Effects of Issue- and Character-Based Narrative Political Ads on Ad Evaluations

Narrative Persuasion

Narratives are the stories that we tell and hear every day. Hinyard and Kreuter (2007, p. 778) defined narratives as “coherent” stories with characters and identifiable structural features that include beginnings, middles, and endings. Given the prevalence of narratives in ads, movies, and books, it is not surprising that many political ads are based on narratives (Vafeiadis et al., 2018). Narratives are different from argument- or statistics-based messages in several ways. First, narratives present information using the real or likely experiences of individuals. As such, they have characters and plots. These unique features make them more interesting and more likely to engage readers or viewers through storylines than many non-narrative messages. Furthermore, the purpose of narratives is often to entertain and inform, making their persuasive intent less likely to be contested by recipients (Dal Cin et al., 2004). To date, research has provided strong evidence in support of the effectiveness of narratives in changing attitudes, engaging audiences, and reducing counterarguing (see Green, 2006).

The primary mechanisms underlying the persuasiveness of narratives include transportation, empathy, and reduced counterarguing. Transportation has been conceptualized as a mental state where one’s “attention, imagery, and feelings are focused on story events” (Green, 2006, p. 164). Scholars have found that when individuals are transported into a narrative, they are likely to be persuaded by the narrative, and then develop narrative-consistent beliefs and attitudes (Dal Cin et al., 2004; Green & Brock, 2000).

Narratives can also lead to emotional involvement with characters through empathy. Derived from the Greek word empatheia, empathy means feeling into or sharing the thoughts and feelings of someone else (Campbell & Babrow, 2004). Bae (2008), for example, demonstrated that narratives influenced individuals’ empathic responses and their subsequent issue involvement. Similarly, Mazzocco et al. (2010) found that empathy mediated the impact of narratives on issue attitudes.

A narrative’s ability to involve readers in the storylines also means that readers are less motivated to engage in message-related counterarguing, which refers to the generation of rebuttals and opposing arguments (Bilandzic & Busselle, 2013). When the persuasive intent of a message is detected, people tend to become defensive by generating refutations that run counter to the persuasion goals of the message. The effects of a narrative to inhibit counterarguing depends on viewers’ absorption into the narrative and identification with its characters (Slater & Rouner, 2002). Transported individuals are likely to allocate their cognitive and affective resources to the story event. As a result, they are not likely to scrutinize persuasive information in the message. Therefore, scholars generally consider the suppression of counterarguing as one of the key mechanisms underlying a narrative’s impact (Bilandzic & Busselle, 2013).

On the basis of the aforementioned discussions, we expect that narrative political ads will have an impact that is similar to what has been found in prior research, albeit in a different context. In other words, those exposed to the narrative political ads will likely report higher scores on transportation and empathy and lower scores on counterarguing than those who viewed the non-narrative political ads. We also expect that transportation, empathy, and counterarguing will mediate the effect of narrative political ads on ad evaluations. Our two hypotheses can thus be summarized as follows:

Message Valence and Focus in Political Advertising

In addition to narratives, political candidates and their surrogates use other message strategies to improve the effectiveness of political advertisements. Most prominent among them are the tone (or valence) and focus of messages. Regardless of the format, the tone of political ads can be either negative or positive (Garramone, 1985). While positive ads usually support the featured candidate, negative ads are often designed to attack rival candidates. The use of negativity in political advertising is common. In general, academics and practitioners both believe that negative political ads can be advantageous in competitive races (Perloff & Kinsey, 1992).

According to Lau (1985), negative ads may work because any negative information about a candidate is inherently more salient and eye-catching to voters. Furthermore, negativity about candidates is often a harbinger of some threats or risks. As such, voters are likely to be vigilant against any unfavorable information, and automatically give it more weight than favorable information when forming evaluations and perceptions of candidates. In a meta-analytic review of research on negative advertising, Lau et al. (2007) revealed that negative ads were generally found to have a noticeable effect on memory and campaign knowledge, and modest decrease in affect for the target of these ads.

