsider
their default attitude toward the candidate (p. 66).

If communication behaviors are intervening processes in the relationship

between political emotions and both political knowledge and participation, then

emotions represent a necessary but not sufficient condition for learning and engaging in

campaign activity. Clearly, people who follow political news and programs and discuss

with others what is going in the election are not blank slates that process information in

a vacuum of emotions (Civettini & Redlawsk, 2009). Furthermore, communication

behaviors are not constant throughout the campaign period. As the National Annenberg

Election Survey (NAES) has shown (Jamieson & Kenski, 2006; Stroud, 2010), exposure to

campaign news varies over time and changes following particular events. The same is

true with talking about politics. According to NAES studies on the 2000, 2004 and 2008

elections, after party conventions and presidential debates there is a spike in the

frequency of political discussion reported by respondents (Hardy & Scheufele, 2005;

Kenski, Hardy, & Jamieson, 2010). These communication dynamics reveal that people

are responsive to the political environment. The type of response they enact, however,

is governed by the disposition and surveillance systems. Therefore, emotions should be

21
considered as antecedents of communication behaviors in the affective citizen

communication model. Put another way, people who will follow campaign media

content and discuss it with others must first feel something about the campaign that will

motivate them to seek out more information through the media or through others.

Greater attention to news content and exposure to political programs, in turn,

should lead to higher levels of political discussion, knowledge and behavior. Media use

often sparks informal political conversations, particularly during election campaigns

(Gamson, 1992; Just et al., 1996; Walsh, 2004). Media use also makes news content

available in memory. Without media use, the information about the candidates policy

stands cannot be learned because it is not brought to conscious awareness (Eveland,

2001; Eveland, Shah, & Kwak, 2003). Media use also leads to participation by making

information about campaign activities available and comprehensible (Boyle &

Schmierbach, 2009; Moy, McCluskey, McCoy, & Spratt, 2004; Sei-Hill & Miejeong, 2005;

Zhang & Chia, 2006). In fact, a variety of political acts require information, such as

where to vote, where to donate money and how to do it (Lemert, 1977, 1984).

Lastly, political talk should directly affect learning and participation. When

people talk about public affairs, they are more likely to mobilize and engage in political

activities, particularly during election campaigns (for an overview, see Delli Carpini,

Cook, & Jacobs, 2004). This is because conversations involve not only exchanges of

information but also interpretive frameworks that help to process that information

(Schmitt-Beck, 2008). By allowing people to grapple with ideas, elaborate arguments

and reflect upon the information acquired, conversations are a rich form of political

information (Huckfeldt & Sprague, 1995). From a behavioral perspective, political

discussions lower the costs of acquiring information and can motivate individuals to

22
learn and participate more often (Just et al., 1996; Klofstad, 2007; McClurg, 2003;

Neuman, Just, & Crigler, 1992).

The affective citizen communication model builds on the findings discussed

above by positing a three-step causal chain of processes. As shown in Figure 2.3, in the

first step and in line with affective intelligence, the political context activates individuals'

disposition and surveillance systems, as signaled by feelings of enthusiasm, anxiety or

anger towards political figures. In the second step, these emotions lead individuals to

maintain, increase or abandon their consumption of informational media (i.e., news use,

debate viewing) as well as the discussion of campaign information with others.

Communication processes, in turn, directly influence individuals' cognitive and

behavioral engagement with the campaign, which in this study are constrained to

political knowledge and political participation.

Figure 2.3 Structural Paths Predicted by Affective Citizen Communication

Media Use
(News, Debate
Viewing)

Emotions
Involvement
(Positive,
(Knowledge,
Negative)
Participation)

Discussion
(Politics, Public
Affairs)

23
 It is not clear, however, if communication fully or partially mediates the

relationship between emotions and markers of political involvement. As

aforementioned, emotions have been found to influence knowledge and participation

directly or, at least, through other mechanisms not considered here. To represent this

possibility of direct effects, emotions and involvement are also connected through an

arrow in Figure 2.3.

REFINING THE AFFECTIVE CITIZEN COMMUNICATION MODEL

Positive vs. Negative Emotions

Certainly, the affective citizen communication model presented in Figure 2.3 is a

simplistic representation of more complex relationships. First, it could be argued that

not all emotions operate the same way. Watson and colleagues (Watson, Clark, &

Tellegen, 1988; Watson & Tellegen, 1985) argued that emotions could be described

along two, orthogonal dimensions, positive and negative, which is consistent with

affective intelligence's disposition and surveillance systems, respectively. In this case,

the affective citizen communication model would look like Figure 2.4.

24
Figure 2.4 Affective Citizen Communication with Positive and Negative Emotions

Media Use
(News, Debate
Viewing)

Positive
Emotions
(Hope, Pride)

Involvement
(Knowledge,
Participation)

Negative
Emotions
(Anxiety, Anger)
Discussion
(Politics, Public
Affairs)

Viewing emotions as a two-dimensional concept of positive and negative affect

leads to the question of which type of emotion is a more powerful driver of media use,

political discussion, learning and campaign participation. In this case, the expectation

from affective intelligence theory is quite clear: negative emotions should have a

stronger effect than positive emotions. While it is possible that enthusiasm and hope

could lead voters to support more decisively a particular candidate and participate

more, the empirical evidence to date shows that there is a strong negativity bias in

terms of the mobilizing effects of emotion (Geer, 2006; Ito, Larsen, Smith, & Cacioppo,

1998; Marcus, Neuman, & MacKuen, 2000, p. 90). The expectation is based on the fact

that positive emotions could stem from satisfaction with the status quo, that is, no

threats are perceived and thus no specific action is being required (Valentino, Brader,

Groenendyk, Gregorowicz, & Hutchings, 2011). Negative emotions, on the other hand,

motivates more effortful and systematic processing of information, but, most

25
importantly, directs attention to new information (Pratto & John, 1991). For instance,

Weber (2008) found that anxiety increases information seeking and political

engagement directly, while anger boosted participation through increased political

efficacy.

The negativity bias notwithstanding, there is evidence that positive emotions can

also have a significant, independent effect on communication behaviors if these

emotions spark increased motivation to follow campaign news. In a summary of existing

empirical evidence on affective intelligence, Brader (2006a) reported that 11 of the 21

tests conducted by six independent studies assessing the effects of enthusiasm on

political interest and information-seeking variables (e.g., attention, time spent reading

campaign information, etc.) yielded a positive, significant relationship. At the same time,

several scholars have found that enthusiasm spurs political action in as much as anxiety

and anger does (Brader, 2005, 2006b; Valentino, Brader, Groenendyk, Gregorowicz, &

Hutchings, 2011).

Furthermore, there is reason to suspect that not all negative emotions can

impact communication and involvement variables equally. For instance, it is quite clear

that anxiety leads to heightened interest in and focus on threatening stimuli (LeDoux,

1996), both in laboratory experiments and in surveys. If that is the case, then one should

expect that fear and anxiety be closely related to media use and discussion and, by

extension, to learning effects. By the same token, anxiety should lead to less

participation, considering that individuals would be less likely to commit themselves to a

specific course of action. Would the same trends be applicable to anger, another

important negative emotion? Previous research has found that anger is more closely

related to approach behaviors, rather than to avoidance behaviors (Huddy, Feldman, &

Cassese, 2007). That is, angry people are more likely than non-angry people to take
26
action. This contention was validated by the work of Valentino and colleagues (2011)

using experimental data, panel surveys and ANES cumulative data sets. These scholars

found that anger boosted participation in a more consistent fashion than fear did,

because

anger arises when threats are attributable to a particular source and the
individual feels that she has control over the situation, while anxiety is triggered
when an individual is less certain about the cause and does not feel in control (p.
160).

If anger has more mobilizing power than fear but a weaker effect on

informational needs as represented by media use and discussion, then anger could lead

to more participation and have a minimal, or even negative, effect on knowledge.

The previous discussion on the differential effects of emotions on political

knowledge and political participation should make it readily apparent that it is not

possible to rely on existing research to posit in advance (1) the specific structure of

political feelings, and (2) the sign of the relationship between emotions and political

involvement variables. This question, necessarily, needs to be addressed empirically,

with the data at hand -a recommendation that, somewhat ironically, the same authors

of affective intelligence have advanced (see Marcus, MacKuen, Wolak, & Keele, 2006).

Because the current study is based on the ANES 2008-2009 Panel and the NAES 2004

Debates Panel, I conducted an exploratory factor analysis via principal component

analysis of the emotion items in order to clarify right away the structure of emotions

elicited by candidates Obama and McCain in 2008 and by candidates Bush and Kerry in

2004. The results of this analysis, detailed in Chapter 3 and in tables A.3 and A.4 in the

Appendix section, yielded two orthogonal dimensions: positive emotions (i.e., hope and

pride in the candidates) and negative emotions (i.e., fear and anger towards the

candidates). These results are consistent with affective intelligence's disposition and
27
surveillance systems and, thus, for the remaining of the dissertation I will refer

interchangeably to positive emotions or feelings of enthusiasm and negative emotions

or feelings of anxiety. The sign of the relationship between these two emotions clusters

and communication and political behaviors, in turn, are the basis of the empirical results

explained in Chapter 4.

News Use, Debate Viewing and Personal Communications

In addition to the different constructs used to examine the role of emotions, the

communication variables incorporated in the model can also be further broken down

into more specific constructs. Habitual exposure to news in traditional and online media

is different from exposure to major political media events, such as party conventions

and televised debates. The notion of intra-media mediation (Holbert, 2005) posits that

different forms of media use have a complementary function. For instance, habitual

news use can lead to debate viewing. Debate viewing, in turn, has been found to

provide fresh material for political discussions (Landreville, Holbert, & LaMarre, 2010)

and exert strong effects on a variety of political outcomes, including political knowledge

(Benoit, Hansen, & Verser, 2003). Accordingly, the affective citizen communication

model can be tested using the following structural model (see Figure 2.5):

28
Figure 2.5 Affective Citizen Communication with Habitual and Particular Media Use

Debate
Viewing

Positive
Emotions
(Hope. Pride)

News Involvement
(Knowledge.
Media Use Participation)

Negative
Emotions
(Anxiety. Anger)

Discussion
(Politics. Public
Affairs)

The different structural path models described above presume that exogenous

factors related to emotions, communication behaviors, political knowledge and

campaign participation have already been taken into account. These variables derive

from existing models of political behavior and political knowledge, such as the civic

voluntarism model (Verba, Schlozman, & Brady, 1995) and the opportunities-

motivation-ability model (Delli Carpini & Keeter, 1996; Luskin, 1990), and from the

antecedents of communication behaviors identified by the communication mediation

model. While the influence of these exogenous forces was excluded from the previous

discussion to simplify the presentation of the affective citizen communication model, I

elaborate on them in the methods section (Chapter 3).

29
HYPOTHESES

The hypotheses of the present study are implicit in the structural models

presented in the previous figures. However, to make these hypotheses more explicit,

each structural path described in Figure 2.5 will be formally tested.

Direct Effects

HI: Emotions will be positively related to news media use controlling for

demographics and political orientations.

H2: Emotions will be positively related to debate viewing controlling for

demographics and political orientations.

H3: Emotions will be positively related to political discussion controlling for

demographics and political orientations.

H4: News media use will be positively related to political knowledge controlling

for demographics, political orientations and emotions.

H5: News media use will be positively related to political participation controlling

for demographics, political orientations and emotions.

H6: Debate viewing will be positively related to political knowledge controlling

for demographics, political orientations and emotions.

H7: Debate viewing will be positively related to political participation controlling

for demographics, political orientations and emotions.

H8: Political discussion will be positively related to political knowledge

controlling for demographics, political orientations, emotions, news media use and

debate viewing.

30
 H9: Political discussion will be positively related to political participation

controlling for demographics, political orientations, emotions, news media use and

debate viewing.

Indirect Effects

H10: The relationship between emotions and political knowledge will be

mediated, at least partially, by news media use, debate viewing and political discussion.

Hll: The relationship between emotions and political participation will be

mediated, at least partially, by news media use, debate viewing and political discussion.

H12: The relationship between news use and political knowledge will be

mediated, at least partially, by political discussion.

H13: The relationship between news use and political participation will be

mediated, at least partially, by political discussion.

H14: The relationship between debate viewing and political knowledge will be

mediated, at least partially, by political discussion.

HIS: The relationship between debate viewing and political participation will be

mediated, at least partially, by political discussion.

31
 Chapter 3: Methods

DATA SETS

The hypotheses derived from the affective citizen communication model will be

examined using the American National Election Study (ANES) 2008-2009 Panel Study

and the National Annenberg Election Survey (NAES) 2004 Debates Panel Study. As

aforementioned, there are several advantages to relying on separate data sets,

including: (1) checks on the consistency of the theorized model across different

campaign cycles; (2) variation in the time-span to measure effects; and (3) robustness of

the substantive findings to alternative operationalization of variables. Furthermore,

both the ANES and NAES studies were designed as panel surveys, allowing for stronger

causal inference than cross-sectional surveys. In addition to measuring individual

differences, panel data permits measuring change overtime in a set of variables among

the same individuals (Finkel, 1995). 1 In addition, by gauging variables at two different

points in time, alternative specifications to the proposed model can be tested and

compared to find the best-fitting structure between key variables.

Because the surveys cover different election cycles, namely, the 2004 and 2008

U.S. presidential elections, the findings of the dissertation gain in generalizability. For

instance, the Obama-McCain election campaign of 2008 was quite unique in that it was

the first election without incumbents in the primaries since 1928, was the most costly in

1 Nevertheless, as explained in Chapter 1, panel data presents some limitations for causal inference. In
addition to the problem of measurement error, there is always the possibility of reactivity when repeating
the same questions over time (e.g., changes in political knowledge induced by previous exposure to the
questionnaire, not because of other meaningful intervening process). Also, variations in the content of
both media and interpersonal communication (e.g., high news coverage of the campaign before wave 1,
followed by low news coverage during the interval between wave 1 and wave 2) may hinder the
measurement of actual change in key outcome variables. Further discussion of these limitations is
provided in Chapter 5.
32
terms of campaign spending, and saw an unprecedented level of mobilizing efforts by

campaign volunteers (Kenski, Hardy, & Jamieson, 2010). Furthermore, one of the

candidates based his central campaign message on hope—a central emotion of affective

intelligence's disposition system—and the debacle of the financial sector in September

of 2008 certainly increased voters' anxiety levels. Thus, the use of data collected during

the Bush-Kerry election campaign of 2004 offers a check on the consistency of the

findings obtained for the exceptional 2008 election. Lastly, by relying on data sets with

somewhat different operationalizations of variables, the study is able to cancel out

some of the measurement weaknesses inherent to the use of a single survey. For

instance, the ANES survey gauged news media use based on exposure measures only,

whereas the NAES survey asked both exposure and attention items. Likewise, the NAES

study gauged negative emotions only. ANES, in contrast, had items on both negative and

positive emotions.

Overview of the ANES 2008-2009 Panel Study

The 2008-2009 ANES Panel Study is a series of political and non-political surveys

of a representative sample of U.S. adult citizens. Respondents were recruited by

telephone and completed up to 21 surveys over the Internet each month between

January 2008 and September 2009. A first cohort was recruited in late 2007 using

random-digit-dialing (RDD) methods and offered $10 per month to complete surveys on

the Internet. Those without a computer and Internet service were offered a free web

appliance and free Internet service for the duration of the study. The second cohort was

recruited the same way in the summer of 2008 and asked to join the panel beginning in

September 2008. Before the first monthly survey, most respondents also completed an

online profile survey consisting primarily of demographic questions. To limit panel

33
attrition and conditioning effects, only 10 of the 21 monthly surveys contained

questions about political topics prepared by ANES. The panelists answered the ANES

questions in January, February, June, September, October, and November 2008, and in

January, May, July, and August 2009 (further details are available at DeBell, Krosnick, &

Lupia, 2010).

The present study will use data collected during September, October and

November 2008 (waves 9, 10 and 11, respectively), comprising respondents from the

first and second cohorts. 2 These three waves were selected because they cover the

official campaign period of the 2008 election as well as the post-election period.

Furthermore, the items for measuring most of the variables of the affective citizen

communication model were asked during these waves only, which prevented the use of

earlier and/or later waves. The number of completed interviews for the three waves

used in the study varied from 2,586 to 2,665. The estimated response rates (using the

American Association of Public Opinion Research, 2008, RR3 calculation) for waves 9,

10, and 11 were 26%, 26%, and 27%, respectively. 3 These rates, although relatively low,

are very similar to those reported by other organizations using RDD samples, such as the

Pew Research Center (Pew Internet & American Life Project, 2009) and, as will be

explained shortly, to the NAES surveys. The retention rate across waves was relatively

high: 93% of wave 9 respondents also completed wave 11.

A comparison of the demographics of the ANES sample with known population

parameters indicates some important differences when using the unweighted estimates

2 Specifically, interviews were conducted between September 3 and October 2, 2008 (wave 9), between

October 2 and November 3, 2008 (wave 10), and between November 5—just after Election Day—and
December 15, 2008 (wave 11).
3 Alternatively, the minimum response rate (AAPOR's RRl) for waves 9, 10 and 11 was 16%, 16% and 17%,
while the maximum response rate (AAPOR's RR5) for each of the same waves was 46%, 46% and 47%,
respectively.
34
(see Table A1 in the Appendix). In general, respondents were older and more educated.

However, few biases remain in the sample when using weighted estimates. In this case,

of the 29 statistics included in the comparison, 18 are within 2 percentage points, 7 are

between 3 and 6 percentage points, and 4 exceed 7 percentage points. Therefore,

following the recommendation of the principal investigators of the ANES 2008-2009

Panel (DeBell, Krosnick, & Lupia, 2010), all statistical analysis will be conducted using the

appropriate weights supplied with the original data files. 4

Overview of the NAES 2004 Debates Panel Study

The NAES 2004 Debates Panel is a two-wave survey of a representative sample

of U.S. adults initially interviewed for the NAES 2004 National Rolling Cross-Section

Study (RCS) before the first presidential debate of September 30, 2004, and re-

interviewed after the last debate of October 13, 2004. Respondents for the RCS were

recruited via RDD techniques and interviewed by telephone. The response rate for the

RCS was in the 22 to 25% range (Winneg, Kenski, & Adasiewicz, 2006, p. 21). The

Debates Panel consists of 1,248 completed interviews out of 3,013 respondents

contacted originally for the RCS, yielding a cooperation rate of 41%. The fielding period

for the pre-debates survey was September 20, 2004 to September 29, 2004, while the

fielding period for the post-debates survey was October 14, 2004 to October 24, 2004

(further details are available at Annenberg Public Policy Center, 2006).

The current study will use a subsample of respondents who were asked a series

of questions about negative emotions towards the candidates, knowledge of the

candidates' issue stances and political participation during the pre- and post-debates

survey phases. Using a split-sample methodology, two thirds of respondents [n = 850)

4Specifically, the cumulative late panel weight for Wave 11 will be employed. This weight is labeled
WGTL11 in the SPSS file provided in the ANES Web site, http://www.electionstudies.org
35
were randomly assigned to complete questions on emotions and knowledge; a further

random one third [n = 415) was selected to complete items on participation in campaign

activities. The complete sample was asked questions on news media use, debate

viewing and frequency of political discussion. Although the loss of statistical power due

to a reduced sample size is unfortunate, there was no other choice considering that

emotions, political knowledge and participation are key variables of the affective citizen

communication model. On the other hand, the retained sample size [n = 415) was still

large enough to accommodate the sample requirements of the statistical techniques

used in the dissertation.

A comparison of the demographics of the NAES Debates Panel sample with U.S.

Census data shows few biases in the sample when using weighted estimates (see Table

A.2 in the Appendix). Of the 23 statistics included in the comparison, 8 are within 2

percentage points, 12 are between 3 and 6 percentage points, and only 3 exceed 7

percentage points. In general, the average respondent in this panel was more educated

and more likely to be female than the average member of the population. Importantly,

the income distribution of the sample was remarkably similar to population parameters.

Therefore, all statistical analysis of the NAES data set will be conducted using post-

debates weighted estimates.

DEFINITION OF KEY VARIABLES

The conceptual diversity on what scholars mean by political knowledge and

political participation has had the unfortunate consequence of preventing a

straightforward comparison of existing research quantifying how much citizens know

about politics and how much they participate in political activities. For instance, Delli

Carpini and Keeter's (1996) index of general political knowledge is relatively easy to

36
calculate—it is the sum of correct answers to a list of political questions. Luskin's (1987)

measurement of political sophistication, in contrast, involves a complex formula to

estimate the organization of political cognitions. Using one approach over the other can

lead to different conclusions about how knowledgeable American citizens are.

Consequently, in this section I will review the strengths and weaknesses of

different conceptual definitions and measurement approaches to political knowledge

and political participation, followed by my own explanation of how these two important

political variables will be empirically examined. In addition, I will delve into the

conceptual and methodological definitions of the three main predictor variables of the

affective citizen communication model: emotions, media use and political discussion.

While there is less controversy over the meaning and operationalization of emotions

and political discussion in the literature (but see Eveland, Hively, & Morey, 2009;

Marcus, MacKuen, Wolak, & Keele, 2006), the same is not true of media use. Medium or

content, behavior or self-report, exposure, attention, retention or knowledge—all these

aspects need to be settled before pursuing research on the antecedents and

consequences of media use.

Political Knowledge

As is the case with other political variables, political knowledge is a

multidimensional concept. While the number of dimensions varies, most researchers

identify at least two. One refers to the quantity of stored political cognitions; the other

refers to the organization of stored political cognitions (Luskin & Bullock, 2004). These

two dimensions have been known under different rubrics: factual vs. structural

knowledge (Eveland, Marton, & Seo, 2004), denotative vs. connotative knowledge

(Graber, 2001), and differentiation vs. integration (Neuman, 1981), to name a few. The

37
important distinction here is that knowledge as organization represents a higher-order

type of knowledge than knowledge as quantity. The latter refers to having learned facts

and concepts about politics. The former, instead, means having learned the meanings

and inferences behind those facts and conceptual statements. In the language of

information-processing theory, factual knowledge refers to the number of nodes in

human memory; structural knowledge refers to the density of links and connections

between these nodes (Graber, 2001).

Despite the notable conceptual and operational differences between structural

and factual political knowledge, there is substantial evidence suggesting that the

quantity of political information a person holds is highly correlated with both how well

he or she has organized it and how accurate it tends to be (Eveland, Marton, & Seo,

2004; Luskin, 2003; Luskin & Bullock, 2004). This makes sense. Since memory works by

organized nodes of information (or schemas; see Graber, 2001), it is unlikely that an

individual can have a large but disorganized system of political knowledge. As a

consequence, factual measures of political knowledge may serve as rough proxies of

more sophisticated measures of knowledge. Thus, in this dissertation, measures of

factual political knowledge will be used, with the assumption that they are positively

correlated with structural political knowledge.

The challenge, of course, is to define the type of factual knowledge questions to

be used. In their oft-cited work on political knowledge, Delli Carpini and Keeter (1996)

studied three dimensions: the rules of the game, the substance of politics, and people

and parties. However, these three categories are far from being exhaustive. A review of

the literature on political knowledge during election campaigns shows that studies have

focused on knowledge about the candidates' issue stances (Drew & Weaver, 1991,

1993, 1998, 2006; Weaver & Drew, 1995, 2001), knowledge of the candidates'
38
biographical background (Chaffee, Zhao, & Leshner, 1994; Feldman & Price, 2008) and

knowledge of presidential campaign endorsements (Holbert, 2005). The choice of

questions should be defined in terms of the purpose these questions serve. In the case

of a study trying to uncover the learning effects of affective and communicative

processes occurring at a specific time frame such as an electoral campaign, it is

preferable to use items for which the correct answer depends upon recent exposure to

information. With general civic questions, one need only remember civics lessons and

high school textbooks. With domain-specific questions—to use the terminology of Delli

Carpini and Keeter (1993)—news use, political discussion, and general attention to

campaign materials become the more important source. Therefore, domain-specific, not

general, knowledge items will be used in the operationalization of political knowledge.

Settling for domain-specific questions, however, implies defining the number of

domains to be asked about. There is some debate over whether or not to distinguish

among different domains of factual questions. The literature on issue publics (Carmines

& Stimson, 1989) indicates that some groups of the population are especially informed

about particular domains, be they the environment, civil rights or religion. In this

context, measuring political knowledge requires asking about a variety of topics and

issues. This is why a valuable approach is to rely on questions assessing candidates' issue

stands. By inquiring about different policy realms, the responses of individuals who

belong to different issue publics have a higher likelihood of being taken into

consideration.

The review of the existing literature on political knowledge has addressed

differences in both conceptual and empirical approaches. While not exhaustive, this

effort in concept explication is a necessary step for providing a proper definition and

operationalization of political knowledge as applied in the current study. An oft-cited

39
definition of political knowledge is Delli Carpini and Keeter's (1996): "The range of

factual information about politics stored in long-term memory" (p. 10). While such a

definition has the advantage of distinguishing political knowledge from values, attitudes

and behaviors, it is too broad. Nevertheless, one can elaborate on Delli Carpini and

Keeter's definition by specifying that political knowledge should be useful information

on which to base decisions to participate and vote. In formal terms, the concept

advanced here could be summarized as follows:

Political knowledge is the range offactual information about politics stored in

long-term memory that is useful for citizens to make decisions that represent
their interest.

In the context of electoral campaigns, the most important decision for citizens is

to vote in accordance with their own preferences and priorities (Lau, Andersen, &

Redlawsk, 2008). Arguably, knowing the stances of the candidates in the issues that are

important for voters should facilitate decision-making. Thus, political knowledge will be

operationalized as knowledge of the issue positions of the main presidential candidates.

By using candidate-issue knowledge, we are employing a domain-specific approach,

which is the preferable approach when studying electoral campaigns and the role

played by information sources, such as the news media and social networks. Because

this study relies on survey data, multiple-choice, close-ended questions will be used to

compute an index of political knowledge. 5 Detailed information as to the

operationalization of knowledge is provided in the measures section below.

5 Settling for close-ended questions that tap factual knowledge about specific policy domains does not

resolve an important problem of assessing political knowledge, and that is the problem of coding. In the
last few years, political scientists have debated over what to do with those respondents who chose "don't
know" categories in multiple- choice knowledge questions. Some researchers, such as Delli Carpini and
Keeter (1996) recommend the encouragement of don't know responses. Others, such as Mondak (2001),
have argued vehemently against this practice, arguing that it contaminates the efforts to obtain valid
knowledge indexes. For instance, it has been found that respondents' propensity to guess and willingness
to give correct answers is not constant across the population (Mondak & Halperin, 2008). Other research,
40
Political Participation

Political participation is a slippery concept; several interpretations have been

given of what exactly is meant by this construct. Some researchers identify political

participation with electoral activities, such as voting and working for political parties

(e.g., Conway, 1985). Recognizing that participation goes beyond elections, others have

included in their measures activities such as working for the community and attending a

protest (e.g., Verba, Schlozman, & Brady, 1995). Even processes such as media use and

news attention have been identified as markers of participation (e.g., Zaller, 1992). The

heterogeneity of conceptual definitions has led, inevitably, to a cacophony of empirical

measures of participation and thus to contradictory results regarding the antecedents

and consequences of political participation.

