Elections: How do they project a winner with just a small % of districts reporting?

I’m sorry if this has been previously addressed, but I didn’t see it in a search.

Driving home tonight, I heard that Clinton was projected the winner in KY with just 15% reporting. How do they do this? Admittedly, it’s not a surprise, but isn’t it some sort of mathematical possibility to know how 100% of the districts will vote when 85% haven’t been counted?

Further, 15% actually sounds like a lot of counted votes when compared to other races I’ve seen. I seem to recall recently that Jon Stewart made a joke when he pointed out that a winner was projected with 0% reporting. How does this work, exactly?

Jon Stewart’s quip was probably just a joke.

But, let’s say that districts #1,2,3,4 and 5 have voted in the same way as the state’s winner in all previous elections. Then if Candidate A wins in those districts, even though they might be only a small fraction of the whole, the projections will declare that candidate a winner, expecting history to repeat itself.

Fraught with danger, obviously, as all extrapolation. But the number of times they are right allows them to get the jump on the news competition. And when they’re wrong, you get a pix of a jubilant Truman holding a newspaper saying Dewey wins.

I don’t have an answer but Isaac Asimov wrote Franchise which took the concept to it’s ultimate conclusion.

From Wikipedia:
In the future, the United States has converted to an “electronic democracy” where the computer Multivac selects a single person to answer a number of questions. Multivac will then use the answers and other data to determine what the results of an election would be, avoiding the need for an actual election to be held.

Statistics are very powerful and counter-intuitive to most people. It only takes a very small sample size to figure out the outcome of a race. The key is to figure out the right sample and that can be hard. As long as you do that however, a sample size of say, 1000 people could almost always determine the outcome of the presidential votes for all 300,000,000 people in the U.S. The absolute number of the people in the sample doesn’t really matter no matter what the population size as long as the sample gives an accurate representation of the people in the population. As long as those 15% were evenly distributed among Kentucky voters then they already know the outcome.

Not just a joke. They were showing actual news show footage which included projected winners with zero percent of the vote counted. Colbert’s pointed this out before as well.


These predictions are not based on actual votes at all, but on the results of exit polling.

West Virginia was indeed called for Clinton with 0 votes counted. With a 40 percentage point margin over Obama, the exit polling would have made it obvious that the state was a landslide for her.

It’s only when the polling is close that the news media hold off and wait until sufficient votes come in from certain precincts preselected for their usefulness in gauging outcomes.

Statistical techniques are so advanced these days that no race is ever called wrong, except for the very few that become newsworthy because the actual vote difference is so small that even a one percent error bar can swamp the reality.

I don’t think they project it based on the reported results. They project it based on polling done at the voting stations.

Pretty much, some elections are no contest. This was one of them. The news would be that HRC lost. Or that Obama lost Ore.

Thank you all for responding. After I wrote the question, it dawned on me that the announcement may be based on exit polls. So, although only 15% of districts are reporting official results, exit polls, from across the entire State, show that HRC won KY (for example).

It seems to me, though, that it would make more sense if they explained that reasoning, instead of just leaving ignoramouses like me scratching our heads.

Often they do, when they first call the race. Invariably, if a race is called the instant the polls close, it’s based on exit polls. If it can’t be called based on exit polls, the networks will have to wait until a reasonable sample of the vote is in, although that sample can be fairly small. Of course, if you tune in late and see that the race has already been called, you have no way of knowing whether it was based on exit polls, actual results, or a combination.

Of course, for very close elections, even 100% sampling might not be enough (see Florida)


In California, for example, we had a special election with several ballot measures on it, as proposed primarily by the Republican Governor. Republicans were more likely to vote YES and Dems, NO.

I was watching the results as they came in to the Secretary of State website. After a small percentage of the vote was in, I could see which counties were voting which way.

The vote tally was about 50-50 on several of the measures. However, I could see that NONE of the precincts reporting were from the heavily Democratic Los Angeles and San Francisco areas.

Thus, if the vote is about 50-50 on a ballot measure with only a few precincts reporting, and all the precincts reporting are from the Republican counties, it was easy to predict that once the juggernaut LA and SF Dem votes are considered, the ballot measures would go down in flames.