Abstract

Since Citizens United, elections have witnessed an increasing number of candidate-focused advocacy ads (that are not sponsored by candidates). The literature addressing the effects of the increase in advocacy ads has largely focused on their financial dominance, leaving unclear the potential impact that such advocacy ads have on the decision-making processes of voters. In this research, we employ the framework of attitudinal ambivalence to explore experimentally the psychological influences of campaign ads by comparing the impact of candidate-sponsored and advocacy ads on viewers. Based on two studies, we find evidence that viewers perceive and evaluate campaign-related advocacy ads in much the same way as candidate-sponsored ads. However, we find evidence that PAC-sponsored ads generate higher levels of ambivalence as compared to candidate ads, which renders the measured link between attitudes and voting unpredictable.

Keywords: Ambivalence, Felt Ambivalence, Latent Ambivalence, Attitudes, Voting, Candidate Ads, Advocacy Ads, Persuasion.

Introduction

For the past five decades, United States campaign finance laws—beginning with the Federal Election Campaign Act (FECA 1971) and including the Bipartisan Campaign Reform Act (BCRA 2002) have focused on three things: limits on contributions, limits on expenditures, and disclosure of contributors. These laws have been counterbalanced by court decisions reaffirming the primacy of freedom of expression, particularly for those unaffiliated with campaigns. Since Citizen’s United (2010), those interested in the outcome of elections have increasingly donated to outside groups that avoid the restrictions imposed on candidates and traditional Political Action Committees (PACs). As a result, election spending has skyrocketed since 2010, fueled largely by outside organizations. For example, in presidential elections, outside spending increased from $142.3 million in 2008 to $2 billion in 2020 [1]. While in the decade before Citizen’s United, outside groups outspent candidates in only 15 races, in the subsequent decade that number increased eight fold to 126. Although the majority of justices in Citizen’s United acknowledged the government’s interest in disclosing the names of contributors to organizations involved in outside spending, the erosion of BCRA coincided with a decline in disclosures from 97% in 2004 to 40% in 2012 [2].

Despite increased spending on issue advocacy, little understanding exists regarding the differential persuasive effects of advocacy ads compared to candidate-sponsored ads. The literature exploring the information processing influence of issue advocacy ads on voters since Citizens United is scant, and the available work has focused largely on the magnitude of financial resources spent by Super PACs during campaigns [3,4]. For example, recent research on advocacy ads has focused on such areas of inquiry as the influence of tone, target, and issue ownership [5] and interest group spending and why the groups spend on television ads [6].

However, very little research examines the role of such ads on cognitive inconsistency, or ambivalence, in a voter’s mind. Voters experience ambivalence when they hear contradictory evidence about candidates—an experience that increases as they are exposed to more ads sponsored by outside groups. In contrast to the bulk of research addressing advocacy ads, our research focuses, instead, on the psychological influence that both candidate-sponsored and advocacy-sponsored ads have on a voter’s experienced ambivalence [7], and, through it, on intended voting behavior.

The literature has well established that political ads persuade voters during elections [8-10]. Both candidates and advocacy organizations employ ads to disseminate persuasive messages to potential voters [11]. The communication strategies employed in advocacy ads are similar to those in traditional candidate-sponsored ads [4]. Additionally, candidate ads, advocacy ads promote both favorable and unfavorable messages about candidates, their rivals, and political issues.

Advocacy ads are much more likely to be negative than candidate sponsored ads. Prior literature has examined and demonstrated the impact of positive and negative ads during elections on voter attitude [10-17]. By influencing voter attitude, the message create positive feelings [18], thoughts [19], and attitudes [20,21]. Voters dislike negative ads and tend to hold candidates responsible for sponsoring them. Potentially, this backlash might not occur if voters understand that advocacy ads are sponsored by outside groups rather than candidates [22,23].

Advocacy ads tend to dominate electoral information sources, potentially having a greater influence on voter persuasion [24-26]. Election related ads are required to identify their sponsors (candidate names for campaigns and organization names for PACs). While the objective of both is to persuade voters to go to the polls and vote for the supported candidate [9], advocacy ads must omit “vote for” and other electioneering terminology. Given the similarity between the objectives and delivery methods of the ads, empirical evidence examining the differential effects of advocacy ads on voters’ information processing is limited. Controlling for party affiliation, Siev and Petty [27] found that ambivalent attitudes promote support for extreme political violence in campaigns. We did not explore this topic, but it certainly points to the importance of understanding voter ambivalence, and these studies aim to do precisely that.

Our research seeks to contribute to the literature by exploring whether voters process ads differentially depending on their sponsor: candidate or interest group. Because of prior literature in social psychology [28-30], we are particularly interested in the role of ambivalence in voter decision-making. Attitudinal ambivalence refers to a mental state where a voter simultaneously accesses both positive and negative reactions toward a given candidate [28,29,31,32,33]. We use Priester and Petty’s [28] notion of ambivalence as a cognitive response to being faced simultaneously with both positive and negative stimuli (ad valence) regarding a subject (for us, a candidate). We explore whether voters process the two ad types differently and whether such differences impact voter attitudes and behaviors, specifically through the mechanism of attitudinal ambivalence.

Although partisanship and candidate knowledge have traditionally been identified as the major drivers guiding voter decisions, Lavine [34] argued that during elections, attitudinal ambivalence toward candidates tends to be more important and influential. For example, an individual could hold both positive and negative thoughts about a given candidate: A voter may like the candidate’s social policy positions and dislike their fiscal policy positions. If both these opposite-valence reactions become simultaneously accessible during the decision-making process, the individual will experience a cognitive inconsistency leading to feelings of tension or discomfort. Confronted with this conflict, the individual seeks to either reduce or avoid the discomfort [35]. Hence, understanding the effects of ambivalence would generate a better understanding of voter behavior. Both advocacy ads and candidate-sponsored ads deliver both positive and negative messages during elections, but no research currently explores the combined effects of valence and type of ads on attitudinal ambivalence in the political domain. Hence, understanding the effects of ambivalence would generate a better understanding of voter behavior.

