Media Exposure – What Election Polls Missed
By Todd Murphy, Vice President, Universal Information Services
Monday morning quarterbacking is the easiest job in the world, especially if you have data analysts like I do at Universal Information Services. With the Trump v. Clinton Presidential election in the rear view mirror by a week, it appears obvious why Donald Trump won over Hillary Clinton. Don’t worry, I’m not going to go anywhere near the politics of this issue, but only point out why media placement, message points, and share of voice media exposure is at least a better indicator of a future outcome than polling.
Past Performance Can Indicate Future Outcomes
At Universal Information Services we’ve been using media exposure as a predictor of outcomes since the Obama vs. Romney race. What we have found is that unless there is some extraordinary circumstance that causes the media exposure for one candidate to eclipse the other, there appears to be a strong correlation between media exposure and an election outcome. An extraordinary media event would be something so scandalous as to essentially render the candidates opportunity impossible.
Yes, the 2016 Presidential Election was extraordinary, in fact, I would argue unprecedented. However, both candidates had an approximately equal number of detracting message points to effectively balance the playing field. In other words, neither candidate was able to shoot themselves in the foot so badly they couldn’t recover. Frankly, short of killing someone on stage at a debate, I’m not quite sure what issues would have hindered either of them.
What Can Share of Voice Tell Us?
Oct. 9 – Nov. 8, 2016
Although Universal Information Services provides deep analysis of news and media exposure, this post is looking at only a few relatively simple measures. Our PR measurement tools let us generate a comparison of media exposure between two or more subjects. In this case, we retroactively analyzed the volume and type of coverage both Presidential candidates, Clinton and Trump, received in the 30 days leading up to the election on November 8th. What was revealed was that Trump was receiving more media exposure, across all media types, for that period. And although some call share-of-voice a vanity metric, the impact from message quantity does factor in to the overall potential for behavioral change. The more one is exposed to a message the greater the probability of that message affecting behavior. How much message volume impacts behavior requires a deeper dive, but this is one small piece of evidence pointing towards a Trump victory.
Why Does the Type of Media Outlet Matter?
Oct. 9 – Nov. 8, 2016
Quality of a media outlet, credibility, viewership/impressions, and ease of amplification are several factors that determine the velocity a message might gain through placement in one type of outlet over another. For example, a live interview on Fox News television reaches millions of viewers. The link to that story also has the potential to reach millions of people because of the committed viewer base that Fox News has. In tandem, a single story on Fox News and their website could travel wider and impact more people than a blog post might from a political website in Iowa. This is not a reflection of the quality of the information contained within this story, but rather the magnitude a media outlet can impact the public. More stories on the more prevalent, higher rated outlets, the greater the probability a message has of impacting a voter. Here we see Donald Trump surpass Hillary Clinton in each of the five measured types of media outlets. Another piece of the pie pointing to a Trump victory.
How Can News Aggregation Platforms Help?
Oct. 9 – Nov. 8, 2016
Analyzing the volume of messages accumulating within broad media platforms, or points of aggregation, can easily be used to validate the above measure of media type value. In the above chart we looked at five platforms that pull stories and content from a variety of sources. Again, we see that Trump, in green, surpasses Clinton on all five platforms. Here we are simply using the diversity of an aggregation platform to validate what the data is telling us from other measures. Accurately spotting trends, knowing how to normalize disparate data, and having professional analysts to render valid insight are just a few things a reliable media measurement service can do when analyzing media messages. In the case of the Trump v. Clinton surprise, it becomes clearer that maybe the Trump victory was no surprise at all.
As mentioned at the beginning of this post, we’ve studied the predictive qualities monitored media results can have. We’ve seen media measurement more accurately predict the outcomes of a race than polling can. The 2016 presidential election was different for us. When you look at a candidates messaging, and their ability to impact behavior, it may be easiest to think of it in terms of energy. From a physics perspective, the potential to retain or release the energy of their message is arguably equal in both candidates at the beginning of a race. When you add the media as a message catalyst, that internal energy, enthalpy (H), the equilibrium of the two candidates can change and affect the impact of one candidate over another. It is not uncommon for us to use physics when analyzing message impact.
But as you know, presidential candidates, especially this cycle, are not a static thermodynamic system. Although there is much hot air in each (sorry, editorial comment), I’m sure nobody would argue that either candidate were analogous to “ideal gases”. Science aside, how do you explain the gulf between the expectation of the presidential outcome and the actual outcome? All the analysts are looking closely at this race, so every theory is interesting. Please share your theories.
As much as I’d love to be able to totally agree with you that media exposure is a predictor of outcomes, since it would be oh-so-great for the PR industry, I do need to point out that share of voice is far more frequently correlated with negative outcomes. The Trump brand and hotels are seeing 20% decline in revenue, and I’m sure Samsung, Volkswagen and Chipotle would wish for a slightly lower share of voice. What is important is “Share of Desirable Voice” — i.e. coverage that conveys the messages you want to get across. Clearly Trump was better at getting his messages across to his target audience than everyone else.
I’m glad you read the post, Katie! As you know so well, pointing out only a few measures in a post is far from a comprehensive measurement. Too often we see analysts exclude volume of mentions, often because they can’t track comprehensively, in favor of a smaller sample of PESO placements. With the ability to look at all media mentions, I took a look at the inherent energy of a message (candidate) and pointed out that these measures, only these measures, pointed to a Trump victory and the polls missed that. I’m not sure it is true to say a command of Share of Voice more often correlates to negative outcomes. Our holistic framework shows that desirable or undesirable outcomes are a product of the effort, not the numbers analyzed. In other words, analysis is an objective view of the effort, regardless if it is what the client wants to see. But I know I’m preaching to the choir here, or rather the Queen. Hope to see you at #WMIC16.
Hello my friend, I think I might take a step backward and analyse on this.
Share of voice was 15% of the entire win, meaning the data had a lot of negative noise.
Share of desirable voice (SOdV), I believe, is the way to go, if we must talk about SOV of SOI in our measurement program, as it tells you the actual SOV, separating the negative, neutral from the positive – and put all desirable on the table to compete.
My team separated the both parties negative mentions and neutral mention from the entire data tracked. The one with the highest SOV drop down in the SOdV, which tells us, there is a need for separation when analysing SOV.
I completely agree that looking at a the share of desirable voice would be appropriate, if this election had followed a normal trajectory. The point of the post is that negative posts and even neutral posts seemingly had no impact. Objectively speaking, in this election both candidates had so much negative noise, and when looking at consumer generated media, even false news.
This election cycle broke from tradition, nearly everyone failed to accurately predict the outcome, and what we are seeing in the data almost supports the ancient idea that “all news is good news”. In other words, say what you want, when you want, where you want, and it appears the average news consumer might only be hearing name/brand reinforcement. I don’t think you can evaluate the media results like a traditional brand or service, there is something new or different going on with this outcome. This post merely points to the possibility that the potential energy of greater exposure could be the difference. The objective wasn’t to fill hotel rooms, it was to put one name in front of another.
Could it be that when an electorate is so focused on an outcome that transcends the actual candidates, exposure does have an effect? Simply, name reinforcement. If masses decide that negative or false news is not the fault of the candidate, it may actually have a hardening effect on your choice. Could it be that trying to apply the idea of a desired outcome, in this scenario, is immaterial? The formula in this case might be, “I don’t want A or B, but it sounds like A is more inline with the outcome I want”?