Skip to main content

Study finds Twitter polls are poor representation of public opinion

Study finds Twitter polls are poor representation of public opinion

Share this story

Twitter trends 640
Twitter trends 640

Polling Twitter users is a risky method of gauging public opinion, according to a new study from Pew Research. The research group found that results can differ with other public polls — often to a wide degree. Unsurprisingly it turns out Twitter users are a liberal bunch. There's perhaps no better evidence of that than reaction to the first presidential debate last year. Pew found that 77 percent of Twitter users somehow thought President Obama emerged the victor despite his widely-panned performance, whereas most public sampling favored Mitt Romney. Twitter's user base also leaned left in reacting to a federal court ruling that found a gay marriage ban in California to be unconstitutional. Twitter's own methodology showed a similar liberal slant during election season.

"Often it is the overall negativity that stands out."

Still, conservative sentiment has also surged at times, however. Compared with national trends, Twitter users were far more critical of Obama's second inaugural address as well as his most recent State of the Union speech. Occasionally Twitter's community trends closely with national opinion; such was the case in a poll gauging reaction to the Supreme Court's landmark ruling on the president's healthcare law. Why the disparity? Pew believes that people who keep up with news via Twitter are "very different demographically from the public." Researchers attribute this largely to Twitter's popularity among younger people, describing the platform's reach with adults as only "modest."

That's not to say there's nothing interesting to take away from Twitter's millions of users. "Overall, the reaction to political events on Twitter reflects a combination of the unique profile of active Twitter users and the extent to which events engage different communities and draw the comments of active users," the research firm says, but ultimately warns that it's simply not the best way of assessing public opinion. For that, you're better off with traditional means of polling.