Identifying the types of activities that fall under the rubric of political

participation requires, first of all, a normative definition. As Teorell (2006) noted, the

literature has been dominated by three different conceptualizations of political

participation:

however, has not found support for this finding and, actually, has argued that promoting don't know
responses is a better approach (Sturgis, Allum, & Smith, 2008). Even if one decides to encourage or
discourage don't know responses, there is the issue of what do to with incorrect responses when creating
summated indexes of political knowledge. Some researchers distinguish between not answering or not
knowing the correct response to a question from providing incorrect answers, so as to correct for
propensity to guess and for misinformation—holding incorrect factual information and believing in it
(Kuklinski, Quirk, Jerit, Schwieder, & Rich, 2000). Thus, incorrect answers are sometimes subtracted from
correct answers, while missing values and don't knows are scored as zero. While there are merits for this
approach, common practice in political communication research is to just compute correct answers while
leaving all other options, incorrect or missing, as zero (e.g., Drew & Weaver, 2006; Eveland & Thomson,
2006). Because this dissertation is based on a secondary analysis of existing ANES and NAES surveys, there
were constrains to the scoring approach of political knowledge indexes. For instance, the knowledge
questions of the ANES Panel Study relied on Mondak's recommendations by encouraging respondents to
guess and not providing a don't know response choice. Thus, by common practice and by force, this
dissertation will compute knowledge scores by counting only correct answers—all other responses will be
scored as zero.
41
1. Participation as influence: This view stems from the classic work of Verba and Nie

(1972) and Milbrath and Goel (1977) and views participation as "an instrumental act

through which citizens attempt to make the political system respond to their will"

(Teorell, 2006, p. 789). Based on notions of democratic representation and

government responsiveness to public opinion, this view understands participation as

"activity that has the intent or effect of influencing government action—either

directly by affecting the making or implementation of public policy or indirectly by

influencing the selection of people who make those policies" (Verba, Schlozman, &

Brady, 1995, p. 38). The best example of participation as influencing attempts is

voting: when people vote, they choose politicians who are supposed to deliver what

they want in policymaking (Wlezien & Soroka, 2007). However, other types of

campaign behavior, such as attending rallies by candidates and donating money to

political parties, are also indicators of political participation as influencing.

2. Participation as decision making: This view stresses citizen behavior that results in

direct policymaking, rather than indirect policymaking via influence on government

officials. In representative democracies, the realm of direct decision making by

citizens is somewhat constrained to local issues, so active participation in local

affairs is what participation as decision making is mostly about. Barber's (1984)

definition of political participation falls within this paradigm: "[P]olitics in the

participatory mode (...) is self-government by citizens rather than representative

government" (as cited in Teorell, 2006, p. 790). The important distinction here is

that participation is regarded as direct involvement in the policymaking process, not

delegation into a representative body. Belonging to a city council, working in

neighborhood associations and, in general, participating in institutions that have the

power to decide policies, are indicators of participation as decision making.

42
3. Participation as deliberation: Based on the idea of deliberation and deliberative

democracy (Cohen, 1989; Dryzek, 2000; Fishkin, 1991; Gastil, 2008; Gutmann &

Thompson, 1996; Habermas, 1996), some political theorists have construed political

participation as involvement in a collective process of discussion. Through

interpersonal (and intrapersonal) processes of reasoning and argument exchange,

people who participate are able to express their public concerns, learn about

important issues affecting their communities and, eventually, reach some kind of

decision on how policymakers should move forward (Conover & Searing, 2005;

McLeod et al., 1999). The best example of participation as discussion is the

Deliberative Polling project, which creates a formal instance of deliberation for

citizens (Fishkin & Luskin, 2005; Luskin & Fishkin, 2002).

As Teorell (2006) noted, participation as influencing attempts has dominated the

literature on political participation in the U.S. and elsewhere, to the point that textbooks

(Brady, 1999) and major reviews (Delli Carpini, Cook, & Jacobs, 2004; Milbrath & Goel,

1977) rely on this definition exclusively. The current study will not depart from this

tradition. This is warranted for two reasons. First, citizens' involvement in direct

decision-making processes is relatively infrequent, at least when compared to citizens'

participation in elections and other traditional modes of influencing policymaking.

Second, citizens' deliberative behaviors, even of the informal kind such as talking about

politics with friends and family, is treated in the affective citizen communication model

as an antecedent of participation, not a constituent form of it. Hence, the participation

as deliberation model is inapplicable to this study. The formal definition adopted here,

then, stresses that

Political participation is the breadth of political activities carried out by citizens to

influence the selection of people who make government policies.

43
 From a normative point of view, citizens' participatory behavior, like trust, is

regarded as vital for the good functioning of democratic systems. When people

participate, they have a voice in public affairs, can hold authorities accountable and

empower themselves to act on their own behalf (Burns, Schlozman, & Verba, 2001).

Still, participation, like trust, is not automatically conducive to democratic governance.

Those who support the "limited citizenship" model, for instance, argue that too much

participation can obstruct and complicate governance by delaying, politicizing and

oversimplifying policy problems that require skilled leadership and expertise (damson,

2001, p. 56). Most scholars, however, agree that the problem of most democracies in

the 21 st Century is not of an excess of participation but of stagnation or outright decline

of it, particularly among young cohorts (Putnam, 1995b).

The most traditional approach to measuring political participation in survey

research is to dummy-code a host of participatory behaviors and, subsequently,

combine the discrete behaviors into an index. Some researchers separate participation

into voting and everything else (Bimber, 2001), or subdivide the participation index into

acts inside and outside of the political system (Kim, Wyatt, & Katz, 1999), acts that have

a potential of public confrontation or not (Mutz, 2002), and several studies examine

individual participation acts as distinct dependent variables (Verba, Schlozman, & Brady,

1995). As Dylko (2010) noted, indexes gauge the breadth of activities in which

respondents have taken part—regardless of their specific nature. Therefore, in the

current study participation will be measured as an additive index of political campaign

activities. This is warranted for two reasons. First, it is consistent with previous work

that has tracked political participation back to emotions (Brader, 2005; MacKuen,

Wolak, Keele, & Marcus, 2010; Parsons, 2010b; Valentino, Gregorowicz, & Groenendyk,

2009; Wolak, MacKuen, Keele, Marcus, & Neuman, 2003). Second, it is in line with the
44
measurement approach to participation adopted by scholars working with

communication mediation models (Cho, Shah, McLeod, McLeod, & Scholl, 2009; Gil de

Zuniga, Puig-i-Abril, & Rojas, 2009; Gil de Zuniga & Valenzuela, Forthcoming; Shah et al.,

2007; Valenzuela, Kim, & Gil de Zuniga, Forthcoming).

Emotions

There are a variety of labels used in political communication to refer to affective

processes. Terms such as affect, moods, emotions and feelings are sometimes used

interchangeably (e.g., Holbert & Hansen, 2006; Neuman, Marcus, MacKuen, & Crigler,

2007). However, these terms are not exact synonyms. A review of the distinctions made

by Brader (2006b, p. 51), Parsons (2010a, pp. 34-36), and Thoits (1989, pp. 318-319),

yields the following definitions:

1. Mood: A diffuse, long-lasting positive or negative mental state that is not

attributable to a specific attitude object. For instance, political scientists speak of a

"liberal policy mood" (Stimson, 1999) to describe American voters' positive

dispositions towards liberal governmental policies (e.g., abortion, gun control, gay

rights, etc.)

2. Emotion: A specific mental and/or physiological (e.g., increased heart rate)

disposition triggered in response to the perception or appraisal of an external

stimulus. It is generally attributable to a specific attitude object (e.g., when a voter is

anxious at the prospects of Candidate A winning the election).

3. Feelings: Refers to the awareness and experience of emotions as well as basic

human drives (e.g., pleasure, pain, fatigue). Contrary to moods but in a similar

fashion as emotions, feelings are typically targeted at a specific attitude object (e.g.,

45
 when a voter says: "Candidate A makes me feel anxious"). In other words, feelings

are the subjective interpretation of emotions.

4. Affect: A catch-all term that encompasses phenomena such as moods, emotions,

feelings and basic human drives (e.g., pain and pleasure). Scholars in the social

sciences tend to use the term affect in opposition to the term cognition, as in the

scholarly debate between the primacy of affect and the primacy of cognition

(Lazarus, 1984; Zajonc, 1984).

Considering that this project deals with responses elicited by specific attitude

objects, namely, presidential candidates, I will avoid the term mood. Feelings, however,

will be used interchangeably with emotions, as the existence of an emotion is a

necessary condition for a feeling. Affect will also be employed when referring to the

general influence of emotions and feelings on communication behavior, political

learning and political participation.

An important issue that remains to be addressed is how to operationalize

emotions. Social scientists still debate about the proper structure of emotion. Early work

on social and political psychology arranged emotions using a single valence dimension,

such as positive-negative, like-dislike or approach-avoidance (e.g., Brady & Sniderman,

1985). Other researchers have advanced two-dimensional models of emotions

(Cacioppo, Berntson, Klein, & Poehlmann, 1997; Russell, 1980; Watson & Tellegen,

1985), although there is disagreement on the characteristics of these two dimensions.

Russell (1980) has argued that one dimension determines the valence of emotional

experience while the other determines the level of arousal. Such a structure of emotions

allows researchers to distinguish, say, the effects of being terrified from being afraid

(i.e., more negative and aroused), in addition to the more obvious distinction between

being afraid and being elated. Another dual model of emotions, often associated with
46
the work of psychologists Tellegen and Watson (Tellegen, Watson, & Clark, 1999;

Watson, Clark, & Tellegen, 1988; Watson & Tellegen, 1985; Zevon & Tellegen, 1982),

posits that each channel performs a unique function. One dimension gauges the novelty

or threat of environmental stimuli, while the other evaluates the performance of

familiar routines and behaviors. Contrary to the valence-arousal model of emotion,

these two dimensions are viewed as orthogonal to each other. In this sense, affective

intelligence theory clearly sides with Tellegen and Watson's model of the structure of

emotions because it posits that the surveillance and disposition systems operate in

parallel (a similar argument is made by Valentino, Hutchings, Banks, & Davis, 2008).

Considering that this dissertation is based on affective intelligence theory, a

natural choice is to create separate measures for positive and negative emotions,

treating both as orthogonal to each other. Nevertheless, previous work has found that

emotions towards political candidates do not always fit this dual pattern. For instance,

an analysis of the emotional reactions to then president Bill Clinton found that anger

and aversion constituted a separate dimension from enthusiasm and anxiety (Marcus,

MacKuen, Wolak, & Keele, 2006). Other work has found that the effects of anger and

fear on both political knowledge and participation are quite different (Huddy, Feldman,

& Cassese, 2007; Valentino, Banks, Hutchings, & Davis, 2009). Thus, as was mentioned in

Chapter 2, the issue of measurement of emotions was tackled empirically using

exploratory factor analysis techniques (i.e., principal component analysis), the results of

which are addressed in further detail in the measures section below.

News Media Use

Theories of media effects rest on the assumption that people are influenced by

media messages, either directly through exposure to media content or indirectly

47
through people's conversations about it. In the context of political communication

research, news reception is of particular interest, making its empirical measurement a

critical step in studies that examine various outcomes such as political knowledge and

political participation. Considering the central role assigned to this concept, its lack of

consistent measurement is somewhat surprising. Exposure to, attention to, reliance on,

and relative preference for news—all these have been used to gauge individuals' use of

news media (Chaffee & Schleuder, 1986; Drew & Weaver, 1990; McLeod & McDonald,

1985; Price & Zaller, 1993; Prior, 2007; Slater, 2004).

Part of the multiplicity of instruments to capture citizens' news consumption is

due to the fact that experiencing news requires a behavioral component (exposure), a

cognitive component (attention), and, in many instances, an attitudinal component,

such as a motivation to follows news. In McGuire's (1972) information processing

model, exposure is a prerequisite for attention, comprehension, acceptance and

retention of media messages. Thus, exposure is a necessary but not sufficient condition

for media effects. And, yet, there is evidence that the different dimensions of news

media use are strongly correlated with each other (Eveland, Hutchens, & Shen, 2009).

People who are more exposed to news tend to also pay more attention to it and,

consequently, learn and retain more media information. In fact, several scholars rely on

these relationships to justify the use of exposure items only. Thus, to the extent that the

ANES and NAES surveys contained measures of exposure and attention to media

information about the campaigns, both will be taken into account.

One common way of measuring media exposure is through survey items asking

global assessments of exposure, such as time spent online or frequency of watching

television. This is problematic because these measures miss completely the specificity of

exposure to actual media content, with the subsequent problem of underestimating

48
effect sizes. Putnam's infamous indictment of television as the culprit of civic

disengagement in America (Putnam, 1996) and Kraut et al.'s (1998) dismissal of the pro-

social role of the Internet are good examples of the pitfalls of relying on global

assessments of exposure. A better approach, in this case, is asking for exposure to

specific media content (news, entertainment, etc.), following the approach of the uses

and gratification tradition (Katz & Gurevitch, 1974). Existing research shows that

informational uses of media have a pro-social effect, increasing people's political

knowledge and participation (Shah, Cho, Eveland, & Kwak, 2005). Thus, exposure to

campaign news or public affairs news will be used in the current study, rather than time

spent using a particular medium.

Compared to news exposure, the measurement of news attention is not as

straightforward. This is because "attention is a covert mental activity occurring within

the 'black blox' of a person" (Chaffee & Schleuder, 1986, p. 77). In the case of surveys,

attention has been gauged by directly asking people their level of attention to news

and/or public affairs content using Likert-type scales. As it happens, this is also the

approach adopted by both the ANES and NAES surveys.

A final issue with measurement of news use refers to the combination of

exposure and attention. Should these measures be channel- specific (text vs. audiovisual

content), medium-specific (cable TV vs. national TV networks, or print newspapers vs.

online news) or content-specific (generic categories such as "news" or "foreign affairs"

vs. particular news events, such as "presidential elections" or "the Iraq War")? While

previous work has compared these different approaches (e.g., Chaffee & Schleuder,

1986), I believe this issue is dependent upon the specific purposes of the research. If the

project seeks to identify the contribution of the news media in general to individual's

political involvement, as is the case in this dissertation, then generic measures of news
49
exposure and attention may suffice. Conversely, if the interest is on the impact of the

coverage of a specific issue such as health reform on individuals' behaviors, then

attention and exposure to particular news content would be more appropriate.

Therefore, I will measure overall exposure and attention to news combined because a

central purpose of the study is to gauge the overall effect of emotions on information-

seeking in the media.

Political Discussion

According to Conover and colleagues (Conover & Searing, 2005; Conover,

Searing, & Crewe, 2002; Searing, Solt, Conover, & Crewe, 2007), there is (1) structured

deliberation, which usually takes place in formal, public settings, following specified

rules and procedures, such as Congress, juries and deliberative polls; (2) informal public

discussion, which is more informal and less structured but also takes place in public

settings, such as in political parties, interest groups, work, churches, school boards and

town hall meetings; and (3) casual political talk, which is also informal and unstructured,

but takes place in private settings, such as a family dinner and with friends.

This dissertation deals mainly with research that has taken place in the latter two

settings (i.e., a focus on the individual or interpersonal level). Specifically, I will borrow

the definition advanced in the Encyclopedia of Political Communication (Schmitt-Beck,

2008, p. 341) and conceptualize political discussions as:

Episodes of political conversation and discussion that take place between the
non-elite members of a political community.

Two problems arise here. First, what is a political conversation? If I go to the

supermarket and then I discuss with my wife the rising cost of produce, does that count

as political discussion? Or do I need to blame government to make it political? Second,

50
what counts as discussion? Mansbridge (1999, p. 214) refers to a "snort of derision" in

front of friends in response to a television character's sexist behavior to be "a political

act." There is reason to suspect that scholars' definition of politics differs from that of

most people. In this regard, Walsh (2004, pp. 38-41) found that the participants in her

study thought of politics as consisting of elections, elected officials and political parties,

and did not recognize political discussion even when they were engaging in it. It could

be argued that political discussion is whatever it means to people, which would be fine if

perceptions of being involved in a political discussion were of interest to researchers.

But, as Eveland and colleagues (2009) noted, if the actual act of being involved in a

political discussion is the variable of interest, this response is insufficient. On the other

hand, what qualifies as discussion or not is also of importance.

Therefore, I will also borrow from Rafaeli's (1988) concept of interactivity, which

refers to the extent to which communication transcends reaction and involves

reciprocal exchange messages (i.e., messages must recount the relatedness of earlier

messages). Thus, neither shouting at the TV in response to a sexist message nor a mere

answer to a question about party ID would count as discussion. Rather, political

discussion will be measured as the frequency in which respondents have engaged with

other people in conversations about public affairs, including elections, government and

news.

MEASUREMENT OF VARIABLES

The affective citizen communication model as it will be tested here contains the

four groups of theoretical endogenous variables explained in the previous section and a

host of exogenous variables included as controls. Drawing from studies on affective

intelligence (Brader & Valentino, 2006), uses and gratifications (Eveland, 2004; Shah,

51
Rojas, & Cho, 2009), political knowledge (Delli Carpini & Keeter, 1996; Luskin, 1990) and

political behavior (Rosenstone & Hansen, 1993; Verba & Nie, 1972; Verba, Schlozman, &

Brady, 1995), the antecedents common to emotions, communication and involvement

revolve around three classes: (1) resources (e.g., education, income, self-efficacy); (2)

incentives (e.g., interest in politics); and (3) identity (e.g., party identification).

Therefore, these factors have been included in the analyses as controls to strengthen

claims that the relationships between the variables of the affective citizen

communication model are not spurious.

All variables were measured in both the ANES and NAES surveys, though their

operationalization was not identical. Consequently, the following two sections describe

the measurement of the exogenous variables and theoretical constructs, separately for

each data set.

Variables from the ANES 2008-2009 Panel Study

Demographics

The analyses on the ANES data set included seven exogenous variables

measured at wave 9—the first of the waves of ANES data used in the current study. The

measures of age (M = 47.47 years, SD = 16.89 years) and gender (52.5% females) were

relatively straightforward. Education levels were assessed on a 5-point scale ranging

from "no high school diploma" to "graduate degree" (M = 2.88, SD = 1.12, Mdn = 3.00).

Respondent's income was measured using a non-linear 19-point scale which increased in

$2,500 increments between incomes of below $5,000 and $14,999, in increments of

$5,000 for incomes between $15,000 and $39,999, in $10,000 increments for incomes

between $40,000 and $59,999, in increments of $15,000 for incomes between $60,000

52
and $99,999, and in increments of $25,000 for incomes higher than $100,000 (Mode =

$60,000 to $74,999, Mdn = $50,000 to $59,999).

Political Orientations

In addition to demographics, three important political orientations measured

also at Wave 9 were included as exogenous in the statistical analyses. Internal political

efficacy was tapped with the following question: "How much can people like you affect

what the government does?" Responses were measured on a 5-point scale, ranging

from "a great deal" to "not at all" (reverse coded, M = 2.65, SD = 1.04). For strength of

partisanship, respondents' party identification was measured using the typical 7-point

scale ranging from strong Democrat to strong Republican, with the midpoint being true

independent (the other categories were weak Democrat, independent leaning

Democrat, independent leaning Republican, weak Republican). This item was folded into

a 4-point scale, ranging from "no partisanship" to "strong partisanship" (M = 1.90, SD =

1.06). Habitual interest in politics was measured on a 5-point scale, ranging from

"extremely interested" to "not interested at all" (reverse coded, M = 2.56, SD = 1.03).

Measures of Emotions

For the first component of the affective citizen communication model, emotions,

several measures were constructed using questions from waves 9 and 11 about the

intensity with which respondents felt several emotions about the presidential

candidates Barack Obama and John McCain. Respondents were asked how angry,

hopeful, afraid and proud had each candidate made them feel using a 5-point scale

ranging from "extremely" to "not at all," with the midpoint being "moderately." For

easier comparison with the items on emotions from the NAES survey, responses were

53
recoded to range from 0 to 1, with higher values for more intense feelings of emotions.

Descriptive statistics of these items are presented in Table 3.1.

Table 3.1 Descriptive Statistics for Emotion Items in the ANES 2008-2009 Panel

M SD N Min Max
Wave 9 data
Emotional reaction to Obama
Angry .20 .31 2,304 0
Hopeful .37 .35 2,304 0
Afraid .31 .35 2,304 0
Proud .35 .35 2,303 0
Emotional reaction to McCain
Angry .17 .28 2,304 0
Hopeful .33 .30 2,304 0
Afraid .23 .30 2,303 0
Proud .39 .34 2,304 0
Wave 11 data
Emotional reaction to Obama
Angry .18 .30 2,309 0
Hopeful .47 .37 2,310 0
Afraid .28 .34 2,310 0
Proud .45 .38 2,306 0
Emotional reaction to McCain
Angry .16 .27 2,309 0
Hopeful .30 .29 2,308 0
Afraid .18 .28 2,308 0
Proud .40 .33 2,305 0

In order to have single measures for each emotional reaction, scores of total

anger, hope, fear and pride were computed for all respondents by averaging each pair

of emotion across both candidates (for a similar methodological choice, see Marcus &

MacKuen, 1993). Subsequently, these four measures of emotion were subjected to an

54
exploratory factor analysis, to uncover the structure that best fit the data. This exercise

was conducted separately for both Wave 9 and Wave 11 data. In both waves, the

rotated solution suggested the existence of two orthogonal dimensions (details are

displayed in tables A.3 and A.4 in the Appendix section). 6 More specifically, positive

emotions loaded strongly on one dimension, while negative emotions loaded equally

strongly on another dimension, a structure that matches the dual-channel structure

advanced by Watson et al. (Watson & Clark, 1997; Watson, Clark, & Tellegen, 1988;

Watson & Tellegen, 1985, 1999). Affective intelligence theory also provides a theoretical

rationale for conceptualizing positive and negative emotions as orthogonal to each

other, following the nature of the disposition and surveillance systems. In this case, the

two factors indentified in the principal component analysis can also be labeled

enthusiasm and anxiety, respectively.

Considering the results of the exploratory factor analysis, a scale of negative

emotions was constructed averaging items on anger and fear (Cronbach's a = .77, M =

.23, SD = .18 for Wave 9; Cronbach's a = .81, M = .20, SD = .18 for Wave 11), while

positive emotions was the average of measures of pride and hope (Cronbach's a = .81,

M = .36, SD = .17 for Wave 9; Cronbach's a = .83, M = .40, SD = .19 for Wave 11).

6 Based on the recommendations of previous work on political emotions (Brader & Valentino, 2006;
Marcus, MacKuen, Wolak, & Keele, 2006), the results of both unrotated and rotated solutions of principal
component analysis are presented in the Appendix in Table A.3 and Table A.4, respectively. The two
unrotated solutions presented in Table A.3 replicate the same pattern, with the emergence of two factors
with eigenvalues greater than 1.0. Clearly, the first factor represents a valence dimension, with positive
and negative emotions loading highly but with opposite signs. The second factor contributes nearly as
much explained variance as the first one, but in this case all four variables load on it in the same direction.
Arguably, the second factor may represent arousal, that is, the intensity with which respondents feel
something toward candidates Obama and McCain (for a similar choice of factor labels, see Brader &
Valentino, 2006). The structure of emotions suggested by the unrotated solution consistent with the
valence-arousal model identified by Russell (Russell, 1980; Russell & Bullock, 1985; Russell, Weiss, &
Mendelsohn, 1989). However, I will rely on the rotated solution presented in Table A.4 because it is the
most consistent with affective intelligence theory.
55
Measures of Communication Behavior

Habitual news media use was an index (M = 14.14, SD = 6.27 for Wave 9; M =

14.02, SD = 6.31 for Wave 10) constructed by averaging four items tapping the number

of days in a typical week that the respondent was exposed to news on television, radio,

the Internet, and printed newspapers, not including sports. Debate viewing was

measured in wave 11 as the number of McCain-Obama debates watched by the

respondent on television (range 0 to 3, Mode = 1, Mdn = 2). 7 For political discussion,

respondents were asked the frequency with which they talked "about politics with

family or friends" in a typical week. The response scale was identical to the scale of the

items on news media use [M = 3.17, SD = 2.11 for Wave 9; M = 3.22, SD = 2.17 for Wave

11).

Measures of Political Involvement

The two final endogenous or outcome variables were political knowledge and

political participation. A scale of candidate issue knowledge combined 20 items

measuring respondents' knowledge of the issue positions of both McCain and Obama.

Each correct response was coded as 1, while all other responses (i.e., don't knows,

incorrect responses, refusals) were coded as 0. 8 Subsequently, the items were added

and divided by 20 so that the scale run from 0 to 1(Cronbach's a = .82, M = .54, SD = .23

7 The three debates took place on September 26, October 7 and October 15, 2008.
8 The ANES staff followed Mondak's (2001) recommendations and explicitly instructed respondents to
provide their best estimate even if they were not completely sure about the best answer. Thus, a "don't
know" response choice was not available to respondents. In any case, to check that the results of the
analyses are not an artifact of using a dichotomous scoring scheme, another scale was constructed from
items that were recoded to correct for guessing. That is, correct responses were scored as 1, incorrect
responses were scored as -1, and all other responses (refusals, missing, etc.) were scored as 0. Preliminary
analyses revealed that there were no substantive difference between the "truncated" index and the
"corrected for guessing" index. However, because endogenous variables will be standardized to a 0 to 1
scale in the multivariate analyses, the truncated scale explained in the text will be used since it does not
contain negative scores.
56
for Wave 9; Cronbach's a = .81, M = .58, SD = .22 for Wave 11). The issues were same-

sex marriage, federal income taxes, government funding of prescription drugs, health

care, trial of terrorist detainees, government wiretapping, illegal immigration, global

warming, and environmental regulations of automakers. The list of issues combined

candidates' stances on thoroughly discussed topics (e.g., taxes, illegal immigration) with

more obscure issues (e.g., regulations for automakers, wiretapping), providing a rich

measure of respondents' political knowledge.

Campaign political participation was measured as an additive scale of

dichotomous items measuring whether respondents had ever: (1) joined a protest,

march, rally, or demonstration; (2) signed a political or social petition on the Internet;

(3) signed a political or social petition on paper; (4) given money to an non-religious

organization concerned with political or social issues; (5) attended a meeting to discuss

political or social concerns; (6) invited people to attend a meeting about political or

social issues; and (7) distributed information or advertisements supporting a political or

social interest group. Affirmative responses were coded as 1 and negative responses

were coded as 0. Subsequently, the items were added and divided by 7 to have a scale

running from 0 to 1(Cronbach's a = .75, M = .39, SD = .29 for Wave ll). 9

Variables from the NAES 2004 Debates Panel Study

Demographics

The analyses of the NAES Debates Panel set includes the same exogenous

variables used with the ANES Panel, in this case all measured in the pre-debates wave.

9 While these participation items were asked in Wave 11, they were not asked in Wave 9, as all the other
variables detailed above. This does not represent a problem for the cross-sectional analyses using Wave
11 data only. However, for the longitudinal analysis, I had no other choice than to exclude participation
from the analysis.
57
Age (M = 47.17, SD = 16.12) was measured in years and gender (55.9% females) was a

dummy variable, with males as the reference category. Education was gauged on a 9-

point non-linear scale ranging from "grade 8 or lower" to "graduate or professional

degree", with the median category being "some college, no degree." To make the scale

of the variable comparable to the measure of education used in the ANES data set, it

was recoded into a 5-point scale ranging from "less than high school" to "graduate

degree" [M = 3.06, SD = 1.19, Mdn = 3.00). Income was measured using a non-linear 9-

point scale which increased in $5,000 increments between incomes of below $10,000

and $34,999, in increments of $15,000 for incomes between $35,000 and $49,999, in

$25,000 increments for incomes between $50,000 and $149,999, and in increments of

$50,000 for incomes higher than $150,000 (Mode = $50,000 to $74,999, Mdn = $35,000

to $49,999).