During the past few decades, scholarly research has also focused on examining the effects of different types of campaign ads on voter responses (Lau et al., 2007). Most relevant to our present study is the research on negative political advertising that tends to focus on political issues or candidates’ personal qualities (character). Shyles (1984) defined those ads focusing on candidates’ character as image ads and those focusing on candidates’ positions on political issues as issue ads. An analysis of political ads in prior elections suggests that while most of the ads emphasized issues or policies, a significant number of them focused on candidates’ character (Benoit, 2001).

While some prior research has found that individuals responded differently to either issue- or character-focused political ads (see Shen, 2004), others found no significant differences between issue- or character-focused ads on their impact on ad and candidate attitudes (Dardis et al., 2008). Part of the purpose of the present study is therefore to see whether the use of narratives in political ads focusing on either issues or characters will affect individuals’ subsequent ad evaluations. In addition, we want to examine if and how narrative ads might affect message evaluation via transportation, empathy, and reduced counterarguing. Although prior research has examined valence and message focus or frames within political messages (De Vreese, 2003; Lagerwerf & Yu, 2017), their combination with narratives in political ads has not been explored. We thus formulate a corresponding research question as follows:

Design and Participants

To test the hypotheses and answer the research question, we conducted an experiment with a 2 × 2 × 2 between-subjects factorial design. The three factors were message format (non-narrative vs. narrative), message valence (negative vs. positive), and message focus (issue vs. character). A total of 275 participants from a large US-based university were recruited for the study. About 79.6% of them (N = 219) were female participants, and 20.4% (N = 56) were males. Their ages ranged from 18 to 25 with an average age of 19.97 years (SD = 1.06). With respect to self-identified race, 220 (80%) identified as White/Caucasian, 21 (7.6%) as Asian American, 20 (7.3%) as Hispanic, 12 (4.4%) as other, and two (0.7%) as African American. Participants also reported affiliation with Democratic (36.0%) and Republican (27.3%) parties, followed by Independent (17.8%) and other (4.0%) parties. A smaller portion of participants reported being unaffiliated with a political party (14.9%).

Procedures

We invited participants to the study by sending them a link to our Qualtrics-based questionnaire. After they read the consent form, respondents were informed that they would view a political ad for a special election campaign and then complete a questionnaire. One of the eight political ads was randomly shown to each participant. Afterward, respondents were asked to fill out a short survey designed first to check the manipulation of independent variables and then measure transportation, empathy, counterarguing, and ad evaluations. Standard demographic questions were also asked about gender, age, party affiliation, and race. Upon completion, each participant was thanked and received extra credit for their participation.

Stimuli

The stimulus ads used in our study resembled a political advertisement for a special election for US Congress. Ads were largely based on longer-length textual political advertising flyers that are distributed in regular mail, email, or on social media platforms. Stock imagery of a senior male candidate standing behind a podium, along with the slogan “Your Vote Counts” and an icon graphic of a hand submitting a vote in a ballot box, were consistent across stimuli. Other design elements such as colors, fonts, and layouts were also held constant across conditions. “Steven Porter” was presented as the candidate’s name and was described as a professional accountant prior to running for public office. The general location for the special election was described as occurring in the state where the university is located. However, the specific district for the election was not listed to avoid selection bias with the sample, since local participants might have been familiar with election activities of some local districts. Additionally, the source for each advertisement was a fellow district resident with a generic name, and the sponsor for each ad was described as “Paid for by Citizens for Progress.” Each one of these informational controls was unchanged across conditions.

The eight conditions were manipulated by altering headlines and information within the body of the ads. Narrative ads contained core characteristics such as a plot with a beginning, middle and end, elements of suspense, and candidate Steven Porter. Narratives were shared from the first-person perspective of the fellow district resident. Non-narrative conditions were also authored by the same district resident and manipulated by presenting several bulleted statements with no plots. Finally, the character condition was altered through information related to the candidate’s honesty and integrity, while the issue condition was manipulated through information about the candidate’s issue positions and policy stance on the environment and public health (see the Appendix for examples). The stimuli went through several rounds of reviews and edits by the researchers. Every effort was made to make them look like realistic and authentic political ads.

