Raising Hope: Hope Inducement and Voter Turnout

Abstract

Politicians routinely invoke hope on the campaign trail, presumably because they believe that inducements of hope attract supporters and impel citizens to the polls. Social psychologists and political scientists similarly posit that activating positive emotions like hope and other powerful psychological mechanisms has the capacity to stimulate prosocial behavior, like voting in elections. In this study, I subject these claims to empirical scrutiny by designing and implementing a series of randomized field experiments to examine whether inducing hope raises electoral participation. Overall, I find little evidence that hope affects voting behavior, but I acknowledge the null effects may reflect the [im]potency of the treatment.

ACKNOWLEDGMENTS

The author is grateful to Meiko Lin for assistance and to Andrew Delton, the editor, and anonymous reviewers for helpful comments.

Notes

¹The sample was restricted to single-voter households to minimize interference between units. Such interference can potentially result in a violation of the Stable Unit Treatment Value Assumption, a fundamental assumption in the analysis of such experiments, and add bias to the estimated treatment effects. The approach of restricting the sample to single-voter households has been commonly adopted in field experimental studies of this sort (Panagopoulos 2011).

^a Test statistics generated using a one-way analysis of variance to evaluate whether mean turnout levels differ across categories of random assignment. In all cases, I cannot reject the hypothesis of equal means at conventional significance (p < .05) levels, implying balance across groups. ^b Vote propensity gauged using a composite index using available vote history data denoting the total number of two recent general elections (for Charlestown, WV: November 2000 and November 2002; for Herndon, VA: November 2008 and November 2006; range = 0–2). Mean levels reported above.

²Lerner and Keltner (2001, p. 153) used a variant of this approach in their study to induce fear or anger; subjects were instructed to provide truthful and detailed answers to two open-ended questions: One questions asked participants to describe three to five things that made them most angry (fearful), and a second question asked them to describe in more detail “the one situation that makes you, or has made you, most angry (afraid).” Subjects were instructed to write down their responses so that someone reading it could even get angry or fearful just by learning about the situation. The authors report successful inducement of the respective emotions using this approach. In the current adaptation of this manipulation approach, subjects are not asked to write a lengthy narrative, but they are asked to undertake a similar cognitive task that presumably results in successful induction of the desired emotion (in this case, hope). I acknowledge it is conceivable that the adapted approach that simply asks subjects to engage in reflection but does not reinforce it via written description may be inferior and may or may not result in successful manipulation. Such potential limitations are discussed next.

³All treatments employed in each of the experiments I report in this article invoked notions of civic duty, potentially confounding the effects of hope inducement with civic duty. Even as prior research has showed that civic duty appeals delivered via mail are generally ineffective (Green & Gerber, 2008), I acknowledge this potential limitation.

⁴The postexperimental manipulation check was conducted August 25 and 26, 2013; additional data collected March 12 and 13, 2014.

⁵Details about the manipulation check and a complete questionnaire are available upon request.

⁶I acknowledge that some subjects assigned to be treated may not have been successfully contacted, but reliable estimates of contact rates for direct mailings are unavailable. Thus I report ITT effects throughout, noting these are likely conservative estimates of the treatment effects. Taking contact rates into account would only magnify the treatment effects I report.

^a Covariates include Gender and Vote propensity (gauged using a composite index using available vote history data denoting the total number of two recent general elections: for Charlestown, WV: November 2000 and November 2002; for Herndon, VA: November 2008 and November 2006; range = 0–2). Mean levels reported above. See Table 1 for details.

⁷Because I employ similar procedures to analyze the replication study reported next, the empirical strategy is presented here in greater detail but only summarized next.