Elections are breeding grounds for ambivalence [27]. Prior literature has demonstrated that increased ambivalence tends to reduce voting and political participation [36]. Interestingly, the current advertising pattern may lead to a situation where voters may face increased ambivalence. Essentially, campaigns deploy ads favorable toward their candidates while, simultaneously, competing candidates and advocacy groups deploy negative ads. Because advocacy ads are not sponsored by the candidates, the candidates are buffered from voter backlash against negative ads [23]. Voters try to deal with these conflicting messages using compensatory processes by weighing both positive and negative information to derive an overall evaluation. For example, according to the multi-attribute model, if an individual has three positive and two negative attitudes, the person might compensate for the two negatives with the positives [20]. The formed overall evaluation will then guide subsequent decisions [28,29,32].

This balancing suggests that voter preference might be weakened by the presence of conflicting messages. In order to strengthen the preference, campaigns and other organizations increase the frequency of ads, intending to create univalent attitudes toward candidates. For example, Shaw [37] examined the presidential elections of 1988, 1992, and 1996, and found that an increase in political advertising led to an increased vote share for the sponsoring candidate. The advertising strategy assumes that increasing the frequency of positive ads will lead to a higher number of positive thoughts to compensate for contrasting stimuli and lead to an overall positive evaluation of a candidate, thereby winning the voter’s favor.

Since both sides seek to increase their advertising effectiveness, the conflicting ads can prompt voter ambivalence. Within a given election cycle, many voters remain ambivalent and are unable to vote in line with their evaluations of candidates. After studying the data on House elections between 1990 to 2000, Basinger and Lavine [38] reported that 30% of voters were ambivalent toward their preferred party, and, more critically, voted opposite to what the overall measured evaluations would have predicted. Furthermore, the ambivalence literature suggests that increased levels of ambivalence toward candidates leads to increased instability in voters’ evaluations of candidates [34,39].

In the context of elections, ambivalence has higher significance because of consequences to individual voters and political campaigns. From the perspective of voters, ambivalence creates discomfort due to the increased salience of the cognitive inconsistency between the positive and negative reactions toward the same candidate or party [28,30,40,41]. People dislike inconsistency as it creates feelings of discomfort [30]. Hence, feelings of ambivalence are undesirable for voters because they create feelings of discomfort and are highly aversive [30]. From a political campaign’s perspective, increased ambivalence toward a candidate is detrimental to their campaign because it creates instability in evaluations [34,39]. Ambivalence negatively correlates with attitudes [28,29] and weakens the ability of attitudes to guide behavior. Finally, the process interrupts the impact of attitudes on voting behavior [30,36,42,43].

In light of the preceding literature, we ask three research questions: RQ1: Does watching advocacy ads prompt differential information processing?

RQ2: Does watching an advocacy ad generate increased ambivalence toward candidates?

RQ3: Does the ambivalence literature apply to voters’ information processing in guiding their electoral choice?

As discussed previously, advocacy ads provide indirect information about elections in contrast to the direct electioneering of candidate (“vote for me”) ads. Prior literature suggests that the structure of incoming information guides its processing [44]. The structure of advocacy ads tends to generate an issue-based cognitive framework to facilitate learning and help with intra-dimensional comparison of candidates [11]. Because advocacy ads should focus on issues, voters could perceive them differently from candidate ads. The literature suggests that advocacy ads focus primarily on issues [3]. Their use of evidence makes them more credible, trustworthy, and persuasive [11,25]. Because most of this literature predates Citizens United, we predict a different outcome in our studies, which will be a contribution to the extant literature.

The electoral motivation of candidate-sponsored ads is clear, obvious, and usually self-serving for the candidate [45]; however, the motivation is less obvious for advocacy ads. Because candidate sponsored ads have a clearer focus than advocacy ads, we predict that voters process advocacy ads more thoroughly, given their higher credibility and relevance. Thus, we propose that the valence of the advertisement will have an impact on evaluations of and preferences for candidates. Watching positive ads will increase positive attitudes about the candidate, thereby increasing voter preference for the candidate.

We expect that ambivalence enters into the evaluation process by way of expectation confirmation theory. The literature suggests that when we are exposed to information that contradicts our expectations, we tend to be ambivalent about the subject of that information. In general, in the face of purely positive information, consumers feel ambivalent because they keep wondering what negative information might be missing [46].

Employing the expectation-confirmation paradigm, we anticipate two ways in which advocacy ads will differentially influence voter ambivalence compared to candidate ads. Although issue ads could be either positive or negative [47], usually during elections they are negative. Hence, viewers will expect interest groups to sponsor negative issue ads. Thus, negative issues ads will confirm expectations leading to lower levels of ambivalence [46]. Hence, we propose that consuming an unfavorable issue advocacy ad toward a candidate will lead to lower experienced ambivalence as compared to negative candidate ads.

Given that advocacy groups tend to emphasize dissenting positions between themselves and candidates, viewers are unlikely to expect positive issue ads. In this situation, consuming a positive issue ad will lead to expectancy-disconfirmation. Hence, we propose that watching an issue advocacy ad favoring a candidate would lead to increased levels of experienced ambivalence as compared to a candidate ad favoring a candidate.

H₁: Ad valence stimulates voter emotion (positive or negative), which subsequently influences voter preference for a candidate.

H₂: Positive advocacy ads provoke higher levels of ambivalence than candidate-sponsored ads.

H₃: Regardless of gender, ambivalence interrupts the link between preference and the behavior of voting.

Based on these predictions, we pursue two objectives using two randomized experiments. First, we investigate the differential effect of advocacy versus candidate ads. Study 1 analyzes the differential persuasion path for these two ad types. Then, in Study 2, we rule out gender of candidate as a confounding factor. The methodological details of both studies, along with their results, follow.

Materials and Methods

We conducted two studies using essentially the same materials and approaches. The major difference between Studies 1 and 2 is that Study 1 compared candidate ads and issues-advocacy ads. Study 2 looked at advocacy ads and the impact of gender on intended voter behavior.

Study 1

We used an experimental research design to examine whether individuals process advocacy ads differentially than candidate sponsored ads. We used an online Qualtrics sample (N=532) representative of the United States’ voting population based on gender, race, and ethnicity. This sampling assured strong external validity, while random assignment to experimental conditions helped demonstrate causality.