Political Orientations

There were three political orientations treated as exogenous variables in the

analyses. Internal political efficacy was measured on a 5-point scale ranging from

"strongly agree" to "strongly disagree" with the statement "politics seems too

complicated to understand what is going on" (M = 3.21, SD = 1.56). For strength of

partisanship, respondents were asked about their party identification. Four response

choices were given: Republican, Democrat, independent, and something else. These

responses were recoded into a dummy variable, where Republicans and Democrats

were coded as 1, and independents and something else were coded as 0 (60.9%

identified with a party). Habitual interest in politics was a measure of frequency of

following what is going on in government and public affairs using a 4-point scale,

ranging from "most times" to "hardly at all" " [M = 3.21, SD = 0.86).

58
Measures of Emotions

As for the first component of the affective citizen communication model,

emotions, the NAES survey contained measures gauging negative feelings only.

Specifically, the survey asked respondents the intensity with which George W. Bush and

John Kerry made them feel angry, uneasy and afraid using a 5-point scale ranging from

"yes, very [angry/uneasy/afraid]" to "does not make me [angry/uneasy/afraid]." The

middle category was "little." In a similar fashion as the recoding of items on emotions

from ANES data set, responses were recoded to range from 0 to 1, with higher values

for more intense feelings of emotions. Descriptive statistics of these items are

presented in Table 3.2.

Table 3.2 Descriptive Statistics for Emotion Items in the NAES 2004 Debates Panel

M SD N Min Max
Pre-debates wave data
Emotional reaction to Bush
Afraid .31 .42 800 0 1
Uneasy .42 .44 799 0 1
Angry .37 .43 801 0 1
Emotional reaction to Kerry
Afraid .24 .37 790 0 1
Uneasy .40 .43 793 0 1
Angry .21 .35 800 0 1
Post-debates wave data
Emotional reaction to Bush
Afraid .31 .42 795 0 1
Uneasy .42 .44 794 0 1
Angry .36 .43 796 0 1
Emotional reaction to Kerry
Afraid .25 .39 793 0 1
Uneasy .37 .43 793 0 1
Angry .21 .37 794 0 1

59
 In order to have single measures for each emotional reaction, scores of total

fear, uneasiness, and anger were computed by averaging each pair of emotion across

both candidates. In line with the structure of emotions examined using the ANES data

set, the three negative emotions gauged by the NAES formed a single dimension when

conducting an exploratory factor analysis (see Table A.5 in the Appendix section). Thus,

a scale of negative emotions was developed by averaging the measures of fear,

uneasiness and anger (Cronbach's a = .72, M = .32, SD = .20 for the pre-debate wave;

Cronbach's a = .75, M = .32, SD = .19 for the post-debate wave).

Measures of Communication Behavior

There were three endogenous communication variables measured at both the

pre- and post-debate waves, including media use, debate viewing and political

discussion. Habitual news media use was an index created using measures of exposure

and attention to news content, as suggested by previous research (Chaffee & Schleuder,

1986; Eveland, Hutchens, & Shen, 2009). Exposure was an average of four items tapping

the number of days in the previous week that the respondent was exposed to national

network news, cable news channels, local television news, and a daily newspaper.

Responses were recorded on an 8-point scale ranging from 0 to 7 days. For attention to

news, respondents were asked to report how much attention they paid to stories (for

national network or cable television news and local television news) and articles (for

newspapers) about the campaign for president in the past week. A 4-point response

scale was used for these items, ranging from "a great deal" to "none." Subsequently, the

three items of attention were averaged. To create a single scale of news media use,

measures of exposure and attention were standardized into a common metric running

60
from 0 to 1 and averaged (Cronbach's a = .78, M = .53, SD = .23 for the pre-debate

wave; Cronbach's a = .82, M = .49, SD = .25 for the post-debate wave).

Debate viewing was measured in the post-debate wave as amount of exposure

to the first Bush-Kerry debate of September 20, 2004 using a 4-point scale ranging from

"yes, all" to "did not watch" (range 1 to 4, Mode = 4, Mdn = 3). 10 An index of political

discussion was created by averaging respondents' answers to questions about the

frequency with which they discussed politics with family and friends and with co­

workers in the previous week. The response scale was identical to the scale of the items

on exposure to news media (M = 2.66, SD = 1.98 for the pre-debate wave; M = 2.78, SD

= 2.02 for the post-debate wave).

Measures of Political Involvement

The last two endogenous variables were related to political knowledge and

political participation. Candidate issue knowledge was a scale of seven items measuring

respondents' knowledge of the policy stances of Bush and Kerry. Respondents were

queried about tax policy (e.g., tax cuts for companies), social issues (e.g., abortion),

health care (e.g., government health insurance), and Social Security (e.g., investment of

pension funds in the stock market). Responses were coded as 1 if they were correct, 0

otherwise. As was the case with the measure of political knowledge used with the ANES

data, the items were added and the divided by 7 so that the scale ran from 0 to 1

(Cronbach's a = .65, M = .57, SD = .28 for the pre-debate wave; Cronbach's a = .63, M =

.70, SD = .26 for the post-debate wave). Although this measure covers fewer issues than

10 The post-debate questions of news media use and political discussion queried respondents engaging in
either activity over "the previous week." A majority of all respondents in the debates panel data set were
interviewed within a week after the third and final Bush-Kerry debate, and a still sizeable percentage was
interviewed within a week after the second Bush-Kerry debate. Consequently, the only way to maintain a
consistent temporal ordering of the pre-debate, debate viewing, and post-debate measures while
retaining a decent sample size was to work with the first debate viewing measure only.
61
those covered in the ANES surveys, it combines salient topics of the campaign with less

prominently discussed topics.

Campaign political participation was also measured as an additive scale of

dichotomous items measuring if during the presidential campaign respondents had: (1)

gone to any political meetings, rallies, speeches, dinners or events in support of a

presidential candidate; (2) given money any of the presidential candidates; (3) done any

type of work for a candidate; (4) tried to influence the way other people vote; and (5)

worn a presidential campaign button, put a campaign sticker on the car, or displayed a

sign on their property. Subsequently, the scale was divided by 5 so as to have a measure

with a 0 to 1 range (Cronbach's a = .62, M = .19, SD = .23 for the pre-debate wave;

Cronbach's a = .66, M = .21, SD = .24 for the post-debate wave).

STATISTICAL ANALYSIS

Structural Equation Modeling

The affective citizen communication model advanced in the current study is the

result of a series of hypothesized relationships between affective, communication and

political involvement variables in the form of a structural equation model (SEM) (see

Figure 2.5 from Chapter 2). Although each path could be examined separately through a

simple multivariate regression model, by testing simultaneously the structural paths

between variables one can gain a better understanding of the multiple processes of

effects (i.e., direct and indirect paths of influence) theorized. As Holbert (2006) noted,

SEM forces researchers to "see the whole board" by focusing on the study of processes

rather than discrete concepts. SEM is able to examine the process that links several

variables by relying on full-information estimation (e.g., maximum likelihood) rather

than partial-information estimation (e.g., OLS regression-based path analysis). Thus, it is

62
particularly appropriate for testing the communication model advanced here, which

proposes that emotions have a direct as well as an indirect effect on political knowledge

and participation through various communication behaviors.

Within the field of communication, three approaches to SEM have been

common: observed variable (OV), latent composite (LC), and hybrid (HY) (Holbert &

Stephenson, 2002). The OV approach uses observed (or manifest) variables only, be they

single-item measures or scales and indexes combining several indicators. 11 As such, it is

similar to regression path analysis (Wright, 1918, 1921) with the added advantage that it

employs full-information estimation. The LC approach also relies on single-item

measures or indexes but specifies all variables as latent (i.e., not manifest). The HY

method is also based on latent variables but, contrary to the LC approach, does not use

indexes; rather, each observed variable loads individually on its respective latent

variable.

In general, the greatest advantage of the LC and HY approaches over the more

simplistic OV technique is that they account for measurement error by combining the

structural model with a measurement model (via confirmatory factor analysis).

Nevertheless, this study will adopt an OV approach. This decision is justified on both

theoretical and empirical grounds. First, it allows for a simpler statistical model, with

fewer paths to estimate, aiding the graphic presentation of the results and the

substantive interpretation of it. Second, only a single indicator was available for some of

the variables tested in the model, which prevents the estimation of measurement error

1 1 Strictly
speaking, all scales are indices but not all indices are scales. While both employ multiple items
of measurement, indices combine items without concern about their intercorrelation (e.g., the Consumer
Price Index used to gauge inflation). Scales, on the other hand, reflect a latent concept and thus the items
used to construct the scale must be intercorrelated (e.g., the Big Five personality scale by Goldberg,
1990).
63
and factor variance. Lastly, previous research in political communication, particularly the

literature on communication mediation models (e.g., Jung, Kim, & Gil de Zuniga,

forthcoming; Shah, Cho, Eveland, & Kwak, 2005), has also used an OV approach. In fact,

the communication sciences in general rely on OV models (Holbert & Stephenson,

2002).

Assessing Model Fit in SEM

Model fit in SEM is assessed through a variety of statistical tests. In this

dissertation, the model fit statistics used will be those reported by Mplus 6.0 (Muthen &

Muthen, 1998-2010), a statistical modeling software. Specifically, two absolute fit

indices—the chi square goodness-of-fit test and the standardized root mean squared

residual (SRMR)—and five incremental indices—the comparative fit index (CFI), the

Tucker-Lewis index (TLI), 12 the root mean squared error of approximation (RMSEA),

Akaike's information criterion (AIC) and Bayesian information criterion (BIC)—will be

included for every estimated model. According to SEM experts (Benter & Chou, 1987;

Bollen & Long, 1993; Holbert & Stephenson, 2002; Hoyle, 1995; Hu & Bentler, 1999;

Kaplan, 2000), a non-significant chi-square is preferable because it demonstrates that

the model is not a null model. However, the chi-square test is sensitive to sample size.

To correct for this shortcoming, it is advisable to calculate the discrepancy level, defined

as the ratio of chi square to degrees of freedom of the model. In this case, values of 5 or

below are considered acceptable (Kline, 1998). A good fit is also indicated by a SRMR

value of .09 or less, a RMSEA index of less than .06, and CFI and TLI values close to .95 or

higher. The AIC and BIC impose different penalties on models that include more

structural paths, with the AIC more generous and BIC more conservative. Because both

12 The TLI index is also known as the non-normed fit index (NNFI).
64
AIC and BIC are not absolute measures, they are used to compare the fit of two or more

models estimated from the same data set. In this case, the model with the smaller AIC

and BIC is to be preferred.

Analytical Strategy

To test the proposed relationships, several structural models will be estimated.

Before fitting the theorized model to the data sets, a residualized covariance matrix

among the key endogenous variables (i.e., emotions, communication, and involvement

variables) will be created with a partial correlation matrix controlling for the exogenous

variables (i.e., demographics and political orientations). By using the residualized

covariance matrix as input in the estimation of the structural models, the subsequent

analyses will already be taking into account the influence of demographics and political

orientations. 13

Subsequently, each data set will be analyzed using two different modeling

strategies, a cross-sectional model and an auto-regressive model. The cross-sectional

model will relate individual differences in emotions, media use, discussion, issue

knowledge and campaign participation based solely on one wave of data (see Figure 3.1

for a graphical representation). With the ANES data, this model means using variables

measured at Wave 11 only (with the exception of news exposure, measured at Wave

10); with the NAES data, in turn, this means using the post-debate wave variables only.

The cross-sectional model does not take advantage of the panel design of the surveys

used in the study. Nevertheless, as Shah et al. (2005) explained, it is useful for two

13For reference, the partial correlation tables for the ANES and NAES data sets are displayed in Table A.6
and Table A.7, respectively, in the Appendix section. Because the nature of this study is to test the
relationships hypothesized by the affective citizen communication model, the influence of control
variables will not be discussed in the text. Nevertheless, Table A.8 and Table A.9 in the Appendix detail the
direct effects of demographics and political orientations on the variables of interest via regression
analysis.
65
reasons. It allows me to connect the current research with existing literature that has

relied mainly on cross-sectional analyses, and, importantly, it serves as a baseline for

comparing the performance of the model that does take into account the panel design

of the surveys—the auto-regressive model—explained shortly.

Figure 3.1 Theorized Model Using Cross-Sectional Data

Debate
Viewing

Positive
Emotions
(Hope. Pride)
Political
Knowledge

News
Media Use
Political
Participation

Negative
Emotions
(Anxiety. Anger)

Discussion
(Politics, Public
Affairs)

In a nutshell, the auto-regressive model takes into account the dynamic nature

of the data by relating aggregate change estimates generated by lagging first wave

variables on their second wave counterparts for all endogenous variables in the model

(Finkel, 1995). With the ANES data, this model will be estimated with lagged Wave 9
66
variables on their Wave 11 counterparts. Likewise, with the NAES data set, the post-

debates wave variables will be lagged on their pre-debates wave counterparts —with

the obvious exception of debate viewing, which was measured only in the second wave

of both surveys. The auto-regressive approach sets a more stringent test of the affective

citizen communication model because it takes into consideration both temporal stability

and covariance of the key endogenous variables. Furthermore, it has been used by

previous scholars in political communication (Binder, Dalrymple, Brossard, & Scheufele,

2009; Shah, Cho, Eveland, & Kwak, 2005) and has several advantages over the cross-

sectional model:

"This approach aims to explain the unexplained variance among endogenous

Wave 2 variables while accounting for stability in these variables over time.
Estimates of change are derived across the sample rather than within each
individual. As a result, error variances are generally reduced, producing more
stable, albeit potentially less sensitive, estimates of gains or losses" (Shah, Cho,
Eveland, & Kwak, 2005, p. 543).

The specific structure of the appropriate auto-regressive model, however, needs

further clarification because it can take several forms. Synchronous models consider the

relationships among variables within waves, controlling for the values of those variables

in a previous wave. An example of this model would be a structure in which knowledge

at time 2 is predicted by news use at time 2, controlling for knowledge at time 1 and

news use at time 1. Lagged models, in contrast, involve relationships among variables

across waves. For instance, a time 2 measure of knowledge is predicted by a time 1

measure of news use, controlling for time 1 knowledge. Previous research has found

greater support for the claims of a synchronous influence of communication behaviors

on political knowledge and civic engagement than for a lagged influence of these

variables on both outcomes (see Eveland, Hayes, Shah, & Kwak, 2005b; Shah, Cho,

67
Eveland, & Kwak, 2005). For the influence of emotions, however, it is not clear which

structure is more appropriate. If emotions have a short-term effect on communication

behaviors, a synchronous model may perform better than a lagged model. On the other

hand, if emotions have a long-lasting or cumulative effect, a lagged model may

outperform a synchronous model.

To decide which structure is more appropriate, I tested the synchronous model

shown in Figure 3.2 below with two other possible lagged models, using both the ANES

and NAES panels. The first one was a lagged auto-regressive model that estimated

relationships among time 1 measures of emotions, time 1 media use and political

discussion, and time 2 political knowledge and participation, controlling, of course, for

time 1 political knowledge and participation. The second lagged auto-regressive model

estimated relationships among time 1 emotions, time 2 media use and political

discussion, and time 2 political knowledge and participation (again, controlling for time

1 measures of involvement). The goodness-of-fit statistics revealed that neither of the

lagged models fit the ANES and NAES data well (see Table A.10 and Table A.11 in the

Appendix). The synchronous model, in contrast, exhibited a very good fit and, thus, will

be used for the auto-regressive specification estimated in Chapter 4.

68
Figure 3.2 Theorized Model using Panel Data

69
 Although the synchronous auto-regressive model depicted in Figure 3.2 has a

superior performance than the lagged models, it does not resolve the issue of causality.

Therefore, I will also test an alternate causal ordering of the variables using a

synchronous model as well (see Figure 3.3 below). In this case, the directionality of

effects will flow from communication behaviors to emotions and, subsequently, to

political knowledge and campaign participation. By comparing the goodness-of-fit

statistics of the proposed and alternative models, which I will do using both the cross-

sectional and auto-regressive approaches, one can gain a better sense of the match

between the proposed and observed relationships among variables across data sets.

Simply put, if the theorized structure has a better fit than the alternative structure, the

results should be interpreted as further support for the affective citizen communication

model. 14

14 Certainly, I could test for additional alternative causal models, such as one running from involvement to
emotions to communication. However, there is strong empirical evidence that media use and discussion
frequency are causal antecedents of involvement (Aarts & Semetko, 2003; Cho, Shah, McLeod, McLeod, &
Scholl, 2009; Rojas, 2006). Less clear is the direction of the relationship between feeling emotions and
engaging in communication behaviors, which justifies using this particular alternative model to contrast it
with the proposed theoretical model of this dissertation.
70
Figure 3.3 Alternative Model using Panel Data

71
Indirect Effects

With SEM, researchers can examine three types of effects: direct, indirect, and

total effects. Usually, direct effects receive the most attention, in part because the

structural paths of a model estimated via SEM represent the direct influence of one

variable on another (Holbert & Stephenson, 2003). Nevertheless, indirect effects, that is,

the effects of one variable on another as that variable's effect operates through one or

more intervening variables, are also relevant. Classic theories of communication, such as

the two-step flow (Katz & Lazarsfeld, 1955) and diffusion of innovations (Rogers, 2003),

are based on the notion that media effects are indirect (i.e., via opinion leaders or early

adopters, respectively).

The affective citizen communication model advanced in the current study posits

that the effects of emotions on political knowledge and participation are both direct and

indirect, through communication processes. At the same time, the effects of news

media use and debate viewing on knowledge and participation are also direct and

indirect, through political discussion. Oftentimes, indirect effects are labeled mediating

relationships, that is, these terms tend to be used interchangeably. In the case of the

affective citizen communication model, those familiar with Baron and Kenny (1986)

would argue that communication variables mediate-at least partially—the effects of

emotions on political knowledge and participation.

The most popular method to assess mediation in the social sciences is Baron and

Kenny's (1986) causal step approach. However, this approach has several shortcomings

that make its use in this study less than optimal. Most importantly, it has low statistical

power and cannot quantify with precision the indirect effect being tested (Hayes, 2009).

Therefore, to test for indirect (or mediating) effects in the affective citizen

communication model, the study will use Bollen's (1987) delta method, which produces

72
point estimates and standard errors for any indirect effect in a structural model

estimated via maximum likelihood. 15 Mplus has a built-in delta method function for

estimating indirect effects. Therefore, there is no need to use additional macros or

scripts to compute them, as would be the case with SPSS and =other popular statistical

packages.

15Bootstrapping offers an alternative method to estimate indirect effects, and has been advocated by
several methodologists to be the most robust method to date to examine multiple mediator models
(Hayes, 2009; MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002; Preacher & Hayes, 2008).
Nevertheless, it has been found that the delta method produces the same results than the bootstrapping
technique when using large samples, such as the ANES and NAES studies employed here (Bollen & Stine,
1990).
73
 Chapter 4: Results

INTRODUCTION

The affective citizen communication model advanced in this dissertation

proposes that the effects of emotions triggered by political candidates on knowledge of

the candidates' stands on issues and on campaign participation are largely mediated by

communication variables, including news media use, political discussion and debate

viewing. The current chapter presents the empirical tests of these structural

relationships, which were specified in hypothesis form in Chapter 1. The tests are based

on statistical analyses of the two data sets explained in detail in Chapter 3, the ANES

2008-2009 Panel and the NAES 2004 Debates Panel.

Due to the variety of tests for each data set, some repetitiveness in the reporting

of the results will be unavoidable. However, to present the findings in a clearer manner,

the results will be organized by survey, beginning with the ANES and ending with the

NAES. Within the analysis of each survey, I will refer to the tests of the theorized

affective citizen communication model examined through structural equation modeling.

As explained earlier, the model will be estimated using two specifications: cross-

sectional and auto-regressive. Subsequently, the theorized model will be compared to

an alternative model—also outlined previously in Chapter 3—where communication

behaviors lead to feeling emotions about the candidates, and these processes together

predict learning candidates' issue stands and engaging in campaign activities. This

exercise will also be conducted using both cross-sectional and auto-regressive

specifications. Goodness-of-fit statistics will be used to assess if the affective citizen

communication model represents a stronger causal model than its alternative.

74
 Lastly, I will review the research hypotheses outlined in Chapter 2 in light of the

statistical analyses and discuss how well the data supported each of them.

ANALYSES OF THE ANES 2008-2009 PANEL STUDY

Cross-Sectional Model

In order to identify the best fitting cross-sectional model, the estimation

proceeded in three stages. In the first step, all emotion variables were set to have a

direct influence on both communication and involvement variables, and all

communication variables, in turn, were set to have a direct influence on the outcome

variables (this is called a "saturated" model, depicted earlier in Figure 3.1).

Subsequently, all non-statistically significant paths from emotions to the outcome

variable were removed so that indirect effects of emotions through communication

variables could be examined. This model tests if communication behaviors partially or

fully mediate the relationships between emotions and the political involvement

variables considered here (this is the "trimmed" model). Likewise, in the trimmed model

all non-statistically significant paths from news media use and debate viewing on the

outcome variable were removed to assess possible mediation by political discussion.

Lastly, a final model was estimated in which all non significant structural paths were

removed (this is the "final" model). 16 Table 4.1 shows the goodness-of-fit statistics of

these three models. Because it is the most parsimonious (i.e., has the lowest AIC, BIC,

16 This step-by-step process was borrowed from Shah and colleagues (2005), who argued that these final
models
"fit the data best and explain the same amount of variance in the criterion variable (...) as the
saturated and trimmed models while providing the most parsimonious solutions. Given the
performance of these trimmed models on a number of criteria, including their empirical fit across
measures and their relative parsimony compared to other specifications [they are accepted] as
(...) final models" (pp. 544-545).
75
and discrepancy level), the results will be explained using the estimates of the final

model only.

Table 4.1 Comparison of Theorized Cross-Sectional Models using the ANES 2008-2009
Panel

Saturated Trimmed Final

Model Model Model
AIC 2,834.74 2,832.00 2,826.73
BIC 2,977.99 2,963.79 2,941.33
f — 1.26 1.99
df — 2 5
p value — n.s. n.s.
//# — .63 .40
CFI — 1.00 1.00
TLI — 1.00 1.00
RMSEA — .00 .00
SRMR — .00 .01
Notes: All models were estimated using Wave 11 data only (n = 2,276). The saturated
model has all proposed structural paths outlined in Figure 2.5. Thus, relative goodness-
of-fit statistics cannot be computed. The trimmed model has all non-significant emotion
effects on knowledge and participation removed. The final model has all non-significant
paths removed.

The concurrent relationships among emotions, communication and involvement

estimated by the final model are shown in Figure 4.1. Overall, this model fitted the data

extremely well, yielding a chi-square value of 1.99 with 5 degrees of freedom (RMSEA =

.00, CFI = 1.00, TLI = 1.00, SRMR = .01). The variables included in this model accounted

for 5.6% of the variance in political knowledge, 4.2% in campaign participation, 5.2% in

debate viewing, 9.4% in political discussion and 0.7% in news media use.

76
Figure 4.1 Final Cross-Sectional Model using the ANES 2008-2009 Panel

Debate
Viewing
08 06
Negative 14
Emotions

Political
Participation
06

News
.14
Media Use

.08 -.05
Political
Knowledge
Positive
09 .22
07
Emotions
.17
07

Political
Discussion

Notes: Standardized coefficients displayed here are at least significant at p < .05

The relationships observed in Figure 4.1 support the view that both positive and

negative feelings towards Obama and McCain in the 2008 election were positively

related to communication behaviors. Specifically, negative emotions towards both

candidates such as fear and anger were associated with watching the televised

presidential debates, whereas positive emotions, namely, pride and hope, were

predictive of both greater news media use and more frequent political discussions with

family members and friends. These communication behaviors, in turn, were positively

related to greater knowledge of where the candidates stood on the issues and also to

greater involvement in campaign activities. The only exception to this trend of positive

influences was the negative path from news to knowledge, which was significant once

77
all controlling variables and other relationships in the model were taken into

consideration.

Does this finding suggest that news use spurred issue ignorance in the 2008

election? A full answer to this question requires a look at the indirect effects of news

use on knowledge, which are displayed along all other indirect effects in the model in

Table 4.2 below.

78
Table 4.2 Direct, Indirect and Total Effects in the Final Cross-Sectional Model using the
ANES 2008-2009 Panel

News Debate Political Political Political

Media Use Viewing Discussion Knowledge Participation
Positive
Emotions:
Direct .08 — .16 .07 —

Indirect — .02 .01 .03 .02

Total .08 .02 .17 .09 .02
Negative
Emotions:
Direct — .08 — — .06
Indirect — — .02 .01 .01
Total — .08 .02 .01 .07
News
Media Use:
Direct — .21 .09 -.05 .06
Indirect — — .05 .05 .04
Total — .21 .14 — .10
Debate
Viewing:
Direct — — .22 .11 .14
Indirect — — — .04 .02
Total — — .22 .15 .16
Political
Discussion:
Direct — — — .17 .07
Indirect — — — — —

Total — — — .17 .07

Note: Standardized coefficients displayed here are at least significant at p < .05. Indirect
and direct effects might not add up to total effects due to rounding error and
nonsignificant indirect links.

Actually, the relationship between news media use and political knowledge is

quite complex. As shown in Table 4.2, consuming news is positively related to debate

viewing and political discussion frequency, both of which are strongly associated with

79
greater knowledge of where Obama and McCain stood on the issues of the campaign.

Thus, two mechanisms seem to be working in opposite directions simultaneously: the

direct effect of news is negative, but its indirect effect is positive. Because both of these

effects are of the same magnitude, the total effects of news on knowledge are null. This

particular, if not surprising, finding is not—as will be shown later—unique to the ANES

cross-sectional analysis. Therefore, we will elaborate further on it in Chapter 5.

Another important finding from the results of this cross-sectional model refers to

the indirect effects of positive and negative feelings on citizens' involvement in the

campaign. As was expected, communication variables fully mediated the effect of

enthusiasm on McCain and Obama on participation in campaign activities. At the same

time, watching the debates and discussing about the campaign more frequently fully

mediated the relationship between anxiety and learning the candidates' issue stances.

Likewise, discussion frequency partially mediated the relationship of both debate

viewing and news use on political knowledge and participation.

Auto-Regressive Model

As explained in Chapter 3 (see footnote 8), political participation was measured

in the ANES data set in Wave 11 only. Therefore, of the two involvement variables, only

political knowledge will be used in the estimation of the auto-regressive model using the

ANES panel.

Table 4.3 contains a summary of the goodness-of-fit statistics for each of the

three auto-regressive models: saturated, trimmed and final. As was the case with the

cross-sectional analysis, the best performing model was the final model, with all

nonsignificant paths removed. While the discrepancy level exceeded 5, (x2 = 182.03, df =

80
35), all other tests were within acceptable ranges (RMSEA = .04, CFI = .99, TLI = .96,

SRMR = .02). Consequently, the results that follow refer to the final model only.

Table 4.3 Comparison of Theorized Auto-Regressive Models using the ANES 2008-2009
Panel

Saturated Trimmed Final

Model Model Model
AIC 10,135.62 10,134.00 10,124.86
BIC 10,347.64 10,328.83 10,285.31
f 174.79 179.17 182.03
df 26 29 35
p value < .001 < .001 <.001
//# 6.72 6.18 5.20
CFI .98 .98 .98
TLI .95 .95 .96
RMSEA .05 .05 .04
SRMR .04 .04 .04
Notes: All models were estimated using Wave 9 and Wave 11 data (n = 2,276). The
saturated model has all proposed structural paths outlined in Figure 3.1. The trimmed
model has all nonsignificant emotion effects as well as nonsignificant media use effects
on knowledge removed. The final model has all nonsignificant paths removed.

