Despite an inescapable torrent of opposition from popular tech blogs, Twitter users, and city mayors against Chick-fil-A, the self-avowed anti-gay marriage restaurant enjoyed record-breaking sales yesterday. Had I just gazed the world through my Twitter feed, I would think Chick-fil-A was on the verge of bankruptcy…and also that Ron Paul was president, gay marriage was legal, and President Obama didn't have a decent chance of losing the election.
Yet, with all its flaws, Twitter released a new political tracking tool yesterday, claiming that it could help predict the popularity of each presidential candidate, and received much fanfare in the mainstream press. But a recent review of research on the predictive powers of social media made it clear that no one on planet Earth can reliably predict political outcomes from Twitter. "It can be concluded that the predictive power of Twitter regarding elections has been greatly exaggerated," writes computer science professor, Daniel Gayo-Avello.
We've written about this extensively before, but thought we'd post another quick explanation in light of recent news:
1. Young people, who often dominate social media, have a bigger bark than bite: If no one under the age of 30 had voted for Obama in '08, he still would have won every state but two. In other words, young people don't vote in significant numbers. And, the duel protests between conservatives and liberals at Chick-fil-A are a perfect representation of this problem: conservatives vowed to eat at the restaurant on August 1st and liberals are staging a “Same Sex Kissing Day” 2 days later. Kissing Day will get attention (and probably be more entertaining), but Chick-fil-A will still be too busy counting cash from August 1st to notice. Even if Twitter's tool accurately reflected popular belief, it wouldn't reflect popular action.
2. Conservatives Are likely late tech adopters: Liberals are likely more influential on Twitter, in part, because conservatives are late technology adopters. The same psychological tendencies that scare many Republicans away from social change also affect their willingness to try out new communication tools. Social media political trailblazers have almost universally been Democrats: Howard Dean pioneered online fundraising, Obama popularized online coordination, and The Huffington Post (owned by TechCrunch parent company, Aol) was an early adopter of reader-generated blogging. It's true that some conservatives, such as Congressman Darrell Issa, are e-government pioneers. But, in aggregate, the risk-aversion associated with conservative beliefs bleeds into their technology prowess.
Twitter's experimentation is laudable, but their claims are misleading. “The trend in Twitter Political Index scores for President Obama over the last two years often parallel his approval ratings from Gallup, frequently even hinting at where the poll numbers are headed,” explains Twitter's announcement. Bull-pucky.
There's natural variation in Twitter chatter and political polls, which often swing wildly away from what the nation actually believes (which is why the best campaign indices average polls together). It is almost a mathematical certainty that Twitter chatter may occasionally preced changes in opinion polls, but we won't know which ones to believe, making it useless for actionable intelligence.
In short, the Twitter Political Index is a big clunky toy. A more honest approach would be to open up their algorithm for public scrutiny (like Nate Silver has done at the NYTimes). But, until someone finds out how to extract an accurate view of the world from social media chatter, know that there's more voters who would “dislike” your links of pro-gay marriage news on Facebook than actually like them.