From the Secret Ballot to the Public Vote: Examining Voters’ Experience of Political Discussion in Vote-by-Mail Elections

Abstract

Considerable research on political discussion has focused on identifying its antecedents and outcomes. The rise of voting by mail provides an opportunity to examine the subject in a new context—one in which voters discuss their views and electoral choices with others while filling out their ballots. We explored the possibility that conventional predictors of political engagement would predict who partakes in such discussions. Past research also suggested that those voters most likely to report changing their minds as a result of discussion would perceive their discussants as holding contrary views and higher levels of political sophistication. We further hypothesized that less politically engaged voters would seek out discussants they rated as more knowledgeable than themselves, whereas the more politically sophisticated voters would seek out like-minded discussants. Past research also suggested that the least partisan voters would be those most likely to report disagreement in their absentee discussions. To test these hypotheses, we analyzed telephone survey data from two elections conducted in Washington State. Results showed that the factors that predict traditional forms of political participation and discussion do not explain who engages in discussion during vote-by-mail elections. We also found that independent voters were more likely to talk with ideologically divergent discussants, whereas less knowledgeable citizens sought discussants who knew more about politics than they did. Many voters reported that these discussions shaped their vote choices, with the highest rates of perceived influence coming from those who viewed their discussion partners as more knowledgeable and more ideologically divergent.

Keywords:

• political discussion
• political conversation
• civic engagement
• political knowledge

Notes

1. Several U.S. states have instituted vote-by-mail elections, in some cases replacing polling-place elections entirely. Many others have relaxed rules for absentee voting, effectively creating a combination of vote-by-mail and polling-place elections. At the time of this study, Washington was transitioning to mail elections, but a large majority of voters were already voting by mail.

2. Analyses of the combined 2006–2007 data employed a dummy variable to control for potential election year effects on the dependent variables. This predictor never reached significance. In addition, to control for differences in the election context and electoral process between those years, a regression analysis on these combined 2006–2007 data included a series of interaction terms between this election year dummy variable and each independent variable used in the model. The results of that analysis are described later.

3. These 2007 data also gave a window into who these discussion partners are: The vast majority of people, 79.6%, said they spoke with a spouse or partner. But other social and familial connections also seem to be important: 57% of people reported speaking with a housemate, 51.3% reported talking with a friend, 31.5% reported speaking with a child, and 31.1% said they talked with a coworker. Smaller proportions of respondents reported speaking with a fellow member of a community, political, or religious group (23.5%) or a sibling (13.8%).

4. This dichotomous variable was coded in the opposite direction of the agreement scale variable in order to match the direction of the dichotomous knowledge imbalance variable mentioned earlier. That is, both dichotomous variables indicate that either a voter chose to expose herself to an intellectual challenge (in an ideological opponent or more knowledgeable discussion partner) or chose not to do so.

5. There were substantial associations between the state and national political knowledge items on the 2006 survey, with r values ranging from .154 to .362, all significant at a p < .01 level. Scale reliability values dropped to around .5 if the state-level items were deleted, suggesting that they were important to the political knowledge scale. This is in line with Delli Carpini and Keeter (1996), who found that most kinds of political knowledge are related.

6. We conducted an additional version of our analyses using a version of the political knowledge variable comprising these two common items from the 2006 and 2007 surveys, and the results were nearly identical. These two items covered both state and national politics, asking the respondent which political party had the most seats in the Washington State Senate and the U.S. House of Representatives, respectively.

7. This additive index of political participation includes standard measures of political behavior.

8. Absentee voters and discussers were also compared to the remaining voting population in the sample (883 respondents who had not yet voted). Only minor differences could be seen between the three groups: Other voters were somewhat younger than either absentee group, and closer in average income to absentee discussers, though closer to other absentee voters in level of partisanship. Though the pool of non-absentee voters was somewhat limited in 2006–2007, there were still sufficient numbers of survey respondents from several counties across the state to warrant comparison with absentee voters and absentee discussers.

9. Though a dichotomous dependent variable would typically call for a logistic regression, we opted for a linear regression for this variable to yield a more straightforward comparison with the linear regressions on the political participation index.

10. Analysis using a two-item version of political knowledge yielded the same results: knowledge and partisanship were the only independent variables reaching significance in the regression analysis, and the interaction between political knowledge and election year was also significant.

11. Because of the relatively low number of cases for this portion of the study (N = 53), we reduced the number of predictors to include sex, education, political interest, political knowledge, and partisanship. The income and ethnicity items had the most missing data and were removed.

12. This might seem obvious, but we should note that this is based on voters’ perceptions of discussants’ knowledge. That is, someone with a high knowledge index could be talking to someone they see as more knowledgeable. Likewise, the result on disagreement and partisanship seems obvious, but is not a foregone conclusion: The perception of political disagreement is more general than the voting shifts that can occur on specific ballot choices.

Additional information

Notes on contributors

Justin Reedy

Justin Reedy is Assistant Professor, Department of Communication, and Research Associate, Center for Risk & Crisis Management, University of Oklahoma.

John Gastil

John Gastil is Professor and Head, Communication Arts and Sciences and Political Science, Penn State University.

Patricia Moy

Patricia Moy is the Christy Cressey Professor of Communication, Department of Communication, University of Washington.