The over-time relationships among emotions, communication and involvement

estimated by the final model are depicted in Figure 4.2. In Wave 11, this structural

model accounted for 30.9% of the variance in positive emotions, 35.9% in negative

emotions, 54.7% of news media use, 4.6% in debate viewing, 32.6% in political

discussion and 30.6% in political knowledge. 17

17For Wave 9, the model explained 1.8% of the variance in news media use, 4.7% in political discussion,
and 1.7% in political knowledge.
81
Figure 4.2 Final Auto-Regressive Model using the ANES 2008-2009 Panel

82
 Temporal stability was relatively strong for all variables. In other words, past

behavior in terms of news media use, political discussion and participation in campaign

activities was a strong predictor of subsequent behavior. Likewise, positive and/or

negative feelings towards McCain and Obama as well as knowledge of their issue

stances at the beginning of the official campaign period were the best predictors for

feelings toward and knowledge of these figures after Election Day.

As for the endogenous relationships, several of the structural paths of this model

are consistent with the previous cross-sectional model. As shown in Table 4.4, even

after accounting for prior levels of the variables, Wave 11 positive emotions were a

significant positive predictor of Wave 11 news media use (for full results with Wave 9

variables, see Table A.12 in the Appendix). Likewise, the estimates of change in Wave 11

consumption of campaign news were positively associated with watching more

presidential debates, discussing politics and the election more frequently, and knowing

more about the candidates issue stances, even when accounting for the lagged effects

of discussion and knowledge. As was the case with the previous model, the more

debates respondents said they watched, the more days they discussed politics with

friends and family members, even when accounting for the sizeable effect of past

political discussion. Lastly, debate viewing and unexplained variance in the Wave 11

measure of discussion had a positive influence on Wave 11 political knowledge, again,

even after including the influence of prior political knowledge and discussion.

83
Table 4.4 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using
the ANES 2008-2009 Panel

News Media Debate Political Political

Wave 11 Use Viewing Discussion Knowledge
Positive
Emotions:
Direct .03 .16
Indirect .004 .03 .02
Total .03 .16 .03 .02
Negative
motions:
Direct .06
Indirect .003
Total .06 .003
News
Media Use:
Direct .13 .08 -.04
Indirect .02 .02
Total .13 .11
Debate Viewing:

Direct .16 .12

Indirect .01
Total .16 .12
Political
Discussion:
Direct .05
Indirect
Total .05
Note: Standardized coefficients displayed here are at least significant at p < .05. Indirect
and direct effects might not add up to total effects due to rounding error and
nonsignificant indirect links. Full results with Wave 9 effects are available in Table A.12
in the Appendix.

Nevertheless, there were three notable differences detected between the

results of the cross-sectional model and the synchronous auto-regressive model. First,

84
using the auto-regressive approach, it was found that there was a direct link between

feelings of hope and pride in candidates Obama and McCain as measured in Wave 11

and watching the televised debates. Second, the influence of Wave 11 positive emotions

on Wave 11 political knowledge was fully mediated by communication behaviors once

lagged knowledge was included in the equation. And third, there was no significant

relationship between Wave 11 negative emotions and debate viewing. I will return to

the meaning of these differences in Chapter 5.

Comparison with Alternative Model

To further examine the performance of the theorized model, I run additional

tests of an alternative causal ordering of the key variable clusters for both the cross-

sectional and auto-regressive approaches. In the alternative specification, the

directionality of effects flows from communication behaviors to emotions and,

subsequently, to political knowledge and campaign participation (refer to Figure 3.3).

As seen in Table 4.5, reporting final models (i.e., with all nonsignificant paths

removed), the theorized relationship between feeling emotions and subsequently

engaging in communication behaviors, which in turn lead to knowledge and

participation, has a better fit than its inverse, of communication behaviors leading to

feeling emotions and both jointly leading to involvement. The differences in goodness-

of-fit statistics are dramatic across cross-sectional and auto-regressive models,

particularly in regards to discrepancy levels and the AIC and BIC fit indices. This evidence

provides more support for the affective citizen communication model.

85
Table 4.5 Comparison of Theorized Model with Alternative Model using the ANES
2008-2009 Panel

Cross-Sectional Auto-Regressive
Theorized Alternative Theorized Alternative
Model Model Model Model
(Emotions -> (Comm. -> (Emotions -> (Comm. ->
Comm. Emotions -> Comm. Emotions ->
Involvement) Involvement) Involvement) Involvement)
AIC 2,826.73 16,097.24 10,124.86 10,597.60
BIC 2,941.33 16,223.31 10,285.31 10,746.58
f 1.99 4.08 182.03 265.17
df 5 5 35 34
p value n.s. n.s. < .001 < .001
//# .40 .82 5.20 7.80
CFI 1.00 1.00 .98 .96
TLI 1.00 1.00 .96 .94
RMSEA .00 .00 .04 .06
SRMR .01 .01 .04 .05
Notes: The cross-sectional models use Wave 11 data only (n = 2,276). The auto-
regressive models use Wave 9 and Wave 11 data (n = 2,276). The theorized cross-
sectional model refers to the final model with all non-significant paths removed
illustrated in Figure 4.1. The theorized auto-regressive model refers to the final model
with all non-significant paths removed illustrated in Figure 4.2. The alternative models
have also all their non-significant paths removed..

86
ANALYSES OF THE NAES 2004 DEBATES PANEL STUDY

Cross-Sectional Model

Table 4.6 displays the goodness-of-fit statistics of the saturated, trimmed and

final cross-sectional models using data from the NAES 2004 Debates Panel. As was the

case with the ANES survey, the final model was the most parsimonious and with the

best fit of all three (y 2 = 6.29, df= 4, RMSEA = .04, CFI = .99, TLI = .96, SRMR = .02). The

variables included in this model accounted for 15.5% of the variance in political

knowledge, 19.1% in campaign participation, 9.4% in debate viewing, 15.0% in political

discussion and 1.0% in news media use.

Table 4.6 Comparison of Theorized Cross-Sectional Models using the NAES 2004
Debates Panel

Saturated Trimmed Final

Model Model Model
AIC 372.69 372.76 370.98
BIC 451.28 447.42 433.85
f — 2.07 6.29
df — 1 4
p value — n.s. n.s.
//# — 2.07 1.57
CFI — .99 .99
TLI — .93 .96
RMSEA — .05 .04
SRMR — .02 .02
Notes: All models were estimated using the post-debates data only (n = 376). The
saturated model has all proposed structural paths outlined in Figure 2.5. The trimmed
model has all non-significant emotion effects on knowledge and participation removed.
The final model has all non-significant paths removed.

87
 Figure 4.3 below illustrates the direct structural paths between the variables

included in the final cross-sectional model.

Figure 4.3 Final Cross-Sectional Model using the NAES 2004 Debates Panel

Debate
Viewing
22
10

.35

Political
Participation

Negative News
.10 .12
Emotions Media Use

Political
Knowledge

.23
.29
.13

Political
.24
Discussion

Notes: Standardized coefficients displayed here are at least significant at p < .05

In this case, feeling negative emotions towards then-president Bush and

candidate Kerry was positively associated with greater exposure and attention of

campaign news. News media use and debate viewing, in turn, were strong predictors of

the frequency with which respondents discussed about the campaign with their family

and friends. As for the outcome variables, both debate viewing and interpersonal

discussion exhibited direct, significant associations with political participation and

political knowledge.

88
 As is manifest in Table 4.7 below, even after considering communication

behaviors, negative emotions towards candidates Bush and Kerry had a noticeable

direct relationship with political participation. Still, there was a significant indirect effect

of negative feelings operating mainly through political talk. In regards to political

knowledge, the contribution of feelings of anger, anxiety and uneasiness were

completely indirect, operating mainly through discussing more frequently.

Lastly, there was evidence that discussion frequency mediated the relationship

between media use variables and involvement variables, as debate viewing and news

media use had indirect effects on knowledge and participation via frequency of

conversations about the campaign.

89
Table 4.7 Direct, Indirect and Total Effects in the Final Cross-Sectional Model using the
NAES 2004 Debates Panel

News Media Debate Political Political Political

Use Viewing Discussion Knowledge Participation
Negative
Emotions:
Direct .10 — .24 — .22
Indirect — .03 .03 .04 .08
Total .10 .03 .26 .04 .30
News Media
Use
Direct — .31 .23 — —

Indirect — — .04 .14 .11

Total — .31 .26 .14 .11
Debate
Viewing
Direct — — .12 .35 .10
Indirect — — — .02 .04
Total — — .12 .36 .13
Political
Discussion
Direct — — .13 .29
Indirect — — — — —

Total — — — .13 .29

Note: Standardized coefficients displayed here are at least significant at p < .07. Indirect
and direct effects might not add up to total effects due to rounding error and
nonsignificant indirect links.

Auto-Regressive Model

When comparing the fit of the saturated, trimmed and final auto-regressive

models, all three had adequate discrepancy levels and CFI values above .95 (see Table

4.8). Nevertheless, the final model outperformed the other two in terms of relative fit,

with lower AlC, BIC and RMSEA values, and a higher TLI value. Furthermore, it was more

90
parsimonious, as all nonsignificant structural paths were removed. As a consequence,

the findings that will be discussed below refer to the final model only.

Table 4.8 Comparison of Theorized Auto-Regressive Models using the NAES 2004
Debates Panel

Saturated Trimmed Final

Model Model Model
AlC 243.79 241.82 237.34
BIC 391.12 385.37 346.90
X2 76.95 76.98 87.18
df 26 27 34
p value < .001 < .001 < .001
x2/# 2.96 2.85 2.56
CFI .96 .96 .96
TLI .92 .93 .94
RMSEA .08 .08 .07
SRMR .06 .06 .07
Notes: All models were estimated using pre- and post-debates data (n = 323). The
saturated model has all proposed structural paths outlined in Figure 3.1. The trimmed
model has all non-significant emotion effects on knowledge and participation removed.
The final model has all non-significant paths removed.

The relationships among emotions, communication behaviors, political

knowledge and participation estimated by the final model are depicted in Figure 4.4. In

the post-debates wave, this structural model accounted for 48.3% of the variance in

negative emotions, 57.6% of news media use, 10.5% in debate viewing, 45.7% in

political discussion, 27.4% in political knowledge, and 65.3% in political participation. 18

18In the pre-debates wave, the model explained 10.1% in political discussion, 2.4% in political knowledge
and 20.8% in political participation.
91
Figure 4.4 Final Auto-Regressive Model using the NAES 2004 Debates Panel

/if .s .
If\

92
 As could be expected from the short time-span between waves, the stability of

the variables before and after the presidential debates between Bush and Kerry was

strong. This was particularly true for negative emotions, news media use and political

participation. This sets an extremely conservative test for the synchronous relationships

among key variables in the post-debates wave.

As can be gleaned from Table 4.9, the structural paths of this model are

consistent with the cross-sectional model detailed earlier. Feelings of anger and anxiety

towards Bush and Kerry were positively related to discussion frequency, even after

accounting for the lagged effects of emotions and discussion. Similarly, negative

emotions were associated with participation in campaign activities, even after including

previous participation and emotions in the model. The strong effects of news media use

on debate viewing, as well as the relationship between news and political discussion,

were significant when controlling for prior levels of these variables. The contribution of

watching the Bush-Kerry debates on knowing the candidates' issue position was-just

like in the cross-sectional analysis—among the most robust effects detected in the final

auto-regressive specification. This relationship was evident even after taking into

account pre-debate knowledge of candidates' issue positions. On the other hand, the

influence of discussing political affairs with friends and family members on both

knowledge and participation was also positive and significant, even when accounting for

prior levels of these variables. Lastly, the positive indirect effect of news consumption

on learning the candidates' issue positions through discussion frequency was significant,

even when including lagging effects of knowledge.

93
Table 4.9 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using
the NAES 2004 Debates Panel

Post-
Debates News Debate Political Political Political
Wave Media Use Viewing Discussion Knowledge Participation
Negative
motions:
Direct .10 .09
Indirect .02
Total .10 .10
News Media
Use:
Direct .32 .16 -.13
Indirect .13 .03
Total .32 .16 .03
Debate
Viewing:
Direct .31
Indirect
Total .31
Political
Discussion:
Direct .16 .15
Indirect
Total .16 .15
Note: Standardized coefficients displayed here are at least significant at p < .05. Indirect
and direct effects might not add up to total effects due to rounding error and
nonsignificant indirect links. Full results with pre-debates wave effects are available in
Table A.13 in the Appendix.

The similarity of the results with the cross-sectional model, however, is part of

the story. The auto-regressive approach identified a number of differences that are

noteworthy. After taking into consideration lagged news media use, negative emotions

were not significantly related to post-debate media use. This also caused all indirect

effects of negative emotions on debate viewing, political discussion and political

94
knowledge through news media to be nonsignificant. Another important difference was

the negative direct effect of post-debate news media use on post-debate knowledge

after accounting for pre-debate knowledge. This result brings up the complex

relationship between news use and learning candidates' issue positions found

previously in the ANES data set; while news use by itself can cause people to

inaccuracies in candidates' issue stances, it can also spark learning so long as it gets

people to talk about the election. The fact that these inverse relationships are of the

same magnitude makes the total effect of news use on knowledge null. A third major

difference relates to the contribution of debate viewing on both political discussion and

participation. In this case, there was no significant direct or indirect effect detected,

contrary to what was found in the cross-sectional analysis. I will return to the

significance of these relationships in the discussion section (Chapter 5).

Comparison with Alternative Model

To further address the issue of causality, I tested alternative causal ordering of

the variables of the affective citizen communication model, separately for cross-

sectional and auto-regressive approaches. As may be recalled, in the alternative

specification the effects flow from communication to emotions to political knowledge

and campaign participation (refer to Figure 3.3).

As seen in Table 4.10, both theorized and alternative models adequately fit the

data. All discrepancy levels are well below 5, the CFI and TLI values are .94 or higher,

and the highest RMSEA value is .07. However, in both cross-sectional and auto-

regressive specifications, the AlC and BIC values are significantly lower for the theorized

structure than the alternative. This means that the affective citizen communication

95
model, as tested with data from the NAES 2004 Debates Panel, is more parsimonious

and thus preferable to the alternative model.

Table 4.10 Comparison of Theorized Models with Alternative Models using the NAES
2004 Debates Panel

Cross-Sectional Auto-Regressive
Theorized Model Alternative Theorized Model Alternative
(Emotions -> Model (Emotions -> Model
Comm. (Comm. -> Comm. (Comm. ->
Involvement) Emotions -> Involvement) Emotions ->
Involvement) Involvement)
AlC 370.98 1,378.13 237.34 1,159.81
BIC 433.85 1,437.07 346.90 1.269.36
f 6.29 8.28 87.18 88.07
Df 4 5 34 36
p value n.s. n.s. < .001 < .001
//# 1.57 1.66 2.56 2.45
CFI .99 .99 .96 .96
TLI .96 .96 .94 .94
RMSEA .04 .04 .07 .07
SRMR .02 .03 .07 .07
Notes: The cross-sectional models use the post-debates data only (n = 376). The auto-
regressive models use pre- and post-debates data [n = 323). The theorized cross-
sectional model refers to the final model with all non-significant paths removed
illustrated in Figure 4.3. The theorized auto-regressive model refers to the final model
with all non-significant paths removed illustrated in Figure 4.4. The alternative models
have also all their non-significant paths removed.

INTEGRATING RESULTS ACROSS DATA SETS

The results presented up to this point provide a comprehensive examination of

the affective citizen communication model and its implied relationships across key

clusters of variables. As explained in Chapter 2, the model was broken down into 15

96
research hypotheses, which presents some difficulty in summarizing the tests for these

hypotheses. In order to facilitate the synthesis and substantive interpretation of the

multiple analyses employed in the study, Table 4.11 provides a measure of the

proportion of hypotheses supported across data sets and structural models. In addition,

Figure 4.5 presents en edited version of the original model depicted in Figure 3.1

showing only the major, significant paths.

97
Table 4.11 Proportion of Hypotheses Supported using the ANES 2008-2009 Panel Sample and the NAES 2004 Debates Panel

ANES 2008-2009 Panel NAES 2004 Debates

Panel
Final Final Auto- Final Final Auto-
Cross- Regressive Cross- Regressive
Sectional Model Sectional Model
Model Model Overall
Hypotheses:
1: Emotions -> News Use Yes Yes Yes No 3/4
2: Emotions -> Debate Viewing Yes No No No 1/4
3: Emotions -> Discussion Yes Yes Yes Yes 4/4
4: News Use -> Knowledge No No No No 0/4
5: News Use -> Participation Yes — No No 1/3
6: Debate Viewing -> Knowledge Yes Yes Yes Yes 4/4
7: Debate Viewing -> Participation Yes — Yes No 2/3
8: Discussion Knowledge Yes Yes Yes Yes 4/4
9: Discussion -> Participation Yes — Yes Yes 3/3
10: Emotions News, Debate, Discussion -> Knowledge Yes Yes Yes No 3/4
11: Emotions -> News, Debate, Discussion Participation Yes — Yes Yes 3/3
12: News -> Discussion -> Knowledge Yes Yes Yes Yes 4/4
13: News Discussion Participation Yes — Yes Yes 3/3
14: Debate Viewing -> Discussion Knowledge Yes Yes Yes No 3/4
15: Debate Viewing -> Discussion -> Participation Yes — Yes No 3/4
Proportion of hypotheses supported:
Count 14/15 7/9 12/15 7/15 40/54
Percentage (93.3%) (77.8%) (80%) (46.7%) (74.1%)
Figure 4.5 Significant Paths of the Affective Citizen Communication Model

1 analysis

2 analyses

3 or 4 analyses

Debate
Viewing

Posith^
Emofons
(Hope, Pride)
Political
Knowledge

News

Political
Partidpst ion

Negative
Emotions
(Anxiety. Anger)

Discussion
(Politics. Public
Affisfts)

Across data sets and models, there was considerable support for the hypotheses

derived from affective intelligence theory, as feelings towards presidential candidates

were directly correlated to news media consumption and frequency of political

discussion. The direct relationship between emotions and debate viewing, however, was

supported in the ANES cross-sectional analysis only. The hypotheses derived from

communication mediation were also supported in a consistent fashion across surveys

99
and models. The positive effects of news media use on both political knowledge and

political participation were found to be fully mediated by political discussion. A similar

trend occurred with watching the televised debates between the main presidential

candidates, although in this case the mediation by political talk was partial. Notably,

face-to-face discussions with friends and family members about political affairs were a

more immediate cause for political involvement than media use variables. Lastly, and in

clear support of the affective citizen communication model advanced in this study, in all

but one test emotions were found to have (in addition to direct effects) significant

indirect effects on political knowledge and political participation that operated through

a host of communication behaviors, including news use, debate viewing and political

discussion. Summing all tests of hypotheses presented in Table 4.11, it can be concluded

that the null was rejected, on average, three-quarters of the time. This may be

interpreted as substantial support for the affective citizen communication model.

100
 Chapter 5: Discussion and Conclusions

Let's not forget that the little emotions are the great captains of our lives and we
obey them without realizing it. (Van Gogh)

The emotions aren't always immediately subject to reason, but they are always
immediately subject to action. (William James)

INTRODUCTION

Feelings towards public figures running for office are one of myriad sources

voters rely on to decide what to do during a political campaign. Issues, values, identities

and socialization are oftentimes key determinants of paying attention to media, talking

about the campaign, learning candidates' policy stands and getting involved in political

activities. And for a substantial portion of voters, campaigns do not stir their emotions

at all; if anything, they trigger apathy and indifference. Yet, there is consistent evidence

showing that presidential candidates elicit feelings of enthusiasm and anxiety—if not

outright anger—on the American voter (Crigler, Just, & Belt, 2006; Marcus, Neuman, &

MacKuen, 2000; Neuman, Marcus, MacKuen, & Crigler, 2007; Roseman, Abelson, &

Ewing, 1986; Valentino, Brader, Groenendyk, Gregorowicz, & Hutchings, 2011). In this

sense, the elections studied in this research are typical. In the 2008 campaign, ratings of

hope and pride in Barack Obama were in the upper 40s (on a scale from 0 to 100,

averaging across Democrats and Republicans). In the 2004 election, negative feelings

towards George W. Bush's were in the near 40s as well. This dissertation has attempted

to elaborate on the individual-level consequences of these emotions. In so doing, it has

provided a model that goes beyond the simple claim that feelings cause political

knowledge and campaign participation. Rather, it has suggested—and empirically

examined—the various communication channels by which emotions exert an influence

101
on political involvement. Thus, the study has focused on the process and mechanisms by

which emotions lead to learning and participating.

In this chapter, I will first discuss the tests of the theorized affective citizen

communication model presented in Chapter 4. I will also elaborate on the strengths of

the study as well as its limitations and how these shortcomings may be overcome in

future research. Subsequently, I will suggest how the affective citizen communication

model may be expanded and used to address other phenomena of interest for

communication scholars. The convergence between the proposed model and existing

communication theories, including agenda setting, priming and partisan selective

exposure, will also be addressed. Lastly, I will discuss what the results of this study tell

us about current debates on citizen competence.

DISCUSSION OF FINDINGS

This project relies on the theoretical insights of robust bodies of work in political

communication, namely, affective intelligence (Marcus, Neuman, & MacKuen, 2000) and

citizen communication mediation (Cho, Shah, McLeod, McLeod, & Scholl, 2009; Eveland,

2001; McLeod et al., 2001; Shah, Cho, Eveland, & Kwak, 2005). Building on these two

paradigms, I have argued that feelings towards presidential candidates lead individuals

to engage in a variety of communication behaviors, such as consuming news, discussing

politics with fellow citizens, and watching the televised debates between the main

candidates. These communication practices, in turn, have a more immediate effect on

two important political outcomes: learning where the candidates stand on the issues of

the campaign, and participating in a variety of political activities, such as donating

money to a candidate and attending rallies.

102
 To test the proposed relationships, I used survey data from the 2004 and 2008

U.S. presidential elections, collected independently by the National Annenberg Election

Survey (NAES) 2004 Debates Panel and the American National Election Study (ANES)

2008-2009 Panel, respectively. Using structural equation modeling to estimate both

concurrent and over-time relationships among variables, I found considerable support

for the theorized model. In particular, I found that when voters feel enthusiastic about

the candidates, they consume more news about the campaign and are more likely to

watch the debates. On the other hand, when they report negative feelings, such as fear

and anger, they tend to discuss politics with their friends and family members more

frequently. These communication behaviors, in turn, are related to each other. Habitual

news use is predictive of exposure to major programs such as the presidential debates,

and both types of media use spur political talk in a similar fashion.

Notably, people's emotions, media use and political discussion are all related to

how much they know about the issue policies of the candidates and also to the breadth

of their participation in campaign activities. Sometimes this influence is straightforward,

as when watching the debates and engaging in political conversations lead directly to

greater knowledge, or when having negative emotions towards the candidates spurs

political action. Oftentimes, however, the process is indirect, such as when negative

feelings trigger learning the candidates' issue stances if only because they motivate

discussing the campaign with others on a more frequent basis. Likewise, people who

feel stronger emotions towards the candidates are more likely to be involved in the

campaign, both cognitively and behaviorally, due to the communication behaviors

sparked by their feelings. In this sense, among all the indirect mechanisms of influence

detected in the study, the most complex refers to the role of news use on knowledge;

once demographics, political orientations and other components of the proposed model
103
are taken into consideration, news consumption has a negative direct effect, but a

positive indirect effect, on learning. I will come back to this finding shortly.

While the specific evidence supporting the links advanced by the affective citizen

communication model was detailed in Chapter 4, six patterns emerge across data sets

and structural models. First, the structure of emotions towards candidates yielded the

existence of two orthogonal dimensions, positive and negative. This finding is consistent

with most survey-based research of affective intelligence theory, and is in contrast to

the discrete view of emotions posited by appraisal theories and some experimental

work. One could think of a number of reasons for this. Perhaps it's a methodological

artifact; surveys may not be as sensitive as other methods to gauge political emotions. It

could also be a timing issue; by the time the surveys employed here were fielded,

emotions towards candidates were already crystallized into distinct valence dimensions,

with most voters answering emotions questions using the familiar approach-avoidance

or like-dislike framework. Therefore, a future direction for research is to map out the

conditions under which a third dimension of emotion arises, which would help sorting

out the issue of the proper structure of affect in political campaigns.

Second, the paths flowing from emotions to communication variables are

weaker than the paths flowing from communication variables to knowledge and

participation. To facilitate comparisons, Table 4.12 displays average standardized

coefficients weighted by sample size. These averages are computed only for the

variables that were measured in both the ANES and NAES data sets and were included in

the cross-sectional and auto-regressive specifications.

104
Table 4.12 Average Direct, Indirect and Total Effects Weighted by Sample Size

News Media Debate Political Political

Use Viewing Discussion Knowledge
Negative
motions:
Direct .01 .03 .05 —

Indirect — .002 .01 .01

Total .01 .04 .06 .01
News Media
Use:
Direct — .19 .10 -.05
Indirect — — .03 .05
Total — .19 .14 .01
Debate Viewing:
Direct — — .17 .14
Indirect — — .02
Total — — .17 .16
Political
Discussion:
Direct — — — .11
Indirect — — —

Total — — — .11
Notes: Cells report average standardized coefficients that were significant at p < .05
obtained from each of the four final structural equation models presented in tables 4.2,
4.4, 4.7 and 4.9, weighted by sample size.

Several explanations are possible for this trend. First, it may be a measurement

artifact. Simply put, self-reports of frequency of media use and discussion may be more

accurate relative to self-reports of intensity of feelings. This is an unlikely possibility,

however, because all emotion variables were measured with additive scales exhibiting a

strong internal consistency and high test-retest correlations (i.e., the path coefficients

from one wave to another for positive and negative emotions were in the .6 to .7

range). Another explanation lies on the central role played by communication processes

105
in U.S. politics. Journalistic organizations—be they legacy media or new outlets—exert a

considerable degree of control over the electoral process and over campaign

communications (Blumler & Gurevitch, 2001). Furthermore, candidates devote

considerable resources to advertising in the media, particularly television. Even the 2008

Obama campaign, notorious for its novel use of social network sites and digital

technologies (Dickinson, 2008), spent nine times more money on broadcast and print

media than it did on the Internet (Center for Responsive Politics, 2011). It should not

come as a surprise, then, that what voters learn and do in regards to electoral politics is

shaped far more by their attentiveness to media messages and the amplification of

these messages through informal conversations, than by the emotions elicited by

candidates in the first place.

A third prominent trend of the results presented in Chapter 4 is the complex

relationship between habitual news media use and political knowledge. As

aforementioned, whereas consuming news was directly related to inaccuracies in

reporting candidates' issue policies, it was indirectly associated—via debate viewing and

political discussion—to a more accurate understanding of the candidates' issue stances.

And this contradictory relationship holds across surveys and structural models. Again,

there are a number of possible explanations. The most straightforward possibility is

that, as some previous research has suggested (e.g., Craig, Kane, & Gainous, 2005;

Hansen & Benoit, 2007; Weaver & Drew, 2001), exposure to news in the mainstream

media hinders, indeed, objective knowledge of the candidates' positions on the issues of

the campaign. This argument is based on the decades-old finding that professional news

outlets focus considerable attention on the "horse race," polls, campaigns' strategic

moves and candidates' gaffes, instead of providing a more substantive, issue-oriented

coverage of politics (Benoit, Stein, & Hansen, 2005; Cappella & Jamieson, 1997; Iyengar,
106
1991; Patterson, 1993). This should not be interpreted as news consumption being

detrimental to all forms of political knowledge. There is a vast literature suggesting that

news media exposure and attention is positively related to recognition of public figures,

awareness of domestic and international affairs, and knowledge of candidates'

biographical background (Druckman, 2005; Eveland, 2001; Price & Zaller, 1993). But

when it comes to issue knowledge, the role of the press may be insignificant, if not

outright negative as was found here.