The study employed a 2 (valence: positive vs. negative) X 2 (ad type: candidate vs. advocacy) between-subjects design. As shown in the Appendix, we employed ad valence as a between-subjects factor. That is, each participant was randomly assigned to either a positive valence condition or a negative valence cell. For the stimulus material, we selected video ads from past congressional elections (see Appendix Table 11 for candidate names). Most of these videos are no longer available, but Table 13 in the Appendix contains a hyperlink to the positive advocacy ad for Katie Hill and the negative advocacy ad against Catherine Cortez Masto as examples these kinds of ads.

We assigned participants randomly to either a positive or a negative ad valence block. Within each ad block, participants received two stimuli: a candidate ad and an issue ad. For each stimuli, participants first viewed a picture of the candidate and reported their likelihood of voting for the candidate (Candidate Evaluation) using the scale from 0 (not at all likely) to 100 (highly likely). Next, participants viewed a video ad after which they again rated their likelihood of voting for the candidate on the same scale. Finally, participants responded to a series of questions measuring latent ambivalence, felt ambivalence, homophily, attitude towards the ad, and portrayal of the ad (see Table 12 in the Appendix for items in the Ad Attitude variable). The ambivalence literature distinguishes between an attitude and the resulting behavior when it argues that ambivalence weakens the link between attitude and behavior. This distinction parallels the voting behavior literature that separates two different decisions made by voters: Voters decide both which candidate they prefer and also whether they will actually go to the polls and vote.

We measured the impact of the ads on voting behavior by subtracting the first Candidate Evaluation from the second. Complicating the decision process is the participant’s evaluation of the ad. We measured Ad Evaluation with a 100-point feeling thermometer. To model the decision-making process fully, we needed to model all three aspects: candidate evaluation, ad evaluation, and the resulting behavioral change (in voting likelihood).

Measures

Change in Voting Likelihood. As described above, for this variable we subtracted the pre-viewing voting likelihood from the post-viewing score. Hence, a positive difference indicates that the likelihood of voting for the candidate increased after watching the ad. We model this change as a function of Candidate Evaluation and Ad Evaluation.

Candidate Evaluation. This is a standard 100-point feeling thermometer scale measuring warmth toward the candidate depicted in the ad. We modelled Candidate Evaluation with two major variables. First, Homophily: We expect viewers to evaluate a candidate more warmly when they perceive themselves to be similar to the candidate. We adapted the Homophily measure from McCroskey, et al. [48] and Ahif et al. [49]. Participants rated whether they perceived the candidate to be similar to them in general, as well as in thinking, behavior, social class, and culture using scales anchored from 1 (strongly disagree) to 5 (strongly agree). We calculated a Homophily index by taking the average of scores on the five items. Higher values indicate greater homophily toward the candidate. Second, Portrayal of Candidate: When an ad portrays a candidate more positively, we expect the viewer to evaluate the candidate more warmly. Participants rated the ad’s portrayal of the candidate as positive or negative, using a scale of 1 (very negative) to 5 (very positive).

Candidate Evaluation =Homophily Index+ Portrayal of Candidate+Pos.Emotion+Neg. Emotion

Ad Evaluation. Like candidate evaluation, this is a standard 100-point feeling thermometer scale measuring warmth toward the ad. We modeled ad evaluation as a function of three different variables. First, Attitude toward Ad: We expect that positive attitudes toward the ad (in contrast to negative attitudes) will result in a more positive evaluation of the ad. We measured the attitudes toward the ad using six seven-point semantic differential scales anchored by 1 (negative attitudes) to 7 (positive attitudes) as detailed in Table 12 of the Appendix. We averaged the scores on these items to derive an overall score of attitude toward the ad. Second, the variable for Evidence: We expected that ads using more evidence would be evaluated more positively. In a series of dummy variables, we asked participants whether the ad included different kinds of evidence, comprising details, numbers, news stories, or other sources of information. We calculated the evidence variable by totaling the number of types of information used in the ad, for a possible score of 0-4. Third, perceived Interest Group Sponsor. We expected that viewers would evaluate ads more negatively when they thought the sponsor was an interest group. We measured this variable with an item that asked, “How likely is it that the ad was paid for by an interest group?” with possible responses ranging from 1 (Strongly Disagree) to 5 (Strongly Agree).

Ad Eval.= Attitude toward Ad+Evidence+Interest Group Sponsor+Pos.Emotion - Neg.Emotion

Emotions¹. Complicating these relationships are the emotions produced while watching the ad. We measured emotional responses to the ad on a scale anchored from 1 (strongly disagree) to 5 (strongly agree). Positive Emotions included happy, pleased, excited, enthusiastic, and satisfied; while Negative Emotions included resentful, depressed, irritated, angry, worried, uncomfortable, and nervous. We averaged the positive and negative emotional responses separately to yield two variables (with scores ranging 1-5). Because these can influence all aspects of the decision-making process, we included them in all three models (change in voting likelihood, candidate evaluation and ad evaluation).

Ambivalence. The literature suggests that ambivalence breaks the normal link between attitude (for us candidate evaluation) and behavior (for us change in voting likelihood) in a pathway beginning with latent ambivalence and leading through felt ambivalence. We measured latent ambivalence toward each candidate using the formula from Zemborain and Johar [50]:

Latent Ambivalence= [(Positive+Negative)-2(abs|Positive-Negative|)+1]

In this equation, P is participant placement of how positive their thoughts are toward the candidate from 1 to 4 (not at all positive to extremely positive) and N is how negative their thoughts are from 1 to 4 (not at all negative to extremely negative). The calculated measure of latent ambivalence ranges from 1 to 9 with 9 indicating the highest level of latent ambivalence.

We measured felt ambivalence with the Priester and Petty [28] scale. In response to three items, participants indicated how conflicted, mixed, and indecisive their reactions were toward the candidate on a scale ranging from zero (feel no conflict, feel no indecision, one-sided reaction) to 10 (feel maximum conflict, feel maximum indecision, feel completely mixed reactions). We calculated the felt ambivalence index by computing the average scores for the three items.

Study 2

Given prior literature on stereotypes suggesting that voters perceive female candidates differently than males [51,52], in Study 2 we explore whether the type of ad or the gender of the candidate leads to the observed difference in felt ambivalence, as is suggested by Bauer [53]. We employed both male and female candidates for both valences in this study of advocacy ads.

