Stay Updated on Developing Stories

It's not magic, it's math. Here's how CNN makes election projections

(CNN) If you've ever watched CNN on election night, you'll know that Wolf Blitzer announces that the network has projected this or that candidate will win a particular race.

The process that leads to Blitzer saying those words is careful and complicated.

It involves both real-time results and information from exit polls. CNN, NBC, ABC and CBS work with the polling firm Edison Research in what is known as the National Election Pool for results and exit polling data. Fox News and the Associated Press have a separate arrangement.

CNN's Brian Stelter recently interviewed Washington Bureau Chief Sam Feist on "Reliable Sources" about how CNN projects races and how the process will be different this year. A transcript of that conversation, edited slightly for length, is below.

Separately, I also spoke with Jennifer Agiesta, CNN's director of polling and election analytics, who runs the network's decision desk. Keep reading for her views.

This year is different

BRIAN STELTER: This is -- Sam, this is, what, your eighth presidential election at CNN, right?

SAM FEIST, CNN WASHINGTON BUREAU CHIEF: Right. 1992 was the first one.

STELTER: There's a lot that hasn't changed in terms of the decision desk process. What is the biggest X factor this year that makes you and your colleagues think we need to explain this more about how the election works?

FEIST: This is going to be an election like no other. You've heard that over and over. But I'm not sure that the counting or reporting of the votes are going to be a whole lot different. In fact, I think there's every reason to believe it's going to be orderly.

View Trump and Biden head-to-head polling

Obviously, this year, because of the mail-in ballots, as a result of the pandemic, it could take a little bit longer, right? It takes longer to count mail-in ballots. They have to be opened. They have to be processed. Some states don't begin processing mail-in ballots until Election Day, so it could take a little longer.

But I really believe that if we don't have a winner on election night, there's a very good possibility that we're going to know the answer on Wednesday or Thursday because the vast majority of votes will have been counted by then.

How do we get election returns?

STELTER: Let's get into the weeds about the counting. So the local precincts in the states, they count -- they count the votes. And then what happens with the news media, thousands of individuals decentralized across the country that are getting that data and feeding it to you and your colleagues of the decision desk?

FEIST: That's right. Local officials across the country at the county, city, township level, or sometimes at the state level, they count and publicly report the votes. And then all across the country, the National Election Pool and the Associated Press, for that matter, send out reporters to learn about and report the votes.

The votes come back to a central tabulating center for us, and then we of course, report them to the -- to the viewers. And that happens throughout the night, and then it will continue happening and always does after election night is over because votes continue to come in. Mail in and absentee votes continue to be received. And that will happen for days after. And then when all the votes are counted, we report all the results, and it's as simple as that.

STELTER: There's some misconceptions about how this works. For example, the exit polls, there are tens of thousands of interviews with voters after they vote to get a sense of why they voted the way they did. But, you know, networks do not use exit polls alone to make projections and close races. So, I think we should -- we should debunk that myth right now.

Also on screen here, the networks do not compete to be first to announce projections. Now, I feel like in the year 2000, there was a concern that that did happen. What about 2020?

FEIST: So, you were right that in 2000, I think that there may have been some competition between networks to be first. But I have not seen that as long as I've been doing this. I've been doing this working with the decision team since 2004, and there is no race to be first. There's a race to be right, which is not a race at all. It's far more important to be right than to be first.

And I really can assure you that the decision teams at the networks are not in a competition with each other. They're largely in sync with each other. One network may be slightly ahead in one state, one network may be slightly ahead in another state, but it's really the votes that drive the decisions. When there's enough votes in a particular state to give the decision team the confidence that that person is going to win, then they can announce a projection. So, you will not see a race this year, nor should you, and that's a good thing for the public.

STELTER: I think it's interesting that there are two different systems, two different groups doing all of this. And that's new in the past few years. You have the National Election Pool, including CNN, and then this kind of competitor on the market, AP Votecast. Maybe it's a good thing, though, this year to have two different groups assembling the results, because it's kind of a check and balance and might give people more confidence in the results.

