Turnout, registration, and what a margin measures

Turnout is a ratio. The numerator is a count of people who voted, or of ballots counted. The denominator is whatever group the author chose: the voting-age population, the citizen voting-age population, registered voters, or active registered voters. Change the denominator and the percentage changes, even when the ballots do not. The Census Bureau’s voting and registration figures come from a household survey. Respondents say whether they voted. Some people say they voted when they did not. A state canvass counts ballots and, separately, may report how many people were registered. Those are different instruments. Quoting one and comparing it with the other, without saying so, is a common way to invent a trend.

A margin is simpler and still easy to abuse. It is the gap between two candidates, in votes or in percentage points, in one contest, in one election. A two-point margin in a governor’s race says nothing automatic about a House district inside that state. A presidential margin from 2024 says nothing automatic about a Senate race in 2026. The people who turn out are not a fixed list. Campaigns spend money to change it. Weather, a ballot measure, and a competitive race at the top of the ticket change it too. The past margin is still worth reading, because it is the record of what happened. It is not a prior that this site plugs into a model.

Where a past figure appears on a race preview, it is tied to a canvass or to a compilation such as the Federal Election Commission’s election-results series. If this site has not copied the figure from that source, the preview says so and links the office that publishes the canvass. Leaving a number out is a choice against invention. It is not a claim that the election was unremarkable.

Registration is not turnout. A state can add registered voters and see a lower share of them vote, or the reverse. Deadlines, automatic registration, and same-day registration change the denominator close to the election. Comparing a registration total from September with a turnout percentage from November without reading the definitions is how two honest charts start an argument. The voting-deadlines page is about dates. It is not a registration total.

On election night, “percent reporting” is not turnout. It is a statement about how much of the count is in, and the statement is only as good as the definition on the official file. High turnout does not pick a winner. It changes how many ballots the canvass must include. The pages on mail ballots, provisional ballots, and certification describe that work. The winner is the candidate the certificate names, after that work, under the rules for that office.