Can Facebook Predict Election Outcomes?
It’s a fun game to try to predict the outcome of a future event – football games, NCAA tournaments, coin tosses. I can predict the judges’ scores on “Dancing With the Stars” with 90%+ accuracy. After doing it for several seasons in a row, I got bored of that game. I couldn’t find a bookie who would take my money.
Apparently, God has told Pat Robertson who the next President will be, but Pat isn’t talking.
Predicting the outcome of elections, especially U.S. Presidential elections, is big business. Everyone gets in on it. And if you have an opinion that you can spout in 3-second soundbites, 24-hour news channels will put you in a little box next to another guy in a similar box who disagrees with you and will let you duke it out.
But, what if there was a way to track people’s opinions and more accurately predict their behaviors when voting? Well, that’s the Holy Grail, isn’t it? That’s what polling is all about. But, that only ever gets us so far, especially in the U.S. where the ever-present and ever-in-the-majority “undecided” voter gets all the effort and attention from every candidate.
Enter Facebook. And Twitter. And YouTube.
Two websites, Sociagility and AllFacebook, have analyzed metrics from those three sites in the days leading up to the Iowa Caucus. According to their analyses, only Facebook metrics would have predicted the right outcome.
AllFacebook’s approach, in particular, tallied Facebook “likes” and “talking about” information for each candidate, the idea being that, the ratios of how much candidates were being discussed and “liked” were an indicator of their position in the eventual caucus outcome. Let’s collapse their chart a bit and look at Facebook, the only social media outlet in their study that came close to predicting the outcome correctly.
It is an interesting after-the-fact analysis of a single event. But the data they have tallied leads to some other questions when compared to the final outcome.
For example, one huge hole in the election results in terms of Facebook activity is Rick Santorum. He finished in a statistical tie with front-runner Mitt Romney. But his Facebook activity was outrageously down compared to other candidates. In fact, the only candidate trailing Santorum in Facebook mentions and likes was Huntsman.
The final caucus tally put the candidates in this order:
If one had looked at Facebook metrics as a predictor, the race would have been forecast as:
Romney’s lead, or at least a top-three finish, was predicted widely. Ron Paul finishing in the top three was also commonly expected. It is generally acknowledged that Huntsman would be dead last no matter what (for now). So, that left huge questions about the other four candidates. All we really had to go on was the current direction of their inertia in media and popular culture.
And that is what I think we get out of looking at the Facebook data. It is a peek at who is talking, but not about what they are saying. And, for Facebook particularly, we need to look at the quality and content of what is being said.
Bachmann was being talked about a lot. But, there are tons of crazy-eyed Bachmann photos being traded around. She made outrageous statements and gaffes almost daily. She was a walking joke. And Facebook trades jokes.
What about Santorum and his conspicuous absence from Facebook activity? Well, what do you get when you Google “Santorum”? And who wants that kind of NSFW material being passed around with kids all over Facebook? Who wants to explain anal froth? Let’s just not talk about it, shall we?
Huntsman doesn’t have much in the way of controversy or gaffes. He hasn’t been much of a topic of conversation since he admitted that he actually listens to scientists.
Perry, Gingrich and Bachmann all have two things in common: 1) they all have said things that are worth poking fun at, and 2) they were all three on their way down.
What this all adds up to is that “talking about” and even “likes” do not add up to favorability. They simply mean activity. Herman Cain is not even in this race anymore. He’s gone away quietly. But if he showed up on Maury tomorrow for a paternity test, he’d be charting on Facebook again. How would that skew the election analysis? Perhaps analysis of this type should not be done by some campaign consultant, but by your average Redditor.
In the end, elections are complex things. People build careers around analyzing and advising on the topic, just like jury selection specialists. There are social factors, religious factors, racial factors, economic factors, etc. Candidates are advised on which tie to wear, how to smile, what to eat. It extends into legislation and activities that will hopefully affect future election results through re-districting, ballot initiatives, voter fairness/suppression.
Facebook and that ilk tells us what we talk about. And, usually, it just ends up being cats. Now, if Jon Huntsman started doing funny stuff with cats in public before the New Hampshire Primary…