What's surprising about Pew Research's study suggesting that sentiment on Twitter often doesn't track with public opinion isn't that this happens — it's that the extremes are fairly ideologically balanced.
That's Pew's main point: "At times the Twitter conversation is more liberal than survey responses, while at other times it is more conservative. Often it is the overall negativity that stands out." (That last point is particularly unsurprising.) But in looking at eight political moments, Pew's analysis suggests that Twitter opinion was more liberal than the public three times, more conservative three times, and the same twice. That's odd.
There are a few deserved caveats up front. It is fair to question the methodology of the survey, which analyzed the "tone" of tweets and looked for markers indicating the topic of discussion. As some have pointed out, if you were watching a debate and simply tweeted, "I can't believe that jerk said that," the system wouldn't have included your comment (such as it is).
And, of course, we shouldn't expect a self-selected group of people to be a perfect model of America. Or, as science writer Ed Yong put it:
BREAKING: Non-random slice of the population turns out to be unrepresentative. pewresearch.org/2013/03/04/twi…— Ed Yong(@edyong209) March 5, 2013
But what we should expect is that Twitter would reflect Twitter — and Twitter is a more liberal place than the rest of the country.
In the middle of last month, Pew Internet (affiliated with Pew Research) released data on the demographics of American Internet users on various social networks, including Twitter. That data assessed the likelihood of various demographic groups to use the social network, finding, for example, that 17 percent of men use the service compared to 15 percent of women. That's a subtle distinction, but an important one: It means that men are over-represented on Twitter, however slightly, given that the country has more women in it.
The difference becomes more stark when considering age and ethnicity. Twitter is used by 26 percent of black Internet users, compared to 14 percent of whites. That's a massive difference. Likewise, 27 percent of people aged 18 to 29 use Twitter versus 10 percent of those aged 50 to 64. Our chart shows that both groups are largely over-represented — and both were much more likely to support the re-election of President Obama.
That leaves us with three possible reasons that Twitter sentiment was nonetheless so "balanced." First, the black population of America is substantially smaller than the white population, meaning that even an overrepresentation of nearly two-to-one makes only a small difference. Second, it's entirely possible that young people, for example, weren't tuning into the presidential debates with the same eagerness as their parents. And, third, politics can inspire organized response and groups determined to swamp a Twitter conversation can do so without much effort.
To that last point, Pew included one graph which suggests that broad sentiment, not tied to specific events, more closely reflected the political bias of the platform. It's this one, looking at sentiment over the last few weeks of the campaign.
Opinions of Romney skewed far more negative than those regarding the president. Of course, in early November, the American public felt the same way.