![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
Abstract
The proportion of votes cast before election day has risen steadily over the last two decades. Previous research asked how early voting has impacted voter participation. In this article, we ask how early voting has affected the flow of information to voters through the mass media. By increasing the number of days voters are able to vote, are we also increasing the number of days that candidates and campaigns continuously disseminate campaign-related information to the news media? Is news coverage of campaigns quantitatively and qualitatively different when opportunities to vote early are available and utilized? Our expectation is that early voting significantly influences the volume and nature of campaign news coverage. We study the effects of early voting on campaign news coverage of gubernatorial and Senate races in 2006 and 2008. Our findings reveal that the volume and content of campaign news coverage is significantly influenced by early voting.
[Supplementary material is available for this article. Go to the publisher's online edition of Political Communication for the following free supplemental resource: Appendix for Early Voting and Campaign News Coverage—Alternative Model Specifications.]
Notes
An earlier version of this article was presented at the annual meeting of the American Political Science Association, Washington, DC, September 2010. The authors are grateful for the helpful comments of panel participants. Course time releases and a portion of the data collection for this project were made possible by the 2009–2010 Louisiana Board of Regents ATLAS Grant LEQSF(2009-10)-RD-ATL-14. The data collection for this study was also funded in part by a 2009 Summer Stipend Grant awarded by LSU's Council on Research for Funding and departmental funding from the Department of Political Science and the Manship School of Mass Communication. Special thanks go to Barry Burden for a careful reading of the article and extremely helpful feedback. Any errors remain the authors.
1. Also quoted in CitationStein and Vonnahme (2010).
2. See the Pew Center for Excellence in Journalism report “Near Campaign's End, It's All About Numbers” (http://www.journalism.org/node/13498).
3. Included in the count of absentee/mail-in votes are overseas absentee ballots from U.S. citizens, as allowed by the Uniformed and Overseas Citizens Absentee Voting Action (1986).
4. Several states, including Colorado, have permanent mail-in voting where the voter is automatically sent a mail-in ballot before every election and is not required to make a separate request for each election.
5. The media markets without in-person early voting include Indianapolis, Louisville, Boston, Greenville, North Carolina, Seattle, and Spokane. The media markets with in-person early voting include Portland, Maine, Kansas City, St. Louis, Charlotte, Greensboro, Raleigh-Durham, and Portland, Oregon.
6. Alaska, Colorado, Iowa, Louisiana, Montana, Nevada, New Mexico, and Texas.
7. Arizona, Delaware, Mississippi, Nebraska, New Hampshire, Oregon, Rhode Island, and Washington. Oregon has vote by mail.
8. Our sample of daily newspapers in each state ranges from 20% of the state daily newspaper population to 68% of the state daily newspaper population; the mean level of daily newspaper sample to population representation is 44%. Ideally, we would have the whole population. Unfortunately, campaign news coverage for several newspapers would have required on-site data collection or prohibitively expensive news services.
9. The mean electoral margin for 15 senatorial elections was .17, with a standard deviation of .15, a minimum of .01, and maximum of .45. The mean electoral margin for eight gubernatorial elections was .14, with a standard deviation of .13, minimum of .02, and a maximum of .42.
10. This variable is coded using the question “In your opinion, what is the primary focus of the story?” (CitationHayes, 2008; CitationKaplan et al., 2005). Answer choices were strategy coverage, issue coverage, coverage of candidate traits, ad watch coverage, horserace coverage, and other. Along with their initial training on what each type of news story looks like, coders are provided a coding help sheet to which they can continually refer while coding. Intercoder reliability is calculated using Cohen's kappa, which approaches 1 as coding is perfectly reliable and goes to 0 when there is no agreement other than what would be expected by chance (CitationStemler, 2001). A kappa coefficient of 0.61 indicates a reasonable level of agreement (CitationStemler, 2001). For the content data collected thus far, the kappa score for news story focus is .66.
11. There is some variation among these states in the implementation of early voting. States with early voting have different requirements for the number and location of early voting sites as well as different numbers of days of early voting.
12. The EAC's surveys of state voter turnout by mode of voting are subject to some error, specifically due to nonreporting by jurisdictions within the states. Since our measure of total vote cast before election day is assessed at the state level (as opposed to substate jurisdictions such as counties and municipalities), we were able to verify the EAC's counts of early vote by querying the Web sites of the each state's secretary of state. Our queries confirmed the EAC counts of votes cast before election day for all but two states. The offices of the secretaries of state in Iowa and Montana, both states with in-person early voting, do not report in-person and mail-in votes cast separately. For these states, the EAC reports votes cast in person early with mail-in votes under the heading “Absentee Voting.” The inclusion of our second measure of early voting (i.e., in-person early voting) is an additional check on the reliability of our measure of early voting.
13. Circulation might be an alternative measure of capacity to produce political news coverage. CitationDunaway (2008), however, finds a nonsignificant relationship between circulation and the incidence of substantive campaign news coverage. But corporate ownership structure is a significant and negative predictor of substantive campaign news coverage. Differences among newspapers might also lead some readers to question our state-level analysis—given the focus on local newspaper coverage—and suggest a county- or newspaper-level analysis. However, our primary research question of interest regards the effect of early voting and related strategic campaign behavior on news coverage of campaigns. Though we use two indicators of early voting, neither is appropriate for use at the county level. The allowance for in-person early voting is adopted at the state level, and once adopted at the state level counties generally must comply with this allowance—thus, our institutional variable for early voting does not vary across counties within states. And while the behavioral measure of early voting would vary within states and among counties, the statistics on early voting at the county level are not uniformly accessible or reliable across jurisdictions. Finally, we recognize that our observations may not be independent within newspapers, and therefore we also ran our statistical models with the standard errors clustered at both the state and newspaper levels; the results remained substantively unchanged across the models.
14. An alternative measure of competitiveness is the pre-election survey margins between the two major party candidates. Using the average of the last five pre-election surveys in each race, we calculated a competitiveness score that is the absolute difference between the average polling numbers for each of the two majority party candidates. These data are compiled at pollster.com (http://www.pollster.com/polls/2008senate/). The correlation between the pre-election survey measure of competitiveness and the post-election vote margin between the contesting two-party candidates is .91. Not surprisingly, when we substitute the pre-election survey measure for the post-election vote margin, our findings for measures of total campaign news stories and the proportion of news stories by type of story do not change.
15. Because gubernatorial and U.S. Senate races were held concurrently in some of our election states, we generated a “two races” dummy variable and ran specifications of each model with it included. Inclusion of this variable in our analysis did not influence the results, and the coefficient was nonsignificant in all specifications; when we ran models with this dummy variable, the results remained unchanged. These findings are available upon request.
16. In each model, we cluster our standard errors on state because the use of multilevel data can sometimes present problems due to serial dependence within clusters and heteroskedasticity across clusters. There are various methods for dealing with such problems. Here the data are clustered by state using the Huber/White/Sandwich estimation (CitationHuber, 1967; CitationWhite, 1980); it adjusts the variance-covariance matrix to correct for heteroskedasticity and serial dependency. Because observations may not be independent within newspaper, we also ran the models with standard errors clustered on newspaper; the results remained substantively unchanged across the models.