What Both Parties Should Take Into the 2028 Cycle

Speaker Mike Johnson will not run to lead House Republicans from the minority. Chuck Schumer is set to become Senate majority leader. Inside three weeks both parties will have settled on a story about why 11/03/2026 happened, and the story is what the next two years of budgets get built on.

The ground everybody is standing on is clear enough. Democrats won the House and the Senate. Republicans went in holding the House 219 to 214 with two vacancies and 53 Senate seats. Seven Republican held Senate seats went: Kansas, Iowa, Texas, Ohio, Maine, North Carolina and South Carolina. It is the first time in twenty years that a Republican president has lost both chambers.

The argument about why has already started. JD Vance said the war in Iran was the obvious reason, and the President called that largely true while pointing at a late economic message. PBS tied the result to prices, the war and corruption. Those explanations are not variations on a theme. They imply different spending for two years.

Picking an explanation before testing it

Three readings are circulating and they are genuinely different claims.

The first says conditions moved against the party in power for reasons specific to this year, the war above all, and that conditions will move again. The second says prices and the economy did it, and the message was wrong. The third says the map was never where either party's ratings said it was, and an ordinary shift landed on seats nobody was defending.

Those lead to three different plans. The first says wait and keep your powder dry. The second says rebuild the economic argument. The third says fix the measurement, because every dollar in the cycle was allocated against a list that was wrong before the first ad ran.

They are testable against each other, and the tests are not exotic. A uniform environment effect shows up as a roughly constant swing across states regardless of local circumstance. A candidate or campaign effect shows up as a gap between the Senate result and the other statewide races on the same ballot in the same state. A composition change shows up only in the voter file, and it is the slowest to confirm and the longest lasting. Running all three comparisons takes a few weeks and settles an argument that otherwise gets settled by whoever talks loudest in November.

For the party that won: what comes attached to a wide majority

Three problems come with the result. None of them are grounds for complaint. They are the arithmetic of winning broadly.

Geography. Seats won where a party is not normally competitive have to be defended where that party is not normally competitive. Kansas had not elected a Democratic senator in almost a century and South Carolina had not since 1998. Those seats were taken once, under midterm conditions, and they come up again under conditions nobody has measured.

Turnout mix. Midterm and presidential electorates differ in size and composition. A majority assembled in a midterm does not transfer automatically to a presidential year in either direction, and 2026 says very little about 2028 without further work.

Expectations. A result this wide sets a reference point, and every later result gets measured against it, including results that would look strong against any other benchmark. That is a communications problem and a recruitment problem at the same time, because recruitment responds to perceived chances and perceived chances anchored on a peak are usually wrong.

The task is to work out which of the newly won seats are structurally competitive going forward and which were a single cycle, then to be honest internally about the difference. Treating all of them as permanent is the standard route from a wide majority to a narrow one.

For the party that lost: diagnose before you respond

Seven Senate seats across four regions is not a pattern that local explanations can carry, and the House losses ran from California's 22nd to Pennsylvania's 1st to Nebraska's 2nd. The sequence that produces a useful answer is unglamorous. Five steps.

Audit the ratings before the campaigns. Write down what the pre election target and defense lists were, before anybody revises them, and set them against what turned out to be competitive. If seats came off the board that were never on the list, the failure happened in measurement, upstream of every tactical decision. That is cheaper to fix than a persuasion problem. It is also the step most likely to get skipped, because it implicates whoever is running the review.

On election morning Cook still carried six Senate toss-ups and Sabato's Crystal Ball carried none. Seven Republican held seats changed hands.

The ratings against the result

Senate seats either rater called a pure toss-up on election morning, set against the seats that changed hands.

Decompose the result before picking a theory. A seat changes hands through some mix of vote switching, uneven drop off among previous voters, and new participation. A called race does not say which. Precinct returns matched to the voter file do. Choose a response before that and you are guessing at which of three different problems you have.

Separate national from candidate specific. Comparing each Senate result against the other statewide contests on the same ballot in that state is a straightforward way to estimate how much was environment and how much was the particular race. North Carolina, where a former governor ran, is a good place to start.

Do not conclude that a group of voters has left for good. One midterm is one observation. Reading a durable realignment into it is as likely to be wrong as reading a temporary protest into it, and it is the more expensive error, because it talks a party out of competing somewhere.

Restart recruitment immediately. The candidates who run in 2028 are identified in 2027, and a leadership contest between Jim Jordan and Steve Scalise will absorb the winter. That calendar does not pause for an autopsy.

Four things that hold whatever the explanation turns out to be

Fund measurement where nobody has rated the race yet. The seven Senate flips included states nobody was watching closely. Both parties now have evidence that the real map is wider than their lists assumed, which makes the marginal research dollar worth more in an unrated seat than in a well covered one.

Treat research as a running series. A single reading taken in September answers a question that was already settled in March. A series running from the previous year informs the decisions that are still open.

Build the local data capacity. Thousands of contested local races will run between now and 2028. The results are public, badly formatted and mostly unread. Collecting them pays off across cycles, which is exactly why it keeps not getting done.

Write the competing explanations into the plan. Name all three, name the evidence that would kill each one, and put a date on when you will check. A 2028 plan built on a single story inherits an error that nobody will remember putting there, and 2026 was expensive enough to make that an unaffordable habit.