Does the Content of Political Appeals Matter in Motivating Participation? A Field Experiment on Self-disclosure in Political Appeals

A related body of literature examines whether the content of television advertisements matters. For example, Brader (2006) finds that emotional appeals used in television advertisements are more likely to motivate participation than appeals that lack emotional content. Research debating whether negative advertising affects turnout (e.g. Ansolabehere and Iyengar 1994; Ansolabehere and Iyengar 1995; Finkel and Geer 1998) has also examined whether the tone of the advertisements makes them more or less motivating. These television advertisements are a particular type of political appeal because they can use visual cues that other types of appeals cannot. Get-out-the-vote (GOTV) campaigns, door-to-door canvassing, and phone banking are examples of widely used organizing techniques that cannot vary visual content. More research is necessary on the motivating characteristics of written, verbal, or interpersonal appeals.

Research on survey design confirms this idea, as written appeals which are personalized with the target’s name or other personal information are more likely to generate compliance than impersonal written requests (Dillman 2007, 152).

A second experimental condition designed to test the effect of personal threats was also implemented alongside this study of self-disclosure. The sixty subjects receiving this condition heard an appeal that highlighted the personal effect of proposed policy changes. The appeal said, “The quality of your drinking water could be endangered by the administration's proposed rollback…” However, an examination of manipulation checks showed that this appeal did not have the intended effect. The literature on threat argues that personal, specific threats should arouse greater anxiety in respondents, prompting a stronger emotional reaction and greater likelihood to take action (e.g. Marcus 2002; Brader 2006). Subjects hearing this appeal, however, were no more likely to feel worried (t₂₁₅ = 1.3) or angry (t₂₀₀ = 0.73). This condition may not have worked because the threat was not sufficiently different from the implicit threat in the control condition. Because this condition was completely independent of the other conditions, and had no effect on the results presented here, it is not discussed further. A helpful reviewer suggested, however, running the analyses treating these subjects as an additional control group. If we do that, t-tests show that subjects receiving the self-disclosure appeal are still more likely to find the interviewers likeable (t₂₁₅ = 2.36) and more likely to make the donation (t₂₁₇ = 2.40) than the combined control group. A saturated logit model mirroring the analysis in Table 2 also shows that subjects receiving the self-disclosure appeal are statistically significantly more likely make the donation (p = .009). In sum, the results are identical whether or not this group is included as part of the control group.

I also ran the model controlling for political interest and standard demographic variables (age, income, education, gender, and minority status) and found that the results did not change. As the t-tests in Table 1 show, the treatment and control groups did not statistically differ on these dimensions. Thus, even when those variables are included in a multivariate analysis, the effect of the disclosure appeal still holds. Neither political interest nor any of the demographic variables were statistically significant and the coefficient on the disclosure appeal hardly changed (b = .765, s.e. = .397).

The predicted probabilities are calculated by estimating the effect of hearing the disclosure appeal for each interviewer in each hour and averaging across them. The effect ranged from 17.3% for interviewer 3 in hour 1 to 19.6% for interviewer 1, 2, and 4 in hour 2.

This phrase is borrowed from Sidney Verba, “Political Equality: What is It?”, Herbert Munro Lecture, Stanford University, May 9, 2002.