Measurement

Manipulation Check

A series of independent-samples t tests confirmed the success of our manipulations. Participants who viewed the narrative ads, M = 5.79, SD = 1.33, SE = .11, indicated that the ad messages were more narrative than those who viewed non-narrative ads, M = 3.63, SD = 1.47, SE = .13, t(273) = 12.82, p < .001, Cohen’s d = 1.40. Further analyses indicated that these results were consistent when considered across two other manipulated conditions. In other words, narrative messages were perceived to be more story-based than non-narrative messages regardless of the issues or valence. In addition, participants who were exposed to the positive ads, M = 5.47, SD = 1.29, SE = .11, and participants who were exposed to the negative ads, M = 2.32, SD = 1.29, SE = .11, differed from each other on their agreement regarding perceived message valence, t(273) = 20.18, p < .001, Cohen’s d = 2.43. The manipulation check for message focus was also successful. Participants who read issue-focused ads indicated that the ads focused on the candidate’s policy positions, M = 5.07, SD = 1.69, SE = .14, while those who read character-focused ads indicated that the ads focused on the candidate’s character, M = 2.34, SD = 1.64, SE = .14, t(273) = −13.60, p < .001, Cohen’s d = −1.64.

Hypotheses Testing

To test H1, we ran 2 × 2 × 2 ANOVAs with message format, valence, and focus as fixed factors and transportation, empathy, and counterarguing as dependent variables, respectively. Table 1 presents the means and standard deviations of the key variables for each experimental condition. Results revealed that the main effects of message format on transportation and counterarguing were significant. Specifically, participants who viewed the narrative political ad, M = 4.31, SE = .10, were more likely to be transported into the story than those who viewed the non-narrative counterpart, M = 3.83, SE = .10, F(1, 267) = 11.90, p < .001, η_p² = .04. Furthermore, participants who viewed the narrative political ad, M = 3.44, SE = .09, were less likely to generate counterarguments than those who viewed the non-narrative ads, M = 3.71, SE = .09, F(1, 267) = 4.60, p = .03, η_p² = .02. H1(a) and H1(c) were supported. However, those in the narrative conditions did not report more empathy than those in the non-narrative condition, F(1, 267) = .06, p = .81, thus rejecting H1(b).

Table 1 Means and standard deviations by experimental conditions

	Non-narrative				Narrative
	Negative		Positive		Negative		Positive
	Issue N = 35	Character N = 35	Issue N = 34	Character N = 33	Issue N = 32	Character N = 37	Issue N = 37	Character N = 32
Note. Cell numbers are means and standard deviations (in parentheses) for each experimental condition. Means in the same row sharing the same subscripts are different from each other at p < .05.
Transportation	3.85a (1.38)	3.72b (0.98)	3.78 (1.18)	3.98 (0.99)	4.48a (1.17)	4.48b (1.12)	3.97 (1.13)	4.31 (1.16)
Empathy	4.02 (1.44)	3.34a (1.24)	3.55 (1.22)	3.63 (0.95)	3.74 (1.39)	4.03a (1.32)	3.57 (1.30)	3.34 (1.23)
Counterarguing	3.70 (1.17)	4.04a (1.09)	3.68b (1.17)	3.40 (0.68)	3.74 (1.00)	3.47a (0.80)	3.20b (1.00)	3.35 (1.12)
Ad evaluations	4.37 (0.82)	3.64a (1.13)	4.54 (0.75)	4.64 (0.67)	4.08 (0.66)	4.22a (0.84)	4.55 (0.81)	4.27 (1.26)

Table 1 Means and standard deviations by experimental conditions

View as image HTML

To test H2, we performed a mediation analysis using Model 4 in Process macro (Hayes, 2018) with message format (non-narrative vs. narrative) as the independent variable and ad evaluations as the dependent variable while controlling for message valence and message focus. Transportation, empathy, and counterarguing were entered as parallel mediators. Results demonstrated significant indirect effects of narrative political ads on ad evaluations via transportation, B = .06, BootSE = .04, 95% CI [.001, .137], as well as via counterarguing, B = .10, BootSE = .05, 95% CI [.012, .204], and thus H2(a) and H2(c) were supported. However, empathy did not significantly mediate the impact of the narrative ads on ad evaluations, B = .00, BootSE = .01, 95% CI [−.026, .034], and thus H2(b) was rejected. Figure 1 presents all the path coefficients.

