GET OUT AND TALK ABOUT IT: INTERPERSONAL COMMUNICATION AS A MEDIATOR IN GET OUT THE VOTE CAMPAIGN SUCCESS By Rachel Carolyn Barry A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Communication – Master of Arts 2021 ABSTRACT GET OUT AND TALK ABOUT IT: INTERPERSONAL COMMUNICATION AS A MEDIATOR IN GET OUT THE VOTE CAMPAIGN SUCCESS By Rachel Carolyn Barry Get out the vote (GOTV) campaigns using different messages on myriad channels have promoted voter turnout, however, their effectiveness can vary widely and the mechanisms of treatment effectiveness remain unclear. By conducting a GOTV field experiment followed up by an online survey, this research aimed to better understand GOTV campaign effects and to determine whether and how interpersonal communication might play a significant role in campaign success. This study also considered the effects of conversation characteristics such as relational closeness, discussion agreement, and overall valence on the predicted relationship of campaign-driven discussion leading to pro-voting attitudes and voter turnout. Results found the treatment to be unsuccessful in driving turnout, and even negatively impacted some outcome variables. However, individuals who discussed voting more frequently had more positive attitudes toward voting and were more likely to vote. Furthermore, though cumulative campaign message exposure showed no direct effects on outcome variables, it did have indirect positive effects on voting attitude and voter turnout only through the mediating role of interpersonal communication. ii This thesis is dedicated to those continuing the fight for universal voting rights at home and abroad. iii ACKNOWLEDGEMENTS First, I would like to acknowledge my fellow get-out-the-vote collaborators, without whom this data would not have been collected and this study could not have happened: Dustin Carnahan, Dan Bergan, Sarah Reckhow, Kjerstin Thorson, Nazita Lajevardi, and Ana Bracic – thank you for allowing me be part of the GOTV team and for persevering as plan A, B, C, D (did we ever make it to Z?) fell through. To my wonderful advisor, Sandi Smith, thank you for being the perfect mix of inspirational, calming, critical, and kind. Additionally, thank you Monique Turner and (again) Dustin Carnahan for being a part of my thesis committee. Time is truly valuable, and I appreciate each of you acknowledging the value of my efforts by supporting me and taking the time to read this work. To all three of you, the guidance you have offered me throughout my master’s experience has been precious. For it, I am a better thinker and more skilled researcher. Additionally, thank you to my coders – Nick Mondragon and Kat Ray. If at first you don’t succeed, try try again. Thank you both for sticking with me and being flexible throughout the coding process. Finally, professional accomplishments would not be possible if not for personal support. A huge shout out to my partner, Brandon Wade, and my cohort champions, Molly Carlisle and Luc Savoie. I might have survived without you, but it would not have been pretty. iv TABLE OF CONTENTS LIST OF TABLES ……………………………………………………………... vi LIST OF FIGURES ……………………………………………………………... vii INTRODUCTION ……………………………………………………………... 1 LITERATURE REVIEW ……………………………………………………………... 3 Interpersonal Communication as a Mediator………………. 3 GOTV Campaigns………………………………………..... 4 Information…..…….………………………………… 5 Implementation Plan…………………….…………… 6 Message Exposure……………………….…………... 7 Interpersonalization………………………………………... 8 Communication Characteristics as Moderators……………. 11 METHOD RESULTS DISCUSSION APPENDICES REFERENCES ……………………………………………………………... 14 Design…………………………………………………….... 14 Participants……………………………………………….... 14 Procedure………………………………………………....... 14 Information Treatment………………………….......... 15 Implementation Plan Treatment………………........... 16 Measures………………..........………………...................... 17 Discussion Frequency………………..........…………. 17 Discussion Characteristics……………….................... 17 Partner Agreement………………....................... 17 Partner Closeness………………......................... 17 Conversation Valence……………….................. 17 Voting Attitude………………..................................... 18 Voter Turnout………………....................................... 18 Cumulative GOTV Campaign Exposure…………….. 18 Demographics………………..........………………..... 18 ……………………………………………………………... 19 ……………………………………………………………... 23 ……………………………………………………………... 27 APPENDIX A: Tables & Figures..………………………… 28 APPENDIX B: Treatment Materials………….…………… 34 APPENDIX C: Codebook…………………….…………… 38 ……………………………………………………………... 40 v LIST OF TABLES Table 1. Demographic Frequencies……………………………………………………. Table 2. Summary Statistics of Outcome Variables…………………………………… Table 3. PROCESS Analysis Summary for Voter Turnout with Experimental Campaign Exposure……………………………………………….. Table 4. PROCESS Analysis Summary for Voter Turnout with Cumulative Campaign Exposure….……………………………………………… Table 5. Pre-Collapsed Group Mean Comparisons of Outcome Variables by Treatment….………………………………………………………………………… Table 6. Codebook for Comments……………………………………………………… 29 30 31 32 33 39 vi LIST OF FIGURES Figure 1. Interpersonalization Model of Voting ……………………………………….. Figure 2. Interpersonalization Model of Voting with Experimental Campaign Exposure .…………………………………………………………………… Figure 3. Interpersonalization Model of Voting with Cumulative Campaign Exposure ………….………………………………………………………… Figure 4. Sample Breakdown from Student Census to Final Analysis…………………. Figure 5. Information Treatment Letter….……………………….…………………….. Figure 6. Implementation Plan Treatment Letter….……………………………………. Figure 7. Implementation Plan Treatment Cue to Action….…………………………… 11 20 21 29 35 36 37 vii INTRODUCTION Get-out-the-vote (GOTV) campaigns, coordinated efforts to increase voter turnout, are conducted around the world using myriad media. Some studies of these campaigns in a variety of contexts have shown them to increase voter turnout (e.g., Enos, Fowler, & Vavreck, 2014; Jones et al., 2017; Middleton & Green, 2008, Green & Gerber, 2019). Contrastingly, other GOTV efforts have shown meager or non-significant results (e.g., Cardy, 2005; Nickerson, 2008), and this inconsistency is not readily understood. One reason for the unpredictability might be the success (or failure) of GOTV appeals in promoting interpersonal communication around voting, which has been shown to influence voting behavior (e.g., Pattie & Johnston, 2000). Pattie and Johnston (2000) invoked the phrase, “Those who talk together, vote together” (p.63) to describe how interpersonal political communication can influence who people vote for, but less is known about whether political discussions can actually influence the decision to vote. The influence of interpersonal communication on outcomes of persuasive campaigns is supported in other contexts, such as health (Jeong & Bae, 2018). But to date, the aspects of the conversation that lead to the voter attitude change and turnout are unclear. As part of an actual GOTV campaign taking place at a large, Midwestern university, this research aimed to better understand the impact GOTV campaign interventions have on campaign-relevant discussion and, subsequently, whether that discussion influences the campaign’s overall success in driving up voter turnout. A review of the GOTV literature is provided along with work that suggests a connection between interpersonal communication and voter turnout, informed by the social diffusion model. Working from this literature, a series of predictions and research questions involving interpersonal communication as a mediator in the relationship between GOTV campaign exposure and voter turnout are proposed. These 1 relationships were observed using data collected from a direct mail GOTV field experiment accompanied by a post-test survey among first-time voters, allowing insights into interventions’ impact on campaign-relevant discussion, attitudes towards voting, and turnout. Considering that discussions can vary greatly in ways other than just frequency, message characteristics known to affect persuasive outcomes, such as partner agreement, closeness, and conversation valence, are also evaluated. Results are analyzed, future directions are discussed, and implications and limitations are addressed. 