How can it be decided that a candidate has won a state even though not all the votes have been reported?

Question

I am currently watching the 2024 USA election right now, and am currently noticing something strange: Even though not all of the votes have been reported (in really any state), it is still apparently really clear who has won said state enough to award that candidate electoral votes.

For example, at 8:42pm CST, with Wyoming:

^{Description: At 8:42 pm CST, Wyoming reporting 0% of the votes, however Donald Trump has won the state with - votes and - % of the vote total, compared to Kamala with - votes and - % of the vote total}

And from another situation I found odd at 8:10 pm CST:

^{Description: At 8:10 pm CST, it appears that Kamala Harris had won Maryland (which had reported only 1% of the votes) with only 35.9% of the votes, even though Trump at the time led by 25.3%.}

My question: How is this possible to determine the winner of a state this early? Personally I would expect a lot more of the votes to be counted (maybe ~95% of the total) to be able to determine a winner in a state.

Answer 1

The state isn't won, it's just called. This means that although not all the votes have been counted, the news organization (the Associated Press in this case) thinks it's sufficiently clear that ___ has won the state. In other words, if Trump were to win Maryland (which is still possible), then the Associated Press would be ready to metaphorically eat its hat.

Different news organizations call at different points, as you can see if you search up other coverage of the election (e.g. Bloomberg).

Answer 2

For instance

How the AP called Pennsylvania

Trump was leading by 175,000 votes at the time the race was called. Trump cut into Harris’ margins in Philadelphia and its suburbs and expanded his share of the vote elsewhere in the state compared to his performance against Biden four years ago. Harris underperformed Biden’s 2020 performance in almost every county.

Alas I can't tell you how AP called MD that early. Presumably more based on historical data than anything else. They have some general description of the process, but it wasn't enlightening in the latter (MD) respect.

And yeah, it's going to vary between news organizations how they do this. Fox News was among the first (if not the first) to call PA for Trump about an hour before that. They have a pretty long piece but that (alas) explains even less the mechanics of how they made that call.

Answer 3

How is this possible to determine the winner of a state this early?

Those are projected (or predicted) winners, but not the official winners. There's a big difference. News organizations make those calls of projected winners. Due to competition between news organizations, they want to make those predictions as early as possible (but without getting embarrassed). There have been some famous cases of news organizations getting embarrassed by making those calls prematurely. The only consequences are having to apologize in the case those early calls were incorrect.

There is nothing official regarding those news organization predictions. The Constitution gives the 50 states and the District of Columbia a good amount of time to declare their results official. With regard to electing the President, the Constitution merely mandates that the states and the District meet on the first Monday after the second Wednesday of December in their state capitol and send a Certificate of Ascertainment that must be received by the Archivist of the United States by the fourth Wednesday in December.

Even then, it's not official. The official election of the President happens on January 6, which is when the electoral college results are officially counted by Congress. Given modern society's compulsive need for instantaneous results, no one wants to wait until January 6 (which is about two months after the election) to know who will be the next President.

That's where news organizations come in. On Election Day, voting precincts report preliminary results along with estimates of how many votes remain uncounted. News organizations pounce on these. News organizations will call an election if the difference between votes for the leading candidate the votes for second place candidate exceed the outstanding uncounted votes. (They typically call it before that.) It doesn't matter that the preliminary results are just that, preliminary. It doesn't matter that the estimated outstanding votes are just that, estimates. Occasionally news organizations do get eggs on their faces. But they get more egg on their faces when a competing news organization make what eventually turns out to be a correct call prior to the news organization that waited too long.

Answer 4

Others have already answered the "called" vs "actually won" aspect. I'll just make a related mathematical argument, which demonstrates that it is not necessary for all votes to be counted to declare an election.

If A and B are the top two candidates in an election, then A has won the election if:

uncounted_votes < votes(A) - votes(B)

That is because, even if all uncounted votes went to candidate B, their total count votes(B) + uncounted_votes is still less than votes(A), which mean A will win the election in all cases.

(Note: This only holds for FPTP system)

Answer 5

Technically, we won't know until the State as counted the votes, certified the election, nominated its Electors, the Electors have cast their vote, the State has sent the votes off to the Archivist, the Archivist has delivered the votes to Congress, and Congress has counted the votes.

That will happen on January, 6^th 2025.

However, it is very likely we won't need to wait that long. There are many different degrees of certainty with which the results in a State can be called long before January, 6^th 2025.

Absolute certainty

All ballots have been counted.

If all ballots have been counted and one candidate has more votes than the other, it is generally safe to assume that this candidate has "won" the State. There are, of course, always exceptions, e.g., the candidate could die the next morning, Faithless Electors could vote for someone other than the winning candidate, etc. But typically, these concerns are ignored.