A second possibility is related to the assumption of linear effects between news

media use and political knowledge. Previous research has documented the existence of

curvilinear associations between media exposure and political outcomes, including

political learning (e.g., Eveland, 1997; Valenzuela, 2009; Zaller, 1992). It may well be that

knowledge of candidates' issue positions increases as news use increases, but at some

level, knowledge starts decreasing as news use increases more. According to this logic,

news junkies may fare worse than regular news users on tests of issue knowledge

because of information overload (Graber, 1984), or because they also happen to have

other characteristics that cause them to be misinformed. For instance, political

psychologists have long contended that some voters are "motivated reasoners," in that

they use a variety of strategies to explain away information that is incongruent with

their prior knowledge (Kunda, 1990; Lodge & Taber, 2000; Redlawsk, 2002). To the

degree that citizens who are most exposed to news are also most biased in their

information processing, their knowledge of where the candidates stand on the issues

will be sketchy at best. 19

19 To test for the possibility of curvilinear relationships, I re-run the cross-sectional models using both the
ANES and NAES surveys including two measures of news media use: the original variable—now
representing linear effects—and the square of the variable—now representing quadratic effects. The
estimations yielded insignificant effects on political knowledge for both measures of news consumption.
107
 A fourth regular pattern in the findings of the study pertain to the central

mediating role played by personal communications on the influence of emotions on

both political knowledge and political participation. More specifically, negative emotions

towards the candidates were found to predict how often respondents' talked about

politics. There are a number of explanations available from previous research that help

understanding the nature and importance of this finding. It could be that a negative

affective state signals that some action needs to be taken in order to obtain a positive

outcome and prevent a negative one (Schwarz & Clore, 1996). This is particularly true of

anger, which has been dubbed an "approach" emotion because it impels people to a

behavioral response (Carver & Harmon-Jones, 2009). If that is the case, then political

discussion may offer an efficient route to political action—and, indeed, the results in

Chapter 4 confirm a strong link between frequency of discussion and political

participation. Relatedly, previous research has found that anger has a strong

motivational component (Eagly & Chaiken, 1993; Nabi, 2002). To the extent that

political talk is more likely among individuals with higher political interest, a likely

outcome of being angered at political figures is to share with others those feelings as

well as coordinating political action through personal communication. Anxiety—the

other component of negative emotions studied in this dissertation—can also lead to

talking about politics more frequently, albeit through a different mechanism. This

assertion is based on previous work by Valentino and colleagues (2011), who gathered

data from surveys and randomized experiments to compare the differential effects of

anger and anxiety. Their results suggest that anxiety—but not anger—boosts political

However, the sign of the coefficients for the squared term was consistent with an inverted U-shape curve.
In other words, had this coefficient achieved significance, it would have suggested that issue knowledge
was highest for those in the moderate category of news use and lower for those at either end of the
distribution of the variable.
108
behavior that is less costly, such as information-seeking and opinion expression. In this

sense, interpersonal political communication may offer an inexpensive way to obtain

political information.

The fifth noteworthy trend in the results of this dissertation also pertains to the

mediating role of political discussion, though in this case as a channel for the effects of

media use on political involvement. It has long been argued that communication within

social networks is essential for media effects on political engagement (e.g., Berelson,

Lazarsfeld, & McPhee, 1954; Katz & Lazarsfeld, 1955; Lazarsfeld, Berelson, & Gaudet,

1948). This is because conversations involve not only exchanges of information but also

interpretive frameworks that help to process media messages (Eveland, 2004; Feldman

& Price, 2008; Scheufele, 2002; Schmitt-Beck, 2008). By allowing people to grapple with

ideas, elaborate arguments and reflect upon the information acquired, conversations

are a rich form of political information (Huckfeldt & Sprague, 1995). Specific to the

measure of political discussion employed here, Graber (2001) argued that greater

frequency of discussion increases the probability of incorporating into one's current

understanding of a topic a greater number of facts. The relationship between political

discussion and participatory behavior is also a finding consistent with a long line of

research in political communication.

The mechanism assumed to account for this effect is (...) the more people
interact with one another within a social context, the more norms of
participation will be transmitted, and the more people will be recruited into
political activity (Mutz, 2002, p. 839).

Of course, finding that people who often talk about the campaign happen to

know more about the candidates and are more active than their less politically talkative

peers says nothing about the specific participatory effects of like-minded versus

heterogeneous social networks—a hotly contested topic in political communication (cf.,

109
Eveland & Hively, 2009; Feldman & Price, 2008; Huckfeldt, Johnson, & Sprague, 2004;

McClurg, 2006; Mutz, 2006; Nir, 2011; Valenzuela, Kim, & Gil de Zuniga, Forthcoming).

However, considering that the measure of discussion across the ANES and NAES surveys

was based on personal communication with family and friends—the so-called "strong

ties" (Granovetter, 1973)—it could be argued that the results refer to the effects of

homogenous, rather than diverse, social networks. The lack of measures of the level of

agreement and disagreement among discussion partners, however, makes this claim

speculative at best.

Lastly, a prominent finding from the analyses refers to the televised debates

between major-party presidential candidates. On the one hand, there was evidence that

both positive and negative emotions were significant predictors of watching the

debates. This could mean that people who have strong feelings towards the candidates

are more likely than their apathetic counterparts to tune in to the debates. If that is the

case, then this finding would be another evidence for the motivating force of political

emotions argued earlier. As for which type of emotional set, positive or negative, was

more predictive of debate viewing, the results were not clear cut. This is, indeed, a

methodological problem. The NAES survey was designed specifically to examine in

depth the role of the debates on voters' attitudes, knowledge and involvement.

Unfortunately, it only measured negative emotions. The ANES survey, in contrast,

gauged both types of emotions but the time-span between interviewing waves was not

designed to measure with precision the role of the debates. Thus, I cannot compare if

the influence of positive effects on debate viewing that was manifest in the analyses of

the ANES survey was unique to this data set or not. Nevertheless, it is not far-fetched to

say that some level of enthusiasm with the campaign is necessary for people to tune in

to the debates.
110
 It has also been argued that debate viewers are a self-selected population, in

that they are generally more interested in political affairs than the general audience

who tunes into the news, oftentimes, for non-political stories (Hansen & Benoit, 2007).

To the degree that political interest is positively related to intensity of feelings and

attitudes towards the candidates (Bizer, Visser, Berent, & Krosnick, 2004), it may well be

that viewers of the televised debates come with strong positive feelings towards their

preferred candidate and equally—if not stronger—negative feelings for the opposite

candidate. Thus, both sets of emotions will be manifest prior to debate viewing.

On the other hand, following the debates had a unique influence on political

discussion and political involvement, particularly on knowledge of the candidates' issue

positions. This is to be expected; debates are a unique media event in the American

political landscape and, as such, they still command large audiences. Each debate

between Obama and McCain on 2008 drew, on average, 57 million viewers (Nielsen

Media Research, 2008). For the 2004 debates between Bush and Kerry, the average

viewership hit 53 million. Contrast these figures with the 36 million viewers who sat to

watch the Clinton-Dole debates of 1996—and this was before the Internet age fully

blossomed, distracting many from political campaigns altogether ever since (Prior,

2007). With these impressive levels of attention, it is to be expected that debates have a

sizeable effect on the audience on a range of political outcomes, especially when

compared to habitual news media use. Prior research is consistent with such an

expectation. For instance, a meta-analysis estimating the influence of various forms of

communication found that debates were significantly correlated to increased levels of

issue knowledge in more than two-thirds of the 31 studies sampled. Political discussion

and newspaper exposure, in turn, produced significant results in half of the studies,

111
while the proportion for television news was, at one third, significantly lower (see

Hansen & Benoit, 2007).

The learning potential of debates is due, in part, to the opportunities afforded by

the debates for candidates to lay out their policy views. Contrary to common wisdom,

most of the utterances made by the candidates on televised debates relate to policy,

not to personal character (Benoit et al., 2002). Compared side by side, the proportion of

issue content on both newspapers and television news is significantly lower than the

proportion of issue content available during the presidential debates (Hansen & Benoit,

2007). It is reasonable, therefore, to find that citizens' campaign issue knowledge

increases from watching debates.

Taken together, the trends reported in this study have implications for current

scholarly debates on citizen competence, that is, the ability with which citizens can

make a meaningful, informed decision when they are asked to choose the leader of the

Executive power. Here, it has been shown that passionate politics is positively related to

desirable outcomes, such as knowing the issue policies of the candidates and getting

involved in campaign activities. Rather than distracting citizens from deliberative

behaviors, positive and negative emotions are closely related to information seeking in

the media and more frequent informal conversations about politics. These behaviors, in

turn, make for a more informed and participatory electorate. In this sense, the

implications of this study suggests that emotion is not only an unavoidable fact of

political campaigns in the U.S., but a desirable fact altogether.

STRENGTHS OF THE STUDY

The conclusions reported throughout this project are bolstered by the

employment of different data sets across election years, as well as by the comparison of

112
a theorized structural model with an alternative causal model using a cross-sectional

approach relating individual differences and a longitudinal approach relating change

over time within individuals. These multiple tests of the affective citizen communication

model were necessary in order to provide more convincing evidence of the causal

relationships between emotions, communication and political involvement. Further, it

has been argued that replication is a necessary step for the advancement of science, as

it allows for generalizing the finding of particular effect as well as improving the

methodology for detecting it (King, 1995). In this regard, the variety of tests detailed in

Chapter 4 fulfill to some extent the need for replication. The unique characteristics of

the 2008 campaign (e.g., no incumbent, first African-American major-party candidate,

worst financial crisis since 1929, etc.) would have made suspicious any claim of

generalizability obtained from the analysis of the ANES surveys. The fact that there was

an important degree of consistency with the findings reported using the NAES surveys

for the 2004 election reduces this concern. Therefore, the employment of panel data

and the multiple replications of the affective citizen communication model are a

strength of this project.

Another strength is that, while grounded on existing theories of political

communication, the study covers new theoretical territory. Most scholars in the

affective intelligence tradition have assumed a direct influence of feeling emotions

towards political objects (e.g., candidates and issues) and individuals' learning of, and

engagement with, political affairs. In Brader (2006a), I would argue that there is

sufficient evidence pointing out that not all voters are equally affected by their

emotions, and even if feeling enthusiasm, fear or anger at candidates is common, levels

of issue knowledge and participation in political activities are not particularly high. Thus,

something else beyond emotions must be at play between emotions and involvement.
113
The model advanced here provides (a very partial) map of the process of becoming

informed and engaged with presidential campaigns through emotions. In particular, it

suggests that the communication behaviors triggered by positive and negative feelings

towards candidates shapes, in part, how much they learn and participate over the

course of a campaign. On the other hand, scholars in the tradition of the communication

mediation model have implicitly assumed that the psychological predictors of media use

and personal discussion refer to political interest and surveillance motivations. Here, I

present the case for emotions as unique determinants of these communication

behaviors and provide ample evidence for it. Therefore, the project is at one time an

exploration of new territory as well as a return to known territory. It opens up new

communication-oriented processes for affective intelligence theory and traces citizen

communication mediation back to emotions. It confirms that emotions are conducive to

learning and political action, and that media effects on citizen engagement tend to

operate through political discussion. Thus, borrowing McCombs' (1992) metaphor of

agenda-setting researchers, the affective citizen communication model is the outcome

of both explorers' and surveyors' work.

A third strength of the research refers to the external validity of the findings.

Experimental work on the political effects of emotions is vast and growing (e.g., Brader,

Valentino, & Suhay, 2008; Gross, 2008; McClain, 2009; Miller, 2007; Miller & Krosnick,

2004; Valentino, Brader, Groenendyk, Gregorowicz, & Hutchings, 2011; Valentino,

Hutchings, Banks, & Davis, 2008). The difficulty of manipulating various variables at the

same time, however, has translated into research that mostly tests direct effects of

political emotions. At most, researchers have included self-reports of one mediating

mechanism (e.g., political interest, informational utility, etc.) A thorough experimental

test of the affective citizen communication model, in contrast, would require

114
manipulating simultaneously several variables—some of which are not amenable to

experimentation. For instance, the few published experimental studies aimed at

detecting a causal role of emotions on informal political talk have failed to produce

significant results due to problems in the manipulation of treatment conditions (see

Kim, 2009; McClain, 2009). On the other hand, even if one were to come up with a solid

experimental design to test this study's theorized model, all we would know is that the

proposed relationships between emotions, communication behaviors and political

involvement can happen—not that they, indeed, happen—in real-world contexts. In this

sense, the findings obtained from nationally representative surveys are better equipped

to inform us about the ability of the proposed model to adequately describe the

affective and communication processes that citizens experience over the course of a

presidential campaign.

METHODOLOGICAL LIMITATIONS

Having argued for the use of observational research to test the affective citizen

communication model does not mean that the study is impermeable to the limitations

imposed by surveys. A first problem deals with causal attribution. So as not to repeat

the earlier point on the utility of experiments versus surveys, I will concentrate on some

of the inherent problems of relying on the panel design of the ANES and NAES studies to

assess causality, as well the shortcomings of the structural equation models employed

to test the affective citizen communication model.

A serious disadvantage of panels refers to the possibility of conditioning effects

by repeating the same questions over time, that is, an initial interview may affect a

subsequent interview so much that observed changes may result from nothing but the

previous interview. To these conditioning effects, one must add the problem of attrition

115
or mortality effects, which occurs when a portion of the original respondents is not

reinterviewed and the sample becomes increasingly unrepresentative of the population.

In either case, noted Bartels, "the observed (...) change [in the variables of interest] in

the surviving panel sample may provide a biased estimate of the corresponding (...)

change in the relevant population" (Bartels, 1999, p. 3). Panel mortality is less of a

concern for the ANES panel than it is for the NAES panel, because the re-contact rate for

the former was, at 93%, much higher than the re-contact for the latter (41%). Still, all

the analyses presented here employed weights to adjust for the probability of being

selected for the surveys, relieving somewhat the concerns over lack of

representativeness due to panel attrition.

Less clear is the magnitude of possible conditioning effects. One common

strategy to study these effects is to compare the responses from the panel sample with

responses from a cross-sectional sample that has not been exposed to previous

interviews. Bartels (1999) conducted such an analysis for the ANES panels of 1992 and

1996 and concluded that conditioning effects were rare in most cases. A similar exercise

does not seem to be available for the NAES panel. Therefore, additional analyses to

tease out panel effects are warranted.

A second limitation of the research refers to the structural equation models

employed for the longitudinal analyses. Chapter 4 presents the results of synchronous

models, where the direction of influence runs one way from emotions to knowledge and

participation via communication. The choice of a synchronous approach was borne out

by the data; the lagged models had worse fit than the synchronous models. Of course, it

goes without saying that a model with good fit is not necessarily a correct model—in

fact, all models are wrong because they are an over-simplification of a complex reality

(Shoemaker, Tankard, & Lasorsa, 2004). For this reason, a more crucial test for the
116
causal attributions made by the theorized affective citizen communication process was

provided by the estimation of another synchronous model with an alternative causal

ordering of the variables, with the proposed model exhibiting a better fit with the data.

On the other hand, previous research using two-wave panels to assess the effects of

communication variables on political knowledge and participation has also found that

the synchronous approach outperforms the lagged approach (see Eveland, Hayes, Shah,

& Kwak, 2005a, 2005b; Shah, Cho, Eveland, & Kwak, 2005). Nevertheless, the issue of

lagged versus synchronous models when examining the influences between political

feelings, media messages, personal communications and involvement could be further

explored with panel data comprising more than two waves.

The ability to draw causal conclusions is also limited based on the time lag

between waves. Previous work in political science and political communication is mixed,

at best, in terms of the ideal time-lag for estimating the effects of emotions.

Experimental work on the influence of emotions and communication assumes, by the

very nature of the method, short-term effects. On the other hand, survey-based

research considers longer time spans. For instance, in their seminal work on affective

intelligence, Marcus and MacKuen (1993) used the ANES 1980 Panel, which interviewed

respondents every two months and a half between February and September. This lag-

length is longer than that between waves 9 through 11 of the ANES 2008-2009 Panel

employed in the current study. The time lag between pre- and post-waves of the NAES

2004 Debates Panel was even shorter; a few weeks for most respondents. Certainly, the

lack of consensus on proper lag specification is a problem for most panel research

designs (Finkel, 1995), not only for work on political emotions and communication. Still,

it is a problem that merits some caution in the interpretation of the results because it is

possible that some of weaker or null effects found among some variables—particularly
117
between affective and communication variables—could be more robust or significant

with the specification of a proper time lag. Again, the recommendation is for future

research based on panel surveys to test for different time lags when testing the

relationships among the variables of the affective citizen communication model.

Lastly, there are limitations in the measurement of the variables which need to

be further addressed. As with any secondary analysis of surveys, the researcher has to

work with questions collected by another organization for perhaps completely different

purposes. The lack of measures of news attention in the ANES 2008-2008 Panel is

unfortunate, considering the well-known validity problems of self-reports of news

exposure items (Bartels, 1993; Prior, 2009a, 2009b). Likewise, the lack of measures for

positive emotions elicited by candidates Bush and Kerry in the NAES 2004 Debates Panel

prevents a direct replication of the effects encountered for feelings of enthusiasm in the

ANES data. On other hand, the items to measure emotions are far from exhaustive.

Using a longer list of emotions terms, some researchers have identified an additional

dimension, labeled aversion, that is distinct to the typical enthusiasm and anxiety

dimensions (e.g., Marcus & MacKuen, 1996). Perhaps employing a longer list of affective

terms could have yielded a different picture than the dual model found in the study.

Nevertheless, the list of items on emotions measured did form a structure consistent

with affective intelligence theory and the dual-model of affect advanced by Watson and

colleagues (1988) using the PANAS list of 20 emotion terms. This issue, of course, is as

much a methodological quandary as a theoretical one because appraisal theories of

emotion (Roseman, Wiest, & Swartz, 1994; Scherer, 1999; Smith & Kirby, 2001) assert

that each emotion has a discrete effect on individuals' attitudes, cognitions and

behaviors. Thus, a fruitful venue for future research could be to explore the predictive

power of alternative theoretical models of affect.

118
EXPANDING THE AFFECTIVE CITIZEN COMMUNICATION MODEL

The theoretical model thus advanced and tested posits that emotions citizens

feel towards presidential candidates lead them to information-seeking behaviors,

including news media use, debate viewing and interpersonal political discussion, which

in turn lead to political learning and campaign participation. This model can certainly be

expanded to incorporate additional influences that have been left out in the original

formulation. Furthermore, the model can be used to explain additional phenomena than

markers of political involvement. I will briefly outline some of these refinements to, and

alternative applications of, the model.

Bringing in the Moderators

The effects outlined in the affective citizen communication may not be constant

across different segments of the electorate. A variety of moderating variables could be

incorporated into the model to account for these differences. In the context of a

political campaign, variations in the motivating force of emotions may well be

conditional on the relevance of the election itself and on individuals' perceived ability to

undertake political action. 20 People who perceive the election to be personally

irrelevant tend to rely less on the media because they have less need for orienting cues

(McCombs & Weaver, 1973; Weaver, 1980). On the other hand, individuals with little or

no interest in political affairs—another dimension of relevance (Bizer, Visser, Be rent, &

Krosnick, 2004; Lee, 2005)—are less likely to be familiar with the candidates (Kiousis &

McCombs, 2004), which is strongly associated with more intense feelings towards them

(Miller, 2007). Thus, the relationships between the different components of the

20 These variables, as may me remembered, where included as controls in the analyses due to their
known relationship with media use, political discussion, issue knowledge and political participation. Here,
I am arguing for treating them as variables of interest, instead of serving as controls.
119
affective citizen communication model may be stronger among individuals who perceive

the election as relevant compared to individuals who perceive it as irrelevant. Future

research could bring in relevance as a moderator of the relationship between emotions

and information-seeking behaviors as well as between communication and political

involvement.

Self-efficacy or internal political efficacy—to use the label preferred by political

scientists (Morrell, 2003; Niemi, Craig, & Mattei, 1991)—has been found to moderate

the relationship between anxiety and political participation (Rudolph, Gangl, & Stevens,

2000). It has also been found to be a causal factor in developing feelings of anger, which

in turn relates to more active participation in campaign activities (Valentino,

Gregorowicz, & Groenendyk, 2009). These studies suggest that internal political efficacy

can play various roles in the process linking feelings with political involvement, as both

an antecedent and moderator. Both roles could be re-examined in the context of the

affective citizen communication model by testing, for instance, a structural model

running from efficacy to emotions and communications, and comparing it to another

model in which efficacy interacts with emotions and, perhaps, communications, to

explain political involvement. Of course, it may well be that self-efficacy plays dual roles

in the whole process. Future research could delve deeper into this issue by

incorporating internal efficacy into the processes outlined in the affective citizen

communication model.

A third factor not considered in the study that could moderate the relationships

posited by the proposed model refers to the composition of individuals' discussion

networks. This project corroborated the literature on the effects of frequency of

personal discussions on both learning and participating in political activities (Delli

Carpini, Cook, & Jacobs, 2004; Eveland, 2004; Just et al., 1996). Yet, it is left to discern
120
what specific attributes of discussions may spur such outcomes. Particularly fruitful has

been the study of how political disagreement among discussants relates to participation

(Brundidge, 2010; Huckfeldt, Mendez, & Osborn, 2004; Mutz, 2006; Valenzuela, Kim, &

Gil de Zuniga, Forthcoming). Thus, future research could investigate the relationship

between emotions and exposure to agreement and disagreement within individuals'

social networks.

There is tentative evidence that people whose feelings towards candidates are

more polarized (e.g., strong positive emotions towards the preferred candidate and

strong negative emotions towards to opposite candidate) tend to talk more with people

who share their political views and less with people who do not share their views

(Parsons, 2010b). This finding raises the possibility that the influence of emotions and

political discussion frequency examined in this study could operate differently at various

levels of discussion disagreement. The existing literature also shows that disagreement

in informal political discussions can regulate the effects of both media use and

discussion frequency on issue knowledge (Feldman & Price, 2008). One can think of a

number of reasons for this findings, such as the confusion and ambivalence brought

about by being exposed to contradictory information through personal communications.

However it may be, network disagreement is a likely candidate for moderator of the

processes implied by the affective citizen communication model.

A fourth contingent factor for the relationships implied by the model refer to

contextual effects. The question here is what difference does it make for the outcomes

predicted by the model country-level characteristics, such as political system, civic

culture and media system. It may well be that emotions in presidential campaigns have

different effects in other types of campaigns, such as referendum campaigns. To the

extent that the affective citizen communication model refers to psychological processes
121
derived from neuroscientific evidence, some level of uniformity in the associations thus

mapped out should be evident. Nevertheless, having an open press, competitive

elections, and freedom of expression seem necessary conditions, too, and these factors

are not constant across countries.

Beyond Political Involvement

The focus of this study has been on the influence of emotions and

communication behaviors on two dimensions of political involvement, knowledge and

participation. However, other political and non-political outcomes could be studied

using the proposed theoretical model. I will sketch some of these possibilities.

The hierarchy of influences model posits that campaigns activate cognitions,

attitudes and behaviors (McGuire, 1989). A notable absence from the current study is

the attitudinal component. For instance, how do emotions and communication aspects

relate to voters' political attitudes? Future research could fill in this gap by addressing

the role of attitudes towards candidates in their dual role of being an antecedent and a

consequence of affective and information-seeking processes. For instance, experimental

work by Way and Masters (1996b) revealed direct and indirect effects of participants'

anxiety on their evaluations of politicians. Thus, it would be apt to test if the affective

citizen communication model can predict differences in the ratings of various public

figures. Reversing this direction of causality could also prove a fruitful venue for

research. Since the early work on affective intelligence theory, scholars have analyzed

alternative mechanisms for explaining the relationship between emotions, learning and

decision-making. Of relevance here is the significance of the association between

emotions and political evaluations. For some researchers (e.g., Ladd & Lenz, 2008,

2011), positive and negative emotions towards candidates are rationalizations of a

122
causally prior evaluative process; liked candidates are more likely to provoke

enthusiasm, while disliked candidates are more likely to provoke anxiety and anger.

Others scholars rely on experimental evidence to support a view that the effect of

emotions on attitudes is stronger than the reverse (Brader, 2011; Marcus, MacKuen, &

Neuman, 2011). Of course, it may well be that—as it happens with processes of media

selectivity and media effects (Slater, 2007)—there is a reciprocal, reinforcing

relationship between feelings and attitudes elicited by political figures. However it may

be, the previous discussion should make it readily clear that attitudinal outcomes, such

as evaluations of candidates, are one way of moving forward the affective citizen

communication model.

In addition to new political outcomes, the proposed model could be tested to

explain non-political phenomena. Instead of knowledge of the candidates' issue

positions, one could test knowledge of local news, social issues, scientific discoveries,

businesses or any other domain of news and current events likely to trigger strong

feelings in the public. In addition to campaign participation, future research could delve

into civic participation, or in participation in issue-specific activities (e.g., a health

campaign). Once the targets of emotional appraisal are properly identified, it should be

a simple matter to adapt the wording of the media and interpersonal communication

variables to reflect the specific domain under study.

CONVERGENCE WITH OTHER COMMUNICATION THEORIES

The current work has been informed by affective intelligence theory and

communication mediation processes. These paradigms offered solid theoretical ground

on which to expect a relationship between emotions, media use, personal discussions

and political involvement. The implications of the relationships established by the

123
proposed model, however, need not be restricted to affective intelligence and

communication mediation. In the following paragraphs, I will point out possible areas of

convergence with prominent theories of communication (namely, agenda setting,

priming, and selective exposure).

Agenda Setting and Priming

As mentioned in Chapter 1, agenda-setting theory as developed by McCombs

and colleagues (McCombs, 2004; McCombs & Shaw, 1972; Shaw & McCombs, 1977;

Weaver, Graber, McCombs, & Eyal, 1981) is one of the most influential theories of

media effects. What started as a metaphor to describe the role of the news media in

defining the salience of issues for the public now comprises five different domains, and

scholars worldwide have found new applications for the theory (e.g., Coleman &

McCombs, 2007; Huck, Quiring, & Brosius, 2009; Matthes, 2008; Shehata, 2010; Stroud

& Kenski, 2007; Valenzuela, Forthcoming). Within these domains, three seem to be most

compatible with the affective citizen communication model.

First, it is a well-known fact that the narrative structure of news can exert

considerable effects on the public's emotional response to issues and public figures

featured in the news. In fact, experimental researchers usually manipulate emotional

states in the lab by exposing participants to media messages (e.g., Valentino, Banks,

Hutchings, & Davis, 2009). One way in which agenda setting deals with the effects of

specific news narratives on audiences is through the notion of "attributes," those

characteristics and traits that describe and define objects in the news. While some

attributes are emphasized, others receive less attention, and many receive no attention

at all. Just as objects vary in salience, so do the attributes of each object. Thus, for each

object there also is an agenda of attributes. The question of interest here is which

124
attributes in the news are more likely to produce enthusiasm, anxiety, anger and other

feelings on the audience. Existing research on attribute agenda-setting has focused

mostly on cognitive (e.g., personal qualities of candidates), attitudinal (e.g.,

thermometer ratings of candidates), and behavioral effects (e.g., vote choice) (Golan &

Wanta, 2001; McCombs, Lopez-Escobar, & Llamas, 2000; Son & Weaver, 2006;

Valenzuela & Correa, 2009). To my knowledge, the only study that has explicitly

incorporated emotions into attribute-agenda setting is Coleman and Wu's (2010), and

its results were promising: there was a significant, positive relationship between the

content of TV visuals about the 2004 U.S. presidential candidates and the public's

negative emotional responses towards Bush and Kerry. More research is needed to

examine the attributes in the news that arouse stronger feelings and compare these to

people's emotional responses—adding emotional effects to the process of attribute

agenda setting. For instance, Tan and Weaver (2010) found that different sources in the

news evoked different affective responses among the public, a study that opens the way

to examining a host of news story attributes. This type of research endeavor would have

the added benefit of tracing the origins of emotions, which in the current study were

conceptualized as exogenous.