2 LITERATURE REVIEW Interpersonal Communication as a Mediator Interpersonal communication is often noted as being difficult to define (Cappella 1987; Southwell & Yzer, 2009), but for the purpose of this research it is conceptualized with an interactional approach such that “each person must affect the other's observable behavior patterns relative to their typical or baseline patterns” to be considered interpersonal communication (Burleson, 2010, p.6; Cappella, 1987). McLeod, Scheufele, and Moy (1999) acknowledged the import of interpersonal discussion when it comes to impact on political engagement. Shah et al. (2007) extended this work, finding that interpersonal political discussion mediates the relationship between media exposure and civic participation. Furthermore, Cho and colleagues (2009) advanced Shah et al.’s O-S-R-O-R (orientations-stimuli-reasoning-orientations-responses) model with their additions of media reflection and political knowledge to the theoretical framework. They argued against the idea of passive citizens, and suggested interpersonal interaction plays a critical role in the outcomes of campaigning by the political elite. However, their work focuses primarily on prolonged exposure to myriad media as opposed to a single-exposure or directed campaigns. Though the role of interpersonal communication as a mediator appears clear, analysis of its relationship with voting attitudes and behavior is less abundant, and experiments on this topic that can point to causal claims seem to be missing from the literature completely. Previous findings have shown that interpersonal communication plays a moderating role in which candidate people choose to vote for (Cho, 2005) and other studies have shown a correlation between political discussion and voting preference (Pattie & Johnston, 2000). Mutz (2002) found that certain types of conversation, specifically frequent cross-cutting (discussing with someone 3 you disagree with) can lead to a decrease in voting. This logically indicates that less polarizing discussion should be beneficial to voter turnout, but stops short of demonstrating that more can be better. That is, can talking about voting actually get people to vote? And, if so, how can we get people talking? One potential way is through exposure to GOTV messages. GOTV Campaigns GOTV campaigns have utilized a variety of channels such as door-to-door canvassing (e.g., Middleton & Green, 2008), mail (e.g., Cardy, 2005), text (e.g., Dale & Strauss, 2009), and social media (e.g., Aldrich et al., 2016; Haenschen, 2016) to increase voter turnout. For the context of this study, turnout refers to the act of voting in an election and does not take into account candidate choice or if the individual votes in person or via absentee ballot. Even in high profile elections when the media is saturated with information, results of GOTV campaigns have shown significant increases in treated publics when it comes to voter turnout (Middleton & Green, 2008). Though effects of these campaigns are typically small (Green & Gerber, 2019), recent U.S. elections have shown a few percentage points can be consequential for election outcomes, even in national contests. In-person treatments have been shown to be the most effective channel in promoting voter turnout (Green & Gerber, 2019); but in an increasingly virtual world, research has sought to identify the effectiveness of technologically oriented strategies. However, evidence suggests these efforts to be limited in their influence and efficiency. Although email is highly cost- effective, studies have shown this method to be highly ineffective particularly when trying to get youth to vote (e.g., Nickerson, 2008). Despite the higher relative costs associated with phone banks (see Green & Gerber, 2019 for a review), GOTV campaigns executed over the phone have been shown to be similarly ineffective (Green & Gerber, 2000) and suffer from additional 4 barriers that might make contact difficult, especially for younger demographics (Pew Research Center, 2019). Conversely, non-partisan direct mail voting has been shown to have positive effects on voting behavior (Green & Gerber, 2019). These effects are consistently small, however, highlighting the opportunity to strengthen results by having a clearer idea of the mechanisms with which these campaigns lead individuals to vote. Information A critical aspect of GOTV messages, particularly for new voters, is providing necessary information. Indeed, Lassen (2005) found that more informed individuals are more likely to turnout to vote. Unfortunately, research has found that one in five young people do not believe they should vote because they do not feel they have the appropriate knowledge to do so (Kawashima-Ginsberg & Kiesa, 2019). This could be a contributing factor to the fact that only 50% of youth aged 18 to 24 voted in the 2016 election, as opposed to 65% of those aged 25 and older (Costa et al., 2018). One reason youth may be less likely to vote than their older counterparts is due to the barrier of knowledge when one first votes, such as figuring out how to register and finding his/her/their polling place. Therefore, it is possible that by providing young people with information about the registration and voting processes, they will be more likely to take action. When it comes to GOTV campaigns, however, information treatments typically do not show significant effects on their own, particularly when the information is sent via direct mail (Green & Geber, 2019). In other contexts, such as health, information treatments have been found to increase knowledge without being effective in motivating individuals to actually participate (e.g., Gagliano, 1988; Meropol et al., 2016). With mailers being a high-cost and 5 impersonal form of intervention, it appears necessary to have additional efforts to increase the likelihood of engagement with the materials and, subsequently, increase their effectiveness. Implementation Plan Going one step further than simply providing information is encouraging potential voters to make a plan to vote. Also known as implementation intentions, making a plan to act increases the likelihood of following through (Gollwitzer, 1999). In the GOTV literature, making a plan to vote has shown mixed results with some studies actually demonstrating a decrease in turnout (Green & Gerber, 2019). However, other sources have found success in this approach (Green & Geber, 2019) with Nickerson and Rogers (2016) noting a 4.1% increase in turnout after including implementation intention measures to a scripted intervention; specifically, asking what time the respondent planned to vote, where they would be coming from to do so, and what they would be doing before they voted. Beyond simply making a plan, social pressure has historically been used successfully to encourage individuals to turn out and vote (Green and Gerber, 2019). However, in-person social interaction is not always feasible, and it is important to uncover ways to replicate this type of social pressure. One such way is by encouraging potential voters to create an implementation plan to vote and to display that commitment in their home. This also acts as a reminder to the individual of their commitment to vote, which has been shown to increase turnout (Green & Gerber, 2019). Indeed, Costa et al. (2018) found that reminding potential voters of a previously made plan to vote increased turnout, especially when the individual had never voted before, and pledging to vote in a way that is visible to others is more effective than doing so privately. Specifically in youth GOTV campaigns, sizeable effects have been found when encouraging potential voters to make