FEIST: Yes. I wouldn't really call it a competitor. I would suggest that there -- the two organizations are working in parallel, the National Election Pool and the Associated Press. Each will be independently obtaining the vote count from around the country. And I think that they in some ways provide a check on each other because when the public sees that two independent media consortiums, two independent media outlets report the votes and they're very similar, I think that's a confidence builder and that's something that we need in this election. So, I don't really see them as competitors. I see them as working in parallel and that's good thing.

Have patience, America. This could take a while.

STELTER: Typically 11:00 p.m. is the earliest time -- 11:00 p.m. eastern time -- that an election will be called because the western states have closed their polls. Is there any chance of a projection at 11:00 p.m. on November 3?

FEIST: Yes, of course there's a chance of a projection. It is possible and we tend to make projections early on election night if the race is not close particularly in those battleground states, so it is possible. But it is entirely possible that there won't be a projection on election night.

You know, people forget that in two of the last five elections, we have gone to bed without a president-elect. Everyone remembers 2000 where Florida was the deciding state and then it was too close to call on election night, and we didn't know, and it took another 31 days.

But the very next election was 2004, and in that case, Ohio was the state that was going to be decisive and we did not have enough votes in to project a winner on election night, so we waited. And it was midday the next day when enough results were clear in Ohio that George W. Bush won Ohio and was reelected. So, it is not unusual for elections to not be decided on Election Day, especially this year because mail-in ballots take longer to count. You have to open them, you have to process them, and so it could be that we don't know until Wednesday or Thursday or even later. But I think the vast majority of the votes in the country will be counted by late in the election week, so I believe that we will likely know a winner. It just may not be on election night.

And that's OK. That doesn't mean anything is wrong. The public, the media, the candidates just need to be a little bit patient.

What if a candidate prematurely declares victory?

STELTER: Right. We need to tell people to have patience. A slow count is a safe count. But you know, we don't know what's going to happen in terms of Florida or other key states, what's going to happen with the turnout in those votes. What about let's say it's midnight or 1:00 a.m., and Donald Trump comes out and says I am the winner of the election, and our data does not back that up at all, what will CNN do?

FEIST: If we have not projected enough states for a candidate to get to 270 electoral votes, and a candidate comes out and declares victory, we will make it clear that the facts do not back up that claim of victory. And we'll do it in a number of ways. If you've watched CNN's election night, John King at the magic wall spends an awful lot of election night explaining why we haven't projected a winner in a particular state.

And he goes county by county, reveals how many votes are left to come in, how many votes have been counted, which counties have not reported much votes. This year, of course, we will layer in absentee votes throughout the night in our conversation. And if we're not ready to project the state -- we're not ready to project the state, that doesn't mean that anything is wrong.

And we will make it clear to our viewers and our readers, that there's simply not enough information to make a projection, and that the candidate, if a candidate goes out and declares a winner -- declares victory ahead of time, that they are doing it before the votes have been counted, before -- that is based in fact.

Everyone is looking at the same results, including Fox News

STELTER: Yes, the candidates won't know anything more than the networks. They won't have any magic data that the networks won't have access to.

FEIST: That's correct.

STELTER: Sam, what about Fox? What about Fox News? Is the decision desk at Fox trustworthy given that elsewhere at the network you've got propagandists like Sean Hannity who might try to side with Trump in some sort of election tug of war?

FEIST: All of the networks have excellent decision teams. These are made up of political scientists and data scientists. Jenn Agiesta, who runs the CNN decision team, is our polling director. She's been working with our decision team for many years. And that is true at all of the networks. The public has every reason to have confidence in the decision teams of the networks.

And my advice to everybody, all of the politicians, the partisans, the commentators, the analysts, wait for the projections from the networks and the Associated Press, and don't get out ahead of them. These are the experts that the country has come to depend on over the years.

