The study, published Sept. 2 in the journal Science Advances, measured the persuasive effects of 49 high-profile advertisements from the 2016 presidential campaign on a nationally representative sample of 34,000 people through a series of 59 randomized experiments. Expanding on prior research suggesting that political ads have little impact on voters' preferences, the study shows that those weak effects are consistent irrespective of a number of factors, including an ad's tone, timing, and its audience's partisanship.

"There's an idea that a really good ad, or one delivered in just the right context to a targeted audience, can influence voters, but we found that political ads have consistently small persuasive effects across a range of characteristics," said Coppock, an assistant professor of political science in the Faculty of Arts and Sciences. "Positive ads work no better than attack ads. Republicans, Democrats, and independents respond to ads similarly. Ads aired in battleground states aren't substantially more effective than those broadcast in non-swing states."

Coppock and his co-authors -- University of California-San Diego political scientist Seth J. Hill and UCLA political scientist Lynn Vavreck -- conducted the study throughout the 2016 presidential primaries and general election.

Over 29 weeks, a representative sample of Americans was divided at random into groups and assigned to watch campaign advertisements or a placebo advertisement -- a car-insurance commercial -- before answering a short survey.

The researchers selected ads using real-time, ad-buy data and news coverage of each week's most important ads. They tested ads attacking or promoting Republican candidate Donald Trump and Democratic candidate Hillary Clinton as well as commercials concerning primary candidates, such as Republican Ted Cruz and Democrat Bernie Sanders. They analyzed the ads' effects on survey respondents across several variables, including the candidate, party, or political action committee that sponsored them; whether they were positive or negative in tone; the partisanship of those viewing the ads; the time to Election Day when they aired; whether they were viewed in a battleground state or not; and whether they aired during the primary or general election.

They found that, on average and across all variables, the ads moved a candidate's favorability rating respondents only .05 of a point on the survey's five-point scale, which is small but statistically significant given the study's large size, note the researchers. The ads' effect on whom individuals intended to vote for was smaller still -- a statistically insignificant 0.007 of a percentage point.

Campaigns should carefully consider efforts to tailor advertisements to specific audiences given that the evidence shows that ads' persuasive effects vary little from person to person or from commercial to commercial, the researchers concluded

The findings do not demonstrate that political advertising is always ineffective, Coppock said, noting that the study didn't analyze the influence of an entire advertising campaign.

"TV ads help candidates increase their name recognition among the public, which is extremely important," said Coppock, a resident fellow at Yale's Institution for Social Policy Studies and the Center for the Study of American Politics. "Moreover, the effects we demonstrated were small but detectable and could make the difference between winning and losing a close election."

The Andrew F. Carnegie Corporation, UCLA's Marvin Hoffenberg Chair in American Politics and Public Policy, and J.G. Geer, dean of the College of Arts and Sciences at Vanderbilt University, supported the study.

Story Source:

Materials provided by Yale University. Original written by Mike Cummings. Note: Content may be edited for style and length.

Journal Reference:

1. Alexander Coppock, Seth J. Hill, Lynn Vavreck. The small effects of political advertising are small regardless of context, message, sender, or receiver: Evidence from 59 real-time randomized experiments. Science Advances, 2020; 6 (36): eabc4046 DOI: 10.1126/sciadv.abc4046