A continuing goal in this project was to avoid overstating how sure we can be of the result when considering only “everyone other than WEP.” Each state’s box shows the estimate: who would have won among those voters, and by how much. Because the estimate comes from a survey, it has a 95-percent range, and the box also shows the conservative end of that range: the end least favorable to the estimated winner (that is, the end where the estimated winner’s lead is narrowest, or where the other party is even ahead).
Where even the conservative end shows the estimated winner ahead, the result is settled, and the map shows whether the state flips. Where the conservative end and the estimate differ on which party is ahead, the state is shown as indeterminate (hatched).
The conservative end is rounded away from the estimate, to the next tenth of a percentage point, so that rounding never makes it look closer to the estimate than it is. South Carolina in 2012 is shown to hundredths, because at tenths its conservative end would show a tie.
In some states, the conservative end shows a larger change from the certified result than the estimate does. That is because we are being cautious about who would have won, not about how much the result would change. How small the change could be is shown in “Margin Shift,” the visualization on the following page.
In this view, every elector tile shows all of the state’s electors, in the color of the winner the map shows, which leaves out the following details:
Nebraska 2008, 2020 and 2024 (indeterminate): already one certified Democratic elector each cycle.
Maine 2016, 2020 and 2024 (no flip): already one certified Republican elector each cycle.
Washington 2016 (no flip): already four faithless electors in the certified count.
Hawaii 2016 (no flip): already one faithless elector in the certified count.
Texas 2016 (flip): already two faithless electors in the certified count; the other 36 electors move to the Democratic candidate.