The Great Sorting · The Ballot
Americans once happily voted for a president of one party and a congressman of the other. That habit had a structure — the further down the ballot, the more people crossed over — and over fifty years the structure first widened, then collapsed into a straight party line.
Ask a committed partisan how they voted and the surprises are down the ballot. At the top of the ticket, party loyalty has long been near-total; the room for crossing over opened at the offices most Americans know least — the House seat, the second senator. Tracking how often partisans defect — cast a major-party vote for the other side — at three offices over seven decades turns the abstract idea of "nationalization" into something you can watch happen.
The shape is not a steady decline. It is a rise and a fall.
How often partisans cross over, by office
Weighted share of party identifiers (leaners included) casting a major-party vote for the other party, by office and year.
The opening of the ladder in the 1980s and 90s is the part most people misremember. It was not that Americans were loyal and then defected more; it was that incumbency and candidate reputation let a voter split — a conservative Southern Democrat pulling the lever for a Republican president while keeping a Democratic congressman they had sent to Washington for decades. Hopkins calls the reversal of this an "increasingly United States": as politics nationalized, the local cues that sustained ticket-splitting lost to the party label.1 By 2024 the House-minus-president gap, which peaked above twelve points in 1992, had shrunk to under one.
One group resisted longest: partisans living behind enemy lines. Define a partisan as "stranded" when their state's presidential lean runs more than three points against their own party — a Democrat in a red state, a Republican in a blue one. The stranded defect more than comfortable partisans at every office, and the gap is widest down the ballot.
Cross-pressured partisans cross over more — especially down-ballot
Weighted defection rate, pooled 1976–2024, by office and whether the partisan is stranded in an opposite-leaning state.
The split ticket did not fade evenly. It opened where local knowledge mattered, then closed as the party label swallowed the local.
The behavioral logic and the geographic one tell the same story from two sides. The down-ballot offices were where a voter's local attachments could outrun their national loyalty, so they were where defection ran highest — and where the stranded, with the strongest reason to split, split most. Nationalization is the erosion of exactly that margin. When the House race becomes a referendum on the president, the conservative Democrat and the suburban Republican lose the cover that let them vote two ways at once. The ladder folds into a single rung.
One honest limit frames all of this: the public CES file that powers most county-level work carries no House or Senate vote, so the office ladder cannot be drawn at county resolution. ANES supplies the vote choice but only state geography. The "stranded" here are stranded in a state, not a precinct — a coarser net that, if anything, understates how cross-pressured a blue-county Republican or red-county Democrat really is.
Data. ANES Cumulative File vote-choice items for president, senate, and house, 1952–2024, weighted; state presidential lean computed from MIT Election Lab state returns (full two-party counts, 1976–2024).
Defection. A respondent is a partisan if they identify with or lean toward a party. Defection at an office means casting a major-party vote for the other party there; the denominator is partisans who cast a major-party vote at that office. "Stranded" means the respondent's state leaned more than three points against their party in that year's presidential race.
Honest limits. The CES cumulative file has no down-ballot vote choice, which is why the analysis runs on ANES at the state level rather than the county level the topic invites. Per-office, per-year cells below 50 respondents are suppressed; senate vote is unavailable in a few off-cycle years. Pre-1976 cross-pressure is omitted for lack of comparable state-return coverage.