Washington

Washington does not register voters by party, and its demographics tables prove it by omission: they publish county, age group, gender, congressional district, legislative district and city, and no party table exists. Every other state in this corpus supplies one — Nebraska five columns, Utah nine, Oregon twelve, California ten. Any pipeline with a required party field, any partisan universe built from registration, and any turnout model segmented by party has nothing to work with here for 5.5 million voters.

General-election turnout alternates on a forty-point cycle that has held for decades: 78.95% in 2024 and 84.14% in 2020 against 39.24% in 2025 and 36.41% in 2023. The odd years carry city, school district and special district offices — the local layer — so the offices closest to voters are chosen by the smallest electorate, consistently and predictably. A turnout assumption carried from an even year into an odd one is wrong by a factor of two.

Washington votes entirely by mail and sends a ballot to every active registered voter, so the 426,104 inactive registrations are people who receive nothing. The share varies more than threefold by county — Whitman 13.2% and Asotin 10.0% against Lincoln 4.7% and Mason 4.9% — and the high end is where student and transient populations concentrate. A contactable universe sized from total registration overstates reach by up to thirteen percent in the counties where address churn is highest.

The Census counts 1,890 local governments in Washington: 1,275 special districts, 295 school districts and 281 municipalities. Districts outnumber cities more than four to one, and together the two district tiers are 83% of all local government. These are the bodies elected in the odd years at under 40% turnout, so the largest tier of government by count is also the least observed — and modelling the local layer as cities captures under a sixth of it.

The same workbook reports the electorate by county and, separately, by congressional district, legislative district and city — four different geographies over one population. A city may span counties, a legislative district certainly does, and the tables do not reconcile to each other by simple addition. Choosing the county table and the district table and summing them together double-counts every voter in a split jurisdiction.

Each year in the published series gives registered voters, ballots counted and the rate together — 2025 as 5,100,871, 2,001,425 and 39.24%. That is unusual and it is the right shape: elsewhere in this corpus the denominator has to be reconstructed from a separate registration report filed at a different date, and the two rarely align. A consumer that stores only the percentage discards the ability to recompute it, and one that pairs the rate with a registration figure from another month gets a third number.

Washington administers 39 counties and the Census counts 39 county governments. Across this corpus the two diverge only where consolidated city-counties exist — Colorado, Montana, Nevada and California — and agree in the other six states tested. Washington is a clean case, so a Census reconciliation is safe here while the identical reconciliation is short by one or two in four neighbours, for a reason knowable before the comparison is run.

King County holds 1,451,666 active registrations of the state's 5,136,372 — 28% — and with Pierce and Snohomish the three make up half the electorate. The remaining 36 counties share the other half, and the smallest, Garfield, has 1,768 registrations in total. Any process that samples counties uniformly, allocates equal effort per county, or treats county as a comparable unit is dominated by one jurisdiction eight hundred times the size of the smallest.