Ethan Busby and Dr. David Magleby, Political Science Department
Introduction
Although the term “swing voter” appears in the news frequently, researchers and commentators disagree on the exact definition of this group. Some contend that these voters are individuals who are persuadable or moveable politically. Such moveable voters can be more readily influenced by political rhetoric and campaigning, presumably because they hold less rigid political beliefs. Such persuadability, however, is difficult to measure.
Despite such methodological difficulties, persuadable swing voters play a key role in political campaigns. Numerous candidates openly appeal to moderates, compete over independents, or otherwise reach out to individuals they think are swing voters. But are such voters really more persuadable than other groups? Using a set of survey experiments, this design aimed to determine the relative moveablity of swing voters.
I theorized that non-swing voters tend to be more partisan and to hold stronger political beliefs. As a result, this group of respondents should be less susceptible to experimental manipulations in the survey (differences in question wording), especially in comparison to swing voters. Swing voters, as a group, should be more open to persuasion and manipulation, and I expected to see this in the range of responses given by this group to the experimental manipulations.
Methods
As stated, the purpose of this project is to determine which definitions of swing voters align most closely with persuadability. To fulfill this purpose effectively, I relied on a set of questions from the 2012 Cooperative Congressional Election Study (CCES), a nationally representative survey. In 2012, The Center for the Study of Elections and Democracy at BYU arranged to include a set of questions in this survey, including a feeling thermometer on the 2 major political parties. Those who rated the two major political parties within 15 points of each other on a 100 point scale (following the methodology of previous research) are classified as swing voters. I also compared those who ranked the two major presidential candidates within 15 points of each other on this scale. These survey respondents ranked the political parties and candidates about the same, indicating that they do not heavily favor one over another.
The feeling thermometer captures those who are swing voters in the current election; however, in order to identify those who may usually be swing voters but may prefer one candidate or another in this election, I also compare those who state that they split their votes between the two major parties to swing voters using the feeling thermometer definition.
I used these measures of swing voters to explain how respondents support or oppose the policies mentioned in the experiments on the CCES in 2012. This project uses two of these experiments. Both attribute a controversial government policy—the Patriot Act and requiring health insurance—in one of three ways: to no political figures, to a prominent Republican, and to a prominent Democrat. Subjects were randomly assigned to one of these three treatments. Respondents were then asked to state if they supported or opposed these policies.
Results
Figure 1 shows how these experiments influenced respondents on the CCES.
Figure 1 indicates that the experimental manipulations did not result in changes in respondents’ reactions to either the Patriot Act or requiring health insurance. A closer examination of swing voters and non-swing voters on these experiments reveals that no classification of swing voters responds to this experiment differently than other kinds of votes. No types of voters were meaningfully different from each other in responding to this experimental cue. Perhaps more importantly, no types of respondents seemed to respond to the changes in the questions at all. Different people with different ideologies all seemed to respond to this experiment in the same way. This casts some serious doubt on the quality of the experiment itself.
Conclusions
The results here are completely inconclusive. The data do not show that swing voters are more influenced by experimental cues than other types of voters. Changing the definition of swing voters did not change this at all—all of the various categorizations did not coincide with an increased or decreased effect from the experiments. This may be due to the issues asked in the experiment. It is possible that respondents had already formed their opinions on these political issues, and were not generally influenced by the experiment.
However, this absence of a finding may be due to the experiments themselves rather than anything intrinsic about swing voters and non-swing voters. The experiment did not seem to have a substantive effect on any of the respondents, regardless of party identification or ideology. This may be due to poor question wording or the issues the experiments consider. Further research, with different experiments, is needed to disentangle what these results actually mean. These data and this analysis indicate the importance of careful experimental planning and execution for reliable and meaningful results.