a pledge to vote (Bergan et al., 2021). It is thus anticipated that 6 emphasizing plan creation and sending out a reminder to potential voters will work better than just providing information. Message Exposure As Hornik and Yanovitzky (2003) emphasize in their analysis of anti-drug media campaigns for youth, it is worthwhile to work through and measure various routes of exposure. In the case of GOTV, one such consideration is the pure level of pro-voting messages an individual is exposed to. Green and Geber (2019) indicate that five to six messages is ideal when it comes to number of GOTV messages to maximize turnout effects. This could be due to an increase in information salience, which studies have shown does not directly lead to attitude change but may influence attitude and, subsequently, behavior via the indirect route through interpersonal communication (Berger & Iyengar, 2013; Jeong & Bae, 2018). That is, with each exposure information salience increases, thereby leading to higher accessibility to the message’s information. Consequently, the likelihood of initiating discussion on that topic increases. However, it is unclear from previous research if interpersonal communication’s role as a mediator is dependent on exposure frequency or type. Therefore, both are tested in this study. At face value, both information and implementation plan treatments have the potential to promote pro-voting attitudes, but it is expected that the implementation plan treatment will be more effective to this end. In general, more exposure to campaign messaging will have more positive effects than less exposure. This leads to the first sets of hypotheses: H1a: GOTV campaign message exposure will promote positive attitudes toward voting. H1b: The implementation plan treatment will promote more positive attitudes toward voting than the information treatment alone. A direct relationship on voting is also predicted, so that: 7 H2a: GOTV campaign exposure will promote voter turnout. H2b: The implementation plan treatment will promote voter turnout more than the information treatment alone. Yet, the variation in success of past GOTV efforts with both information and plan interventions leads to the speculation that there are currently unknown variables at work mediating the relationship between these treatments and campaign success. Interpersonalization The positive benefits of personalization in GOTV campaigns are clear. As Green and Gerber (2019) point out, methods that are “utterly impersonal rarely get people to vote” (p. 17). Perhaps it is not just the personalization of the campaign message that matters, but rather the incorporation of message materials into interpersonal conversations that leads people to act. This proposed concept, from this point referred to as interpersonalization, is supported in the interpersonal communication as mediation literature and is specifically defined as campaign- driven interpersonal discussion that has the capacity to change behavior. Cho et al. (2009) emphasize a needed perspective shift from elite campaign effects on voter choice and turnout to the actual communication processes of individuals amongst one another. However, the argument here is not a question of whether they have post-campaign communication or not, but rather if there is a process from campaign exposure to interpersonal communication to changes in attitude towards voting and voter turnout. This is in line with Hornik and Yanovitzky’s (2003) theorizing on social diffusion as a means of campaign exposure. They argue that campaign effects are complex and looking directly from exposure to behavior is overly simplistic. While it is expected that campaigns have some direct effect on behavior, it is further proposed that the more effective route is through 8 interpersonal communication. This is in accordance with the social diffusion model, which posits that “campaign activity generates relevant conversation, which, in turn, affects a person’s normative perceptions and, ultimately, behaviors” (Hwang, 2012, p.124). As noted by Chaffee and Mutz (1988) “mass media often provide grist for the conversation mill and stimulate informal discussions that might not otherwise take place” (p. 21). Furthermore, research supports that variation in campaign message affects the types of resultant conversations (Dunlop et al., 2010). In the GOTV research, Bergan et al.’s (2021) study on the 2018 midterm election found that individuals treated with a social pressure intervention in addition to information (specifically, pledging to vote and sharing that pledge with others), were more likely to discuss the election with a roommate. They even uncovered spillover effects, so that the roommates of individuals in this treatment group were more likely to vote as well, signifying campaign-relevant communication took place between these individuals. One prediction, then, is that campaign exposure should generate discussion, with campaigns encouraging plan implementation resulting in higher frequency of conversation. This leads to a third set of hypotheses: H3a: GOTV campaign exposure will promote discussion about voting. H3b: The implementation plan treatment will promote discussion about voting more than the information treatment alone. Interpersonal communication plays an important role in how far of reach a campaign has and how one processes new information (Katz & Lazarsfeld, 1955; Rogers, 1983). It has also been found to modify the success of health campaigns (e.g., Dunlop et al., 2010; Hafstad et al., 1997, Hwang, 2012; Jeong & Bae, 2018). For example, Dunlop et al., (2010) found that interpersonal discussion could mediate the effects of pro-vaccine messages on attitude and 9 intention. Furthermore, Hwang (2012) provided support for both the individual and social diffusion models when analyzing campaign exposure effects on beliefs as mediated by communication. A meta-analysis by Jeong and Bae (2018) determined health campaigns that generated some campaign-specific interpersonal discussion were 1.28 times more likely to be successful in their efforts. Though a relatively small effect size, GOTV campaigns typically move in inches not feet. Therefore, borrowing from the health literature, these findings suggest interpersonal discussion if applied to a political context could benefit efforts to increase voting. It is therefore predicted that: H4a: Discussion about voting will promote positive attitudes towards voting. H4b: Discussion about voting will promote voter turnout. Though the mechanisms at play between attitude and behavior are oft debated, the link between these constructs is well-established (Kim & Hunter, 1993). Southwell and Yzer (2007) caution that promoting discussion without understanding the underlying attitudes of the audience can have negative effects on a campaign’s overarching goals. It is therefore taken into account, and leads to a fifth hypothesis: H5: Individuals who have a more positive attitude toward voting are more likely to vote. In sum, it is predicted that exposure to GOTV interventions will directly and positively influence discussion frequency, attitude towards voting, and turnout at varying degrees. It is also expected that GOTV interventions will generate increased discussion about voting, subsequently influencing more positive attitudes toward voting which, in turn, will have a positive, direct effect on turnout. While never tested in the context of a GOTV campaign, a recent article (Solovei et al., 2020) found more indirect effects than direct effects when analyzing Dutch non- health related public information campaign exposure effects on awareness, knowledge, attitude, 10 intention and behavior through the mediation of interpersonal communication. That is, discussion influenced campaign-relevant outcomes above and beyond the effects of campaign exposure on those same outcomes. Similar campaign effects may be found outside of the contexts of health and within the United States as well. The following