The Crowdwire


The Archive of Persuasion

We won’t know until after Election Day what impact TV ads had on the outcome. But there’s no question that ads are changing the nature of politics, thanks to huge increases in fundraising levels and the rise of super PACs. The question is how they’re changing it, and it’s not easy to answer.

Enter the Wesleyan Media Project, a continuation of the University of Wisconsin Advertising Project. From 1998-2008, the Wisconsin project studied how candidates, political parties, and special interest groups communicated with voters. Now based at Wesleyan University, the project is still run by three of the original graduate students from Wisconsin, now teaching at other universities.

We interviewed co-director Michael Franz about the project’s methods and findings.

What is the basic mission of the Wesleyan Media Project?

Michael FranzWhat we do is twofold: first, we track and code television ads in real time, which allows us to publicize the intensity and content of the ad wars. This is very hard to do absent the data we have, because campaign finance reports are often not as comprehensive as one would like. Lots of groups do not have to report their spending, for example, and candidate reports to the Federal Election Commission are not detailed enough to know when ads aired.

Second, we archive the data and eventually make it available to scholars for use in their research. Together with the Wisconsin Advertising Project, we have an archive of political ads that stretches back to 1996 and accounts for over 14 million airings.

How does your ad tracking system work?

We buy the raw data from a commercial firm called Kantar Media/CMAG, which has ad trackers in all 210 media markets across the country. These trackers capture and record ads every time they air on local broadcast stations (i.e., local ABC, NBC, CBS, and Fox affiliates) or on national cable.

We get the data on the ad airings weekly, and each individual ad in video form. We then code each ad on nearly 100 dimensions, such as its tone, the issues covered in the ad, and the identity of the sponsor. We even code for whether the ad is likely making an emotional appeal such as humor or fear. By appending our coding to the frequency of ad airings, we can then see what kinds of ads are airing where, by whom, and at what levels.

How does your coding system work? Do you do it by hand?

We code these by human hand, and the tone categories are essentially objective. If an ad only mentions an opponent, it’s a negative ad; if it only mentions a supported candidate, it’s positive; and if it mentions more than one candidate, it’s a contrast ad. Our inter-coder reliability on this measure is very high. For the emotions, we do a lot of training with coders about how to identify those appeals, and as expected, our reliability there is lower. We get strong agreement across coders for emotions like anger, fear, and humor. Pride and enthusiasm are a bit harder to detect or code consistently. We are looking at ways to automate this, and some political science research is arguing for an automated approach over a human coder.

What are your long-term goals for the project? What have you accomplished so far?

Our long-terms goals are to continue adding to the archive in subsequent elections, and also disseminate our results to the broader public. With outside groups increasingly prevalent in federal elections, it is imperative that voters know to what extent wealthy groups and donors are bankrolling persuasion efforts on television. We can speak to that—without any partisan or ideological goal on our part—by simply providing voters with information on who is advertising and at what levels.

To date, scholars have used the ad data from the project in dozens of scholarly articles, book chapters, and books. The data have allowed political scientists to develop and test sophisticated theories about how ads work in convincing voters to vote, and to see if ads have any effects on turnout or knowledge of key issues. Prior to this, the data for testing these effects were indirect and incomplete.

Do you have any hard results to share from this research? Do certain types of ads work better than others?

Great question, and there is no easy answer. It seems pretty clear that ads “work” in the sense that a great number of ads can move votes in your direction. But because many campaigns wind up with an equal number of ads for each side, the net effect is often zero. Travis Ridout and I have a book on this, The Persuasive Power of Campaign Ads, that uses data from the 2000, 2004, 2006, and 2008 elections. And political science has generally reached consensus that ads do not depress turnout; they might even stimulate interest in the election and raise turnout in small amounts. As to what “type” of ad works, the work continues.

What trends have you seen with political ads for the 2012 election?

This presidential race has had more ads than any previous election; both candidates, especially Obama, have aired an incredible number of negative ads; and outside groups, especially in support of Romney, have been extremely active. Even in congressional races, the level of spending from outside groups has been very high. This has been a very interesting election, one that has broken lots of records with respect to political ads.

How does the political ad landscape in swing states compare to other areas of the country?

In the presidential race, it’s simple. If you live in a battleground state, you are probably sick of hearing candidates approve their message by now. If you don’t live in a battleground state, you are being ignored. The only exception is one particularly interesting feature of media markets across the country—they don’t always line up with state boundaries. So some voters in non-battleground states (like Maine or Massachusetts) see ads because of their proximity to battleground states (like New Hampshire, which is covered in part by the Portland, ME and Boston media markets).

Of course, House and Senate races are a lot broader geographically, so voters in lots of places across the country are seeing ads in competitive congressional races.

What is the most surprising thing you’ve learned so far in the 2012 election through this project?

The level of outside spending is astonishingly high. 44% of all pro-Romney ads in October and 33% of all Senate Republican ads in October have come from outside groups. Obama, by contrast, has aired in October about 86% of the ads that support his candidacy. But in comparison to 2008, the level of pro-Obama ads coming from outside groups has risen from about 4% to about 14%. We never expected to see so many congressional candidates, and a major presidential candidate, so dependent on outside spending.

Have you seen anything noteworthy in the Massachusetts Senate race, where the candidates agreed not to allow outside groups to advertise?

That’s what’s interesting. The Brown-Warren pact set up ground rules for trying to prevent outside groups from airing television ads. That’s seemed to work so far. Of course, there are loopholes in the agreement, and outside spending is fairly aggressive in that race, just off the air. We’ve been surprised, though, that the two candidates were able to work out an agreement that made it difficult for groups to air ads during the race. It could conceivably be a model for future races, though it requires both candidates to sign on—and that’s the sticking point.

Does any part of your project involve speaking with the general voting public about their experience with TV ads? If so, what have you found?

We love talking about our work to the general voting public. We generally find that voters express exhaustion at the level of ads on television. Our response, though, is a bit of a defense of such ads. Political ads are still one of the only ways for candidates to speak to a wide spectrum of the public at the same time. Increasingly, voters are consuming less news through television and in print form, and they are getting their news in much more partisan places (especially online). Ads are attempts to speak to all voters, and they can sometimes be the only place for a Republican voter to hear what the Democratic candidate has to say, and vice versa.

Are there any non-political applications for what you’re doing?

One particularly neat application would be to look at the level and intensity of ads for pharmaceuticals. Anyone who watches the nightly national news understands that ads for prescription drugs are almost as ubiquitous as ads for candidates. It would be interesting to compare the frequency of such ads to the demographics of the programs on which they air; and even to see if sales of drugs rise and fall with the frequency of ads.