WASHINGTON — Most Americans are used to sitting down in front of the TV on election night and watching news anchors go state by state, slowly but surely projecting a winner of a given election.
We're used to a giant touchscreen map of the United States with a progress bar of the percentages of precincts reporting results. But rarely, if ever, is the projected winner announced on that Tuesday night.
This year, with unprecedented voter registration and early voting turnout, the experts behind those projections are shifting their processes and preparing for their projections to be more complicated and take longer than ever before.
Gone are the days of 2000, when networks are racing to be the first to declare a winner. This year they're taking their time, and so are the statisticians and analysts behind the scenes running the scenarios.
The Verify team talked to the experts at the height of their fields, and we're breaking down what you need to know about what it takes to declare a projected winner of an election.
Before we dive in, let's make one thing clear: we won't know the real winners of the 2020 elections, from your mayor all the way up to your president, until all the votes are counted and the Electoral College finishes voting on December 14. What you might hear on Election Night and the coming days are projected winners.
- Drew McCoy, President of Decision Desk HQ, the election data collection company that has provided data and predictions to Vox, Buzzfeed, Business Insider, the Atlantic, HuffPost, Reuters, the Baltimore Sun and more.
- Larry Rosin, Co-Founder and President of Edison Research, the election data collection company for ABC, NBC, CBS and CNN.
- A spokesperson for the Associated Press
- Tammy Patrick, a Senior Elections Advisor at the Democracy Fund
What data is used to project the winner of an election?
There are three major groups who project election results: The Associated Press, Decision Desk HQ, and Edison Research.
While each of them have their own teams and their own methodologies, they tabulate their predictions in similar ways. They all use some combination of historical election results, demographics, exit polls, nationwide surveys, voter registrations, voter files and votes as they are reported by localities.
"So what you start with is a lot of homework," Drew McCoy told us. He's the President of Decision Desk HQ, the company which provides election data analyses to a number of major media outlets. He says ahead of the election they spend time understanding each race and each county.
"What you're really looking for is historical trends. Understanding how large your geography is in terms of potential votes, what its partisan leans are, and what it's done historically. You also want to know when you can expect to get votes," McCoy says. "But what you're really looking for when you're calling an election, is you're really calling a loser more than a winner."
Larry Rosin, the co-founder and president of Edison Research, agreed.
When elections officials start reporting votes, analysts are looking for the likelihood that whoever is trailing will come back ahead. They're using historical voter data combined with exit polls and surveys to make that determination.
"Let's say that the Democrat is ahead, and the Republican is down by a certain number of votes, but the overwhelming majority of outstanding voters are in Democratic parts of the state," Rosin explained. "At some point, you're gonna be able to call that race even though you don't have all the votes, because you know that person's lead is only just going to get bigger as the votes come in from the more Democratic part of the state."
Each organization has their own statistical methodology for how to make predictions. Rosin told us Edison Research does not officially project a winner of an election until they determine that the odds that a candidate will fail to catch up are higher than 99.5%.
The Associated Press declined an invitation for an interview, but a spokesperson provided us with a number of resources detailing their election projection process.
In an online blog post, they say "The AP calls a race when we conclude that the trailing candidate will not catch the leader."
McCoy tells us that even will the vast amount of data on hand, some results are impossible to predict. "Sometimes you just have to wait. You do it vote by vote, county by county, because they're too close and you don't want to make a mistake. You don't want to be premature about it."
Tammy Patrick tells us election projection is both a science and an art. While they have a great track record of getting it right, there are a million things to take into account.
"The electorate has changed somewhat, and quite a bit in this moment," Patrick said. "That's why I think it makes election administration and some of the political science behind it both very exciting and exacerbating at times."
Patrick explains that even historical voting and registration data can sometimes fail you.
"Quite frankly, you don't know, even if they're registered one way, what they're going to do when they pull that curtain and it's just the voter and that ballot," Patrick said.
What's different about 2020 compared to previous years?
The U.S. is on track to break voter turnout records. More Americans have voted early and absentee than ever before, and because the voting a tabulation process varies so much from state to state, our experts say the way they receive and report data will have to change.
One thing you might notice has changed on election night: Precincts reporting.