Similar to Study 1, Study 2 included a representative sample of 512 participants from a Qualtrics panel. The study employed a 2 (valence: positive vs. negative) X 2 (gender: male vs. female) mixed-subjects design. As shown in Table 13 of the Appendix, we employed ad valence as a between-subjects factor, with the gender of the candidate as a within-subjects factor. Using random assignment for each valence, participants responded to two stimuli, one male and one female candidate. We randomized the order of presentation of ads (male and female candidates) to prevent gender order effects.

As in Study 1 we selected four advocacy video ads from past congressional elections for the study, with the candidates identified in the Appendix. Participants viewed a picture of the candidate, responded to the voting likelihood measure, watched an ad for the same candidate and finally responded to a series of items including the voting likelihood measure for a second time. In the second condition, all participants followed the same sequence as in the first, but with a candidate of the opposite sex.

For this study, we focused on the final component of the decision making process—the effect of the ad on voting likelihood. We used the measures from Study 1 of voting likelihood, candidate evaluation, ad evaluation, positive emotion, negative emotion, and felt ambivalence.

Results

Study 1: Advocacy Ads and Intention to Vote

We had two major goals. First, we sought to develop a model of voter decision making to evaluate whether voters process advocacy ads differently than candidate-sponsored ads. Second, we sought to determine whether advocacy ads increase voter ambivalence. This is a concern because our review of the literature suggests that such ambivalence could decrease the likelihood that voters will follow through on their candidate evaluation, thus changing their behavioral intention to vote for the candidate.

The first step of the analysis was to examine whether the type of ad (candidate or advocacy) influenced viewer perceptions of it. For parsimony, we analyzed all the measures separately for both positive and negative ads. Unless specified, we analyzed all results by employing a one-way ANOVA using the type of ad as the independent variable. Two separate analyses, presented in Table 1, control for ad valence (positive vs. negative). Regardless of ad type, valence will influence viewers’ responses to the ads by either increasing or decreasing viewer favorability toward the featured candidate. As shown in Table 1, the means for positive ads are higher than those for negative ads. Positive ads increased the likelihood of voting for the portrayed candidate (supporting H1), while negative ads decreased the likelihood (supporting H2). These results support our decision to address valence when analyzing ads.

The question we began to answer with these data is whether viewers perceived advocacy ads differently than candidate-sponsored ads. The general pattern in looking at the means is that candidate ads tended to produce a more positive response than advocacy ads. In comparison to advocacy ads, the means for candidate ads are higher for the main variables that influence voting likelihood: candidate evaluation, ad evaluation, and both positive and negative emotions. The one exception is that for negative ads, ad evaluation is higher for the advocacy ad than for the candidate ad. However, few of these differences are statistically significant and so we cannot reject the null hypothesis for most of them. Furthermore, the differences that are statistically significant have very small values of η² —explaining at most 2% of the variance in the variable. In general, these ANOVAs support previous results we have obtained in pilot studies, indicating that viewers do not perceive a difference between candidate sponsored and advocacy ads.

Given that viewers do not perceive significant differences between candidate and advocacy ads, the question remains of whether viewers process the two types of ads differently. To develop a fuller model of voter decision making, we first looked at the antecedents to the major components of the decision-making process: candidate evaluation and ad evaluation. Using OLS regression, Table 2 shows the antecedent model for candidate evaluation; and Table 3, for ad evaluation. Both tables indicate robust models, with the R2 of the first model indicating that it explains 44.9% of the variance of candidate evaluation while the second model explains 29.6% of the variance in ad evaluation. The magnitudes of the coefficients do not vary by ad type, suggesting that viewers develop these attitudes in similar ways, regardless of the ad’s sponsor.

Each model includes one variable that has a weak relationship with the dependent variable. For the model depicting candidate evaluation, the standardized coefficients for positive emotion indicates a weak relationship. For candidate ads, the relationship is weaker and not statistically significant. Similarly, for the model depicting ad evaluation, the standardized coefficients for interest group sponsor are weak and the relationship for advocacy ads is not statistically significant. With those two exceptions, all of the relationships are statistically significant. In addition, the magnitude of the coefficients is very similar between ad types. We conclude that ad type does not affect viewer processing of these factors in evaluating either the candidate or the ad.

The next step of the analysis was to analyze the direct effects on voting likelihood. Table 4 shows the regression for change in voting likelihood. Like the other models, the regressions indicate that this is a robust model of voter decision-making that does not vary by ad type. All of the variables are statistically significant, most at the 0.01 level. Across the ad types, the magnitude of the coefficients remains similar. The standardized coefficients indicate that candidate evaluation explains the most variance in voting likelihood.

Since none of the three models indicate that ad type has explanatory power, Figure 1 shows the full model of decision-making, undifferentiated by ad type. As shown in the model, the strongest causal path begins at candidate portrayal, which has a strong influence on candidate evaluation, which in turn has a strong influence on change in voting likelihood. Negative emotions and ad evaluation have a countervailing effect, decreasing the change in voting likelihood.

Our last question of Study 1 concerned the role of ambivalence in the process. Usually felt ambivalence is modeled as a function of latent ambivalence. But given the countervailing effects of ad evaluation vis-à-vis candidate evaluation, we chose to include it in the model as well. As shown in Table 5, this decision was appropriate. Although the relationship between latent and felt ambivalence is statistically significant, its coefficient indicates a moderate relationship at best. In contrast, ad evaluation had a much higher coefficient, indicating a strong relationship. In addition, the relationship was much stronger for advocacy ads (β=0.528) than for candidate ads (β=0.329). This gives credence to our concern that advocacy ads increase the ambivalence of voters.

With the increased risk of felt ambivalence for viewers of advocacy ads comes the potential for a break in the link between voter preference and the behavior of voting. The regressions in Table 6 added felt ambivalence to the model (from Table 4) predicting change in voting likelihood. The model for candidate ads remains largely unchanged from Table 4. The coefficients for the original variables maintain the same magnitude while the coefficient for felt ambivalence is close to zero. However, the model for advocacy ads changed markedly. Felt ambivalence now has a statistically significant negative relationship with change in voting likelihood: As advocacy ads increase viewer ambivalence, viewers were less likely to vote for their preferred candidate. The impact of this effect can be seen in the coefficient for candidate evaluation. Whereas in Table 4 the coefficient was 0.440, once we controlled for felt ambivalence, it increased to 0.521. Figure 2 depicts the role of ambivalence. Latent ambivalence combines with the effects of ad evaluation to spur felt ambivalence in viewers, which breaks the linkage between candidate evaluation and the resulting likelihood of voting.

