Heck of a Job, Comey!

I have now begun analyzing the data I gathered about Clinton and Trump ads posted on YouTube throughout the entire election campaign. (For a detailed discussion of the data and research plan see my blog post last Nov.25.)

Participation in social media is voluntary and initiated by citizens. Visits to social media sites often represent expressions of approval and affiliation. Researchers are therefore coming to the conclusion that the extent and intensity of social media activity is a reliable predictor of election results. Just before the 2015 Canadian election, Justin Trudeau had many more Twitter mentions than Stephen Harper or Tom Mulcair. Similarly, I found that the Liberals had a YouTube viewcount for all their election ads of almost 10 million, the NDP 2.5 million, and the Conservatives approximately 2 million.

Moving to the US context, my research assistant totaled up the cumulative viewcounts of Clinton and Trump ads on Election Day. The Clinton campaign posted 226 ads that had a total viewcount of 17.8 million. The Trump campaign posted 96 ads with a total viewcount of almost 20 million. Advantage Trump, 53 percent to 47 percent.

We also totaled likes and dislikes, when the candidates did not disable that function (which they often did). Trump had 89,000 likes and 52,000 dislikes, for a ratio of 1.7 to 1. Clinton had 100,000 likes, but 170,000 dislikes, for a ratio of .6 to 1. The two ratios are almost the inverse of one another. There were a lot more of Clinton’s critics (likely Sanders supporters during the primaries and Trump supporters during the general election campaign) watching her ads and disliking them than was the case with Trump’s critics.

I arrayed the ads for Clinton and Trump in order of the date they were released, going from earliest to latest. I also ranked their ads by viewcount. The ranking is roughly consistent with the Pareto rule, as a small number of ads are responsible for a large share of the viewcount. For the Clinton campaign the top 5 (of 226) ads accounted for 32 percent of all views while for the Trump campaign the top 5 (of 96) accounted for 47 percent of all views.

Clinton’s top five ads were “Role Models”, an ad that replayed Trump’s objectionable speeches to a group of children (1.7 million views); “Captain Khan,” an ad narrated by the father of a Muslim American soldier killed in Iraq (1.3 million views); “Grace,” an ad narrated by the parents of a disabled child (970,000 views); “I Love War,” an ad using Trump’s own words about foreign policy (827,000 views); and “Equal,” an ad supporting gay and lesbian marriage (731,000 views). The main common denominator is that the top four were attack ads and none were narrated by Clinton herself. “Equal” was launched at the outset of the campaign (June 2015), “Grace” in June 2016, “Role Models,” in July 2016, “I Love War” in early September 2016, and “Captain Khan” near the end of the campaign, on October 21, 2016.

Now we turn to Trump’s top five. “Crook” compared Hilary Clinton to Richard Nixon and had 2.6 million views; a short 15-second ad introducing Mike Pence after the Republican convention came second with 2.3 million views; an attack on the “biased” moderators of the second presidential debate had 1.9 million views; “Bad News,” which referenced the FBI’s reopening of the Clinton email investigation, had 1.5 million views; and “Hilary has Failed Every Single Time,” an attack on her record as Secretary of State, had 1.1 million views. Like the Clinton ads, the main common denominator is negativity and the use of voices other than Trump’s. But his most viewed ads had higher viewcounts despite being launched later in the campaign: “Crook” on Sept. 23, “Bad News” on November 2, and “Hilary has Failed Every Single Time” on November 1.

All of the Clinton and Trump top five ads can still be found on YouTube if you search on their titles. The total viewcounts today are greater than those on Election Day, but the Election Day totals are what matter to my research. By clicking on “More” underneath a video, and then clicking on “statistics” a graph shows how its viewcount grew over time. For the two top five Trump ads introduced in the last week of the campaign, the curve rises very steeply.

Looking at the last week of the campaign, starting on November 1, Clinton posted 16 ads with a total viewcount of 1,038,000, while Trump posted 11 ads with a total viewcount of 4,100,000. Trump thus took an 80 percent share of all views of new ads. The extent to which Trump had seized the momentum is startling. Clinton’s most popular ad dealt with Trump’s sexism and misogyny and had a viewcount of only 363,000, while two of Trump’s ads, “Bad News” (1.5 million views) and “Hilary has Failed Every Single Time” (1.1 million views) made the top five for the entire campaign.

These results seem to me to support the view that the election was won in the last week, and that an important factor in Trump’s gaining momentum then was the announcement of the FBI reopening the investigation into Clinton’s emails. So, to paraphrase but invert George W. Bush’s misguided compliment to his hapless FEMA Director after Hurricane Katrina, “heck of a job, Comey!” Bush fired Brownie. Trump re-appointed Comey. From the new president’s perspective, it was a job well-done.

Sandford

1 comment

  1. Sandy, I agree that the election was decided in the last week. However, I’m not sure I agree that Comey’s announcement was significant in shaping the result. (Correlation, not causality)

    I’ve seen some polling data which suggests that few decided voters actually changed their minds in the last week, but the undecideds broke heavily for Trump. This is typical when the electorate is generally dissatisfied with the status quo; the change candidate gains heavily. (e.g. Carter – Reagan in 1980) Since the right tack – wrong track polls were something like 70-30 wrong track this time, that was not surprising.
    In 2012, undecideds broke for Obama over Romney, because there was still a mood of optimism and hope. That had dissipated by 2016, especially in the upper midwest where the election was decided.

    The advertisement data you have put together may suggest that the undecideds were looking for information in that last week, prior to making a final decision.

Leave a Reply

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

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>