Same Neighborhood, Same Industry, Same Job Title — and a Republican Still Has More Republican Coworkers

Go down a coal mine and roughly seven in ten of the people beside you will be Republicans. Walk into a museum and the ratio flips almost exactly: about three quarters of the staff around you lean Democratic. Nobody wrote that into a job posting. No hiring manager asked. And yet the pattern holds, over and over, across 366 metropolitan areas and 31 million American workers.

That number is not a typo. Justin Frake at the University of Michigan’s Ross School of Business, working with Reuben Hurst and Max Kagan, has built what is by some margin the largest picture yet of who Americans actually sit next to at work, and it turns out we are sorting ourselves by party almost as thoroughly as we sort ourselves by gender.

The trick was in the plumbing. Vote choice is secret in the United States but party registration is not, so the team took a national voter file covering roughly 185 million people and welded it onto a database of about 103 million employment profiles scraped from LinkedIn. Matching the two is hard, because there is no shared identifier, nothing so convenient as a date of birth, and people often go by a nickname professionally that bears little resemblance to the name on their registration. The researchers threw two methods at it: classical probabilistic record linkage, plus an embedding-based large language model pass to catch the Bobs who register as Robert. What came out the other end was 45.3 million matched workers, of which 31 million made the final cut.

The average Democrat’s coworkers are 11.7 percentage points more Democratic than the average Republican’s coworkers.

You might reasonably shrug at that. Of course a nurse in Portland has different colleagues from a driller in Odessa; that is geography and industry doing the work, not politics. So the team stripped those out, comparing only workers who live in the same census tract, work in the same industry, and hold the same occupation. The gap shrinks, but it doesn’t vanish: about 2.9 percentage points survives, which is almost exactly the same figure you get when you run the identical analysis on gender (2.8 pp).

The People Doing the Sorting

Who’s driving it? Broadly, the people who care most about politics and have the most freedom to act on it. Political donors are sorted about 27 per cent more than non-donors. Senior executives sit near the top of the distribution as well, at 14.7 pp, which invites an uncomfortable question about whether they are choosing their colleagues or their colleagues are choosing them.

Frake is careful about what this shows, but not especially coy. “For a lot of Americans, work is one of the few places left where they regularly cross paths with people who vote differently,” he said. “What surprised us is how much even that has sorted by party.”

The reason that matters is buried in some fairly old social psychology. Gordon Allport’s contact hypothesis holds that prejudice between groups tends to soften under specific conditions: roughly equal status, a shared goal, cooperation, and the backing of an authority figure. That list describes a workplace almost exactly. It does not describe a neighbourhood, where you and the man across the street can wave for fifteen years and never once cooperate on anything. So if the office is one of the last venues in American life where a Democrat and a Republican have to actually build something together, and if the office is sorting too, that is a wall developing a crack. Politics is not a protected class under federal law or in most states, so unlike race or gender this sorting faces almost no legal guardrails at all.

Two caveats: The sample skews white collar and college educated, because that’s who puts a profile on LinkedIn; the true nationwide figure could sit lower. And the study is descriptive, not causal.

Four Suspects, No Verdict

The paper cannot tell you whether workers are choosing copartisan employers, or employers are screening for copartisan hires, or political minorities are being converted by their colleagues, or simply leaving. All four would produce the same fingerprint in the data, and they are almost certainly all happening at once.

There is one counterintuitive result tucked in here, and it cuts against the standard narrative about polarisation. Republicans get far more exposure to the other side than Democrats do. The typical Republican’s workplace is about half Democratic; the typical Democrat’s is only about a third Republican. Partly that’s arithmetic, since Democrats make up a larger share of the employed workforce, but segregation then magnifies the imbalance. Cross-party contact, in other words, is not just scarce; it is unevenly rationed. And the big knowledge-economy metros come off surprisingly well, with all twenty of the largest MSAs showing less workplace segregation than their residential patterns alone would predict. San Jose, Seattle, Boston, Austin. The office, it seems, still mixes people that the neighbourhood does not.

The other half of Frake’s work makes the stakes plainer. In a separate study linking employment records to turnout for 28 million Americans, the team found that workplaces have their own civic cultures: when your coworkers vote at higher rates, you become more likely to vote yourself, an effect bigger than many door-knocking campaigns and strongest among people who normally stay home. It runs downhill too. Land in a low-turnout office and your own odds of voting fall. “Taken together, the findings suggest the workplace is both one of the last places Americans meet the other party and a force shaping whether they take part in democratic life at all,” Frake said.

What nobody has yet shown is the thing everyone wants to know: whether a politically mixed office actually softens anyone’s contempt for the other side, or whether the two camps simply agree not to discuss it and go back to the spreadsheet. Segregation has barely budged since 2012, roughly 11 points then and roughly 11 points now, which suggests this is structural rather than a passing spasm of Trump-era feeling.

Frequently Asked Questions