How can election winners of states be confirmed, although the remaining uncounted votes are more than the difference in votes?

Question

For the US presidential elections, the Associated Press announced winners for several states although a significant portion of the votes were uncounted.

One example is Illinois where at the moment of writing (12:32 Central European Time) 80% of the votes are counted

• Mr. Biden received 55,1% or 2.881.554 votes
• Mr. Trump received 43% or 2.246.472 votes

that sums up to 5.346.276‬ votes. If 80% of the votes are counted that would make that the total amount of votes is 5.346.276‬/0.8=6.682.845.

That means that 6.682.845 - 5.346.276 = 1.336.569 votes are uncounted. Which means that theoretically Mr. Trump could still get 2.246.472 + 1.336.569 = 3.583.041 votes. And thus could still win the state, although Mr. Biden was already announced the winner.

Although this is statistically unlikely it would not be impossible. And this kind of bothers me, because I took an extreme example but one could imagine less extreme cases where the results could really change.

It would seem reasonable that only if the total number of remaining/uncounted votes is less than the difference one could announce a winner, but apparently they use much more flexible criteria. What are the limits of probability applied when prematurely announcing winners before all votes are counted?

Note that this question is not about the particular example, but about the general principle of officially announcing a winner while theoretically the other candidate could still win.

EDIT based on the given answers the following part is not correct.

Also one could consider possibilities of "legal" "fraud" where some of the votes, would just be counted later. Say During manual counting 10% of votes for Candidate A are put aside for later counting, or votes from particular regions with particular preferences are counted later. In such scenarios without really changing any vote from A to B election results could still be influenced just by postponing a selection of the votes.

EDIT My previous understanding was that the states were called before all votes were counted and that they would not continue to count the remaining votes. But based on the answers this information was only a projection so all votes will still be counted, and there is no fraud except for the influence that preliminary results might have on the voting behavior in states where voting was still open, due to the time difference.

Answer 1

First things first: The projections made by the Associated Press, or by any other news organization, are not official. Those unofficial projections do not determine the makeup of the Electoral College. In Illinois, it is the State Board of Elections that proclaims the official results of elections for presidential electors. Other states do it differently, but it is always a state official / set of state officials who make the official proclamation, and it is usually days after the election. (It can take up to 21 days in Illinois.)

What those unofficial projections made by news organizations can do, when they are wrong, is to reduce the standing of the organization that made the erroneous projection. The AP does not want a repeat of the 2000 election when they projected that Albert Gore would win Florida, but soon thereafter retracted that projection.

That said, projecting Illinois a bit early was easy. The AP, along with other organizations, look not only at the outstanding vote, but also at where in the state those outstanding votes are. Illinois is a reliably blue (Democratic voting) state, and the outstanding votes were in the bluest parts of Illinois. Mr. Trump was never going to make up that 600000 vote deficit in Illinois. Similarly, Mr. Trump was projected to be the winner in Indiana and Kentucky well before the votes in those states were anywhere close to final. Mr. Biden could in theory have come from behind in Indiana and Kentucky. In practice, that would not have happened.

In fact, the Associated Press called the election in several states including Illinois and Indiana, as soon as polls closed in those states. In other words, the AP called the election in some states with zero precincts reporting. Suppose their projection (a guess) in one of those states went the other way. The projections have no official bearing. The projections gone wrong with zero precincts reporting would have only served to make the AP look bad.

Answer 2

One example is Illinois where at the moment of writing (12:32 European time ) 80% of the votes are counted

The 80% is not the percentage of votes, but the percentage of polling places that have reported vote totals. The number of polling places is known (fixed) at the start of the day.

The total number of votes will not, and can not, be known until all polling places have reported results and all mail in / absentee votes have been counted.

See, COUNTING THE VOTE

Q: Does AP use "precincts reporting"?

A: In all races tabulated by AP, we provide details on precincts reporting. This figure is the percentage of precincts in a race from which AP has received and is reporting results. In many states, this figure is a reasonably accurate estimate of the amount of vote counted, particularly in those states that do not separately report results for ballots cast in advance from those cast on Election Day.

Answer 3

Calling an election is like calling a sports game. You take a lot of things into account when calling a game, and deciding that one team has no chance to win.

If it’s halftime, and the 2007 Patriots (NFL) are up 32-3, I’d feel comfortable turning that game off and watching something more competitive. Same goes for Brazil up 2-0 after five minutes in to a game against the Trinidad team in soccer (football). Sure, there’s still plenty of time in the game, but are they really going to win?

Of course, it the US is up 3-1 over Mexico in soccer, odds are they will win, but I wouldn’t call the game. And if the 2013 Lions go up 14-3 over the Packers, I probably would still consider them underdogs to win.

That’s the same thing that’s going on here. The election’s not called in any of these states by anything official - rather, these are the equivalent of sportscasters deciding that one person is going to win and telling you to turn to a different game. Their statistical models say that at this point it’s exceedingly unlikely to for the outcome to be different.