Because regression does not address causal ordering, we chose to verify our model for the role that ambiguity plays in this process using mediation analysis. Because ambivalence operates differently for candidate and advocacy ads, we performed a mediation analysis separately for the two ad types. We conducted the serial mediation analysis with ad valence as the independent (X) variable (coded: -1 for negative valence and +1 for positive valence), and the change in voting likelihood as the dependent (Y) variable. Based on the literature, we employed ad evaluation, candidate evaluation, latent ambivalence, and felt ambivalence as the mediator variables (M).

For the candidate ads, ambivalence did not mediate between candidate evaluation and change in voting likelihood. The analysis produced the following pathway (Process Model 6; indirect effect = 0.1456, SE = .0663 [CI: 0.0412, 0.3000]):

X → AdEvaluation → CandidateEvaluation → Change in Voting Likelihood

In contrast, for the advocacy ads ambivalence mediated the path from X (ad valence) to change in voting likelihood (Process Model 6; indirect effect = -.0199, SE = .0121 [CI: -.0501, -.0033]):

X → AdEvaluation → CandidateEvaluation → Latent Ambivalence → Felt Ambivaldence → Change in Voting Likelihood

Thus, the mediation analysis supported the conclusion that the tension between the effects of ad evaluation and candidate evaluation leads to ambivalence regarding the candidate. However, only for advocacy ads does that ambivalence affect the link between preference and behavior, thereby decreasing the likelihood that the viewer will turn out and vote.

Study 1 adds to the literature in three ways. First, these experimental results are independent of party affiliation. Controlling for party affiliation by randomly assigning participants to ad conditions, we found a more parsimonious model of ad effects upon voter intention. Second, we found experimental evidence demonstrating that advocacy ads elicit a different persuasion process than do candidate ads. Individuals process information about the candidate and the ad in similar ways, regardless of ad type. In support of Hypothesis 1, ad valence stimulates voter emotion, either in the positive or negative direction, which subsequently influences the voter’s preference for the candidate correspondingly in either a positive or negative direction.

However, in translating their evaluation of the candidate into voting intention, a difference occurs. Viewers of advocacy ads, unlike viewers of candidate-sponsored ads, experience ambivalence in a way that breaks the link between perception of the candidate and intention to vote, as suggested by Hypothesis 2. Third, ambivalence plays an important role in individual voter behavior. Using aggregate data, prior literature on advocacy ads focused on their impact on political knowledge and found that ambivalence has a stronger impact on voting behavior than knowledge about the candidate [34]. Supporting this conclusion, but using experimental individual-level data, we add to the literature with our finding that the ambivalence engendered by advocacy ads led to a different process of evaluating the candidates. This finding has implications for understanding the impact of advocacy ads on the behavioral intention of voters. This partially supports Hypothesis 3.

In Study 1, we found evidence that participants who watched a positive advocacy ad experienced higher levels of felt ambivalence compared to those who watched a negative advocacy ad. A limitation of Study 1 was that it employed female candidates for ads with positive valence (both advocacy and candidate) and male candidates for both negative ads. A potential confound could be that the gender of the candidates, rather than the ad sponsor, drove the effects we observed in Study 1.

Study 2: Gender and Valence

In Study 1, we conflated gender and ad valence. Given prior literature on stereotypes suggesting that voters perceive female candidates differently than males [51,52,54] and that female candidates generate ambivalence among voters [53], it is worth exploring how gender influences the role of ambivalence regarding advocacy ads. In Study 2, we seek to disentangle these effects. We explore whether the type of ad or the gender of the candidate leads to the observed difference in felt ambivalence. We employed both male and female candidates for both valences in this study of advocacy ads.

Similar to Study 1, Study 2 included a representative sample of 512 participants from a Qualtrics panel. The study employed a 2 (valence: positive vs. negative) X 2 (gender: male vs. female) mixed subjects design. As shown in the Appendix, we employed ad valence as a between-subjects factor, with the gender of the candidate as a within-subjects factor. Using random assignment for each valence, participants responded to two stimuli, one male and one female candidate. We randomized the order of presentation of ads (male and female candidates) to prevent gender order effects.

The stimulus materials were similar to the prior study. We selected four advocacy video ads from past congressional elections for the study, with the candidates identified in the Appendix. As in Study 1, participants viewed a picture of the candidate, responded to the voting likelihood measure, watched an ad for the same candidate and finally responded to a series of items including the voting likelihood measure for a second time. In the second condition, all participants followed the same sequence as before with a candidate of the opposite sex.

For this study, we focused on the final component of the decision making process—the effect of the ad on voting likelihood. We used the measures from Study 1 of voting likelihood, candidate evaluation, ad evaluation, positive emotion, negative emotion, and felt ambivalence.

In Study 1, we found that when participants viewed advocacy ads, felt ambivalence influenced their decision-making process. Because of the study design, we questioned whether what was operating in that process was the type of the ad or the gender of the portrayed candidate. In Study 2, we sought to clarify the process by modifying the experiment to include only advocacy ads. Within each valence, we included two ads directed toward candidates with a different gender.

Table 7 shows the results of a series of ANOVAs for different dependent variables with gender and ad valence as the independent variables. The general pattern shown is that within each valence, participants tended to respond more favorably to ads presenting female, as opposed to male, candidates. But, as in Study 1, the Eta-Squares indicate that the differences are small, with gender explaining less than two percent of the variance in the given variable. In addition, only one condition yielded a statistically significant difference: For negative ads, the candidate evaluation for female candidates is about nine points higher than for men and that difference is statistically significant at the 0.01 level. With that one exception, the pattern here is similar to Study 1. In general, for advocacy ads, we cannot be confident that viewers perceive ads focused on female candidates differently from those focused on male candidates.