serial mediation is henceforth predicted: H6: GOTV exposure will have a serial indirect effect on voter turnout via voting discussion frequency and attitudes toward voting. Figure 1. Interpersonalization Model of Voting Communication Characteristics as Moderators Finally, if interpersonalization is campaign-driven interpersonal discussion that leads to changes in behavior, one must consider that not all conversations are the same. Previous research on campaigns has called for a better understanding of the mechanisms within the communication 11 that leads to certain outcomes (Gil de Zúñiga et al., 2016). As previously mentioned, Mutz (2002) found that level of agreement can have a direct positive effect on political engagement. It is yet unclear, however whether agreement with one’s partner will affect the predicted mediated interpersonalization model of voting (see Figure 1). More specifically, she proposed that individuals tend to avoid partaking in political behaviors if they feel it puts their personal relationships at risk. Contrastingly, some studies (e.g., Bode et al., 2018) find political discussion to actually increase political engagement. This was found even in discussion amongst coworkers; unlike Mutz (2006), which argues for distinctions in behavioral effects caused by talking with one’s friends and families versus talking with coworkers. This points to something deeper than simply agreeing or disagreeing with one’s communication partner. Rather, the closer you feel to someone may also play a role in the resultant effects of political discussion, though in which way is yet to be fully understood. Perhaps the most established relationship is the one between the valence of the conversation and persuasive outcomes. In the health communication literature, conversational valence has been associated with promoting positive health attitude and behavior change (e.g., Frank et al., 2012; Hendricks et al., 2014; Dunlop et al., 2010). Hendricks et al. (2014) found that fear-based messages regarding binge-drinking led to negative-valence conversations and, subsequently, more negative attitudes toward binge-drinking behavior. Furthermore, Dunlop et al. (2010) reported that individuals who had more positive conversations regarding vaccines had higher behavior intention. In the context of voting, emotions of differing valence (i.e., enthusiasm and anxiety) have been linked to partaking in political deliberation (Marcus & Mackuen, 1993) but the link between the valence of conversation about voting and its influence 12 on turnout is yet to be empirically tested. Therefore, two research questions are ultimately proposed to address each of these potential moderators: RQ1: How do the conversation characteristics of (a) agreement, (b) closeness, and (c) valence affect the proposed relationship between voting discussion frequency and voting attitude? RQ2: How do the conversation characteristics of (a) agreement, (b) closeness, and (c) valence affect the proposed relationship between voting discussion frequency and voter turnout? 13 METHOD Design Potential participants (N = 12,935) were drawn from a census of in-state freshman and sophomore classes at a large, Midwestern university (see Appendix A). Individuals in this sample were randomly assigned to one of three treatment groups: information treatment (n=2090)1, implementation plan treatment (n=4224), and control (n=6621). Once randomly assigned, the treatment was mailed to their current address (via university registrar). Individuals in the control group received nothing. Final sample sizes based on returned post-tests are n = 190 (info), n = 355 (plan), and n = 577 (control) for a total of 1122. Overall response rate was 8.67%, with respective group response rates equaling 9.09 % (information), 8.40% (implementation plan), and 8.71% (control). Participants The final sample who completed the post-test survey (n = 1122) was disproportionately white (76.1%), female (66.4%), and of higher socioeconomic status (53.6% household income of $100,000+). The sample skewed toward liberal, with 62.6% of respondents identifying as at least somewhat liberal and only 21.8% being at least somewhat conservative. 15.6% of the participants declared they were moderate/middle of the road (see Table 1 for more demographic details). Students who completed the post-test were put into a lottery to receive one of five $100 Amazon gift cards as an incentive for their time. Procedure Individuals in both the information and implementation plan treatments received a mailed 1 Due to under printing of materials, 22 individuals in the information treatment did not receive the mailer and were subsequently moved to the control treatment. 14 letter to their current address about one month prior to Election Day, however the design and resources provided on each were altered (see Appendix B)2. The control group did not receive any materials. The post-test (see Appendix C) gauged voting attitude. It also asked if the respondent had talked to anyone they lived with about voting in the 2020 election, and, if affirmative, subsequently measured discussion frequency and partner agreement level, closeness to partner, and conversation valence. Voting behavior was self-reported but is being confirmed using a private vendor voter match service. The vendor matches university registrar data to the state’s voting record. More specifically, the vendor was provided specified student information (e.g., age, gender, race, home address), which they used to match to the state’s voting databases.3 Accuracy of matches will be cross-checked with the registrar list. Finally, though one-shot interventions have been effective in the past, it is beneficial to consider that in a field experiment, participants are not in a vacuum and are very likely to be exposed to pro-voting message elsewhere. Therefore, cumulative GOTV message exposure was also measured via self-report. Information Treatment The information treatment included a mailer (see Appendix B) from the university’s voting organization (a “non-partisan committee of students, faculty, and administrators that informs, empowers, and encourages students to participate in elections”). Information was provided on how to register, apply for an absentee ballot, how to cast a ballot, and how to check 2 Treatments were originally broken up into information letter, information letter with magnet, plan letter, plan letter with magnet, plan letter and follow up message, and plan letter with magnet and follow up message. This was done to better understand the incremental effects of each piece of the treatment. Upon analysis, however, there were no significant differences between groups within a treatment if they had or had not received the magnet on outcome variables and groups were subsequently collapsed into information and plan treatments (see Table 5). 3 Due to contractual hold ups, the vendor match data is not yet available. Once it is received, all analyses referenced in this article will be run again using the voter match results as the outcome variable as opposed to user self-report. 