The second domain in agenda-setting research that seems convergent with the

affective citizen communication model refers to need for orientation, a key

psychological moderator of agenda-setting effects (Chernov, Valenzuela, & McCombs,

2009; Matthes, 2006; Weaver, 1980). As explained earlier, a key defining condition for

individuals' need for orientation is relevance. When people find an issue relevant, they

are more likely to seek information about it on the media, which increases the likelihood

of being influenced by it. Among the various dimensions of relevance, both survey (Evatt

& Ghanem, 2001; McCombs, 1999) and experimental (Miller, 2007) research points to
125
emotional relevance as a major determinant of agenda-setting effects. This evidence is

consistent with the affective citizen communication model, which predicts that stronger

feelings lead to increased information-seeking in the media, with the consequent

learning and behavioral effects described earlier. Thus, researchers interested in the

concept of need for orientation may find it useful to look at the theoretical and

empirical affordances of the model advanced in this dissertation.

A third point of convergence is offered by the consequences of agenda setting

for opinions about political figures and public issues. Among these consequences, media

priming—the influence of the news media on the criteria used to evaluate political

objects—remains the most widely investigated (Domke, Shah, & Wackman, 1998;

Iyengar & Kinder, 1987; Iyengar & Simon, 1993; Miller & Krosnick, 1996; Valenzuela,

2009). A more restricted version of media priming refers to attribute priming, the

phenomenon by which the attributes emphasized in press coverage determine the

weight individuals assign to those same attributes in their political judgments. Sheafer

(2007) studied five elections in Israel and found consistent evidence that the negative

tone with which the media covered the economy had a subsequent effect on

evaluations of performance of the incumbent political party. This finding raises the

question of what explicates this relationship. Is it a cognitive process of attribution of

responsibility for the state of the economy to the incumbent party, or an affective

process of news making voters feel anxious, which in turn causes them to pay more

attention to the economy and thus update their evaluation of the incumbent party? As

the affective citizen communication model posits, there may well be other intervening

processes involved as well, such as political discussion within individuals' social

networks. This example illustrates that one potential area for future research in priming

refers to the emotional basis for voters' political judgments.

126
Selective Exposure

The increasing segmentation of the news market into niche audiences, coupled

with the nearly unlimited choice people have to choose their media preferences has

reinvigorated scholarly interest on selective exposure—the process by which people

deliberately select information channels that match their predispositions and beliefs

(Stroud, 2008). Previous research demonstrates that the way individuals select

information is determined, in part, by their emotional state. For instance, anxiety can

motivate selecting particular content that is useful to address the problem that caused

anxiety in the first place (Valentino, Banks, Hutchings, & Davis, 2009). This is to be

expected. As affective intelligence theory predicts, fear and anxiety lead to increased

attention, processing and recall of information that is related to the source of these

emotions (Boyle et al., 2004; Civettini & Redlawsk, 2009; Redlawsk, Civettini, & Lau,

2007). This perspective offers a more nuanced view on information-seeking behaviors

than the one presented in previous chapters. It may well be that negative and positive

emotions do not lead to news consumption in general as has been stated but, say, to

news consumption in high-choice media platforms only (e.g., cable television news and

online news). Or, perhaps, in some contexts emotions may lead to choosing

interpersonal channels of information over mediated channels, rather than to both

forms of communication as was proposed here. The basis for these expectations would

be that the informational utility of each channel is different for different emotional

states.

Some evidence of the relationship between selectivity processes and preferred

communication modality could also be derived from this dissertation's finding that

negative emotions were more closely associated with political discussion than they were

with habitual news use and debate viewing. If future research replicates this result, the

127
next step would be to theorize about the causes behind the differential effects of

emotions. And selective exposure may provide a useful answer.

A second path of convergence relates to individual-level effects of selective

exposure. Research by Stroud (2011) provides ample evidence that exposure to

politically congenial news outlets leads to political reinforcement and polarization, at

least during U.S. election campaigns (e.g., conservative Republicans are more likely to

watch FOX News, and by doing so, their political opinions become more entrenched and

extreme over time). Thus, a developing area of study is the extent to which exposure to

the partisan press contributes to the polarization of emotions towards political issues.

For instance, a study conducted in Switzerland using a three-wave panel survey

detected significant effects between selective attention paid to political advertising

about immigration and negative emotions towards immigrants (Schemer, 2010). These

results are interesting because they emphasize the causes of political feelings, which —

as explained earlier—were not addressed empirically in the dissertation.

SUMMARY AND CONCLUSIONS

This study applied affective intelligence theory and the communication

mediation model to further our current understanding of the consequences of emotions

on learning and participating in the context of presidential campaigns. An affective

citizen communication model connecting emotions, communication behaviors and

political involvement variables was proposed and tested using secondary analysis of

nationally representative panel surveys from the 2004 and 2008 U.S. elections. The

results of various structural equation models specifying concurrent and longitudinal

relationships yielded strong support for the theorized model.

128
 Affective intelligence posits that learning and behavior are determined by a dual-

system of emotional appraisal. The disposition system, characterized by emotions

related to enthusiasm and/or aversion, regulates individuals' habitual routines. The

surveillance system, activated by fear and anxiety, regulates individuals' information-

seeking behaviors, which lead to learning and involvement. Communication mediation,

on the other hand, proposes that media use affects knowledge and participation

indirectly, by triggering a series of intra- and interpersonal deliberative behaviors, such

as mental elaboration and political discussions with family and friends.

Based on these two bodies of work, the current project theorized that positive

and negative affect towards presidential candidates is related to three information-

seeking behaviors: news media use, debate viewing and political discussion. These

communication processes, in turn, are associated to each other and can, directly and

indirectly, cause issue learning and campaign participation. To test the model, data from

the ANES 2008-2009 Panel and the NAES 2004 Debates Panel studies was used. Overall,

40 of 54 tests of the hypotheses were supported, a proportion of 74.1%. The limitations

of the study notwithstanding, the tests of the model indicate that emotions are

connected to communication in the ways that were predicted, and these two factors

jointly affected political outcomes.

Future research, as was discussed earlier, can refine the model in a number of

directions. One is by incorporating moderating variables, such as relevance and internal

political efficacy. Another is by distinguishing between media platforms and examining

the attributes of political discussion (e.g., agreement and disagreement with discussion

partners). The affordances of the model to agenda setting, priming and selective

exposure—key theories of communication—were also highlighted as possible venues

for future studies. By pursuing these lines of research, we can refine our knowledge on
129
the influence of emotions on political life and better specify the role that media and

personal communications bear in this process.

130
 Appendix

This appendix includes information on the demographic characteristics of the

ANES and NAES samples used in the dissertation, along with a comparison with

population parameters gathered by the U.S. Census (Tables A.l and A.2), as well as

detailed results of the factor analyses of emotions items described in Chapter 3 (Tables

A.3, A.4, and A.5). The partial correlation matrices used to estimate the structural

models (Tables A.6 and A.7) and the multiple regressions estimating the effects of the

control variables (Tables A.8 and A.9) are also displayed. The goodness-of-fit statistics of

the synchronous and lagged auto-regressive structural models described in Chapter 3

are also included in this section (Tables A.10 and A.11). The Appendix ends with tables

showing all direct, indirect and total effects of the estimated synchronous models

(Tables A.12 and A.13).

131
Table A.l Demographic Profile of ANES 2008-2009 Panel Sample and U.S. Census Data

ANES Wave 11 ANES Wave 11 CPS March

Unweighted Weighted 2008
Age
18-29 8.3% 18.0% 21.3%
30-39 15.3% 17.1% 16.4%
40-49 21.6% 20.4% 19.3%
50-59 25.2% 19.1% 18.3%
60-69 19.0% 12.7% 12.6%
70 or older 10.6% 12.6% 12.2%
Gender
Male 42.1% 47.5% 48.0%
Female 57.9% 52.5% 52.0%
Race / ethnicity
White 87.6 83.7% 83.4%
African American 9.2% 12.0% 12.5%
Asian 3.9% 4.0% 3.7%
Native American / Alaska 1.9% 2.4% 1.8%
Pacific Islander 0.9% 1.2% 0.3%
Hispanic 4.9% 7.5% 9.5%
Other 6.5% 8.4% 0.0%
Education
Less than high school 3.3% 9.6% 11.2%
High school diploma 15.6% 30.9% 31.7%
Some college 36.9% 30.7% 29.6%
Bachelor's degree 24.6% 19.3% 18.5%
Graduate degree 19.6% 9.6% 9.0%
Annual household income
Less than $14,999 5.5% 5.9% 8.9%
$15,000 to $29,999 10.7% 12.9% 13.6%
$30,000 to $49,999 21.7% 22.7% 18.0%
$50,000 to $74,999 22.5% 24.1% 19.2%
$75,000 to $99,999 15.3% 14.8% 14.4%
$100,000 or more 24.3% 19.7% 26.1%
Notes: Estimates use the 2,312 respondents who completed Waves 9, 10 and 11 of the
ANES Panel Study 2008-2009. Distribution of race/ethnicity does not at 100% because
respondents could check more than one category.
Source: (Table 17 of DeBell, Krosnick, & Lupia, 2010, p. 79).

132
Table A.2 Demographic Profile of NAES 2004 Debates Panel Sample and U.S. Census
Data

NAES Post- NAES Post- CPS

Debate Debate November
Unweighted Weighted 2004
Age
18-29 9.1% 14.5% 19.9%
30-39 15.7% 21.8% 18.3%
40-49 21.1% 22.2% 21.2%
50-64 31.2% 24.3% 23.8%
65 or older 22.9% 17.2% 16.9%
Gender
Male 41.8% 44.1% 47.6%
Female 58.2% 55.9% 52.4%
Race / ethnicity
White 89.2% 81.8% 80.4%
African American 5.8% 11.1% 13.1%
Asian 1.1% I.5% 4.6%
Hispanic 3.9% 7.8% 14.4%
Other 3.9% 5.6% I.9%
Education
Less than high school 3.9% 7.1% 14.3%
High school diploma 19.5% 32.4% 32.3%
Some college 21.2% 22.0% 19.2%
Bachelor's degree 32.7% 24.6% 25.5%
Graduate degree 22.8% 13.9% 8.7%
Annual household income
Less than $14,999 8.7% 10.3% 13.3%
$15,000 to $24,999 7.9% 7.8% II.6%
$25,000 to $49,999 29.6% 33.3% 25.1%
$50,000 to $74,999 21.8% 21.5% 18.1%
$75,000 to $99,999 13.7% II.4% 12.0%
$100,000 or more 18.3% 15.6% 19.9%
Notes: Estimates use the 1,248 respondents who completed the NAES post-debate
panel survey in October 2004. Distribution of race/ethnicity does not add 100% because
respondents could check more than one category.
Source: Compiled by the author directly from NAES 2004 Debates Panel Study and CPS.

133
Table A.3 Exploratory Factor Analysis of Emotional Reactions to Presidential
Candidates in the ANES 2008-2009 Panel (Unrotated Solution)

Wave 9 data Wave 11 data

1 II 1 II
(Valence) (Arousal) (Valence) (Arousal)
Emotional reaction
Proud .81 .44 .73 .57
Hopeful .80 .45 .72 .58
Angry -.46 .78 -.63 .66
Afraid -.47 .77 -.62 .68

Eigenvalues 1.72 1.60 1.84 1.55

% Variance 43.0% 39.9% 46.0% 38.8%
(Valid cases) 2,303 2,305
Notes: Principal component analysis.

134
Table A.4 Exploratory Factor Analysis of Emotional Reactions to Presidential
Candidates in the ANES 2008-2009 Panel (Rotated Solution)

Wave 9 data Wave 11 data

1 II 1 II
(Positive / (Negative / (Positive / (Negative /
Enthusiasm) Anxiety) Enthusiasm) Anxiety)
Emotional
reaction:
Proud .92 -.01 .93 -.05
Hopeful .92 -.02 .93 -.03
Angry -.01 .90 -.05 .91
Afraid -.02 .90 -.03 .92

Eigenvalues 1.72 1.60 1.84 1.55

% Variance 43.0% 39.9% 46.0% 38.8%
(Valid cases) 2,303 2,305
Notes: Principal component analysis with varimax rotation and Kaiser normalization.

135
Table A.5 Exploratory Factor Analysis of Emotional Reactions to Presidential
Candidates in the NAES 2004 Debate Panel (Unrotated Solution)

Pre-Debate Wave Post-Debate Wave

1 1
(Negative / Anxiety) (Negative / Anxiety)
Emotional reaction
Afraid .81 .85
Uneasy .80 .82
Angry .80 .77

Eigenvalues 1.93 2.000

% Variance 64.4% 66.7%
(Valid cases) 782 786
Notes: Principal component analysis.

136
Table A.6 Partial Correlation among Key Variables in the ANES 2008-2009 Panel

W9 Positive Wll Positive W9 Negative Wll Negative W9 News W10 News Wll Debate
Emotions Emotions Emotions Emotions Use Use Viewing
W9 Positive Emotions 1.00
Wll Positive Emotions .56** 1.000
W9 Negative Emotions -.06** -.08** 1.000
Wll Negative Emotions -.06** -.03 .60** 1.000
W9 News Media Use .14** .07** -.004 .00 1.000
W10 News Media Use .10** .08** .02 -.01 .74** 1.000
Wll Debate Viewing .16** .17** .02 .01 .10** .15** 1.000
W9 Political Discussion .02 -.03 .10** .05* .19** .20** .14**

00*
o
Wll Political Discussion .00 -.001 .09** .17** .21** .24**
00*
o

W9 Political Knowledge .07** .02 .07** .03 .05* .17**

00*
o
Wll Political Knowledge .06 .01 .03 -.02 .01 .21**
Wll Political Participation .03 .02 .12** .07** .08** .10** .12**
Notes: Valid cases (using listwise deletion) = 2,267. Controlling for education, income, age, female, internal political efficacy,
strength of party identification, and habitual political interest.
* p < .05; ** p< .01
Table A.6 Partial Correlation among Key Variables in the ANES 2008-2009 Panel (continued from previous page)

W9 Political Wll Political W9 Political W9 Political Wll Political

Discussion Discussion Knowledge Knowledge Participation
W9 Positive Emotions
Wll Positive Emotions
W9 Negative Emotions
Wll Negative Emotions
W9 News Media Use
W10 News Media Use
Wll Debate Viewing
W9 Political Discussion 1.000
Wll Political Discussion .55** 1.000
W9 Political Knowledge .11** .15** 1.000
00*
o

Wll Political Knowledge .15** .55**

Wll Political Participation .17 .17** .14**
Notes: Valid cases (using listwise deletion) = 2,267. Controlling for education, income, age, female, internal political efficacy,
strength of party identification, and habitual political interest.
* p < .05; ** p< .01
Table A.7 Partial Correlation among Key Variables in the NAES 2004 Debates Panel

W1 Negative W2 Negative W1News W2 News W2 Debate W1 Political W2 Political

Emotions Emotions Use Use Viewing Discussion Discussion
W1 Negative Emotions 1.000
W2 Negative Emotions .70** 1.000
W1 News Use .09 .07 1.000
W2 News Use .09 .11* .76** 1.000
W2 Debate Viewing -.09 -.024 .27** .32** 1.000
W1 Political Discussion .27** .28" .20** .22** .19** 1.000
W2 Political Discussion .25** .28" .17** .31** .22** .66** 1.000
W1 Political Knowledge .15** .18** .08 .18** .14* .10 .17**
W2 Political Knowledge .07 .08 .02 .09 .36** .24** .25**
W1 Political Participation .27** .27** .18** .17** .12* .42** .36**
W2 Political Participation .30** .32** .13* .19** .16** .43** .44**
Notes: Valid cases (using listwise deletion) = 323. Controlling for education, income, age, female, internal political efficacy,
strength of party identification, and habitual political interest. W1 = pre-debates wave. W2 = post-debates wave.
* p < .05; ** p< .01
Table A.7 Partial Correlation among Key Variables in the NAES 2004 Debates Panel (continued from previous page)

W1 Political W2 Political W1 Political W2 Political

Knowledge Knowledge Participation Participation
W1 Negative Emotions
W2 Negative Emotions
W1 News Use
W2 News Use
W2 Debate Viewing
W1 Political Discussion
W2 Political Discussion
W1 Political Knowledge 1.000
W2 Political Knowledge .46 1.000
W1 Political Participation .13 .10 1.000
W2 Political Participation .25 .25 .80 1.000
Notes: Valid cases (using listwise deletion) = 323. Controlling for education, income, age, female, internal political efficacy,
strength of party identification, and habitual political interest. W1 = pre-debates wave. W2 = post-debates wave.
* p< .05; ** p< .01
Table A.8 Regression Analysis of Key Variables on Controlling Variables in the ANES
2008-2008 Panel

W9 W9 W9 Wll
Positive Negative News Debate
Emotions Emotions Use Viewing 21

Block 1: Demographics
Education 2Q*** -.03 .12*** 20***
Income .02 -.03 2Q*** .12***
Age .12*** .01 .33*** .25***
Female .06** -.01 -.003 .02
A R2 (%) 2.9% 0.3% 14.5% 9.4%

Block 2: Political Orientations

Internal Political Efficacy 2g*** -.02 .04* .01
Strength of Party .15*** .09*** .06** 07***
Identification
Habitual Political Interest .17*** ^g*** .33*** .33***
A R2
(%) 10.7% 4.7% 11.9% 9.8%
Total R 2 (%) 13.6% 5.0% 26.4% 19.2%
(Valid cases) (2,288) (2,287) (2,295) (2,293)
Notes: All control variables measured at wave 9. Cell entries show upon-entry
standardized OLS regression coefficients with standard errors in parentheses.
* p< .05; ** p< .01; *** p< .001

21 Considering that debate viewing is a count variable, a Poisson regression was performed in addition to
the OLS regression results shown in the table. There were no significant differences in the results of both
types of regression analysis. Thus, OLS estimates are shown for easier interpretation.
141
Table A.8 Regression Analysis of Key Variables on Controlling Variables in the ANES
2008-2008 Panel (continued from previous page)

W9 Political W9 Political Wll Political

Discussion Knowledge 22 Participation 23

Block 1: Demographics
Education .13*** .26***
Income .08*** .09***
Age 24*** ^g***
Female .01 .03
A R2 (%) 9.2% 13.6%

Block 2: Political Orientations

Internal Political Efficacy .11***
Strength of Party Identification .01
Habitual Political Interest
A R 2 (%)
Total R 2 (%) 20.9%
(Valid cases) (2,287)
Notes: All control variables measured at wave 9. Cell entries show upon-entry
standardized OLS regression coefficients with standard errors in parentheses.
* p< .05; ** p< .01; *** p< .001

22 Political knowledge is also a count variable. Nevertheless, the Poisson regression results were virtually

the same as the OLS regression results.

23 Political participation is a count variable; the results of a Poisson regression revealed no significant
differences with the results of the more familiar OLS regression shown in the table.
142
Table A.9 Regression Analysis of Key Variables on Controlling Variables in the NAES
2004 Debates Panel

W1 W1 W2 W1
Negative News Debate Political
Emotions Use Viewing 24 Discussion

Block 1: Demographics

00
*
Education

o
.12** .04 .06
Income -.06 .07 .09* .19**
Age .15*** .25*** 24*** -.06
Female .03 .03 -.01 -.11**
A R2 3.1% 7.5% 2.7% 7.5%

Block 2: Political Orientations

00

00
*

Internal Political Efficacy

o

o
.01 .13***
Strength of Party Identification .04 .00 .03 .00
Habitual Political Interest 20*** gg*** .31*** .35**
A R2 4.2% 13.5% 12.7% 13.2%
Total R 2 (%) 7.3% 21.0% 15.4% 20.7%
(Valid cases) (775) (783) (790) (793)
Notes: All control variables measured at the pre-debates wave. Cell entries show upon-
entry standardized OLS regression coefficients with standard errors in parentheses.
* p< .05; ** p< .01; *** p< .001

24 Because debate viewing is an ordinal variable with four values, an ordinal regression with
complementary log-log function was also conducted but the results were virtually the same as those
obtained with OLS regression.
143
Table A.9 Regression Analysis of Key Variables on Controlling Variables in the NAES
2004 Debates Panel (continued from previous page)

W1 W1
Political Knowledge 25 Political Participation 26

Block 1: Demographics
Education .35*** .16**
Income .05 .04
Age .09* -.06
Female -.14*** .02
A R2 16.3% 3.8%

Block 2: Political Orientations

Internal Political Efficacy .18*** .18***
Strength of Party Identification .05 .08
Habitual Political Interest .35*** .32***
A R2 17.9% 16.4
Total R 2 (%) 34.2% 20.2%
(Valid cases) (696) (390)
Notes: All control variables measured at the pre-debates wave. Cell entries show upon-
entry standardized OLS regression coefficients with standard errors in parentheses.
* p < .05; ** p< .01; *** p< .001

25 Political knowledge is a count variable, so a Poisson regression was also conducted to check the

consistency of the OLS estimates reported in the table. There were no significant differences.
26 Political participation is also a count variable. Nevertheless, the Poisson regression results were
virtually the same as the OLS regression results.
144
Table A.10 Comparison of Theorized Auto-regressive Models using the ANES 2008-
2009 Panel
Synchronous Lagged Lagged
Model Model 1 Model 2
AIC 10,124.86 1,301.81 -1,134.15
BIC 10,285.31 1,376.31 -1,042.47
x2 182.03 74.69 121.83
Df 35 9 14
p value < .001 < .001 < .001
//# 5.20 8.30 8.70
CFI .98 .95 .97
TLI .96 .89 .95
RMSEA .04 .06 .06
SRMR .04 .04 .04
Notes: All models were estimated using Wave 9 and Wave 11 data (n = 2,276). The
synchronous model estimates the structural paths outlined in Figure 3.1. Lagged Model
1 predicts Wave 11 political knowledge with Wave 9 emotions, Wave 9 communication
variables, and Wave 9 political knowledge. Lagged Model 2 predicts Wave 11 political
knowledge with Wave 9 emotions, Wave 9 and Wave 11 communication variables, and
Wave 9 political knowledge. All models have non-significant paths removed.

145
Table A.ll Comparison of Theorized Auto-regressive Models using the NAES 2004
Debates Panel

Synchronous Lagged Lagged

Model Model 1 Model 2
AIC 237.34 -49.06 -454.86
BIC 346.90 30.27 -371.76
X2 87.18 105.33 149.63
Df 34 12 27
p value < .001 < .001 < .001
//# 2.56 8.78 5.54
CFI .96 .84 .88
TLI .94 .71 .85
RMSEA .07 .16 .12
SRMR .07 .13 .14
Notes: All models were estimated using pre- and post-debates data (n = 323). The
synchronous model estimates the structural paths outlined in Figure 3.1. Lagged Model
1 predicts post-debates political knowledge and political participation with pre-debates
emotions, pre-debates communication variables, and pre-debates political knowledge
and political participation. Lagged Model 2 predicts post-debates political knowledge
and political participation with pre-debates emotions, pre- and post-debates
communication variables, and pre-debates political knowledge and political
participation. . All models have non-significant paths removed.

146
Table A.12 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using the ANES 2008-2009 Panel

Wave 9 Wave 11
News News
Media Political Political Positive Negative Media Debate Political Political
Use Discussion Knowledge Emotions Emotions Use Viewing Discussion Knowledge
Wave Positive
9 Emotions:

Ln
cn
00
Direct .13 —
— — —

Indirect .03 .003 —

.10 .04 .05

Ln
cn
00
Total .13 .03 —
.10 .04 .05
Negative
Emotions:
Direct .10 —
.60 — — —

Indirect —
.01 —
— .09 .01
Total .10 .01 .60 — .09 .01
News
Media Use:
Direct .19 —
.74 — — —

Indirect —
.02 —
— .18 —

Total .19 .02 .74 — .18 —

Political
Discussion:
Direct —
.10 —
— .52 —

Indirect — — —
— — —

Total —
.10 —
— .52 —
Table A.12 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using the ANES 2008-2009 Panel
(continued from previous page)

Wave 9 Wave 11
News News
Media Political Political Positive Negative Media Debate Political Political
Use Discussion Knowledge Emotions Emotions Use Viewing Discussion Knowledge
Wave Political
9 Knowledge:
Direct .53
Indirect
Total .53
Positive
Emotions:
Direct .03 .16
Indirect .004 .03 .02
Total .03 .16 .03 .02
Negative
motions:
Direct .06
Indirect .003
Total .06 .003
News
Media Use:
Direct .13 .08 -.04
Indirect .02 .02
Total .13 .11
Table A.12 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using the ANES 2008-2009 Panel
(continued from previous page)

Wave 9 Wave 11
News News
Media Political Political Positive Negative Media Debate Political Political
Use Discussion Knowledge Emotions Emotions Use Viewing Discussion Knowledge
Wave Debate
11 Viewing:
Direct — — — — — — — .16 .12
Indirect — — — — — — — — .01
Total — — — — — — — .16 .12
Political
Discussion:
Direct — — — — — — — — .05
Indirect

Note: Standardized coefficients displayed here are at least significant at p < .05. Indirect and direct effects might not add up
to total effects due to rounding error and nonsignificant indirect links.
Table A.13 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using the NAES 2004 Debates Panel

Pre-debates Wave
News Media Use Political Discussion Political Knowledge Political Participation
Pre-debates Negative Emotions:
Wave Direct — .25 .15 .17
Indirect
Total - .25 .15 .17
News
Media Use:
Direct — .17 - .09
Indirect
Total — .17 — .09
Political Discussion:
Direct
Indirect
Total
Political Knowledge:
Direct
Indirect
Total
Political Participation:
Direct
Indirect
Total
Table A.13 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using the NAES 2004 Debates Panel
(continued from previous page)

Post-debates Wave
Negative News Debate Political Political Political
Emotions Media Use Viewing Discussion Knowledge Participation
Pre- Neg. Emotions:
debates Direct .70 — —

wave Indirect .22 .10 .28

Total .70 — — .22 .10 .28
News Media Use:
Direct .76 — —

Indirect .23 — .15

Total .76 — .23 — .15
Pol. Discussion:
Direct .61 — .35
Indirect —
— —

Total .61 — .35

Pol. Knowledge:
Direct —
.40 —

Indirect —
— —

Total —
.40 —

Pol. Participation:
Direct —
— .73
Indirect —
— —

Total —
— .73
Table A.13 Direct, Indirect and Total Effects in the Final Auto-Regressive Model using the NAES 2004 Debates Panel
(continued from previous page)

Post-debates Wave
Negative News Debate Political Political Political
Emotions Media Use Viewing Discussion Knowledge Participation
Post- Negative
emotions:
debates Direct — — — .10 — .09
Wave Indirect — — — — — .02
Total — — — .10 — .10
News Media
Use:
Direct — — .32 .16 -.13 —

Indirect — — — — .13 .03

Total — — .32 .16 — .03
Debate
Viewing:
Direct — — — — .31 —

Indirect — — — — — —

Total — — — — .31 —

Pol. Discussion:
Direct — — — — .16 .15
Indirect — — — — — —

Total — — — — .16 .15

Note: Standardized coefficients displayed here are at least significant at p < .05. Indirect and direct effects might not add up
to total effects due to rounding error and nonsignificant indirect links.
 References

Aarts, K., & Semetko, H. A. (2003). The divided electorate: Media use and political
involvement. Journal of Politics, 65, 759-784.