Rosin says all the networks he is aware of are abandoning that terminology, as it's likely not very accurate and makes the reporting of early and absentee votes confusing. Instead, they'll be reporting the percentage of expected votes. McCoy says Decision Desk HQ is doing the same.
"It was a sort of rough shorthand that people understood," McCoy explains, "But it didn't necessarily correlate. To say we have 25% of precincts reporting, that doesn't mean you have 25% of the vote in. Not all precincts are created equal. Some will have 200 voters, and some will have 5,000 voters. So it was always sort of a flawed metric."
McCoy and Rosin both told us that estimating the total voter turnout will be an ongoing process as votes are counted.
"You're constantly adjusting the factors that you would take into account and then judging that against what you're seeing in the real world," McCoy says. "And then how do you take the combination of those things and project forward again."
"That number will adjust all day long and into the night as we get reports of how many people actually voted," Rosin told us. "You know, it's an estimation game based on a lot of surveys and a lot of other indicators."
Another thing that 2020 changed about the process is exit polling.
It's estimated that at least 50% of Americans will have voted early or absentee this year, and if you're only standing outside a precinct with a clipboard on Election Day, you're missing a ton of people.
Instead of using any exit polling, the Associated Press started their own election survey system, AP VoteCast. They use a combination of mail, phone interviews and online surveys to get a better reflection of the American electorate.
Rosin told us Edison Research has always employed telephone surveys and early voting exit polls in some areas, but this year they've had to up the ante.
"It's nothing totally different from what we've done in the past. It's just really the proportions of things that are different now," Rosin says. "The proportion that will count from the early vote will be much higher, because that will be a bigger proportion of the total."
Even with these new changes and processes, they are confident in their systems and their operations.
"It's not going to be neat and contained as it normally is. And again, that's okay," McCoy assured us, "That's the process playing out. And that's what people should be watching for."
What projection mistakes have been made in the past?
The big mistake that comes to the minds of many is the 2000 George W. Bush v. Al Gore race when the major TV networks flip flopped in their projections of the winners.
"There was intense pressure on the networks to be the fastest one," said Rosin. "We called it first, you come here for the fast calls.'"
Rosin says after that election, the sentiment has changed. "There's a lot more patience built into the system and a lot more belief in getting it right."
"Everybody makes mistakes and misses calls, and nobody wants to, but we've all done it," McCoy says. He's looking at a 2018 California race, incumbent Republican David Valadao versus Democrat TJ Cox, as a cautionary tale of an early projection.
"What happened was California collects ballots for a couple days afterwards, and the Republican was leading by eight or nine points on election night. Now that's normally a safe call."
McCoy says all the major outlets projected Valadao as the winner. "As the ballots were counted over the course of a couple of weeks, it became clear that those late breaking ballots were going to be heavily Democratic and enough to actually overturn the election night lead."
The U.S. Constitution left election administration up to the states. That means there are different processes and timelines in place for each one.
While some states like Florida and Maryland started counting absentee ballots before November 3, others like New York, Pennsylvania and Wisconsin don't even start counting until election day. Projections are going to depend on how quickly those votes are counted.
When can we expect a projected winner of the U.S. Presidential race?
Right now, we don't have a way to answer this.
There are a million factors that go into making a projection, but our experts all say it's important to observe the election with patience.
"I think people understanding the state of the race is important. I think people understanding who has won based on a lot of experts looking at a lot of data is important," McCoy says. "But the reality is, just because we call a race, or just because a candidate concedes that doesn't really change the final outcome."
Rosin made a clear point to tell our researchers that even if he projects a winner, it's still just a projection.
The official results won't be known until the Electoral College votes on December 14.
"I could go on TV tomorrow night and say I won, right. It doesn't matter. Saying you won doesn't mean you won." Rosin compared it to breaking into the press box during the seventh inning of a baseball game, grabbing the mic and saying you won. "No, you've got to finish the game."
A common claim online over the past few weeks is that we need to know the results of the elections on the night of November 3. Tammy Patrick says that's never been the case.
"Anyone who has studied elections or pays attention to elections knows that we have never had that answer on election night. It's just a demonstration of a lack of understanding of how elections work in the United States of America."
Here's how WUSA9 will project winners
We will be relying on the data compiled and projections called by the Associated Press and CBS News for the presidential race, U.S. Senate races and U.S. Congress races.