15 the status of an absentee ballot. The university’s voting organization page URL was also provided, indicating students can access it for further information. Five QR codes on the cover letter linked to websites for state voter registration, registration update and ballot tracking, clerk information, absentee ballot request, and drop box locations, respectively. Some mailers also included a magnet, which provided the same information as the letter.2 The magnet showed six steps on “How to vote by mail” including: register to vote, request your absentee ballot, fill out your ballot when it arrives, sign your ballot, return your ballot ASAP, and check the status of your ballot. Three QR codes were presented, one each for registration, requesting an absentee ballot, and checking the status of an absentee ballot. Implementation Plan Treatment Those in the plan treatment received an envelope with voting information mailed to their home address.2 However, in addition to information, the cover letter also included the phrase: “Committing to a vote plan has been shown to make people more likely to follow through and vote. We have included a magnet where you can make a plan to vote this election year. We will check back with you in a few weeks to follow up about your vote plan” and ends with “Please take some time to make a plan to vote.” Some individuals in this treatment also received a magnet, but it was writable and included a checklist with the fillable phrase “I, _________, am making a plan to vote by mail.” The same list of six steps was provided, but with markable checkboxes instead of numbers and the directions “Make a plan. Check the boxes.” The three QR codes were also included. Moreover, those in the implementation plan group received an email two weeks prior to election day reminding them of their plan to vote (see Appendix A). This reminder again provided information on registration, absentee voting, and in-person voting. It encouraged them to still cast a ballot if their original plan had fallen through. 16 Measures Discussion Frequency Discussion frequency (M = 3.89; SD = 0.97) was gauged by a single item asking respondents how frequently they talked about voting with the people they lived with in the month leading up to Election Day. A 5-point scale from “never” to “very often” was utilized. Discussion Characteristics If respondents indicated they did more than “never” talk about voting with someone they lived with, they were subsequently asked three additional questions about the characteristics of the discussion(s) and the partner with whom they discussed voting the most. Partner Agreement. Still thinking of the person with whom they live with and talked about voting with the most, participants indicated the level to which they agreed or disagreed in their conversations about voting on a scale of 1 “mostly disagree” to 5 “mostly agree” (M = 3.92; SD = 1.45) Partner Closeness. Respondents were asked to indicate how close they are with the individual that they live with whom they talked about voting with the most on a scale of 1 “not close at all” to 3 “very close” (M = 2.80; SD = 0.44). Conversation Valence. An open-ended question asking the respondent to describe their most memorable discussion with the individual with whom they talked about voting the most was provided, and responses were coded for positive/negative elements. The unit of analysis was set as the entire comment, and two independent coders were trained (see Appendix C for codebook). To calculate intercoder reliability, after two practice rounds coders received 10% of the sample (n = 80) comments to code reaching a Cohen’s Kappa = .829. Once intercoder reliability was 17 established, coders split the remaining 722 comments. Comments were coded as neutral (32.6%), negative (25.4%), both negative and positive (6.9%) or positive (35.1%). Voting Attitude Three items, each on a 7-point semantic differential scale, measured attitude toward voting in the 2020 election. Items included good/bad, positive/negative, and desirable/undesirable (Osgood, Suci, & Tannenbaum, 1957). Cronbach’s alpha was originally .681, but after dropping the item desirable/undesirable, Cronbach’s alpha was an acceptable .811 (M = 6.50, SD = 0.97). Voter Turnout Voter turnout was, in the short-term, measured by self-report (nVote = 1065, nNoVote = 52). Long-term, use of a private vendor voting match service will be utilized to provide a more accurate measure of voter turnout. Cumulative GOTV Message Exposure Though campaign exposure was manipulated via the field experiment treatments, knowing that it was highly likely that individuals would be exposed to other pro-voting messaging, a single item was used to measure the level of this exposure. Specifically, respondents were asked to indicate the frequency with which they were exposed to pro-voting messaging in the three months leading up to Election Day on a 3-point scale from not at all to very frequently (M = 2.93; SD = 0.27). Demographics To control for sample variance, typical demographic measures were recorded along with political affiliation and party identity (see Table 1). 18 RESULTS Hayes’ (2018) PROCESS macros was used in SPSS 27 to test the predicted direct effects and serially mediated indirect effects, which were estimated with bootstrapped confidence intervals (N = 5,000; see Table 3). When comparing all treated individuals to the control, our GOTV messages surprisingly had a negative effect on attitudes toward voting (b = -0.155, p < .01). Breaking down to the treatment level, both the information (b = -0.191, p <.05) and implementation plan (b = -0.135, p = .05) treatments again showed negative direct effects on attitude. Thus, neither H1a nor H1b were supported. Turning to turnout, when comparing treated individuals to the control there was no evidence of an effect on voter turnout (b = -0.525, p = .177). The implementation plan treatment alone also showed no relationship (b = -0.159, p = .728). However, the information treatment did have a negative effect on turnout. (b = -0.993, p < .05). Subsequently, neither H2a nor H2b were supported. Again, looking at treated versus untreated persons, the field experiment interventions had a direct and negative effect on discussion (b = -0.123, p = 0.05). When turning to those in the treatment groups, the information treatment yielded no significant effect on discussion and the plan treatment had a negative effect that bordered on significance (b= -0.135; p = 0.057). Therefore, no support was found for H3a or H3b. Discussion frequency had positive, direct effects on both attitude towards voting (b=0.160, p <.001) and voting turnout (b=0.602, p < .001) supporting H4a and H4b. Unsurprisingly, and in support of previous literature, individuals who had more positive attitudes toward voting were more likely to vote, (b=0.571, p <.001), in support of H5. When testing the entire model, experimental campaign exposure’s negative effect on turnout was mediated by 19 attitude (b = -.085, 95% CI: -0.167 to -0.019) but when discussion was added to the model this effect was no longer significant. Discussion was not found to significantly mediate the relationship between experimental campaign exposure and voter turnout alone (b = -.072, 95% CI: -0.181 to 0.003) or through attitude (b = -.011, 95% CI: -0.027 to 0.000; see Figure 2). To address the research questions of partner agreement, partner closeness, and conversational valence as moderators, models were run considering each as a moderator to the relationship between campaign-driven discussion frequency and voting attitude (RQ1) and turnout (RQ2). Resultant indices of moderated mediation find no evidence that partner agreement (b= -0.001, 95% CI: -0.006 to 0.004), closeness (b= -0.004, 95% CI: -0.024 to 0.009), or conversational valence at positive (b= -0.007, 95% CI: -0.043 to 0.016) or negative levels (b= -0.002, 95% CI: -0.028 to 0.015)4 moderated the mediated relationship. Figure 2. Interpersonalization Model of Voting with Experimental Campaign Exposure Note. * p <.05; ** p < .01; *** p < .001 4 Forty-eight comments contained both negative and positive thoughts, and of these only one individual had not voted. Therefore, this category was not run in the moderated mediation analyses. 