And any analysts that think they know more than these decision desks that have been doing this for 30 years is wrong. They just don't. And I would have a lot of confidence in the decision desks. And that they will be patient -- they are going to take their time, and they're not going to get ahead of themselves. So, I would give that advice to everybody on election night.

No one knows what will happen

STELTER: Bottom line here is it's about patience, and about not assuming we know what's going to happen. There's a lot of people assuming the worst, and that's not a good idea. But we should also be prepared for a lot of different possibilities. Is that fair?

FEIST: That's correct. That's absolutely right. And we just have to give the -- those -- local election authorities the time to count the vote. In many states, they may have time to do it on election night. In other states, because of state election laws where they can't begin counting absentee ballots until Election Day, just give them time. It may take a day or a few days. Give them time. They will count the votes, and then we'll all know.

Agiesta on the specific things the decision desk considers

WHAT MATTERS: What are the specific metrics you're looking for to project a race? Is there a magic threshold by which you're able to say there's no way a particular candidate can overcome this?

AGIESTA: There isn't any magic involved in projecting races, sadly, it's really all math. There are a number of things we are looking for in each state to have confidence in a projection. Most important is what's been counted: Where are the votes coming from geographically within the state, what types of votes are included in the count, and how much of the total vote does the count represent right now?

If there's a clear lead for one candidate in the current count, but none of the votes from the strongest part of the state for the trailing candidate aren't in yet, that margin likely won't hold up. If instead there is good geographic representation in the vote, that's a point in favor of a projection.

If everything that's been counted is absentee and early votes, or all Election Day votes, there won't be a clear picture of how all the votes will look when both types of vote are included. Some of both are needed for projections in closer races.

And the amount of vote left to count is critical, and may be a more difficult piece of information to track down in 2020. That's more difficult for two reasons. First, the increase in vote by mail, and the number of ballots which may have been mailed in time for Election Day, but are received by election officials afterward. There is no way to know on election night exactly how many of those there are. And second is the decreased value in knowing the number of precincts reporting. There are fewer people voting on Election Day in most places and some states are consolidating precincts, so comparisons of the number of people voting in a particular precinct now to the past are less valuable, and it may be harder to get a good read on Election Day turnout before a county or town is fully reported.

It is also valuable to compare what we know about the vote now to what's happened in a state or county in the past in terms of both turnout and who they're voting for. Whether those patterns remain the same or are changing this year will help to determine when we can make a projection.

What states are key to a presidential projection?

WHAT MATTERS: Is there one specific state you're looking at this year as a bellwether for the presidential race?

Agiesta: It's difficult to narrow it down to one state, but there are generally two types of states that election analysts are looking to this year, and one state within each type that is most critical for determining the president. One set are fast-growing, traditionally Republican, Sun Belt states where Democrats have been gaining ground as the makeup of the population has changed. Arizona, North Carolina and Florida are emblematic of those trends, and within those three, Florida is most critical to the electoral fortunes of either presidential candidate. The other set are Northern states that have been Democratic in recent presidential elections but broke for Trump in 2016. These states -- Michigan, Pennsylvania and Wisconsin -- have sizable populations of White voters without college degrees and economies that have had to change as the economy in America has shifted. In that group, Pennsylvania is the powerhouse in electoral votes.

What about the Senate?

WHAT MATTERS: You're not just looking at the presidential race -- control of the Senate is also up for grabs. Will you be projecting Senate races and is the process different there?

Agiesta: Yes, we will project all those Senate races, as well as gubernatorial contests in the 11 states where those will be held. Projections for downballot races follow the same procedures as presidential races. We're looking at how much we know about all the different types of vote that are out there, where in the state those votes have come from, how they compare to what we know about votes there in the past, and what we know about what's left to count. We need the same type of confidence in that information about Senate and gubernatorial contests as we do for the presidential races.

Outbrain