Gender gap? What gender gap?

by Andrew Heyward and Uzra Khan

Who’s more likely to support Donald Trump on Twitter — men, or women? And what about Hillary Clinton?

It turns out that while Clinton’s support is slightly more balanced between the genders, men dominate both candidates’ supporters on Twitter by a significant margin.

Here at the Laboratory for Social Machines, part of the MIT Media Lab, we’ve developed a collection of algorithms called Electome to analyze the social conversation about the campaign. With access to the full output of Twitter, Electome filters out tweets about the election and classifies them according to issues, candidates, candidates’ supporters, and now — gender. (More about Electome here.)

Electome’s new gender tracking algorithm sorts Twitter users into male, female, or undetermined (which can include individuals whose gender the machine can’t determine as well as — very often — organizations rather than people). It does so by analyzing different aspects of the Twitter account, including profile pictures, descriptions, and the content of tweets.

We wondered if Electome’s new filter would reveal a gender gap between Trump and Clinton supporters. As a proxy for supporters of a candidate, we use unique followers — people who talk about the election and only follow either Trump or Clinton but not the other or any other presidential candidate.

Here’s how Trump’s supporters on Twitter break down by gender through the entire month of August.

As you can see, close to twice as many men as women support Trump and tweet about it.

It’s worth noting that Trump has more unique Trump followers on Twitter than does Clinton — a ratio of about 1.4X. Trump’s unique followers also tend to tweet more than unique Clinton followers, as we’ve observed before.

So given the conventional wisdom about Trump’s and Clinton’s relative appeals, you might well expect to see a higher percentage of women among Clinton supporters.

But here’s Electome’s gender breakdown for Clinton supporters that same month: different from Trump, but by very little, and still overwhelmingly male.

And the heavy dose of testosterone does not just show up among the candidates’ most devoted supporters. Electome’s algorithms show us that Twitter users who talk about this election in general skew heavily male — 54% male to 27% female, still 2-to-1 — no matter how many (or how few) candidates they follow.

There is a gender gap on Twitter overall, but it’s nowhere near big enough to explain this phenomenon. According to the Pew Research Center’s analysis of Twitter demographics, 25% of male Internet users are on Twitter, compared to 21% of female Internet users. (Males and females use the Internet in about equal proportions, according to Pew.)

Electome’s new filter shows that while there is a large gender gap among active Trump and Clinton supporters on Twitter, it’s not between the candidates, but rather between the genders themselves, with men hogging the microphone.

Andrew Heyward is a visiting researcher at the MIT Media Lab’s Laboratory for Social Machines. Uzra Khan, a recent graduate of Harvard’s Kennedy School of Government, is a project manager there. Soroush Vosoughi and Prashanth Vijayaraghavan, researchers at the Laboratory for Social Machines, developed the analytics for this post.