Mainstream news accounts periodically try to explain how widely Twitter divulges from the electorate as a whole. Analyzing a Pew Research poll in April, the Atlantic noted:

Twitter users are statistically younger, wealthier, and more politically liberal than the general population. They are also substantially better educated, according to Pew: 42 percent of sampled users had a college degree, versus 31 percent for U.S. adults broadly. Forty-one percent reported an income of more than $75,000, too, another large difference from the country as a whole. They were far more likely (60 percent) to be Democrats or lean Democratic than to be Republicans or lean Republican (35 percent).

Nevertheless, one senses that candidates and their campaign operations judge reactions to debates, gaffes and revelations about the candidates through the Twitter prism.

Now, Pew is out with a new study emphasizing how tiny a sliver of the electorate Twitter users are. We start with the reminder that only 22 percent of the population tweets. However, even within the Twitterverse only a minority tweet about politics. Just 39 percent of all users mention “national politicians, institutions or groups, as well as civic behaviors such as voting.” So 8.6 percent (39 percent of 22 percent) of the population is in the political Twitterverse. And 97 percent of the political Twitter’s material is produced by a mere 10 percent of users: That is 2.2 percent of the population.

However, it is even worse than that, since Twitter users tend to follow like-minded people. That self-curation essentially means you are talking to yourself, or close facsimiles. A campaign that uses a curated segment of such a small percent of the entire population (registered and unregistered voters) to guide its decisions is bonkers.

Equally bonkers is the tendency of journalists (who I have a sneaking suspicion are over-represented in that tiny population of frequent political Twitter users) to cover the campaign as if Twitter (i.e. they are their peers and equally obsessed political social media users) is representative of the population. And yet that is precisely what a good deal of political reporting looks like.

“Creating a firestorm” often means a lot of angry tweets popped up. “Producing a furious backlash” means a tweet expressing some view or take was “ratioed” (more responses than likes or retweets). This is akin to creating a polling sample consisting solely of campaign staffers (who also, I imagine, are over-represented on political Twitter).

Hence, we get the never-ending shock of political reporters that former vice president Joe Biden is a front-runner or co-front-runner in many polls. The over-interpretation of Twitter also explains why the political press tends to overestimate the impact of debate performances. “Everybody thought he was fabulous!” essentially means a lot of the self-curated segment of 2.2 percent of Americans tweeted approval. If the public seems “slow” to acclimate to new facts or shift positions, it is only because Twitter is the land of snap decisions and instantaneous, deeply opinionated takes. The public is not “slow”; that is reality. Twitter is operating in political hyperspace.

Campaigns and the national political media that cover them should put down the phones (or at least close the Twitter app) for much of the day and observe politics the old-fashioned way — follow quality polling, talk to voters, observe attendance at candidate events and watch local news coverage in early primary states.