To answer RQ1, we used Model 11 in Process macro (Hayes, 2018) with message format as the independent variable, message valence and message focus as the moderators, and ad evaluations as the dependent variable. Transportation, empathy, and counterarguing were entered as parallel mediators. Results showed a nonsignificant three-way interaction on transportation (see Figure 2), and the moderated effect mediated by transportation was not significant, Index = .00, BootSE = .07, 95% CI [−.160, .119]. There was a significant three-way interaction on empathy (see Figure 2). Findings of conditional effects showed that when political ads attacked the competing candidate’s character, the narrative message led to higher levels of empathy than the non-narrative message, B = .69, SE = .30, p = .023. However, the moderated effect mediated by empathy was not significant, Index = −.08, BootSE = .10, 95% CI [−.318, .050]. Lastly, a significant three-way interaction on counterarguing was found (see Figure 2), and the index of moderated mediation was significant, Index = −.42, BootSE = .21, 95% CI [−.857, −.023]. Specifically, when political ads attacked the competing candidate’s character, the narrative ad led to lower levels of counterarguing, which resulted in positive ad evaluations, B = .23, BootSE = .10, 95% CI [.051, .462]. Figure 2 shows the coefficients of three-way interactions on mediators and the path coefficients between each mediator and ad evaluations, the dependent variable.

Limitations

Even with these findings, our study does have its share of limitations. We conducted an experiment for a simulated campaign. While such an approach allowed us to control for the influence of many potentially confounding factors such as partisanship and prior exposure, it may have limited the generalizability of our results. Our dependent variable of ad evaluations does not directly measure the effects of ads during a real election. Future research should also measure other variables such as persuasion outcomes or voting intentions, provided that the research design and sample are relevant to potential voters and examine a real election. Finally, in creating stimuli, we manipulated three factors in each of the ads. This may have inadvertently introduced confounding language that could weaken the internal validity of our ads. For example, we used a district resident as narrator of all messages. The plots of each story revealed contextual clues about the relationship between the narrator and the candidate. Participants may have felt more empathy toward the narrator in the character conditions because she/he had a long-term working relationship with the candidate and was able to offer exclusive information that might not otherwise have been known. Indeed, while the mean of the narrative character-based ad was significantly higher than the midpoint on the narrative manipulation scale, M = 6.31 vs 4.0, t(68) = 19.06, p < .001, Cohen’s d = 1.01, the mean of the non-narrative character-based ads was also above the midpoint on the same scale, M = 4.27 versus 4.0, t(67) = 1.58, p = .06, Cohen’s d = 1.42. Participants thus perceived the non-narrative character-based ads as slightly narrative as well. We suggest that future studies extend or replicate our research by using political organizations, rather than an individual, as the main source of information in non-narrative messages. Another way to avoid potential confounding is to simplify the experimental design by manipulating the message focus and message valence factors separately (and sequentially) rather than crossing the factors. The effects from each manipulation could then be detected for independent occurrence rather than co-occurrence to gain a more accurate assessment of the impact of narrative political ads.

Conclusion

Despite these limitations, our findings provided a more nuanced understanding of the impact of message strategies in political advertising. Our experiment tested limited exposure to a single message and then measured participants’ responses. If such a brief exposure led to meaningful changes in attitudes as the current findings have demonstrated, one could only imagine the potential magnitude of attitudinal impact for a well-executed and well-funded direct mail or email advertising campaign using a similar strategy. The effects of the current study could also be examined through additional modalities, including video. As such, further research into the effect of narrative political advertising is needed in order to better understand the full extent of its impact on political campaigns.