Analyzing US Presidential Campaign Ads

This is a report on an ongoing research project on public sector narrative. With the support of my latest SSHRC Insight Grant, a research assistant and I have been recording and coding all the ads posted on their YouTube channels by the Clinton and Trump campaigns and their associated PACs. We have coded their ads right from the start of their campaigns in late 2015 or early 2016 up to Election Day. We now have a grand total of 226 Clinton ads and 96 Trump ads — every ad either candidate or their PACs posted online.

We coded for 25 categories. Some are narratological, like the fable being referenced (hero, knave, or fool) and the rhetorical source (who is speaking). Others are cinematic, in the sense that we analyze the ads as mini-movies. These include editing (the use of cuts), figural motion, the use of closeups, indoor or outdoor settings, lighting, color palate of the people or objects being shown, and the colors of the candidate’s clothing. Still others come from political science, such as the issue raised, the social or ethnic group(s) to whom appeals were made, and whether the ad was positive, negative, or a contrast.

Finally, we were interested in the social media response as measured by the YouTube viewcount and the number and proportion of likes and dislikes, a feature that the candidates sometimes, but not always, disabled.

Now that the data are in the form of a big Excel spreadsheet, the analysis can begin. The first set of results I am looking for are differences or similarities between the Trump and Clinton campaigns with respect to every variable. Which candidate was more negative? Which candidate devoted more attention to putting forward its own policies? Did they differ in terms of the groups to which they were attempting to appeal and also in the balance between broad and targeted appeals? Did they differ in terms of the issues raised? Did they differ in terms of how they referred to their personal backgrounds and experiences? Did they differ in the use of the various cinematic techniques?

I should make clear that I am looking for differences in the campaigns rather than explanations of why Trump “won” and Clinton “lost.” Analytically, I consider the election a dead heat, in the sense that Clinton handily won the popular vote and Trump handily won the electoral vote.

A second set pf results I will be looking for concern the alignment between the different fables and the cinematic techniques the campaigns employed. Did either Clinton or Trump, or both, associate certain sets of cinematic techniques with certain of the fables?

A third set of results I will be looking for concern the relationship between the social media outcomes (YouTube viewcounts and like/dislike ratios) and the characteristics of the ads for Clinton, for Trump, and for both (by pooling the data). Statistically, this requires multiple regression analysis with viewcounts and like/dislike ratios as dependent variables and the cinematic, narrative, and political science characteristics as independent variables.

Many of these ads were broadcast on television and sent to individuals on Facebook in addition to being posted on the campaigns’ YouTube channels. It is reasonable to think, if an ad catches the voters’ attention on television, they will look at it again on YouTube, and share it with their friends. Thus viewcounts would show the extent to which the ads are embraced and shared.

In closing, I want to thank my research assistant Adam McGrath, a recent graduate of the Co-op Management Program at UTSC, who did excellent work all summer and into the fall (even after beginning full-time work) on the coding. We had very high inter-coder reliability – a good thing – and had a great deal of fun doing the coding.

I will be embarking on this analysis in the next few weeks and awaiting the results with great anticipation. Let the analysis begin!


  1. Clinton’s popular vote win is essentially because of her very large margin in California. Not that it’s particularly relevant, but Trump seems to have won the popular vote if California alone is excluded.

    There is a wrinkle with the California vote this time, which may explain Clinton’s very large margin there. The presidential race was irrelevant there; Clinton was going to win handily, so neither candidate ran much of a campaign. However, there was a statewide race, for Senator, where because of California’s open primary system, two Democrats (and no Republicans) were the candidates. So Democratic voters had an incentive to vote and were persuaded to go to the polls, while Republicans were indifferent. THis probably contributed to a larger margin for Clinton than might otherwise have been expected. (Republicans were only incentivized to vote if there was a competitive House race in their district and there aren’t too many of these.)

  2. Sandy, a question. Are you looking at third party ads, typically done by SuperPACs, supposedly without coordination with the official campaigns? My impression is that these are usually negative ads attacking the opponent.
    Also of interest, although outside your frame: the Bernie Sanders ads, which I thought were terrific and inspiring, much more so than ads for Trump or, mostly, Hillary.

Leave a Reply

Your email address will not be published. Required fields are marked *