Abelson, R. P., Kinder, D. R., Peters, M. D., & Fiske, S. T. (1982). Affective and semantic
components in political personal perception. Journal of Personality and Social
Psychology, 42, 619.

Althaus, S. L. (2001). Who's voted in when the people turn out?: Information effects in
congressional elections. In R. Hart & D. R. Shaw (Eds.), Communication and U.S.
elections: New agendas (pp. 33-54). Lanham, MD: Rowman and Littlefield.

American Association of Public Opinion Research. (2008). Standard definitions: Final

dispositions of case codes and outcome rates for surveys. Retrieved July 8, 2009,
from http://www.aapor.org/uploads/Standard_Definitions_04_08_Final.pdf

Annenberg Public Policy Center. (2006). NAES 2004/2004 National Annenberg Election
Survey: National rolling cross-section codebook. Philadelphia, PA: University of
Pennsylvania.

Atkin, C. K. (1972). Anticipated communication and mass media information-seeking.

Public Opinion Quarterly, 36, 188-199.

Barber, B. (1984). Strong democracy: Participatory politics for a new age. Berkeley, CA:
University of California Press.

Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in

social psychological research: Conceptual, strategic, and statistical
considerations. Journal of Personality and Social Psychology, 51, 1173-1182.

Bartels, L. M. (1993). Messages received: The political impact of media exposure.

American Political Science Review, 87, 267-285.

Bartels, L. M. (1996). Uninformed voters: Information effects in presidential elections.

American Journal of Political Science, 40, 194-230.

Bartels, L. M. (1999). Panel effects in the American National Election Studies. Political
Analysis, 8,1-20.

153
Bartels, L. M. (2006). Three virtues of panel data for the analysis of campaign effects. In
H. E. Brady & R. Johnston (Eds.), Capturing campaign effects (pp. 134-163). Ann
Arbor, Ml: University of Michigan Press.

Bennett, A. (1932). The journals of Arnold Bennett (Vol. 1). London: Cassell and
Company, Ltd.

Benoit, W. L., Hansen, G. J., & Verser, R. M. (2003). A meta-analysis of the effects of
viewing U.S. presidential debates. Communication Monographs, 70, 335-350.

Benoit, W. L., Pier, P. M., Brazeal, L. M., McHale, J. P., Klyukovksi, A., & Airne, D. (2002).
The primary decision: A functional analysis of debates in presidential primaries. :.
Westport, CT: Praeger.

Benoit, W. L., Stein, K. A., & Hansen, G. J. (2005). New York Times coverage of
presidential campaigns. Journalism & Mass Communication Quarterly, 82, 356-
376.

Benter, P. M., & Chou, C. P. (1987). Practical issues in structural modeling. Sociological
Methods and Research, 16(1), 78-117.

Berelson, B., Lazarsfeld, P. F., & McPhee, W. N. (1954). Voting: A study of opinion
formation in a presidential campaign. Chicago: University of Chicago Press.

Bimber, B. (2001). Information and political engagement in America: The search for
effects of information technology at the individual level. Political Research
Quarterly, 54, 53-67.

Binder, A. R., Dalrymple, K. E., Brossard, D., & Scheufele, D. A. (2009). The soul of a
polarized democracy: Testing theoretical linkages between talk and attitude
extremity during the 2004 presidential election. Communication Research, 36,
315-340.

Bizer, G. Y., Visser, P. S., Berent, M. K., & Krosnick, J. A. (2004). Importantce, knowledge,
and accessibility: Exploring the dimensionality of strength-related attitude
properties. In W. E. Saris & P. M. Sniderman (Eds.), Studies in public opinion:
Attitudes, nonattitudes, measurement error, and change (pp. 215-241).
Princeton, NJ: Princeton University Press.

Blumler, J. G., & Gurevitch, M. (2001). "Americanization" reconsidered: U.K.-U.S.

campaign communication comparisons across time. In W. L. Bennett & R. M.

154
 Entman (Eds.), Mediated politics. Communication in the future of democracy (pp.
380-406). Cambridge,UK: Cambridge University Press.

Bollen, K. A. (1987). Total, direct, and indirect effects in structural equation models.
Sociological Methodology, 17, 37-69.

Bollen, K. A., & Long, J. S. (Eds.). (1993). Testing structural equation models. Newbury
Park, CA: Sage.

Bollen, K. A., & Stine, R. (1990). Direct and indirect effects: Classical and bootstrap
estimates of variability. Sociological Methodology, 20, 115-140.

Boyle, M., & Schmierbach, M. (2009). Media use and protest: The role of mainstream
and alternative media use in predicting traditional and protest participation.
Communication Quarterly, 57, 1-17.

Boyle, M. P., Schmierbach, M., Armstrong, C. L., McLeod, D. M., Shah, D. V., & Pan, Z.
(2004). Information seeking and emotional reactions to September 11 terrorist
attacks .Journalism & Mass Communication Quarterly, 81, 155-167.

Brader, T. (2005). Striking a responsive chord: How political ads motivate and persuade
voters by appealing to emotions. American Journal of Political Science, 49, 388-
405.

Brader, T. (2006a). Affective intelligence and beyond: Next steps in research on emotion
in politics. Political Communication Report, 16(3), 1-6.

Brader, T. (2006b). Campaigning for hearts and minds: How emotional appeals in
political ads work. Chicago: University of Chicago Press.

Brader, T. (2011). The political relevance of emotions: "reassessing" revisited. Political

Psychology, 32, 337-346.

Brader, T., & Valentino, N. A. (2006). Identities, interests, and emotions: Symbolic vs.
material wellsprings of fear, anger, and enthusiasm. In W. R. Neuman, G. E.
Marcus, A. N. Crigler & M. MacKuen (Eds.), The affect effect: Dynamics of
emotion in political thinking and behavior (pp. 180-201). Chicago, IL: University
of Chicago Press.

Brader, T., Valentino, N. A., & Suhay, E. (2008). What triggers public opposition to
immigration? Anxiety, group cues, and immigration threat. American Journal of
Political Science, 52(4), 959-978.

155
Brady, H. E. (1999). Political participation. In J. P. Robinson, P. R. Shaver & L. S.
Wrightsman (Eds.), Measures of political attitudes (Vol. 2 of Measures of Social
Psychological Attitudes, pp. 737-801). San Diego: Academic Press.

Brady, H. E., & Sniderman, P. M. (1985). Attitude attribution: A group basis for political
reasoning. American Political Science Review, 79, 1061-1078.

Brown-Kramer, C. R. (2009). The emotional activist: The role of affect in political decision
makingby. Unpublished dissertation, University of Nebraska, Lincoln, NE.

Brundidge, J. (2010). Encountering "difference" in the contemporary public sphere: The

contribution of the internet to the heterogeneity of political discussion
networks. Journal of Communication, 60(4), 680-700.

Bucy, E., & Newhagen, J. (1999). The emotional appropriateness heuristic: Processing
televised presidential reactions to the news. Journal of Communication, 49(4),
59-79.

Bunch, W. (2010). The backlash: Right-wing radicals, high-def hucksters, and paranoid
politics in the age ofObama. New York: Harper.

Cacioppo, J. T., Berntson, G. G., Klein, D. J., & Poehlmann, K. M. (1997). The
psychophysiology of emotion across the lifespan. Annual Review of Gerontology
and Geriatrics, 17, 27-74.

Cacioppo, J. T., & Gardner, W. L. (1999). Emotion. Annual Review of Psychology, 50, 191-
214.

Campbell, A., Converse, P. E., Miller, W. E., & Stokes, D. E. (i960). The American voter.
New York: John Wiley.

Cappella, J. N., & Jamieson, K. H. (1997). The spiral of cynicism: The press and the public
good. New York: Oxford University Press.

Carmines, E. G., & Stimson, J. A. (1989). Issue evolution: Race and the transformation of
American politics. Princeton, NJ: Princeton University Press.

Carver, C. S., & Harmon-Jones, E. (2009). Anger is an approach-related affect: Evidence

and implications. Psychological Bulletin, 135, 183-204.

Carver, C. S., & White, T. L. (1994). Behavioral inhibition, behavioral activation, and
affective response to impending reward and punishment: The BIS/BAS scales.
Journal of Personality and Social Psychology, 67, 319-333.
156
Center for Responsive Politics. (2011). 2008 presidential election: Presidential
expenditures. Retrieved April 1, 2011, from
http://www.opensecrets.org/pres08/expenditures.php?cycle=2008

Chaffee, S. H. (1986). Mass media and interpersonal channels: Competitive, convergent,

or complementary? In G. Gumpert & R. Cathart (Eds.), Inter/media: Interpersonal
communication in a media world (3rd ed., pp. 62-80). New York: Oxford
University Press.

Chaffee, S. H., & Schleuder, J. (1986). Measurement and effects of attention to media
news. Human Communication Research, 13, 76-107.

Chaffee, S. H., Zhao, X., & Leshner, G. (1994). Political knowledge and the campaign
media of 1992. Communication Research, 21, 305-324.

Chernov, G., Valenzuela, S., & McCombs, M. E. (2009, August). A comparison of two
perspectives on the concept of need for orientation in agenda-setting theory.
Paper presented at the annual meeting of the Association for Education in
Journalism & Mass Communication, Boston, MA.

Cho, J., Shah, D. V., McLeod, J. M., McLeod, D. M., & Scholl, R. M. G., Melissa R. (2009).
Campaigns, reflection, and deliberation: Advancing an O-S-R-O-R model of
communication effects. Communication Theory, 19, 66-88.

Civettini, A. J. W., & Redlawsk, D. P. (2009). Voters, emotions, and memory. Political
Psychology, 30, 125-151.

Cohen, J. (1989). Deliberation and democratic legitimacy. In A. Hamlin & P. Pettit (Eds.),
The good polity (pp. 17-34). Oxford: Blackwell.

Coleman, R., & McCombs, M. E. (2007). The young and agendaless? Exploring age-
related differences in agenda setting on the youngest generation, baby boomers,
and the civic generation. Journalism and Mass Communication Quarterly, 84,
495-508.

Coleman, R., & Wu, H. D. (2010). Proposing emotion as a dimension of affective agenda-
setting: Separating affect into two components and comparing their second-level
effects. Journalism & Mass Communication Quarterly, 87, 315-327.

Conover, P. J., & Feldman, S. (1986). Emotional reactions to the economy: I'm mad as
hell and I'm not going to take it any more. American Journal of Political Science,
30, 50-78.

157
Conover, P. J., & Searing, D. D. (2005). Studying "everyday political talk' in the
deliberative system. Acta Politico, 40, 269-283.

Conover, P. J., Searing, D. D., & Crewe, I. M. (2002). The deliberative potential of political
discussion. British Journal of Political Science, 32, 21-62.

Converse, P. E. (1964). The nature of belief systems in mass publics. In D. E. Apter (Ed.),
Ideology and discontent (pp. 206-261). New York: Free Press.

Conway, M. M. (1985). Political participation in the United States. Washington, DC:

Congressional Quarterly.

Craig, S. C., Kane, J. G., & Gainous, J. (2005). Issue-related learning in a gubernatorial
campaign: A panel study. Political Communication, 22, 483-503.

Crigler, A. N., Just, M. R., & Belt, T. (2006). The three faces of negative campaigning: The
democratic implications of attack ads, cynical news and fear arousing messages.
In D. P. Redlawsk (Ed.), Feeling politics: Affect and emotion in political
information processing. New York: Palgrave/Macmillan.

D'Angelo, P., & Kuypers, J. A. (Eds.). (2009). Doing news framing analysis: Empirical and
theoretical perspectives. New York: Routledge.

Dahl, R. A. (1989). Democracy and its critics. New Haven, CT: Yale University Press.

Dahl, R. A. (1999). On democracy. New Haven, CT: Yale University Press.

DeBell, M., Krosnick, J. A., & Lupia, A. (2010). Methodology report and user's guide for
the 2008-2009 ANES Panel Study. Palo Alto, CA and Ann Arbor, Ml: Stanford
University and the University of Michigan.

Delli-Carpini, M. X., & Keeter, S. (1996). What Americans know about politics and why it
matters. New Haven, CT: Yale University Press.

Delli Carpini, M. X., Cook, F. L, & Jacobs, L. R. (2004). Public deliberation, discursive
participation, and citizen engagement: A review of the empirical literature.
Annual Review of Political Science, 7, 315-344.

Delli Carpini, M. X., & Keeter, S. (1993). Measuring political knowledge: Putting first
things first. American Journal of Political Science, 37, 1179-1206.

Delli Carpini, M. X., & Keeter, S. (1996). What Americans know about politics and why it
matters. New Haven, CT: Yale University Press.

158
Dickinson, T. (2008, March 20). The machinery of hope. Rolling Stone, 1048, 36-42.

Domke, D., Shah, D. V., & Wackman, D. B. (1998). Media priming effects: Accessibility,
association, and activation. International Journal of Public Opinion Research, 10,
51-74.

Drew, D., & Weaver, D. (1991). Voter learning in the 1988 presidential election: Did the
debates and the media matter? Journalism Quarterly, 68, 27-37.

Drew, D., & Weaver, D. (1993). Voter learning in the 1990 off-year election: Did the
media matter1 Journalism Quarterly, 70, 356-368.

Drew, D., & Weaver, D. (1998). Voter learning in the 1996 presidential election: Did the
media matter? Journalism & Mass Communication Quarterly, 75, 292-301.

Drew, D., & Weaver, D. (2006). Voter learning in the 2004 presidential election: Did the
media matter? Journalism & Mass Communication Quarterly, 83, 25-42.

Drew, D., & Weaver, D. H. (1990). Media attention, media exposure, and media effects.
Journalism Quarterly, 67, 740-748.

Druckman, J. N. (2005). Media matter: How newspapers and television news cover
campaigns and influence voters. Political Communication, 22, 463-481.

Druckman, J. N., & McDermott, R. (2008). Emotion and the framing of risky choice.
Political Behavior, 30, 297-321.

Dryzek, J. S. (2000). Deliberative democracy and beyond: Liberals, critics, contestations.

Oxford, UK: Oxford University Press.

Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Belmont, CA: Wadsworth.

Erlich, D., Guttman, I., Schonbach, P., & Mills, J. (1957). Postdecision exposure to
relevant information. Journal of Abnormal and Social Psychology, 54, 98-102.

Evatt, D., & Ghanem, S. I. (2001, September). Building a scale to measure salience. Paper
presented at the annual conference of the World Association of Public Opinion
Research, Rome, Italy.

Eveland, W. P., Jr. (1997). Interactions and nonlinearity in mass communication:

Connecting theory and methodology. Journalism & Mass Communication
Quarterly, 74, 400-416.

159
Eveland, W. P., Jr. (2001). The cognitive mediation model of learning from the news:
Evidence from nonelection, off-year election, and presidential election contexts.
Communication Research, 28, 571-601.

Eveland, W. P., Jr. (2004). The effect of political discussion in producing informed
citizens: The roles of information, motivation, and elaboration. Political
Communication, 21, 177-193.

Eveland, W. P., Jr., Hayes, A. F., Shah, D. V., & Kwak, N. (2005a). Observations on
estimation of communication effects on political knowledge and a test of
intracommunication mediation. Political Communication, 22, 505-509.

Eveland, W. P., Jr., Hayes, A. F., Shah, D. V., & Kwak, N. (2005b). Understanding the
relationship between communication and political knowledge: A model
comparison approach using panel data. Political Communication, 22, 423-446.

Eveland, W. P., Jr., Hively, M., & Morey, A. C. (2009, May). Discussing measures of
political discussion: An evaluation of the measurement of network size,
agreement, and disagreement and implications for inferences. Paper presented
at the annual meeting of the International Communication Association, Chicago,
IL.

Eveland, W. P., Jr., & Hively, M. H. (2009). Political discussion frequency, network size,
and "heterogeneity" of discussion as predictors of political knowledge and
participation. Journal of Communication, 59, 205-224.

Eveland, W. P., Jr., Hutchens, M. J., & Shen, F. (2009). Exposure, attention, or "use" of
news? Assessing aspects of the reliability and validity of a central concept in
political communication research. Communication Methods and Measures, 3,
223 - 244.

Eveland, W. P., Jr., Marton, K., & Seo, M. (2004). Moving beyond "just the facts": The
influence of online news on the content and structure of public affairs
knowledge. Communication Research, 31, 82-108.

Eveland, W. P., Jr., Morey, A. C., & Hively, M. H. (2009, August). Beyond deliberation:
New directions for the study of informal political conversation from a
communication perspective. Paper presented at the annual meeting of the
Association for Education in Journalism and Mass Communication, Boston, MA.

160
Eveland, W. P., Jr., Shah, D. V., & Kwak, N. (2003). Assessing causality in the cognitive
mediation model: A panel study of motivations, information processing, and
learning during campaign 2000. Communication Research, 30, 359-386.

Eveland, W. P., Jr., & Thomson, T. (2006). Is it talking, thinking, or both? A lagged
dependent variable model of discussion effects on political knowledge. Journal
of Communication, 56, 523-542.

Feldman, L, & Price, V. (2008). Confusion or enlightenment?: How exposure to

disagreement moderates the effects of political discussion and media use on
candidate knowledge. Communication Research, 35, 61-87.

Finkel, S. E. (1995). Causal analysis with panel data. Thousand Oaks, CA: Sage.

Fishkin, J. S. (1991). Democracy and deliberation: New directions for democratic reform.
New Haven, CT: Yale University Press.

Fishkin, J. S., & Luskin, R. C. (2005). Experimenting with a democratic ideal: Deliberative
polling and public opinion. Acta Politico, 40, 284-298.

Freedman, J. L., & Sears, D. O. (1965). Selective exposure. In L. Berkowitz (Ed.), Advances
in experimental social psychology (Vol. 2, pp. 57-97). New York: Academic Press.

damson, W. A. (1992). Talking politics. New York: Cambridge University Press.

Gastil, J. (2008). Political communication and deliberation. Los Angeles, CA: Sage.

Geer, J. G. (2006). In defense of negativity: Attack ads in presidential campaigns.

Chicago: University of Chicago Press.

Gil de Zuniga, H., Puig-i-Abril, E., & Rojas, H. (2009). Weblogs, traditional sources online
and political participation: An assessment of how the internet is changing the
political environment. New Media & Society, 11, 553-574.

Gil de Zuniga, H., & Valenzuela, S. (Forthcoming). The mediating path to a stronger
citizenship: Online and offline networks, weak ties, and civic engagement.
Communication Research.

Golan, G., & Wanta, W. (2001). Second-level agenda setting in the New Hampshire
primary: A comparison of coverage in three newspapers and public perceptions
of candidates. Journalism & Mass Communication Quarterly, 78, 247-259.

161
Goldberg, L. R. (1990). An alternative "description of personality": The big-five factor
structure. Journal of Personality and Social Psychology, 59, 1216-1229.

Goodwin, J., Jasper, J. M., & Polletta, F. (Eds.). (2001). Passionate politics: Emotions and
social movements. Chicago: University of Chicago Press.

Graber, D. (1984). Processing the news: How people tame the information tide. New
York: Longman.

Graber, D., & Smith, J. M. (2005). Political communication faces the 21st century.
Journal of Communication, 55, 479-507.

Graber, D. A. (2001). Processing politics: Learning from television in the internet age.
Chicago: University of Chicago Press.

Granovetter, M. S. (1973). The strength of weak ties. American Journal of Sociology, 78,
1360-1380.

Gray, J. A. (1985). The neuropsychology of anxiety: An enquiry into the functions of the
septo-hippocampalsystem. New York: Oxford University Press.

Gray, J. A. (1987). The psychology of fear and stress. Cambridge, UK: Cambridge
University Press.

Gray, J. A. (1990). Brain systems that mediate both emotion and cognition. Cognition
and Emotion, 4, 269-288.

Gross, K. (2008). Framing persuasive appeals: Episodic and thematic framing, emotional
response, and policy opinion. Political Psychology, 29, 169-192.

Gutmann, A., & Thompson, D. F. (1996). Democracy and disagreement: Why moral
conflict cannot be avoided in politics, and what should be done about it.
Cambridge, MA: Belknap Press.

Habermas, J. (1996). Between facts and norms: Contributions to a discourse theory of

law and democracy. Cambridge, MA: MIT Press.

Hansen, G. J., & Benoit, W. L. (2007). Communication forms as predictors of issue

knowledge in presidential campaigns: A meta-analytic assessment. Mass
Communication & Society, 10,189-210.

162
Hardy, B. W., & Scheufele, D. A. (2005). Examining differential gains from internet use:
Comparing the moderating role of talk and online interactions. Journal of
Communication, 55, 71-84.

Hayes, A. F. (2009). Beyond Baron and Kenny: Statistical mediation analysis in the new
millennium. Communication Monographs, 76, 408-420.

Holbert, R. L. (2005). Intramedia mediation: The cumulative and complementary effects

of news media use. Political Communication, 22, 447 - 461.

Holbert, R. L. (2006, March 22). Political communication and structural equation

modeling. Paper presented at the Colloquium on Political Communication
Research at the Indiana University, Bloomington, IN.

Holbert, R. L, & Geidner, N. (2009). The 2008 election: Highlighting the need to explore
additional communication subfields to advance political communication.
Communication Studies, 60, 344-358.

Holbert, R. L, & Hansen, G. J. (2006). Fahrenheit 9-11, need for closure and the priming
of affective ambivalence: An assessment of intra-affective structures by party
identification. Human Communication Research, 32, 109-129.

Holbert, R. L, & Hansen, G. J. (2008). Stepping beyond message specificity in the study
of emotion as mediator and inter-emotion associations across attitude objects:
"Fahrenheit 9/11," anger, and debate superiority. Media Psychology, 11, 98 -
118.

Holbert, R. L, & Stephenson, M. T. (2002). Structural equation modeling in the

communication sciences, 1995-2000. Human Communication Research, 28, 531-
551.

Holbert, R. L, & Stephenson, M. T. (2003). The importance of indirect effects in media

effects research: Testing for mediation in structural equation modeling. Journal
of Broadcasting & Electronic Media, 47, 556-572.

Holbrook, T. M. (2002). Presidential campaigns and the knowledge gap. Political

Communication, 19, 437-454.

Hoyle, R. (1995). Structural equation modeling: Concepts, issues and applications.

Thousand Oaks, CA: Sage.

163
Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure
analysis: Conventional criteria versus new alternatives. Structural Equation
Modeling, 6, 1-55.

Huck, I., Quiring, 0., & Brosius, H.-B. (2009). Perceptual phenomena in the agenda
setting process. International Journal of Public Opinion Research, 21, 139-164.

Huckfeldt, R., Johnson, P. E., & Sprague, J. (2004). Political disagreement: The survival of
diverse opinions within communication networks. Cambridge: Cambridge
University Press.

Huckfeldt, R., Mendez, J. M., & Osborn, T. (2004). Disagreement, ambivalence, and
engagement: The political consequences of heterogeneous networks. Political
Psychology, 25, 65-95.

Huckfeldt, R., & Sprague, J. (1995). Citizens, politics, and social communication:
Information and influence in an election campaign. Cambridge, UK: Cambridge
University Press.

Huddy, L, Feldman, S., & Cassese, E. (2007). On the distinct political effects of anxiety
and anger. In W. R. Neuman, G. E. Marcus, M. MacKuen & A. N. Crigler (Eds.),
The affect effect: Dynamics of emotion in political thinking and behavior (pp.
201-230). Chicago: University of Chicago Press.

Hume, D. (1739/1888). A treatise of human nature. Oxford, UK: Clarendon Press.

Ito, T. A., Larsen, J. T., Smith, N. K., & Cacioppo, J. T. (1998). Negative information weighs
more heavily on the brain: The negativity bias in evaluative categorizations. J.
Pers. Soc. Psychol., 75, 887.

Iyengar, S. (1991). Is anyone responsible? How television frames political issues. Chicago:
University of Chicago Press.

Iyengar, S., & Kinder, D. R. (1987). News that matters: Television and American opinion.
Chicago: The University of Chicago Press.

Iyengar, S., & Simon, A. (1993). News coverage of the gulf crisis and public opinion.
Communication Research, 20(3), 365-383.

Jamieson, K. H., & Kenski, K. (2006). Why the National Annenberg Election Survey? In D.
Romer, K. Kenski, K. Winneg, C. Adasiewicz & K. H. Jamieson (Eds.), Capturing
campaign dynamics, 2000 and 2004: The National Annenberg Election Survey
(pp. 1-13). Philadelphia, PA: University of Pennsylvania.
164
Jung, N., Kim, Y., & Gil de Zuniga, H. (forthcoming). The mediating role of knowledge and
efficacy in the effects of communication on political participation. Mass
Communication & Society.

Junn, J. (1991). Participation and political knowledge. In W. Crotty (Ed.), Political

participation and American democracy (pp. 192-212). New York: Greenwood
Press.

Just, M. R., Crigler, A. N., Alger, D. E., Cook, T. E., Kern, M., & West, D. M. (1996).
Crosstalk: Citizens, candidates, and the media in a presidential campaign.
Chicago: University of Chicago Press.

Kaplan, D. (2000). Structural equation modeling: Foundations and extensions. Thousand

Oaks, CA: Sage.

Katz, E., & Gurevitch, M. (1974). Utilization of mass communication by the individual. In
J. G. Blumler & E. Katz (Eds.), The uses of mass communication (pp. 19-32).
London: Faber.

Katz, E., & Lazarsfeld, P. (1955). Personal influence. New York: The Free Press.

Kenski, K., Hardy, B. W., & Jamieson, K. H. (2010). The Obama victory: How media,
money and message shaped the 2008 elections. New York: Oxford University
Press.

Kim, J., Wyatt, R. 0., & Katz, E. (1999). News, talk, opinion, participation: The part played
by conversation in deliberative democracy. Political Communication, 16, 361-
385.

Kim, N. (2009, May). The role of anger and information in deliberation. Paper presented
at the annual meeting of the International Communication Association, Boston,
MA.

King, G. (1995). A revised proposal, proposal. PS: Political Science and Politics, 28, 494-
499.

Kiousis, S., & McCombs, M. (2004). Agenda-setting effects and attitude strength:
Political figures during the 1996 presidential election. Communication Research,
31, 36-57.

Kline, R. B. (1998). Principles and practice of structural equation modeling. New York:
Guilford Press.

165
Klofstad, C. A. (2007). Talk leads to recruitment: How discussions about politics and
current events increase civic participation. Political Research Quarterly, 60, 180-
191.

Kraut, R. E., Patterson, M., Lundmark, V., Kiesler, S., Mukopadhyay, T., & Scherlis, W.
(1998). Internet paradox: A social technology that reduces social involvement
and psychological well-being? American Psychologist, 53(9), 1017-1031.

Kuklinski, J. H., Quirk, P. J., Jerit, J., Schwieder, D., & Rich, R. F. (2000). Misinformation
and the currency of democratic citizenship. The Journal of Politics, 62(03), 790-
816.

Kunda, Z. (1990). The case for motivated political reasoning. Psychological Bulletin, 108,
480-498.

Ladd, J. M., & Lenz, G. S. (2008). Reassessing the role of anxiety in vote choice. Political
Psychology, 29, 275-296.

Ladd, J. M., & Lenz, G. S. (2011). Does anxiety improve voters' decision making? Political
Psychology, 32, 347-361.

Landreville, K. D., Holbert, R. L., & LaMarre, H. L. (2010). The influence of late-night TV
comedy viewing on political talk: A moderated-mediation model. The
International Journal of Press/Politics, 15, 482-498.

Lang, A., & Newhagen, J. (1996). Negative video as structure: Emotion, attention,
capacity, and memory. Journal of Broadcasting & Electronic Media, 40, 460-477.

