The Fear of Losing a Primary Reshapes How Congress Votes

Dan McKee had been governor of Rhode Island for five years, had no scandal attached to his name, and had done the ordinary competent things governors do — balanced budgets, cut ribbons, avoided the papers. Then he became the first sitting governor in eight years to lose his own party's primary. Nobody indicted him. Nobody caught him in a lie. His own voters simply decided they were finished with him, which is a stranger and harder thing to explain than a scandal.

He was not alone. Nine House incumbents lost their party primaries this cycle — six Democrats, three Republicans, and by one count the tally has since climbed to ten. State legislatures saw 164 incumbents fall in the 43 states that have voted so far, a number that touched 40 states and both parties almost evenly: 124 Republican losses to 50 Democratic ones nationwide, once statewide races are folded in. County attorney races, the most local office on most ballots, produced a 62.5 percent incumbent loss rate — five of eight seats changed hands before a single general-election vote was cast.

The strongest response to all this is also the most boring, which is usually a sign it deserves to be taken seriously. Incumbency remains one of the most durable advantages in American politics; 96.4 percent of state legislative incumbents who sought re-election this year still won their primaries. The 2016 cycle saw 198 legislative incumbents lose — more than this year's 174 — and nobody wrote that year's obituary for representative government. Anti-incumbent sentiment tracks perceptions of bad times more than any fixed institutional rot, and those perceptions are famously colored by partisanship: a voter's sense that "things have gotten worse" depends heavily on whether her party holds power. Local scandals, local redistricting quirks, a bad map drawn by the other side's legislature — these explain most individual losses far better than any grand theory of a system in crisis.

Except 2026 was not a redistricting year. The maps that decided these primaries were the same maps that protected these same incumbents two years ago. That excuse, available in 2022 and usable again after 2030, sits idle this cycle. And the losses did not sort themselves by party the way a genuine wave election does — a punishment of the governing side for inflation, or the opposition for overreach. They ran through both parties in the same summer, at the state legislative level and the county attorney level and the U.S. House, which is not the signature of one electorate's grievance. It looks more like a shared appetite.

What the raw numbers miss is the behavior of the 96.4 percent who won. A 2018 study of congressional incumbents found that the credible threat of a primary challenge changes how members vote long before any challenger files paperwork — members "continually adapt to the possibility" of being primaried, adjusting positions and votes to close off the opening. Chuck Schumer, of all people, put the mechanism plainly: primaries are decided by "the vast majority of Americans" staying home and "the 10 percent at each of the two extremes" showing up, which means a member's real constituency on any contested vote is rarely the district. It is the sliver that turns out in June.

That is the finding worth sitting with, more than the loss tallies themselves. Congress does not need a wave of incumbents actually defeated to be reshaped by anti-incumbent sentiment. It needs incumbents who believe they could be. Fear is cheaper than defeat and works just as well on behavior — a legislature full of members quietly recalculating every appropriations vote against a primary electorate that skews to the loudest ten percent on each side is a legislature that has already outsourced its judgment before the roll is called.

I grew up watching a congressman from my part of Ohio survive one bad primary scare and never cast an uncomfortable vote again — not on entitlement arithmetic, not on a base closure, not on anything his district's most engaged voters had opinions about, whether or not those opinions matched the district's interests. He kept his seat for another decade. Nobody ever had to name what changed. That is the part a loss column can't capture: the votes never taken, the amendments never offered, the retirement announced instead of the fight picked.

None of this requires apocalypse. The republic survived 198 primary losses in 2016 and will survive 174 more this year. But an institution built on the assumption that representatives answer to their whole constituency does not need to be destroyed to stop functioning as designed — it only needs enough of its members correctly calculating that the whole constituency isn't who's watching. McKee lost an election. The other 96.4 percent may simply have learned the lesson he taught them for free.