Mathematical certainty

One candidate leads by more votes than there are ballots left to count.

If candidate A leads by 100 votes and there are only 50 ballots left to count, then there is no way that candidate A can lose the election anymore. Even if all the uncounted ballots are votes for candidate B, candidate A will still win by 50 votes.

Statistical likelihood (short-term)

It is statistically unlikely that the uncounted ballots will change the outcome.

Even if there are more outstanding ballots than the difference in votes between the two candidates, generally, an assumption can be made about the distribution of votes in the uncounted ballots.

As a first approximation, it can be assumed that the distribution of votes in the uncounted ballots is roughly similar to the distribution of votes in the already counted ballots. If the margins are big enough, that is good enough. However, if the margins are smaller, that is not a safe assumption to make:

• Smaller counties will sometimes close their polls early and thus start their count early. However, the smaller counties tend to be the rural ones, which tend to lean Republican, whereas the bigger counties tend to be the urban ones, which tend to lean Democrat. So, it is likely that a lot of the votes that are counted early will be Republican whereas the uncounted ones will be Democrat.
• Smaller counties will be able to count their ballots faster than larger counties. Same as above.
• Absentee ballots are more complex to process because the eligibility of the voter is harder to verify than if they were voting in person. Therefore, absentee ballots tend to get counted later than in-person ballots. (In same States, absentee ballots can even only be counted after the polls are closed.) In the last few elections, the Republican candidate has urged their supporters to vote in-person, whereas the Democratic candidate has urged their supporters to vote as early as possible, including by mail. Therefore, it is likely that the absentee ballots will heavily lean Democrat – ballots by active service members (who lean Republican) stationed overseas notwithstanding.

It is perfectly possible to integrate all these points (and more) into a statistical model and compute the confidence level of a prediction. In that case, it is then merely a question of which confidence level a news outlet deems acceptable – some may be willing to accept higher risks than others.

Statistical likelihood (long-term)

However, none of the above actually explains the examples in the OP, where only 1% of ballots were counted, or in the second example, even no ballots at all. What is going on here is a different kind of statistical prediction.

Similar to the above, we can apply statistics not just to the current election, but to trends as well.

The second example in the OP is Wyoming, which was called with no ballots counted. However, Wyoming already voted 70% for Donald Trump in 2020. And in 68% in 2016. And 68% for Mitt Romney in 2012. 65% for McCain in 2008. 68% for Bush in 2000 and 2004. Even Bob Dole pulled in almost 50% in 1996, and that was with a strong third-party candidate. Old Man Bush won with 40% in 1992 (again with a strong third-party candidate) and 60% in 1988. Reagan won with 62% and 70% in 1980 and 1984. And the list goes on and on. Even John F. Kennedy lost to Richard Nixon in 1960.

The last time a Democrat won Wyoming was 1964, the last time before that was Truman in 1948. Given that Wyoming has been staunchly Republican for 60 years, and specifically Donald Trump had already won a landslide victory in the last two elections, there is simply no reason to assume that 2024 would be any different. Donald Trump had more than three times the number of votes as Joe Biden in 2020; even if, by some miracle, Kamala Harris managed to sway two thirds of Donald Trump's voter base to her side, she would still lose.

So, based on this data, it is pretty much 100% safe to call Wyoming for Donald Trump without a single ballot being counted. In fact, Wyoming is so safe for Donald Trump that well over half the people who voted for Trump in 2020 didn't even bother to show up to vote … and he still won by over 40%.

Votes aren't everything

There's more information than just votes.

Last but not least, votes aren't the only datapoint there is. Exit polls at the polling stations as well as polls conducted before the election can be used as well.

Answer 6

I just wanted to expand on the statistical arguments in Jörg's anser.

There is a lot of pre-election polls and historical knowledge of votes down to the precinct level. Now let's assume that a candidate needs to do better than predicted in order to win a state. The first results come in, and they are disappointing: The candidate got even fewer votes than in previous elections and pre-election polls, which already predicted a loss; the remaining precincts would have to vote even more favorable — and more outside predictions! — to bring them the win. After very few additional underperforming precincts the likelihood of all the remaining precincts to bring spectacular surprises, against not only predictions but also against the trend observed in the precincts already counted, is negligible, and the state can be called.

Obviously, in narrow races this call will need more actual evidence than in states that would have needed a spectacular, unpredicted swing to begin with. In some partisan states the pre-election polls or even common knowledge are sufficient to call the race before the first vote is counted.

But the principle for less clear states is: Counted precincts have, after proper statistical evaluation concerning the constituency, a very strong predictive power for the remaining precincts.