![Sample our Humanities journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5939/TOC-Subject-Sample-2021-Humanities.jpg"})
Abstract
Although conventional wisdom suggests that imbalanced message flows are relatively rare in presidential campaigns, this view relies on the assumption that competing campaigns allocate their advertising similarly. In this research, we show that this assumption is false. We combine ad tracking data from the Wisconsin Advertising Project with a unique collection of survey data on the audience for various program genres. Examining advertising in the 2000, 2004, and 2008 U.S. presidential races, we find that the Republican and Democratic candidates distributed their advertising differently across different program genres, reaching different types of voters. A form of microtargeting has increasingly entered into the realm of political advertising buys. We find that who sees certain political ads is more nonrandom than scholars had previously thought, and we find that unbalanced message flows (a precondition for ad persuasion) are more prevalent than conventional wisdom has suggested.
Acknowledgments
We thank Gregg Lindner of Scarborough Research for the use of the survey data, and we thank Bradley Engle for the visualization of the Scarborough data. All conclusions are our own and do not represent those of Scarborough Research. We also thank John McIver for helpful comments on an earlier version of this article.
Notes
1. Figures come from TNSMI/CMAG as reported by CNN.com (CitationPreston, 2007).
2. The 2008 figure comes from TNSMI/CMAG as reported by Broadcasting and Cable (CitationAtkinson, 2008). The 2010 dollar amount comes from Kantar Media/CMAG as reported by Advertising Age (CitationWheaton, 2010).
3. Buying ads on local television instead of on the national networks was first done to any large degree in the 1988 presidential campaign. By 1992, almost all of Clinton's ad buys were on local television.
4. These numbers come from our coding of the Wisconsin Advertising Project data. The trend away from local news broadcasts is not indicative of a shift toward other more traditional ad buy outlets, as the proportion of ads purchased by presidential candidates during talk shows and game shows (both popular genres for political ads) was roughly constant across years.
5. This same mobilization approach also characterizes much of campaigns' increasing Web presence. Because political sites such as partisan blogs and candidate Web sites tend to be visited by core supporters (CitationHindman, 2008), candidates can often tailor messages and mobilization or fund-raising appeals directly to particular user profiles (CitationKaye, 2009).
6. This statement depends, of course, on how you define an independent voter. If you rely on the simple question of whether a survey respondent self-identifies as a Republican, Democrat, or independent, the latter group represents about 30% of the electorate. As political scientists well know, however, many of these respondents lean more heavily toward one of the two major parties (CitationKeith et al., 1992). This leaves about 1 in 10 voters, on average, who are truly independent of any party attachments. But in either case, this slice of the electorate is crucial to the outcome of any close election.
7. Our comparison of the Nielsen data with the TNSMI/CMAG data in the 100 media markets in which they overlapped in 2004 reveals near perfect agreement on the number of presidential ads aired in each market. It should also be noted that both sources of data only track broadcast network purchases and exclude local cable buys. Because cable channels have a more differentiated audience compared to broadcast stations, local cable buys may be even more highly targeted by presidential candidates than broadcast television.
8. We were able to assign most ad airings to a specific genre: 92% in 2000 and 94% in 2004 and 2008. Most of the programs that could not be assigned likely had very few viewers. Many were locally produced, such as Atlanta's Best Homes or New Hampshire's Business, and many others were one-time events, such as coverage of a Puerto Rican Day parade.
9. Although this approach is not ideal, we do not believe that the characteristics of the audience for particular types of television programs would have changed much over the course of 4 years, and indeed our comparison of 2004 with 2008 data confirms that intuition. For instance, the average absolute change across genres in the percentage of viewers who were Democrats between those two elections was 2.1 percentage points the average absolute change across genres in the percentage of viewers who were Black was 1.4 percentage points; and the average absolute change across genres in the percentage of viewers who were religious was 1.2 percentage points.
10. We calculated this by assigning each county in the United States to a media market (CitationGentzkow & Shapiro, 2008). Because some markets cross state boundaries, some of the market's population is located in nonbattleground states. We aggregated the county population by state for each market, getting the percentage of the population that lies in the competitive state.
11. We purchased county-level election data from Dave Leip's Atlas of Presidential Elections (http://uselectionatlas.org/).
12. Court programming was not one of the genres asked about in the 2004 survey, but in the 2008 survey, the audience for court shows was about 22 percentage points more Democratic than Republican.
13. A couple of methodological points are in order. First, we report only those programs during which the candidates combined aired more than 50 spots. While a 3 to 1 ad advantage may look impressive, it does not tell us much substantively about the candidates' relative allocation of ads if only four ads in total were aired during the program. Second, we eliminated those programs that had advertising in only a single media market. This was done so as not to focus in on those local programs that aired only in a market in which one candidate was on the air. For instance, we might observe a 50 to 0 ad advantage for one candidate on a local sports program, but the importance of that is questionable if only one candidate was on the air in that media market.
14. For these models in 2000, we only have data on the top 75 media markets, which explains the smaller sample size in that year than in 2004 and 2008. And because we restrict the analysis to genres with at least 50 airings in the unit of analysis, the variables tapping proportion of Black viewers and religious viewers are dropped because of collinearity.
15. We include a fixed effect for local news because candidates still put between 40% and 50% of their ads in this single genre.
16. Many portray the Bush campaign as abandoning swing voters in 2004 to focus on the base, but it should be noted that CitationRove (2010) challenges that point (pp. 70–74), suggesting that Bush's campaign pursued both base and independent voters.
17. The finding for Black viewership is a bit deceptive. As we demonstrated in , Kerry had huge relative advantages over Bush on shows with largely Black casts. By our coding of shows to genres, however, these programs are assigned to the larger category of comedies, which had Black audiences at about the national average. If we had viewer demographics at the program level, the effect of race might change.
18. Effects were simulated using Clarify for Stata.
19. Microtargeting, as noted, refers to specific research methodologies used by both parties to classify millions of individual voters for the purpose of direct contact through the mail, by telephone, or precinct walking lists. Macrotargeting deals with mass television and radio and looks to increase the cost-efficiency of such “shotgun media.”