20 The cumulative campaign exposure measure indicated the sample was strongly skewed to high message exposure. Concerns that a single mailed intervention would not stand out in such a group, the decision was made to utilize cumulative GOTV message exposure (measured via self- report) to re-run analyses using the entire sample (see Table 4). The logistic regression analysis yielded no direct relationship between cumulative campaign exposure and attitude toward voting (b= 0.193, p = .09). Additionally, cumulative campaign exposure was not significantly related to voting turnout (b= 0.619, p = .191). However, cumulative campaign exposure did have a direct, positive relationship with discussion frequency (b=0.394, p = .001). Figure 3. Interpersonalization Model of Voting Results with Cumulative Campaign Exposure Note. *** p < .001 Looking again at the simple mediation effect, cumulative campaign exposure had a significant, positive indirect relationship (b = 0.219) with turnout via discussion frequency based on bootstrapped confidence intervals (N=5000; 95% CI: 0.042 to 0.469). The effect (b= 0.036) of campaign exposure on voter turnout through discussion frequency and positive attitude also 21 had a confidence interval statistically distinct from zero (95% CI: 0.009 to 0.076), but not when discussion frequency was removed (b= 0.110; 95% CI: -0.026 to 0.278). Though not brought about by experimental conditions, these analyses suggest at least partial support for H1a, H2a, H3a, and H6 (see Figure 3). Addressing the research questions, the cumulative campaign exposure models were run with partner agreement, partner closeness, and conversational valence as moderators. In each case, indices of moderated mediation again suggest no evidence that partner agreement (b= 0.003, 95% CI: -0.009 to 0.018), closeness (b= 0.019, 95% CI: -0.019 to 0.088), or valence at negative (b= -0.012, 95% CI: -0.056 to 0.106) or positive levels (b= -0.043, 95% CI: -0.035 to 0.165) moderated the mediated relationship. 22 DISCUSSION The uncertainty of the COVID-19 pandemic colliding with a general election year demonstrated the importance of having a better understanding of impersonal methods readily available to reach potential voters. With a fully in-person treatment completely derailed at the 11th hour, it became imperative to figure out how best to reach new voters and get them the information they needed prior to Election Day. Though a global pandemic may not be the norm, uncertainty in youth voters is, and with a world forever going more virtual it is useful to investigate how best to engage new voters in meaningful pro-voting messaging without the established advantages of face-to-face interaction. It was expected that exposure to GOTV campaign messages would directly and positively affect voting attitudes and behavior, but that these outcomes would be mediated by interpersonal communication. More specifically, it was predicted that GOTV campaign exposure would increase discussion frequency which, in turn, would lead to more positive attitudes toward voting and subsequently greater turnout. The discussion partner agreement and closeness and conversational valence were considered as potential moderators of this relationship, specifically moderating the relationship between campaign-driven discussion frequency and attitudes toward voting as well as the relationship between discussion and turnout. Pro-voting attitude was anticipated to lead to greater turnout as well. Post-hoc analyses using cumulative GOTV message exposure indicate that pro- voting messaging in general can be beneficial to the cause, but only under certain conditions. That is, increased recall of cumulative GOTV message exposure in the three months leading up to Election Day did relate to higher turnout, but only when mediated by interpersonal discussion. This shows support for the originally predicted model. Specifically, in lieu of personalization of 23 the campaign message, incorporation of message materials into interpersonal conversations can lead people to act through the process of interpersonalization. This echoes findings from health campaigns (e.g., Schuster et al., 2006) as well as the 2020 study by Solovei et al., which concluded that interpersonal communication after campaign exposure increased persuasive outcomes transcending the direct effects of media exposure. This study found that interpersonalization can play a role in voter turnout; and seeing that none of the suggested conversation characteristics played a moderating role, its effect appears to be quite robust. That is, regardless of the level of agreement between discussants, closeness of relationship, or conversational valence, discussion still played a mediating role between campaign message exposure and more positive voting attitudes and greater turnout. This makes it easier for future GOTV campaign creators because they can focus on driving discussion generally as opposed to encouraging a particular type of conversation with specific individuals. Furthermore, though the serial mediation model was significant, discussion frequency also acted alone as a mediator between campaign message exposure and voter turnout outside of its influence on voting attitudes. This is also simpler on campaign design, because it suggests that pre-testing to gauge voting attitude of one’s audience is, though perhaps beneficial, is not completely necessary. Moving forward, variance should be introduced into the sample to see if these relationships hold. For instance, discussion partners in this study were living in the same household. Future studies should observe various living situations between communication partners and see how that effects the relationships shown here. It is also important to note that this study did not test the social diffusion model in its entirety, for to do so the role of perceived norms would need to be addressed (Hwang, 2012). It is highly plausible that the discussion 24 generated from campaign exposure affected individual perceptions of other’s approval of voting (injunctive norm) and turnout (descriptive norm), thereby leading them to conform their own attitudes and behaviors. However, this research is unable to speak to this relationship and subsequently suggests empirically addressing it as a prospective opportunity. This study is not without its limitations. The treatment was a relatively small intervention during a particularly active election cycle; thus, a single mailer may have just not been influential enough to motivate interpersonal discussion. Furthermore, majority of the sample voted and had been highly exposed to pro-voting messaging. This might have been exacerbated by the fact that, though the original sample was a census of the entire in-state underclassman population, the post-test survey was sent out via email where individuals could choose to complete it. Hence, there may be some self-selection bias such that, for example, those more politically interested could have been more likely to complete the survey. It will be beneficial to rerun all analyses once the voter match data previously referenced is available. In the meantime, it must be acknowledged that the treatment did not appear to work and, if anything, actually negatively impacted outcome variables. This could be the case for several reasons. People rely on others to derive meaning from political messaging, consistent with classic work on social influence in mediating the relationship between mass media and political behaviors (e.g., Lazarsfeld et al., 1944). So, it is possible that exposure alone is not enough for individuals to derive meaning from information, and it could even overwhelm them leading to a decreased motivation to engage with the materials. Additionally, the sample was majority female. Longitudinal studies have found women to be generally less expressive of their political attitudes than men (Atkeson & Rapoport, 2003), which could explain some of the negative experimental effects on discussion frequency. The ceiling effect observed in the cumulative 25 campaign exposure measure supports this theory and raises the potential of information overload. Information overload concerns the negative consequences that can arise when an individual consumes high volumes of information (Jacoby, 1984). Studies suggest that individuals protect themselves from becoming “overloaded,” and in doing so may pay less attention to otherwise relevant information or even become irritated toward the information source (Jacoby, 1984; Micheaux, 2011). Indeed, with the high volume of pro-voting message