Given that the differences by gender of the candidate are not statistically significant, the next question is whether viewers process ambivalence differently depending on the gender of the candidate depicted. Table 8 uses the model for felt ambivalence (from Table 5) to determine whether viewers process ambivalence in a similar way. The magnitude of the coefficients has changed from Study 1—for these ads, latent ambivalence explains more of the variance in felt Ambivalence, whereas the reverse was true in Study 1. Furthermore, the R² nindicates that less variance is explained using the current data. However, as in Study 1, both variables are statistically significant at the 0.01 level, which indicates that we can reject the null hypothesis for both. Most importantly, the magnitude of the coefficients is the same regardless of the gender of the candidate. This suggests that the candidate’s gender does not influence the way in which the participant processes the ambivalence provoked by the ad.

The last question is whether gender influenced the likelihood of voting for the candidate. Table 9 applies the final model of Study 1 (from Table 6) to this new dataset, controlling for the gender of the candidate. As in Study 1, candidate evaluation is the major determinant of the change in voting likelihood. In addition, felt ambivalence interrupts the link between attitude and behavior. The magnitude of the coefficients remains consistent regardless of the gender of the candidate, suggesting that the results of Study 1 were not influenced by our research design decision to use female candidates for the positive stimuli and male candidates for the negative stimuli. Rather, the results really were a consequence of the valence of the ads. As shown in Table 10, ad evaluation explained the most variance in voting likelihood for positive advocacy ads, while candidate evaluation explained more variance in change in voting likelihood for negative ads.

Discussion

Study 1

Study 1 adds to the literature in three ways. First, these experimental results are independent of party affiliation. Controlling for party affiliation by randomly assigning participants to ad conditions, we found a more parsimonious model of ad effects upon voter intention. Second, we found experimental evidence demonstrating that advocacy ads elicit a different persuasion process than do candidate ads. Individuals process information about the candidate and the ad in similar ways, regardless of ad type. In support of Hypothesis 1, ad valence stimulates voter emotion, either in the positive or negative direction, which subsequently influences the voter’s preference for the candidate correspondingly in either a positive or negative direction.

However, in translating their evaluation of the candidate into voting intention, a difference occurs. Viewers of advocacy ads, unlike viewers of candidate-sponsored ads, experience ambivalence in a way that breaks the link between perception of the candidate and intention to vote, as suggested by Hypothesis 2. Third, ambivalence plays an important role in individual voter behavior. Using aggregate data, prior literature on advocacy ads focused on their impact on political knowledge and found that ambivalence has a stronger impact on voting behavior than knowledge about the candidate [34]. Supporting this conclusion, but using experimental individual-level data, we add to the literature with our finding that the ambivalence engendered by advocacy ads led to a different process of evaluating the candidates. This finding has implications for understanding the impact of advocacy ads on the behavioral intention of voters. This partially supports Hypothesis 3.

In Study 1, we found evidence that participants who watched a positive advocacy ad experienced higher levels of felt ambivalence compared to those who watched a negative advocacy ad. A limitation of Study 1 was that it employed female candidates for ads with positive valence (both advocacy and candidate) and male candidates for both negative ads. A potential confound could be that the gender of the candidates, rather than the ad sponsor, drove the effects we observed in Study 1.

Study 2

We designed Study 2 to examine whether candidate gender drove the results found in Study 1. Participants still viewed two ads from the same valence (positive or negative). However, this time both stimuli were advocacy ads and we varied the gender of the candidate addressed in the ad. The gender of the candidates did not provoke statistically significant differences between the levels of felt and latent ambivalence. Furthermore, latent ambivalence influenced felt ambivalence without regard to candidate. Finally, advocacy ads induce ambivalence that interrupts the normal link between candidate evaluation and change in voting likelihood—regardless of candidate gender. Supporting Hypothesis 3, regardless of the gender of the candidate portrayed, viewers of advocacy ads are less likely to vote.

Conclusion

Of interest to us was Lavine’s [34] argument that voter ambivalence is a better predictor of voter decisions than voter knowledge and partisanship. Hence, we explored the role of ambivalence in voter decision-making using experimental research designs. This research contributes to the literature in multiple ways. First, we demonstrate that viewers perceive and evaluate campaign-related advocacy ads in much the same way as candidate-sponsored ads.

Second, although viewers do not distinguish between advocacy ads and candidate ads, advocacy ads enter into the process differentially through an increased role for ambivalence. Consequently, the normal linkage between the evaluation of the candidate and the behavior of voting is disrupted. Campaigns are investing millions of dollars in advocacy advertising assuming that such ads will persuade voters. However, our results show that advocacy ads are more likely to increase voter ambivalence and viewers of advocacy ads are less likely to vote.

Third, this finding regarding the underlying connection between viewers’ evaluations (of candidates and ads) and the resulting behavior of voting holds true regardless of the gender of the candidate. Because these results differ from the literature as well as the ongoing trend of higher investment in advocacy advertising, they represents a major contribution to the canon. The valence of the ad makes a much bigger difference on viewers than does the gender of the depicted candidate.

The results are bolstered by our use of the experimental method that presented real congressional ads to a substantial sample of Americans who reflected the demographics of the American electorate. However, using real ads also contained a potential disadvantage: These ads were not fully comparable in terms of content, form or messaging. In the future, it would be helpful to replicate the findings using more controlled ads that keep the persuasive techniques constant and vary only in either valence or candidate gender by creating our own stimuli—which would, of course, bring its own limitations in terms of generalizability.

Even without such replication, the results of this research have important political consequences. The Supreme Court has acted on the assumption that the formal distinctions between candidate and advocacy ads (e.g. sponsor identification and electioneering language) are sufficient for voters to differentiate between them. They are not. When watching election-related advertisements, voters do not consciously differentiate the sponsors.

In addition, election-related advocacy ads insert a dangerous element into the election process: ambivalence. Unlike candidate-sponsored ads, advocacy ads invoke felt ambivalence, which interferes with the natural connection between candidate preference and voting [36]. As a result, unlike candidate ads, advocacy ads decrease the likelihood of voting and through it overall voter turnout. This threat is particularly pernicious given that an increasing proportion of campaign funds are funneled through interest groups, which means that voters are exposed to an ever increasing number of issue ads. The natural consequence is an increase in voter ambivalence and through it a decrease in voter engagement. Thus, the overall impact is harmful to democratic principles.

