The Party Nobody Loves

A political scientist went and checked.

Alex Tolkin, a postdoctoral researcher at Duke, followed 4,941 people across eight surveys between 2016 and 2022 and pulled out the 1,293 who actually moved — partisan to independent, or independent back to partisan. That design matters. Most research on independents compares different people at one moment, which tells you what independents look like but never what happened to them. Tolkin watched the same humans change.

If the coat theory were right, the things predicting the switch would be social. Feeling politics had gotten less civil. Being less willing to talk to the other side. Neither moved — no significant effect, reported in Public Opinion Quarterly.

It's worth being precise about what that does and doesn't establish, because the rest of this leans on it. Both measures are self-reports, and someone actively managing their political presentation is exactly the person least likely to tell a survey that they manage their political presentation. Using survey answers to rule out a theory about how people answer surveys is a genuine limitation, not a technicality. What the nulls license is modest: the obvious social proxies didn't do the work. They don't retire the stigma story. They leave it standing there without the evidence it always assumed it had.

What did move was simpler and much less flattering. People stopped liking their own party. Declining favorability toward your own side was the strongest predictor of the switch, with policy disagreement contributing as well — more strongly among Republicans, which will matter later.

You could object that party identification and party warmth are nearly the same construct measured twice, and you'd be half right. They are separate instruments, and the finding lives in the gap between them: the thermometer falls first, the label comes off after. Strictly, that's a sequence rather than a cause. But a sequence is exactly what you need in order to tell hiding from leaving. Nobody was hiding. They were leaving.

That's not a retreat from the noise. That's a breakup.

And here is where it gets uncomfortable, because you already know how breakups actually work. You don't leave in one motion. You leave in your head first, sometimes years first — the identity revokes its own membership while your body keeps showing up. The job you'd already quit internally for eighteen months before you told anyone. The church you stopped believing in but kept driving to. The friendship you'd already ended in the private ledger where you keep score, while still texting back within the hour.

The gap between when you stop believing and when you stop behaving is where most of a life gets lived.

Political science has known the behavioral half of this since 1992, when Keith and colleagues called it The Myth of the Independent Voter: the large majority of self-described independents lean reliably toward one party and vote almost exactly like the partisans they refuse to be called. Tolkin's study doesn't measure that. It measures why the label comes off.

Setting the two beside each other is an inference, not a finding, and it should be carried that way. Thirty years separate them, and the years in between contain the sharpest documented rise in partisan animosity on record — the leaners of 1992 may not be the leaners of 2022, and nothing in Tolkin's data settles it. But if the old pattern still mostly holds, the picture is the one neither side of the argument wanted: the frustration is real and the exit is mostly symbolic. People genuinely done with their party and still voting for it. Not hypocrisy so much as the ordinary lag between a self-concept and a set of hands.

Notice what "independent" describes. It doesn't name a position. It names a vacancy — a thing you left, not a thing you found. It's the political version of "I'm just not that into labels," which is a sentence people say when the old label stopped explaining them and no new one has shown up. A negative identity is a real identity, but it's rented. It only holds meaning as long as the thing you're not is still standing there for you to not be.

This is the mechanism worth naming. A label works when it does two jobs at once: it signals you to others, and it makes sense of you to yourself. When those come apart — when the signal keeps broadcasting a self you no longer recognize — people drop the signal first, because that's the cheap part. Dropping the behavior costs something. It costs the group chat. It costs being the reliable one at Thanksgiving. It costs the small daily comfort of knowing who your people are before you've read the article.

Which brings back the thread left hanging earlier: policy disagreement predicted the switch more strongly on the Republican side. That asymmetry isn't a footnote to a symmetric story about Americans drifting politely apart. It's the part that leads somewhere worse.

An organization that retains near-total behavioral compliance while losing its affective legitimacy is not a party with disappointed members. It's a shell. And a shell is cheap to capture. Taking over an institution whose members have already left in their heads costs a fraction of what it costs to take one they still love: the defenders are ambivalent, the loyalists are performing, and the people who would have fought hardest for it are the ones who stopped believing first. On that reading the record independent count isn't a mood. It's an unlocked door — and the side where disaffection runs on policy rather than manners is the side where the lock came off first.

The exit stays symbolic, meanwhile, not because people are weak but because there is nowhere to put it. First past the post converts any third option into a donation to the option you like least. In states with closed primaries, registering as an independent means forfeiting a vote in the only election that is usually competitive. The vacancy — a thing you left, not a thing you found — isn't only a psychological condition. It's an accurate description of the ballot.

So the frustration sits there, undischarged, which is what a system with no channel for it should be expected to produce. It's arguably the most honest political data available right now, precisely because it's the one signal nobody was managing for effect. And asking the people holding it to have better character about it is asking them to solve a design problem with a personality.

The question isn't what the independents should do with their frustration. It's what would have to be true for that frustration to be spendable at all — open primaries, ranked ballots, anything that makes leaving cost the party something. Until then they aren't undecided. They're unaddressed. And a party that has stopped needing to be loved is a party that has stopped needing to be good.