seen in the lead up to the 2020 election, individuals may have felt their freedom to choose whether or not to vote was being threatened. This, in turn, may have made participants reactant (Rains & Turner, 2007). An interesting future direction could be evaluating cognitions in addition to affect to glean insight into any reactance from information overload and how that interacts with the overarching model. Regardless, this article provides insight into the processes occurring between GOTV message exposure and voter turnout and offers a better understanding of the mechanisms at play. Future research should utilize stronger GOTV manipulations, and even encourage interpersonal discussion within campaign materials to see if this increases the effectiveness of this strategy. Insofar that campaign messages are successful at promoting discussion, we might expect to see stronger effects of such campaigns on voter turnout out. So, let’s not just get people out to vote – let’s get them out talking about the vote. 26 APPENDICIES 27 APPENDIX A: Figures & Tables 28 Figure 4. Sample Breakdown from Student Census to Final Analysis Table 1. Demographic Frequencies Gender Male Female Non-binary Other Race Black/African American American India/Alaska Native Hispanic/Latinx Asian-American White Middle Eastern/North African Mixed Race Other Political Identity Very liberal Liberal Somewhat liberal Frequency 314 657 11 8 753 42 8 30 97 19 33 7 175 299 151 29 % 31.7% 66.4% 1.1% 0.8% 76.1% 4.2% 0.8% 3.0% 9.8% 1.9% 3.3% 0.7% 17.5% 29.9% 15.1% Cumulative % 31.7% 98.1% 99.2% 100.0% 76.1% 80.4% 81.2% 84.2% 94.0% 96.0% 99.3% 100% 17.5% 47.4% 62.6% Table 1 (cont’d) Moderate/Middle of the road Somewhat conservative Conservative Very conservative Political Affiliation Strong democrat Democrat Lean toward Democrat Independent Lean toward Republicans Republican Strong Republican Family Income Less than $20,000 $20,000-$34,999 $35,000-$49,999 $50,000-$74,999 $75,000-$99,999 $100,000-$149,999 $150,000-$199,999 $200,000-$249,999 More than $250,000 Note: N = 1122 15.6% 10.1% 9.0% 2.7% 17.1% 26.9% 18.9% 12.7% 11.5% 9.1% 3.7% 6.5% 6.7% 6.2% 13.8% 13.2% 23.1% 11.6% 9.3% 9.6% 78.2% 88.3% 97.3% 100% 17.1% 44.0% 62.9% 75.7% 87.2% 96.3% 100% 6.5% 13.2% 19.4% 33.2% 46.4% 69.5% 81.1% 90.4% 100% 156 101 90 27 171 268 189 127 115 91 37 63 65 60 134 128 224 113 90 93 Table 2. Summary Statistics of Outcome Variables Outcome Variable Turnout by treatment Control Information Implementation plan Total Cumulative campaign exposure Yes % Total 557 171 337 49.6% 15.2% 30.0% No 18 18 16 1065 94.9% Mean 2.93 SD 0.27 52 Min 1 30 % Total 1.60% 1.60% 1.42% 4.63% Max 3 Table 2 (cont’d) Discussion frequency Partner closeness Agreement/disagreement Attitude toward voting 3.89 2.80 3.92 6.50 0.97 0.44 1.45 0.97 1 1 1 1 5 3 5 7 Table 3. PROCESS Analysis Summary for Voter Turnout with Experimental Campaign Exposure Predictor (direct effects) Experimental campaign exposure Information Treatment Implementation Plan Treatment Discussion frequency Voting attitude Mediators (indirect effects) Discussion frequency Attitude Discussion frequency and attitude b -0.525 -0.993 -0.993 0.602 0.571 b -0.072 -0.085 -0.011 p 0.177 0.027 0.728 0.001 0.000 -0.181 -0.167 -0.027 95% CI (LL) 95% CI (UL) Moderators of mediation (Discussion & attitude) b 95% CI (LL) 95% CI (UL) 0.003 -0.019 0.000 0.004 0.009 0.015 0.016 0.074 0.118 0.188 0.263 b 95% CI (LL) 95% CI (UL) -0.006 -0.024 -0.028 -0.043 -0.017 -0.080 -0.125 -1.426 -0.001 -0.004 -0.002 -0.007 -0.017 -0.004 0.024 0.007 31 Partner agreement Partner closeness Conversational valence Negative Positive Moderators of mediation (Discussion only) Partner agreement Partner closeness Conversational valence Negative Positive Table 4. PROCESS Analysis Summary for Voter Turnout with Cumulative Campaign Exposure p 0.191 0.001 0.000 95% CI (LL) 95% CI (UL) 95% CI (LL) 95% CI (UL) 0.043 -0.028 0.010 -0.009 -0.020 -0.056 -0.035 0.482 0.278 0.076 0.018 0.068 0.106 0.165 95% CI (LL) 95% CI (UL) -0.208 0.021 -0.275 -0.892 -1.025 0.263 0.416 6.671 Predictor (direct effects) Cumulative campaign exposure Discussion frequency Voting attitude Mediators (indirect effects) Discussion frequency Attitude Discussion frequency and attitude b 0.619 0.559 0.571 b 0.219 0.110 0.036 Moderators of mediation (Discussion & attitude) Partner agreement Partner closeness Conversational valence Negative Positive Moderators of mediation (Discussion only) Partner agreement Partner closeness Conversational valence Negative Positive b 0.003 0.019 0.012 0.043 b -0.070 -0.006 -0.169 -0.041 32 33 APPENDIX B: Treatment Materials 34 Figure 5. Information Treatment Letter Note. Cover letter only. Magnet images not provided due to identifying information (i.e., school logo). 35 Figure 6. Implementation Plan Treatment Letter Note. Cover letter only. Magnet images not provided due to identifying information (i.e., school logo). 36 Figure 7. Implementation Plan Treatment Cue to Action Note. Plan treatment reminder email. 37 APPENDIX C: Codebook 38 Table 6. Codebook for Comments 39 REFERENCES 40 REFERENCES shapes the message. Journal of Consumer Research, 40, 567–579. https://doi.org/10.1086/671345 Examining gender differences in political attitude expression, 1952-2000. The Public Opinion Quarterly, 67, 495–521. social media era: Are digital tools changing the extent, nature and impact of party contacting in elections? Party Politics, 22, 165–178. https://doi.org/10.1177/1354068815605304 Aldrich, J. H., Gibson, R. K., Cantijoch, M., & Konitzer, T. (2016). Getting out the vote in the Atkeson, L., & Rapoport, R. (2003). The more things change the more they stay the same : Bergan, D. E., Carnahan, D., Lajevardi, N., Medeiros, M., Reckhow, S., & Thorson, K. (2021). Promoting the youth vote: The role of informational cues and social pressure. Political Behavior. https://doi.org/10.1007/s11109-021-09686-x Berger, J., & Iyengar, R. (2013). Communication channels and word of mouth: How the medium Bode, L., Edgerly, S., Wells, C., Gabay, I., Franklin, C., Friedland, L., & Shah, D. V. (2018). Participation in contentious politics: Rethinking the roles of news, social media, and conversation amid divisiveness. Journal of Information Technology and Politics, 15, 215–229. https://doi.org/10.1080/19331681.2018.1485607 Burleson, B. R. (2010). The nature of interpersonal communication: A message-centered Cappella, J. N. (1987). Interpersonal communication: Definition and fundamental questions. In Cardy, E. A. (2005). An experimental field study of the GOTV and persuasion effects of partisan Chaffee, S. H., & Mutz, D. C. (1988). Comparing mediated and interpersonal communication Cho, J. (2005). Media, interpersonal discussion, and electoral choice. Communication approach. In C.R. Berger, M. E. Roloff, & D. R. Roskos-Ewoldsen (Eds.), The Handbook of Communication Science (pp. 145–160). SAGE Publications. https://doi.org/10.4135/9781412982818 C. R. Berger & S. H. Chaffee (Eds.), Handbook of communication science (pp. 184-238). SAGE Publications. direct mail and phone calls. Annals of the American Academy of Political and Social Science, 601, 28–40. https://doi.org/10.1177/0002716205278051 data. In R. P. Hawkins, J. M. Wiemann, & S. Pingree (Eds.), Advancing communication science: Merging mass and interpersonal processes (pp. 19–43). Research, 32, 295–322. https://doi.org/10.1177/0093650205275382 41 of pledging to vote on youth turnout. PLoS ONE, 13, 1–16. https://doi.org/10.1371/journal.pone.0197066 conversations about health promoting media messages. Communication Monographs, 77(4), 518–539. https://doi.org/10.1080/03637751.2010.502537 mobilization on the composition of the electorate. Journal of Politics, 76, 273–288. https://doi.org/10.1017/S0022381613001308 Campaigns, reflection, and deliberation: Advancing an O-S-R-O-R model of communication effects. Communication Theory, 19, 66–88. https://doi.org/10.1111/j.1468-2885.2008.01333.x