California

The Report of Registration carries Eligible and Total Registered as adjacent columns for every county — 27,253,826 against 23,155,447 statewide — and prints the percentage. No other state in this corpus supplies the eligible-population layer at all. That inverts the usual problem: elsewhere the argument is over which registration figure to divide by, and here the deeper denominator exists, so a consumer that ignores it discards the only measure of who is missing rather than who turned out.

Placer registers 95.9% of its eligible population, Amador 94.6% and El Dorado 94.5%, while Trinity registers 57.4%, Kings 67.8% and Lassen 68.8%. The statewide 84.96% describes neither end. Any model that applies a statewide registration assumption to a county — to size a contactable universe, or to convert eligible population into likely voters — is wrong by up to twenty-seven points in one direction and eleven in the other.

Registration is reported as Democratic, Republican, American Independent, Green, Libertarian, Peace and Freedom, Unknown, Other and No Party Preference. Unknown and Other are distinct columns with distinct meanings, and No Party Preference is a deliberate status rather than a residual — 5,284,968 registrations, 22.8%, close behind the Republican column. A schema with two parties and an other bucket collapses four different things into one and loses the second-largest affiliation in the state.

The state's election guide sorts offices into voter-nominated — congressional, gubernatorial, statewide, senate and assembly — and nonpartisan, covering the courts and local government. Ballot access then splits three ways: write-in candidates for voter-nominated offices file only at the primary, write-in candidates for nonpartisan office at both primary and general, and independent candidates for either category only at the general. Office class and election stage together determine the route, so a single filing deadline or a single candidate model is wrong for most of the ballot.

Counties began mailing ballots on May 4 for a June 2 primary. Registration closed May 18, but conditional voter registration ran May 19 through June 2, so the roll kept growing to the end. And a ballot postmarked on or before June 2 counts if the county receives it by June 9 — a seven-day tail. So the electorate is not fixed on election day and neither is the count: results on the night are provisional against a roll that was still moving and a ballot pool still arriving.

The Census counts 4,494 local governments in California — the largest total in this corpus — of which 2,949 are special districts and 1,006 are independent school districts, against only 482 municipalities. So districts outnumber cities six to one, and school districts alone outnumber them two to one. Modelling the local layer as municipalities captures about a tenth of California's governments, and modelling schools as a single tier per county is wrong by a factor of seventeen.

California administers elections through 58 counties; the Census counts 57 county governments, because San Francisco is a consolidated city and county classified as municipal. This is the fourth instance of the same pattern in this corpus after Colorado, Montana and Nevada, and it is the largest — San Francisco holds more registered voters than eighteen states have people. A reconciliation against Census unit counts is short by exactly one here, and the missing one is not small.

Los Angeles County alone holds 5,898,798 registered voters against 6,716,770 eligible — more registrations than any state in this corpus except California itself, and more than the entire electorates of Colorado, Idaho, Utah, Nevada, New Mexico, Montana and Wyoming combined. Anything that treats counties as comparable units, allocates equal effort per county, or samples counties uniformly will be dominated by a single jurisdiction whose scale is categorically different from the other 57.