![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
Why are campaign advertisements utilized in congressional campaigns? This study examines the factors that influence if and the extent to which campaign ads are sponsored by candidates. Using the 2008 Wisconsin Advertising Project, the 2010 Wesleyan Media Project data sets, and original data on all U.S. House Candidates in the 2008 and 2010 general election, we investigate the candidate- and district-level factors that influence the use of political advertisements. The results indicate the competitiveness of a race, ad costs, heterogeneous media markets, and campaign contributions influence when and the extent to which candidates expend resources on political ads.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 Indeed, in the 2014 congressional election, over 235 million dollars were spent on nearly 500 thousand political ads.
2 Although the Wesleyan Media Project collected ad data in elections dating back to 1996, the focus herein is limited to the 2008 and 2010 elections for practical reasons. Prior to the Bipartisan Campaign Reform Act (BCRA) of 2002, interest groups were largely unregulated with regard to issue advocacy ad use. BCRA not only targeted soft money in campaign finance, but also established strict criteria to limit the use of issue ads leading up to election day. In the Wisconsin Right to Life vs. FEC (2007) decision, the Supreme Court lifted the Bipartisan Campaign Reform Act’s ban on issue ads in the months leading up to an election. Further, although the Wesleyan Media Project has released the ad data for 2012 and 2014, we have not incorporated these elections in the analysis due to data limitation. To date, we have not completed the data collection on the candidate biography for the 2012 and 2014 elections. Thus, we rely on the analysis of the 2008 and 2010 elections. We recognize the value in analyzing campaign advertising in the post-Citizen United era. This would allow for us to examine if patterns in candidate advertising have changed pre-post Citizens United. Biersack (2018) indicates that although outside third-party spending increased after Citizens United, campaign spending by U.S. House candidate spending exhibits a “flattening of the growth curve after Citizens United.” This suggests candidate ad use may not substantially differ as a results of Citizens United. Of course, this is a topic that should be addressed in future research.
3 For example, based on CMAG estimated ad costs, a candidate in the 2010 election running in the Portland, Maine media market would spend approximately $168,662 on 100 ad spots, while a candidate running in the New York City media market would spend approximately $1,153,760 on the same number of ads.
4 While candidates do invest resources in alternative forms of voter outreach, television advertising remains the dominant tool for voter outreach (Choma 2015). For instance, the Center for Responsive Politics reports that online advertising accounted for just 1.2% of all spending by congressional candidates in 2010.
5 Several sources were used to construct the candidate-level data set. Federal Election Commission (FEC) U.S. House election reports were utilized to construct a complete list of candidates that ran in the general election. The FEC reports provide the name of all candidates, their partisan affiliation, and their vote share in the general election. Next, CQ Weekly Report, The Almanac of American Politics, Politics in America and online resources (candidate campaign and personal websites, Lexis-Nexis, and Newsbank) were used to compile demographic information on each candidate. Finally, in a limited number of cases, personal correspondence with the candidates served to complete the data set.
6 We exclude ads sponsored by political parties, interest groups and hybrid ads–ads co-sponsored by candidates and a political party. The motivation for third-party ad sponsorship is theoretically distinct from candidate sponsored ads.
7 The candidate ad variable ranges from 0 to 7433 (with a mean of 425 and a standard deviation of 918).
8 We utilized data available at: http://www.acsu.buffalo.edu/\∼jcampbel/ For analysis on the accuracy of Cook’s Report see the report by James Campbell at:http://www.acsu.buffalo.edu/\∼jcampbel/documents/CookAccuracySummary.pdf The links were last accessed on April 24, 2018.
9 We constructed the measure by aggregating estimated ad costs to the media market level used in the 2008 Wisconsin Advertising Project and the 2010 Wesleyan Media Project data sets based on data provided by Kantar/Campaign Media Analysis Group (CMAG). While this ad costs measure is not ideal, as ad costs vary not only across markets, but also depending on the sponsor of the ad, we do not have access to information regarding the exact costs for each ad aired. Thus, this measure serves merely as a proxy for the actual costs each candidate paid for each ad aired.
10 These data were culled from the FEC “all candidate summary file,” which provides the total amount of campaign contributions each candidate accrued during the election cycle.
11 It is possible that campaign ads may influence campaign contributions; thus, potentially raising questions regarding the causal direction: do contributions influence ad use or do ads influence contributions? As such, we conducted auxiliary analysis utilizing a modified version of the contribution measure. The modified measure includes only those contributions that were received on or before the primary election. Given that our focus is on ads aired after the primary election, contributions made prior or on the election day cannot be influenced by ads aired after the primary election. To create this measure, we accessed the FEC’s “Detailed Data Files,” specifically the “Contribution by Individuals” and “Contribution from Committees” files. These data include information regarding the transaction date of each contribution a candidate received during an election cycle. To construct the modified contribution measure, we summed the contributions that were received for each candidate on or before the date of the primary election. The results for this auxiliary analysis are presented in the Appendix A. The results are consistent with the results presented in Table 1.
12 The data to create this was culled from the Congressional Quarterly’s Congressional Districts in the 2000s. This source supplies the percent of a district population that falls within specific media markets (i.e. DMAs). For example, Pennsylvania’s 5th District falls across 7 DMAs: 36% in Johnstown-Altoona, PA, 19% in Buffalo, NY, 14% in Pittsburgh, PA, 14% in Wilkes Barre-Scranton, PA, 10% in Elmira, NY, 5% in Erie, PA, and 2% in Harrisburg-Lancaster-York, PA. This district scored a 0.78 on the media market index. The measure ranges from 0 to 0.80, with a mean value of 0.26 and a standard deviation of 0.28.
13 The time indicator is also included to control for any differences in ad use before and after the Citizen United decision handed down in early 2010. This Supreme Court decision allowed third-party actors unlimited spending on political campaigns as long as the activities were independent from political parties or candidates campaign activities.
14 The predict counts presented are based on setting all other covariates to their mean value. The predict count is generated using bootstrapped standard errors and confidence intervals (Long and Freese 2006). All results referred to in the text are statistically significant at 0.05 level.
15 The point estimates presented herein are based on a scenario in which we set the remaining covariates to their mean value.
16 Note, predicted count is contingent on the values of the other variables in the model; thus, the differences may appear muted.
17 Although media index is referenced as statistically significant in Table 1, the predicted counts indicate the difference does not reach a standard level of significance. These predictions are based on the variables used for the fit of the model. In comparing the predicted counts for the minimum and maximum value on the media index measure (while holding all other covariates to their mean), the 95% confidence interval on the change in predicted counts included 0, which indicates the difference in predicted counts does not reach statistical significance.