Cho, J., Shah, D. V., McLeod, J. M., McLeod, D. M., Scholl, R. M., & Gotlieb, M. R. (2009). Costa, M., Schaffner, B. F., & Prevost, A. (2018). Walking the walk? Experiments on the effect Dale, A., & Strauss, A. (2009). Don’t forget to vote: Text message reminders as a mobilization tool. American Journal of Political Science, 53, 787–804. https://doi.org/10.1111/j.1540 5907.2009.00401.x Dunlop, S. M., Kashima, Y., & Wakefield, M. (2010). Predictors and consequences of Enos, R. D., Fowler, A., & Vavreck, L. (2014). Increasing inequality: The effect of gotv Frank, L. B., Chatterjee, J. S., Chaudhuri, S. T., Lapsansky, C., Bhanot, A., & Murphy, S. T. Gagliano, M. (1988). A literature review on the efficacy of video in patient education. Journal of Gil de Zúñiga, H., Valenzuela, S., & Weeks, B. E. (2016). Motivations for political discussion: Antecedents and consequences on civic engagement. Human Communication Research, 42, 533–552. https://doi.org/10.1111/hcre.12086 Gollwitzer, P. M. (1999). Implementation intentions. American Psychologist, 54, 493–503. Gerber, A. S., & Green, D. P. (2000). The effects of canvassing on voter turnout. American Green, D. P., & Gerber, A. S. (2019). Get out the vote: How to increase voter turnout (4th ed.) Haenschen, K. (2016). Social pressure on social media: Using Facebook status updates to (2012). Conversation and compliance: Role of interpersonal discussion and social norms in public communication campaigns. Journal of Health Communication, 17, 1050 1067. https://doi.org/10.1080/10810730.2012.665426 https://doi.org/10.1109/IAdCC.2014.6779372 Political Science Review, 94, 653–663. DOI Brookings Institution Press. https://doi.org/10.5860/choice.194933 increase voter turnout. Journal of Communication, 66, 542–563. https://doi.org/10.1111/jcom.12236 Medical Education, 63, 785–792. DOI: 10.1097/00001888-198810000-00006 42 communication as a mediator of antitobacco campaign effects. Communication Research, 39, 120–141. https://doi.org/10.1177/0093650210389029 432–435. https://www.jstor.org/stable/2488912 campaigns: The case of the national youth anti-drug media campaign. Communication Theory, 13, 204–224. https://doi.org/10.1111/j.1468-2885.2003.tb00289.x influence of emotions on conversational valence and alcohol consumption. Prevention Science, 15, 684–693. https://doi.org/10.1007/s11121-013-0418-2 (1997). Provocative appeals in anti-smoking mass media campaigns targeting adolescents - The accumulated effect of multiple exposures. Health Education Research, 12, 227 236. https://doi.org/10.1093/her/12.2.227 Hafstad, A., Aarø, L. E., Engeland, A., Andersen, A., Langmark, F., & Stray-Pedersen, B. Hendriks, H., van den Putte, B., & de Bruijn, G. J. (2014). Changing the conversation: The Hornik, R., & Yanovitzky, I. (2003). Using theory to design evaluations of communication Hwang, Y. (2012). Social diffusion of campaign effects: Campaign-generated interpersonal Jacoby, J. (1984). Perspectives on Information Overload. Journal of Consumer Research, 10, Jeong, M., & Bae, R. E. (2018). The effect of campaign-generated interpersonal communication Jones, J. J., Bond, R. M., Bakshy, E., Eckles, D., & Fowler, J. H. (2017). Social influence and Katz, E., Lazarsfeld, P.F. (1955). Personal Influence: The part played by people in the flow of Kawashima-Ginsberg, K., & Kiesa, A. (2019). Getting young people to vote: Seven tips for the Kim, M.-S., & Hunter, J. (1993). Attitude-behavior relations: A meta-analysis of attitudinal political mobilization: Further evidence from a randomized experiment in the 2012 U.S. presidential election. PLoS ONE, 12, 1–9. https://doi.org/10.1371/journal.pone.0173851 on campaign-targeted health outcomes: A Meta-Analysis. Health Communication, 33, 988–1003. https://doi.org/10.1080/10410236.2017.1331184 classroom. Social Education, 83, 194–199. https://www.socialstudies.org/system/files/publications/articles/se_8304194_0.pdf mass communications. New York, NY: The Free Press. relevance and topic. Journal of Communication, 43, 101–142. https://doi.org/10.1111/j.1460-2466.1993.tb01251.x Lassen, D. D. (2005). The effect of information on voter turnout: Evidence from a natural experiment. Political Science, 49, 103–118 . https://doi.org/10.1111/j.00925853.2005.00113.x Lazarsfeld, P. F. (1944). The election is over. The Public Opinion Quarterly, 8(3), 317–330. https://www.jstor.org/stable/2745288 43 underpinnings of learning and involvement during presidential campaigns. The American Political Science Review, 87, 672–685. https://doi.org/10.2307/29 Journal of Advertising, 40, 45–65. https://doi.org/10.2753/JOA0091-3367400404 work even in battleground states? Evaluating the effectiveness of MoveOn’s 2004 outreach campaign. Quarterly Journal of Political Science, 3, 63–82. https://doi.org/10.1561/100.00007019 Schluchter, M. D. (2016). Randomized trial of a web-based intervention to address barriers to clinical trials. Journal of Clinical Oncology, 34, 469–478. https://doi.org/10.1200/JCO.2015.63.2257 participation: The role of mass media and interpersonal discussion in local political participation. Political Communication, 16, 315–336. https://doi.org/10.1080/105846099198659 Marcus, G. E., & Mackuen, M. B. (1993). Anxiety, enthusiasm, and the vote: The emotional McLeod, J. M., Scheufele, D. A., & Moy, P. (1999). Community, communication, and Meropol, N. J., Wong, Y. N., Albrecht, T., Manne, S., Miller, S. M., Flamm, A. L., … Micheaux, A. (2011). Managing e-mail advertising frequency from the consumer perspective. Middleton, J. A., & Green, D. P. (2008). Do community-based voter mobilization campaigns Mutz, D. (2002). The consequences of cross-cutting networks for political participation. American Journal of Political Science, 46, 838. https://doi.org/10.2307/3088437 Mutz, D. (2006). Hearing the other side: Deliberative versus participatory democracy. Nickerson, D.W. (2008). Does Email Boost Turnout? Quarterly Journal of Political Science, 2, Nickerson, D. W., & Rogers, T. (2016). Do you have a voting plan? Implementation intention, Osgood, C.E., Suci, G.J., & Tannenbaum, P.H. (1957). The measurements of meaning. Urbana: Pattie, C., & Johnston, R. (2000). “People who talk together vote together”: An exploration of contextual effects in Great Britain. Annals of the Association of American Geographers, 90, 41–66. https://doi.org/10.1111/0004-5608.00183 Rains, S. A., & Turner, M. M. (2007). Psychological reactance and persuasive health Rogers, E. M. (1983). Diffusion of innovations (3rd ed.). New York, NY: The Free Press. Cambridge, MA: Cambridge University Press. 369–79. http://dx.doi.org/10.1561/100.00007032 voter turnout, and organic plan making. 21, 194–199. 10.1177/0956797609359326 University of Illinois Press. communication: A test and extension of the intertwined model. Human Communication Research, 33, 241–269. https://doi.org/10.1111/j.1468-2958.2007.00298.x 44 media campaigns. Annals of the International Communication Association, 31, 420–462. https://doi.org/10.1080/23808985.2007.11679072 campaigns. Communication Theory, 19, 1–8. https://doi.org/10.1111/j.1468 2885.2008.01329.x Rohrbach, L. A. (2006). Intermedia processes in the adoption of tobacco control activities among opinion leaders in California. Communication Theory, 16, 91–117. https://doi.org/10.1111/j.1468-2885.2006.00007.x Campaign ads, online messaging, and participation: Extending the communication mediation model. Journal of Communication, 57, 676–703. https://doi.org/10.1111/j.1460-2466.2007.00363.x Schuster, D. V., Valente, T. W., Skara, S. N., Wenten, M. R., Unger, J. B., Cruz, T. B., & Shah, D. V., Cho, J., Nah, S., Gotlieb, M. R., Hwang, H., Lee, N. J., … Mcleod, D. M. (2007). Solovei, A., & Van Den Putte, B. (2020). The effects of five public information campaigns: The Southwell, B. G., & Yzer, M. C. (2007). The roles of interpersonal communication in mass Southwell, B. G., & Yzer, M. C. (2009). When (and why) interpersonal talk matters for role of interpersonal communication. Communications, 45, 586–602. https://doi.org/10.1515/commun-2020-2089 45