Lang, A., Park, B., Sanders-Jackson, A. N., Wilson, B. D., & Zheng, W. (2007). Cognition
and emotion in TV message processing: How valence, arousing content,
structural complexity, and information density affect the availability of cognitive
resources. Media Psychology, 10, 317-338.

Lau, R. R., Andersen, D. J., & Redlawsk, D. P. (2008). An exploration of correct voting in
recent U.S. presidential elections. American Journal of Political Science, 52, 395-
411.

Lau, R. R., & Sears, D. O. (1986). Political cognition: The 19th Annual Carnegie
Symposium on Cognition. Hillsdale, NJ: Lawrence Erlbaum.

Lazarsfeld, P., Berelson, B., & Gaudet, H. (1948). The people's choice (2nd ed.). New
York: Columbia University Press.

166
Lazarus, R. S. (1984). On the primacy of cognition. American Psychologist, 39, 124-129.

LeDoux, J. E. (1996). The emotional brain: The mysterious underpinnings of emotional

life. New York: Simon & Schuster.

Lee, G. (2005). Agenda setting effects in the digital age: Uses and effects of online
media. Unpublished Ph. D. dissertation, University of Texas at Austin, Austin, TX.

Leigh ley, J. E. (1991). Participation as a stimulus of political conceptualization. Journal of

Politics, 53, 198-211.

Lemert, J. B. (1977). Journalists and mobilizing information. Journalism Quarterly, 54,

721-726.

Lemert, J. B. (1984). News context and the elimination of mobilizing information: An

experiment. Journalism Quarterly, 61, 241-247.

Levendusky, M. S. (2011). Rethinking the role of political information. Public Opinion

Quarterly, 75, 1-23.

Lippmann, W. (1922). Public opinion. New York: Macmillan.

Lodge, M., & Taber, C. S. (2000). Three steps toward a theory of motivated political
reasoning. In A. Lupia, M. D. McCubbins & S. Popkin (Eds.), Elements of reason:
Cognition, choice, and the bounds of rationality (pp. 182-213). London:
Cambridge University Press.

Luskin, R. C. (1987). Measuring political sophistication. American Journal of Political

Science, 31, 856-899.

Luskin, R. C. (1990). Explaining political sophistication. Political Behavior, 12, 331-361.

Luskin, R. C. (2003). The heavenly public: What would the ideal democratic citizenry be
like? In M. B. MacKuen & G. Rabinowitz (Eds.), Electoral democracy {pp. 238-
261). Ann Arbor, Ml: University of Michigan Press.

Luskin, R. C., & Bullock, J. (2004, April). Re(:)measuring political sophistication. Paper
presented at the annual meeting of the Midwest Political Science Association,
Chicago, IL.

Luskin, R. C., & Fishkin, J. S. (2002, March). Deliberation and "better citizens." Paper
presented at the annual Joint Sessions of Workshops of the European
Consortium for Political Research, Turin, Italy.

167
MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A
comparison of methods to test mediation and other intervening variable effects.
Psychological Methods, 7, 83-104.

MacKuen, M., Wolak, J., Keele, L, & Marcus, G. E. (2010). Civic engagements: Resolute
partisanship or reflective deliberation. American Journal of Political Science, 54,
440-458.

Marcus, G. E. (1988). The structure of emotional response: 1984 presidential candidates.

American Political Science Review, 82, 735-761.

Marcus, G. E. (2000). Emotions in politics. Annual Review of Political Science, 3, 221-250.

Marcus, G. E. (2002). The sentimental citizen: Emotion in democratic politics. University

Park, PA: Pennsylvania State University Press.

Marcus, G. E. (2003). The psychology of emotion and politics. In D. O. Sears, L. Huddy &
R. Jervis (Eds.), Oxford handbook of political psychology (pp. 182-221). New York:
Oxford University Press.

Marcus, G. E., & MacKuen, M. (1996). Measuring mood in the 1995 NES Pilot Study.
Report to the National Election Studies Board Based on the 1995 NES Pilot Study
Retrieved April 1, 2011, from
ftp://ftp.electionstudies.org/ftp/nes/bibliography/documents/nes008447a.pdf

Marcus, G. E., MacKuen, M., & Neuman, W. R. (2011). Parsimony and complexity:
Developing and testing theories of affective intelligence. Political Psychology, 32,
323-336.

Marcus, G. E., MacKuen, M., Wolak, J., & Keele, L. (2006). The measure and mismeasure
of emotion. In D. P. Redlawsk (Ed.), Feeling politics: Affect and emotion in
political information processing (pp. 31-46). New York: Palgrave/Macmillan.

Marcus, G. E., & MacKuen, M. B. (1993). Anxiety, enthusiasm and the vote: The
emotional underpinnings of learning and involvement during presidential
campaigns. American Political Science Review, 87, 688-701.

Marcus, G. E., Neuman, W. R., & MacKuen, M. B. (2000). Affective intelligence and
political judgment. Chicago: University of Chicago Press.

Marcus, G. E., Wood, S. L, & Theiss-Morse, E. (1998). Linking neuroscience to political

intolerance and political judgment. Politics and the Life Science, 17, 165-178.

168
Mark, D. (2009). Going dirty: The art of negative campaigning (Updated ed.). Lanham,
MD: Rowman & Littlefield.

Markus, H., & Zajonc, R. B. (1985). The cognitive perspective in social psychology. In G.
Lindzey & E. Aronson (Eds.), The handbook of social psychology (3rd ed., pp. 137-
230). New York: Random House.

Matthes, J. (2006). The need for orientation towards news media: Revising and
validating a classic concept. International Journal of Public Opinion Research, 18,
422-444.

Matthes, J. (2008). Need for orientation as a predictor of agenda setting effects. Causal
evidence from a two-wave panel study. International Journal of Public Opinion
Research, 20, 440-453.

McClain, C. (2009). Debating restrictions on embryonic stem cell research. Politics and
the Life Sciences, 28, 48-68.

McClurg, S. D. (2003). Social networks and political participation: The role of social
interaction in explaining political participation. Political Research Quarterly, 56,
449-464.

McClurg, S. D. (2006). The electoral relevance of political talk: Examining disagreement

and expertise effects in social networks on political participation. American
Journal of Political Science, 50, 737-754.

McCombs, M. (1999). Personal involvement with issues on the public agenda.

International Journal of Public Opinion Research, 11(2), 152-168.

McCombs, M. (2004). Setting the agenda: The mass media and public opinion.
Cambridge, UK: Polity Press.

McCombs, M., Lopez-Escobar, E., & Llamas, J. P. (2000). Setting the agenda of attributes
in the 1996 Spanish general election. Journal of Communication, 50(2), 77-92.

McCombs, M., & Shaw, D. L. (1972). The agenda-setting function of mass media. Public
Opinion Quarterly, 36,176-187.

McCombs, M. E. (1992). Explorers and surveyors: Expanding strategies for agenda-

setting research. Journalism Quarterly, 69, 813-824.

169
McCombs, M. E., & Weaver, D. H. (1973, May). Voters' need for orientation and use of
mass communication. Paper presented at the annual meeting of the
International Communication Association, Montreal, Canada.

McDevitt, M., & Chaffee, S. (2000). Closing gaps in political communication and
knowledge: Effects of a school intervention. Communication Research, 27, 259-
292.

McGuire, W. J. (1972). Attitude change: The information-processing paradigm. In C. G.

McClintock (Ed.), Experimental social psychology (pp. 108-142). New York: Holt,
Rinehart & Winston.

McGuire, W. J. (1989). Theoretical foundations of campaigns. In R. E. Rice & C. K. At kin

(Eds.), Public communication campaigns (2nd ed., pp. 43-65). Newbury Park, CA:
Sage.

McLeod, J. M., & McDonald, D. G. (1985). Beyond simple exposure: Media orientations
and their impact on political processes. Communication Research, 12(1), 3-33.

McLeod, J. M., Scheufele, D. A., Moy, P., Horowitz, E. M., Holbert, R. L., Zhang, W., et al.
(1999). Understanding deliberation: The effects of discussion networks on
participation in a public forum. Communication Research, 26(6), 743-774.

McLeod, J. M., Zubric, J., Keum, H., Deshpande, S., Cho, J., Stein, S., et al. (2001).
Reflecting and connecting: Testing a communication mediation model of civic
participation. Paper presented at the annual convention of the Association for
Education in Journalism and Mass Communication, Washington, DC.

Milbrath, L. W., & Goel, M. L. (1977). Political participation: How and why do people get
involved in politics? Chicago: Rand McNally.

Miller, J. M. (2007). Examining the mediators of agenda setting: A new experimental

paradigm reveals the role of emotions. Political Psychology, 28(6), 689-717.

Miller, J. M., & Krosnick, J. A. (1996). News media impact on the ingredients of
presidential evaluations: A program of research on the priming hypothesis. In D.
C. Mutz, P. M. Sniderman & R. A. Brody (Eds.), Political persuasion and attitude
change (pp. 79-100). Ann Arbor, Ml: Michigan University Press.

Miller, J. M., & Krosnick, J. A. (2004). Threat as a motivator of political activism: A field
experiment. Political Psychology, 25, 507-523.

170
Min, S.-J. (2007). Online vs. Face-to-face deliberation: Effects on civic engagement.
Journal of Computer-Mediated Communication, 12, 1369-1387.

Mondak, J. J. (2001). Developing valid knowledge scales. American Journal of Political

Science, 45, 224-238.

Mondak, J. J., & Halperin, K. D. (2008). A framework for the study of personality and
political behaviour. British Journal of Political Science, 38, 335-362.

Morrell, M. E. (2003). Survey and experimental evidence for a reliable and valid measure
of internal political efficacy. Public Opinion Quarterly, 67, 589-602.

Moy, P., McCluskey, M. R., McCoy, K., & Spratt, M. A. (2004). Political correlates of local
news media use. Journal of Communication, 54, 532-546.

Muthen, L. K., & Muthen, B. 0. (1998-2010). Mplus user's guide (6th ed.). Los Angeles,
CA: Muthen & Muthen.

Mutz, D. C. (2002). The consequences of cross-cutting networks for political

participation. American Journal of Political Science, 46, 838-855.

Mutz, D. C. (2006). Hearing the other side: Deliberative versus participatory democracy.
New York: Cambridge University Press.

Nabi, R. (2002). Anger, fear, uncertainty, and attitudes: A test of the cognitive-functional
model. Communication Monographs, 69, 204 - 216.

Neuman, W. R. (1981). Differentiation and integration: Two dimensions of political

thinking. American Journal of Sociology, 86, 1236-1268.

Neuman, W. R., Just, M. R., & Crigler, A. N. (1992). Common knowledge: News and the
construction of political meaning. Chicago: University of Chicago Press.

Neuman, W. R., Marcus, G. E., MacKuen, M., & Crigler, A. N. (2007). The affect effect:
Dynamics of emotion in political thinking and behavior. Chicago: University of
Chicago Press.

Nielsen Media Research. (2008, October 6). Highest rated presidential debates: i960 to
present. Retrieved April 1, 2011, from
http://blog.nielsen.com/nielsenwire/media_entertainment/top-ten-presidential-
debates-1960-to-present/

171
Niemi, R. G., Craig, S. C., & Mattel, F. (1991). Measuring internal political efficacy in the
1988 National Election Study. American Political Science Review, 85, 1407-1413.

Nir, L. (Forthcoming). Disagreement and opposition in social networks: Does

disagreement discourage turnout? Political Studies.

Parsons, B. (2010a). Passionate political talk: Social networks and the emotional impact
of political discussion. Unpublished Ph.D. dissertation, University of South
Carolina, Columbia, SC.

Parsons, B. (2010b). Social networks and the affective impact of political disagreement.
Political Behavior, 32, 181-204.

Patterson, T. E. (1993). Out of order. New York: Alfred A. Knopf.

Pew Internet & American Life Project. (2009). About our survey methodology in detail.
Retrieved July 10, 2009, from http://people-press.org/methodology/about/

Pratto, F., & John, 0. P. (1991). Automatic vigilance: The attention-grabbing power of
negative social information. Journal of Personality and Social Psychology, 61,
380-391.

Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing
and comparing indirect effects in multiple mediator models. Behavior Research
Methods, 40, 879-891.

Price, V., & Tewksbury, D. (1997). News values and public opinion: A theoretical account
of media priming and framing. In G. A. Barett & F. J. Boster (Eds.), Progress in
communication sciences: Advances in persuasion (Vol. 13, pp. 173-212).
Greenwich, CT: Ablex.

Price, V., & Zaller, J. (1993). Who gets the news? Alternative measures of news
reception and their implications for research. Public Opinion Quarterly, 57, 133-
164.

Prior, M. (2007). Post-broadcast democracy: How media choice increases inequality in

political involvement and polarizes elections. New York: Cambridge University
Press.

Prior, M. (2009a). The immensely inflated news audience: Assessing bias in self-reported
news exposure. Public Opinion Quarterly, 73, 130-143.

172
Prior, M. (2009b). Improving media effects research through better measurement of
news exposure. Journal of Politics, 71, 893-908.

Putnam, R. D. (1996). The strange disappearance of civic America. American Prospect,

24(\A/inter), 34-48.

Rafaeli, S. (1988). Interactivity: From new media to communication. In R. P. Hawkins, J.

M. Wiemann & S. Pingree (Eds.), Sage annual review of communication research:
Advancing communication science (Vol. 16, pp. 110-134). Beverly Hills, CA: Sage.

Redlawsk, D. P. (2002). Hot cognition or cool consideration? Testing the effects of

motivated reasoning on political decision making. Journal of Politics, 64, 1021-
1044.

Redlawsk, D. P., Civettini, A. J. W., & Lau, R. R. (2007). Affective intelligence and voting
information processing and learning in a campaign. In W. R. Neuman, G. E.
Marcus, M. MacKuen & A. N. Crigler (Eds.), The affect effect: Dynamics of
emotion in political thinking and behavior (pp. 152-179). Chicago: University of
Chicago Press.

Reese, S. D., Gandy, O. H., Jr., & Grant, A. E. (Eds.). (2001). Framing public life:
Perspectives on media and our understanding of the social world. Mahwah, NJ:
Lawrence Erlbaum Associates.

Robinson, J. P. (1976). Interpersonal influence in election campaigns: Two step-flow

hypotheses. The Public Opinion Quarterly, 40(3), 304-319.

Rogers, E. M. (2003). Diffusion of innovations (5th ed.). New York: Free Press.

Rojas, H. (2006, June). Orientations towards political conversation: Testing an

asymmetrical reciprocal causation model of political engagement. Paper
presented at the annual meeting of the International Communication
Association, Dresden, Germany.

Roseman, I. J. (1991). Appraisal determinants of discrete emotions. Cognition &

Emotion, 5, 161-200.

Roseman, I. J., Abelson, R. P., & Ewing, M. F. (1986). Emotions and political cognition:
Emotional appeals in political communication. In R. R. Lau & D. O. Sears (Eds.),
Political cognition: The 19th Annual Carnegie Symposium on Cognition (pp. 279-
294). Hillsdale, NJ: Lawrence Erlbaum.

173
Roseman, I. J., Wiest, C., & Swartz, T. S. (1994). Phenomenology, behaviors, and goals
differentiate discrete emotions. Journal of Personality and Social Psychology, 67,
206-221.

Rosenstone, S. J., & Hansen, J. M. (1993). Mobilization, participation, and democracy in

America. New York: Macmillan.

Rudolph, T. J., Gangl, A., & Stevens, D. (2000). The effects of efficacy and emotions on
campaign involvement. The Journal of Politics, 62, 1189-1197.

Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social

Psychology, 39, 1161.

Russell, J. A., & Bullock, M. (1985). Multidimensional scaling of facial expressions:

Similarity from preschoolers to adults. Journal of Personality and Social
Psychology, 48, 1290.

Russell, J. A., Weiss, A., & Mendelsohn, G. A. (1989). Affect grid: A single-item scale of
pleasure and arousal. Journal of Personality and Social Psychology, 57, 493.

Schemer, C. (2010). Reinforcing spirals of negative affects and selective attention to

advertising in a political campaign (No. Working Paper: 45). Switzerland:
National Centre of Competence in Research (NCCR).

Scherer, K. R. (1999). Appraisal theories. In T. Dalgleish & M. Power (Eds.), Handbook of

cognition and emotion (pp. 637-663). Chichester, UK: Wiley.

Scheufele, D. A. (2002). Examining differential gains from mass media and their
implications for participatory behavior. Communication Research, 29, 46-65.

Schmitt-Beck, R. (2008). Interpersonal communication. In L. L. Kaid & C. Holtz-Bacha

(Eds.), Encyclopedia of political communication (Vol. 1, pp. 341-350). Los
Angeles: Sage.

Schwa rz, N., & Clore, G. L. (1996). Feelings and phenomenal experiences. In E. T. Higgins
& A. Kruglanski (Eds.), Social psychology: Handbook of basic principles (pp. 433-
465). New York: Halford.

Searing, D. D., Solt, F., Conover, P. J., & Crewe, I. (2007). Public discussion in the
deliberative system: Does it make better citizens? British Journal of Political
Science, 37, 587-618.

174
Sears, D. 0., & Citrin, J. (1982). Tax revolt: Something for nothing in California.
Cambridge, MA: Harvard University Press.

Sei-Hill, K., & Miejeong, H. (2005). Media use and participatory democracy in South
Korea. Mass Communication & Society, 8, 133-153.

Semetko, H. A., & Valkenburg, P. M. (1998). The impact of attentiveness on political

efficacy: Evidence from a three-year German panel study. International Journal
of Public Opinion Research, 10, 195-210.

Shah, D. V., Cho, J., Eveland, W. P., Jr., & Kwak, N. (2005). Information and expression in
a digital age modeling internet effects on civic participation. Communication
Research, 32, 531-565.

Shah, D. V., Cho, J., Nah, S., Gotlieb, M. R., Hwang, H., Lee, N.-J., et al. (2007). Campaign
ads, online messaging, and participation: Extending the communication
mediation model. Journal of Communication, 57, 676-703.

Shah, D. V., Rojas, H., & Cho, J. (2009). Media and civic participation: On understanding
and misunderstanding communication effects. In J. Bryant & M. B. Oliver (Eds.),
Media effects: Advances in theory and research (3rd ed., pp. 207-227). New York:
Routledge.

Shaw, D. L., & McCombs, M. (1977). The emergence of American political issues. St. Paul,
MN: West.

Sheafer, T. (2007). How to evaluate it: The role of story-evaluative tone in agenda
setting and priming. Journal of Communication, 57, 21-39.

Shehata, A. (2010). Unemployment on the agenda: A panel study of agenda-setting

effects during the 2006 Swedish national election campaign. Journal of
Communication, 60(1), 182-203.

Shoemaker, P. J., Tankard, J. W., Jr., & Lasorsa, D. L. (2004). How to build social science
theories. Thousand Oaks, CA: Sage.

Slater, M. D. (2004). Operational and analyzing exposure: The foundation of media

effects research.Journalism & Mass Communication Quarterly, 81,168-183.

Slater, M. D. (2007). Reinforcing spirals: The mutual influence of media selectivity and
media effects and their impact on individual behavior and social identity.
Communication Theory, 17, 281-303.

175
Smith, C. A., & Kirby, L. D. (2001). Breaking the tautology: Toward delivering on the
promise of appraisal theory. In K. R. Scherer, A. Schorr & T. Johnstone (Eds.),
Appraisal processes in emotion: Theory, methods, research (pp. 121-138). New
York: Oxford University Press.

Son, Y. J., & Weaver, D. H. (2006). Another look at what moves public opinion: Media
agenda setting and polls in the 2000 U.S. election. International Journal of Public
Opinion Research, 18, 174-197.

Southwell, B. G., & Yzer, M. C. (2007). The roles of interpersonal communication in mass
media campaigns. Communication Yearbook, 31, 420-462.

Stimson, J. A. (1999). Public opinion in America: Moods, cycles, and swings (2nd ed.).
Boulder, CO: Westview Press.

Stroud, N. J. (2008). Media use and political predispositions: Revisiting the concept of
selective exposure. Political Behavior, 30, 341-366.

Stroud, N.J. (2010). Polarization and partisan selective exposure. Journal of

Communication, 60, 556-576.

Stroud, N. J. (2011). Niche news: The politics of news choice. New York: Oxford
University Press.

Stroud, N. J., & Kenski, K. (2007). From agenda setting to refusal setting: Survey
nonresponse as a function of media coverage across the 2004 election cycle.
Public Opinion Quarterly, 71, 539-559.

Sturgis, P., Allum, N., & Smith, P. (2008). An experiment on the measurement of political
knowledge in surveys. Public Opinion Quarterly, 72, 90-102.

Sullivan, D., & Masters, R. (1988). Happy warriors: Leaders' facial displays, viewers
emotions, and political support. American Journal of Political Science, 32, 345.

Tan, Y., & Weaver, D. H. (2010). Media bias, public opinion, and policy liberalism from
1956 to 2004: A second-level agenda-setting study. Mass Communication and
Society, 13, 412 - 434.

Tellegen, A., Watson, D., & Clark, L. A. (1999). On the dimensional and hierarchical
structure of affect. Psychol. Sci., 10, 297.

Teorell, J. (2006). Political participation and three theories of democracy: A research

inventory and agenda. European Journal of Political Research, 45, 787-810.
176
Thoits, P. A. (1989). The sociology of emotion. Annual Review of Sociology, 15, 317-342.

Valentino, N., Gregorowicz, K., & Groenendyk, E. (2009). Efficacy, emotions and the
habit of participation. Political Behavior, 31, 307-330.

Valentino, N. A., Banks, A. J., Hutchings, V. L, & Davis, A. K. (2009). Selective exposure in
the internet age: The interaction between anxiety and information utility.
Political Psychology, 30, 591-613.

Valentino, N. A., Brader, T., Groenendyk, E. W., Gregorowicz, K., & Hutchings, V. L.
(2011). Election night's alright for fighting: The role of emotions in political
participation. Journal of Politics, 73, 156-170.

Valentino, N. A., Hutchings, V. L, Banks, A. J., & Davis, A. K. (2008). Is a worried citizen a
good citizen? Emotions, political information seeking, and learning via the
internet. Political Psychology, 29, 247-273.

Valenzuela, S. (2009). Variations in media priming: The moderating role of knowledge,

interest, news attention, and discussion. Journalism and Mass Communication
Quarterly, 86, 756-774.

Valenzuela, S. (Forthcoming). Materialism, postmaterialism and agenda-setting effects:

The values-issues consistency hypothesis. International Journal of Public Opinion
Research.

Valenzuela, S., & Correa, T. (2009). Press coverage and public opinion on women
candidates: The case of Chile's Michelle Bachelet. International Communication
Gazette, 71, 203-223.

Valenzuela, S., Kim, Y., & Gil de Zuniga, H. (Forthcoming). Networks that matter:
Exploring the role of political discussions for online political participation.
International Journal of Public Opinion Research.

Verba, S., & Nie, N. H. (1972). Participation in America: Political democracy and social
equality. New York: Harper & Row.

Verba, S., Schlozman, K. L, & Brady, H. E. (1995). Voice and equality: Civic voluntarism in
American politics. Cambridge, MA: Harvard University Press.

Walsh, K. C. (2004). Talking about politics: Informal groups and social identity in
American life. Chicago: University of Chicago Press.

177
Watson, D., & Clark, L. A. (1997). Measurement and mismeasurement of mood:
Recurrent and emergent issues. J. Pers. Assess., 68, 267.

Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief
measures of positive and negative affect: The PANAS scales. Journal of
Personality and Social Psychology, 54,1063-1070.

Watson, D., & Tellegen, A. (1985). Toward a consensual structure of mood.

Psychological Bulletin, 98, 219.

Watson, D., & Tellegen, A. (1999). Issues in the dimensional structure of affect: Effects
of descriptors, measurement error, and response formats: Comment on Russell
and Carroll. Psychological Bulletin, 125, 601.

Way, B. M., & Masters, R. D. (1996a). Emotion and cognition in political information-
processing. Journal of Communication, 46(3), 48-65.

Way, B. M., & Masters, R. D. (1996b). Political attitudes: Interactions of cognition and
affect. Motivation and Emotion, 20, 205-234.

Weaver, D. (1980). Audience need for orientation and media effects. Communication
Research, 3, 361-376.

Weaver, D., & Drew, D. (1995). Voter learning in the 1992 presidential election: Did the
'nontraditional' media and debates matter? Journalism & Mass Communication
Quarterly, 72, 7-17.

Weaver, D., & Drew, D. (2001). Voter learning and interest in the 2000 presidential
election: Did the media matter? Journalism & Mass Communication Quarterly,
78, 787-798.

Weaver, D., Graber, D., McCombs, M., & Eyal, C. (1981). Media agenda setting in a
presidential election: Issues, images and interest. Westport, CT: Greenwood.

Weber, C. (2008). The emotional campaign. How emotions affect political behavior and
judgment. Unpublished Ph.D. dissertation, State University of New York at Stony
Brook, Stony Brook, NY.

Winneg, K., Kenski, K., & Adasiewicz, C. (2006). Naes datasets, survey procedures, and
content. In D. Romer, K. Kenski, K. Winneg, C. Adasiewicz & K. H. Jamieson (Eds.),
Capturing campaign dynamics, 2000 and 2004: The National Annenberg Election
Survey (pp. 14-42). Philadelphia, PA: University of Pennsylvania.

178
Wlezien, C., & Soroka, S. N. (2007). The relationship between public opinion and policy.
In R. J. Dalton & H.-D. Klingemann (Eds.), Oxford handbook of political behavior
(pp. 799-817). Oxford: Oxford University Press.

Wolak, J., MacKuen, M., Keele, L, Marcus, G. E., & Neuman, W. R. (2003, April). How the
emotions of public policy affect citizen engagement and public deliberation.
Paper presented at the annual meeting of the Midwest Political Science
Association, Chicago, IL.

Wright, S. (1918). On the nature of size factors. Genetics, 3, 367-374.

Wright, S. (1921). Correlation and causation. Journal of Agricultural Research, 20, 557-
585.

Zajonc, R. B. (1984). On the primacy of affect. American Psychologist, 39(2), 117-123.

Zaller, J. R. (1992). The nature and origins of mass opinion. New York: Cambridge
University Press.

Zevon, M., & Tellegen, A. (1982). The structure of mood change: An

ideographic/nomothetic analysis. Journal of Personality and Social Psychology,
43, 111.

Zhang, W., & Chia, S. C. (2006). The effects of mass media use and social capital on civic
and political participation. Communication Studies, 57, 277 - 297.

179
 Vita

Sebastian Valenzuela Leighton was born in Vina del Mar, Chile. Upon completion

of his high school degree from the American School of Quito, Ecuador, in 1995, he

attended Catholic University of Chile (PUC) and pursued a degree in journalism. After

graduating magna cum laude in November 2000, he worked in the economic news

section of El Mercurio newspaper in Santiago, Chile, as a reporter, writer and assistant

editor. In August 2005, he entered the graduate program in Journalism at the University

of Texas at Austin, where he earned a master of arts in May 2007. He began the

doctoral program in the fall of 2007 at the same university. His research interests

include political communication, digital media, media effects and quantitative research

methods. His work has been published or is forthcoming in Communication Research,

Journalism & Mass Communication Quarterly, Journal of Computer-Mediated

Communication, International Journal of Public Opinion Research, International Journal

of Press/Politics, Journalism: Theory, Practice & Criticism, CyberPsychology & Behavior,

International Communication Gazette and Cuadernos de Informacion.

Permanent address: Americo Vespucio N. 1650, Depto. 1201, Vitacura, Santiago, Chile

This manuscript was typed by the author.

180