Declaration of Interest Statement

The manuscript submitted represents original work and has not been previously published or simultaneously submitted elsewhere for publication. The manuscript has been read and approved by all authors. This research was funded by a competitive grant from the Connecticut State Colleges and University system and the Dean of the School of Business at Southern Connecticut State University. We hereby declare that we have no pecuniary or other personal interest, direct or indirect, in any matter that raises or may raise a conflict with the research conducted in this article.

References

1. Open Secrets. (2020). “2020 Election to Cost $14 Billion Blowing Away Spending Records.” Open Secrets: Following the Money in Politics, 28 October. https://www.opensecrets. org/news/2020/10/cost-of-2020-election-14billion-update/ (3 March 2023).View

2. Evers-Hillstrom, K., (2020). Majority of lawmakers in 116th Congress are millionaires.View

3. Pfau, Michael, David Park, R. Lance Holbert, and Jaeho Cho. (2001). “Issue Advertising and The Potential of Inoculation to Combat its Impact on the Democratic Process.” American Behavioral Scientist, 44 (12): 2379-97.View

4. Pfau, Michael, R. Lance Holbert, Erin Alison Szabo, and Kelly Kamninski. (2002). “Issue-Advocacy Versus Candidate Advertising: Effects on Candidate Preferences and Democratic Process.” Journal of Communication 52 (2): 301-15.View

5. Meirick, Patrick C., Gwendelyn S. Nisbett, Matthew D. Jefferson, and Michael W. Pfau. (2011). “The Influence of Tone, Target, and Issue Ownership on Political Advertising Effects in Primary Versus General Elections.” Journal of Political Marketing 10 (3): 275-96.View

6. Hall, Richard L. and Molly E. Reynolds. (2012). “Targeted Issue Advertising and Legislative Strategy: The Inside Ends of Outside Lobbying.” The Journal of Politics, 74 (3): 888-902.View

7. Singh, Amit S., and H. Rao Unnava. (2015). Attitudinal Ambivalence: How Is It Stored in Memory?" In North American Advances in Consumer Research, eds. Kristin Diehl and Carolyn Yoon. Duluth, Minnesota: Association for Consumer Research, 43: 807.View

8. Sides, John, Lynn Vavreck, and Christopher Warshaw. (2022). “The Effect of Television Advertising in United States Elections.” American Political Science Review, 116 (2): 702-18.View

9. Franz, Michael M., and Travis N. Ridout. 2007. “Does Political Advertising Persuade?” Political Behavior 29 (4): 465-91.View

10. Kahn, Kim F., and John G. Geer. (1994). “An Experimental Investigation of Political Advertising on Television.” Political Behavior 16 (1): 93-116.View

11. Groenendyk, Eric W., and Nicholas A. Valentino. (2002). “Of Dark Clouds and Silver Linings: Effects of Exposure to Issue Versus Candidate Advertising on Persuasion, Information Retention, and Issue Salience.” Communication Research, 29 (3): 295-319.View

12. Chang, Chingching. (2001). “The Impacts of Emotion Elicited by Print Political Advertising on Candidate Evaluation.” Media Psychology 3 (2): 91-118.View

13. Meirick, Patrick (2002). Cognitive responses to negative and comparative political advertising. Journal of Advertising, 31 (1), 49-62. DOI: 10.1080/00913367.2002.10673660View

14. Pinkleton, Bruce. (1997). “The Effects of Negative Comparative Political Advertising on Candidate Evaluations and Advertising Evaluations: An Exploration.” Journal of Advertising, 26 (1): 19-29.View

15. Pinkleton, B. E., (1998). Effects of Print Comparative Political Advertising on Political Decision-Making and Participation. Journal of Communication, 48(4). https://academic.oup.com/ joc/article-abstract/48/4/24/4110067View

16. Thorson, Esther, William G. Christ, and Clarke Caywood. (1991). “Effects of Issue-Image Strategies, Attack and Support Appeals, Music, and Visual Content in Political Commercials. Journal of Broadcasting and Electronic Media 35(4): 465-86.View

17. Valentino, Nicholas A., Vincent L. Hutchings, and Dmitri Williams. (2004). “The Impact of Political Advertising on Knowledge, Internet Information Seeking, and Preference.” Journal of Communication 54 (2): 337-54.View

18. Edell, Julie A., and Marian C. Burke. (1987). “The Power of Feelings in Understanding Advertising Effects.” Journal of Consumer Research 14 (3): 421-33View

19. Meyers-Levy, Joan, and Prashant Malaviya. (1999). “Consumers' Processing of Persuasive Advertisements: An Integrative Framework of Persuasion Theories.” Journal of Marketing, 63 (October Special Issue): 45-60View

20. Fishbein, Martin, and Icek Azjen. (1975). Beliefs, Attitude, Intention, and Behavior: An Introduction to Theory and Research. Reading, MA: Addison-Wesley.View

21. Bizer, George Y., Jeff T. Larsen, and Richard E. Petty. (2011). “Exploring the Valence-Framing Effect: Negative Framing Enhances Attitude Strength.” Political Psychology 32 (1): 59- 80.View

22. Dowling, Conor M., and Amber Wichowsky. (2015). “Attacks Without Consequence? Candidates, Parties, Groups, and the Changing Face of Negative Advertising.” American Journal of Political Science, 59 (January): 19-36.View

23. Meirick, Patrick C. (2005). “Political Knowledge and Sponsorship in Backlash from Party-Candidate-Sponsored Attacks.” Communication Reports, 18 (2): 75-84.View

24. Nai, Alessandro, and Juergen Maier. (2023). “Mediatized Campaign Attacks Fuel Affective Polarization if Perceived as Negative: Experimental Evidence with American Voters.” International Journal of Communication, 17 (January): 5246- 67.View

25. Brooks, Deborah Jordan, and Michael Murov. (2012). “Assessing Accountability in a Post-Citizens United Era: The Effects of Attack Ad Sponsorship by Unknown Independent Groups.” American Politics Research, 40 (May): 383-418.View

26. West, Darrell M. (1997). Air Wars: Television Advertising in Election Campaigns 1952-1996. Washington, DC: Congressional Quarterly.View

27. Siev, Joseph J., and Richard E. Petty. (2024). “Ambivalent Attitudes Promote Support for Extreme Political Actions.” Science Advances 14 (June): 1-19.View

28. Priester, Joseph R., and Richard E. Petty. (1996). “The Gradual Threshold Model of Ambivalence: Relating the Positive and Negative Bases of Attitude to Subjective Ambivalence.” Journal of Personality and Social Psychology, 71 (3): 431-49.View

29. Thompson, Megan M., Mark P. Zanna, and Dale W. Griffin. (1995). “Let’s Not Be Indifferent About (Attitudinal) Ambivalence.” In Attitude Strength: Antecedents and Consequences, eds. Richard E. Petty and Jon A. Krosnick. Hillsdale, NJ: Erlbaum, 361-86. View

30. Van Harreveld, Frenk, Joop Van der Plight, and Yael N. De Liver. (2009). “The Agony of Ambivalence and Ways to Resolve It: Introducing the MAID Model.” Personality and Social Psychology Review, 13 (1): 45-61.View

31. Kaplan, Kalman J. (1972). “On the Ambivalence-Indifference Problem in Attitude Theory and Measurement: A Suggested Modification of the Semantic Differential Technique.” Psychological Bulletin, 77 (5): 361-72.View

32. Newby-Clark, Ian R., Ian McGregor, and Mark P. Zanna. (2002). “Thinking and Caring about Cognitive Inconsistency: When and for Whom Does Attitudinal Ambivalence Feel Uncomfortable?” Journal of Personality and Social Psychology, 82 (2): 157-66.View

33. Rothman, Naomi B., Michael G. Pratt, Laura Rees, and Timothy J. Vogus. "Understanding the Dual Nature of Ambivalence: Why and When Ambivalence Leads to Good and Bad Outcomes." Academy of Management Annals, 11, no. 1 (2017): 33-72.View

34. Lavine, Howard. (2001). “The Electoral Consequences of Ambivalence toward Presidential Candidates.” American Journal of Political Science 45 (4): 915-29.View

35. Luce, Mary F., James R. Bettmen, and John W. Payne. (1997). “Choice Processing in Emotionally Difficult Decisions.” Journal of Experimental Psychology: Learning, Memory and Cognitions, 23(2): 384-405.View

36. Çakır, Semih. (2022). "Does Party Ambivalence Decrease Voter Turnout? A Global Analysis." Party Politics, 28 (July): 713-26.View

37. Shaw, Daron R. (1999). “The Effect of TV Ads and Candidate Appearances on Statewide Presidential Votes, 1988-1996.” American Political Science Review, 93 (2): 345-61.View

38. Basinger, Scott J., and Howard Lavine. (2005). “Ambivalence, Information, and Electoral Choice.” American Political Science Review 99 (2): 169-84.View

39. Dentler, Klara. (2023). “Should I Stay or Should I Go? Testing the Influence of Ambivalence on Vote Switching in the German Multi-Party Context.” German Politics and Society, 41 (March): 25-63.View

40. Keele, Luke, and Jennifer Wolak. (2008). “Contextual Sources of Ambivalence.” Political Psychology, 29 (5): 653-73.View

41. Armitage. Christopher J., and Mark Conner. (2000). “Social Cognition Models and Health Behavior: A Structured Review.” Psychology & Health,15 (2): 173-89.View

42. Armitage, Christopher J. (2003). “Beyond Attitudinal Ambivalence: Effects of Belief Homogeneity on Attitude Intention-Behavior Relations.” European Journal of Social Psychology, 33 (4): 551-63.

43. Sedikides, Constantine, and John J. Skowronski. (1991). “The Law of Cognitive Structure Activation.” Psychological Inquiry 2 (2): 169-84.View

44. Mayhew, David. (1974). “Congressional Elections: The Case of the Vanishing Marginals.” Polity, 76 (3): 295–317. View

45. Durso, Geoffrey R. O., Richard E. Petty, Pablo Briñol, Joseph J. Siev, Lucas D. Hinsenkamp, and Vanessa Sawicki. (2021). “Dampening Affect Via Expectations: The Case of Ambivalence.” Journal of Personality and Social Psychology: Attitudes and Social Cognition, 121 (6): 1-23.View

46. Ridout, Travis N., Erika Franklin Fowler. (2012). “Explaining Perceptions of Advertising Tone.” Political Research Quarterly 65(1): 62-75View

47. McCroskey, James C., Virginia P. Richmond, and John A. Daly, J.A. (1975). “The Development of a Measure of Perceived Homophily in Interpersonal Communication.” Human Communication Research 1 (4): 323-32.View

48. Ahlf, Henning, Sven Horak, Andreas Klein, and Sung-Won Yoon. (2019). "Demographic Homophily, Communication and Trust in Intra-Organizational Business Relationships." Journal of Business and Industrial Marketing, 34 (March): 474-87.View

49. Zemborain, Marin R., and Gita V. Johar. (2007). “Attitudinal Ambivalence and Openness to Persuasion: A Framework for Interpersonal Influence.” Journal of Consumer Research, 33 (March): 506-14View

50. Winfrey, Kelly L., and James M. Schnoebelen. (2019). “Running as a Woman (or Man): A Review of Research on Political Communicators and Gender Stereotypes.” Review of Communication Research, 7:109-38.View

51. Huddy, L., and Terkildsen,N., (1993). The Consequences of Gender Stereotypes for Women Candidates at Different Levels and Types of Office, 46(3). https://doi. org/10.1177/106591299304600304View

52. Bauer, Nichole M. (2023). “Gendered Ambivalence: The Structure of Attitudes about Female Candidates.” Journal of Women, Politics and Policy, 45 (4): 375-93.View

53. Herrnson, P.S., Lay, J.C., and Stokes, A. K., (2003). Women Running “as Women”: Candidate Gender, Campaign Issues, and Voter-Targeting Strategies. The Journal of Politics, 65(1). View

54. Otnes, Cele, Tina M. Lowrey, and L. J. Shrum. (1997). “Toward an Understanding of Consumer Ambivalence.” Journal of Consumer Research, 24 (June): 80-93.View



Appendix