The Associated Press has been calling presidential elections since the days of the Pony Express. (Lincoln won the last time horses were needed to get California’s returns.)
It’s a different set of challenges now — but a potential reversion looms of having only limited results being available election night.
Sally Buzbee, the AP’s executive editor, has been on an interview blitz for the last several weeks, explaining how the wire service is planning to hold its place as the gold standard of accurate calls in the presidential race and others.
Buzbee has a lot of experience with the process. She was the AP’s Washington bureau chief before assuming her current position. In an in-house AP promotional video, she described the election night drill as “the single most intimidating part of the job.”
I caught up with Buzbee and posed seven basic questions. They appear below, followed by a brief summary of some other organizations’ plans for election night (and beyond).
Rick Edmonds: Not to put words in your mouth, but is reporting election night results the single most important benefit for your U.S. clients? Critical to AP’s business model?
Sally Buzbee: Providing accurate, fair, nonpartisan journalism of all sorts is the critically important thing that AP does each day. For example, the accountability reporting we did to break the news that President Donald Trump had been on oxygen while sick was critical for the world to know. That matters to a world desperate for factual information, and to customers who depend on us for that strong journalism.
But yes, tabulating the vote count and declaring winners in U.S. elections is one of the most important things we ever do — for our U.S. customers and members, and also for customers and audiences globally who follow U.S. elections so intensely. We do think it is one of the key reasons why so many news organizations and news consumers find AP’s journalism essential.
Edmonds: What are the biggest things that can go wrong or have gone wrong for your projections or for those of others in recent presidential elections?
Buzbee: We only declare that a candidate has won an election when we have determined that the trailing candidate no longer has any path to victory. Thus, we never say we are projecting an apparent winner. We won’t call a race until we are sure who has won. That has led to a pretty extraordinary accuracy rate.
To accomplish that, we have to be careful to look at all possible factors. For example, are the votes from one part of a state still unreleased? How many votes are unreleased? Could that number change the final outcome?
Over the years, more Americans have begun voting before Election Day, whether through mail-in or absentee ballots or through early, in-person voting. Some states (not all but some) count the early vote *after* they count the Election Day vote. The rules vary by state. So over time, we have had to ensure, for example, that we always accounted for those results, before we called races.
That’s one reason why research into state laws and voting patterns is so critical to accurate race calling. We do in-depth research on every state, its election laws, its voting patterns, its geographic or rural/urban patterns, and much more. Our race callers and analysts train heavily, including studying that research, long before they call a race.
Edmonds: What are the biggest adjustments needed to do the job well in 2020 vs. 2016? Some coverage seems to assume it is the potential lateness or delays in the count, but I’m not so sure.
Buzbee: I would say 3 things.
First, as noted above, Americans were already moving more toward advance voting. About 42% of Americans voted before Election Day in 2016. The pandemic has accelerated that trend. It’s certain it will be more than 50% of Americans voting in 2020, and perhaps much more.
We have always had to factor in early voting. But there is more early voting in more states this year. And some states new to large amounts of early voting might count and release votes somewhat more slowly than in 2016 or other past years. So we have to take that into account in our race calling in more states this year.
Second, because of the pandemic and bitter political fights, state laws and practices remained in some flux fairly late in the year. States that had not done much mail-in balloting adjusted to allow people to vote in the pandemic. As we see in the news, there are still court fights going on — just days before the election — over the rules in some states. So it took a lot of research — and still does — to stay on top of all this so our race calling can be accurate.
Third, and critically important: AP decided after 2016 that we would no longer use the traditional exit polls to assist our race calling. They were badly inaccurate in 2016 and they had some issues even before then, which are well documented. Exit polls were built for a different era: when all Americans went to a polling place on Election Day and voted, and survey workers asked them questions afterward.
We made the difficult decision to pull out of the network exit poll consortium. Working with NORC at the University of Chicago, we developed a new methodology and tool called AP VoteCast, which also captures early voters and which has proven highly accurate and robust. We did not develop AP VoteCast for the pandemic: We developed it because we saw the long-term trends. But it has proved a huge blessing given the pandemic.
I have to admit that I offer up a fervent “thank you” pretty much every day … about the fact that AP is not dependent on exit polls any longer this year.
Edmonds: So your calls are now a combination of vote count and modeling based on which votes are in and which remain to be counted.
Buzbee: AP uses its gold standard vote count, AP VoteCast and other analytical tools to declare winners.
In some cases, we are able to use results from AP VoteCast to declare a winner as soon as polls close. In those cases, the results from AP VoteCast — along with our analysis of early voting and other statistics – confirm our expectation that longstanding political trends in these states will hold.
In more closely contested states, we can use the early vote count that starts to come in, and some random sampling of early vote returns that we do ourselves, to confirm what we see in VoteCast, giving us the assurance needed to move ahead with a race call.
In closely contested races, however, we must wait for the vote count to come in. We also make sure we understand where there are gaps in reported vote count results because of early voting or whatever, before we can call such a race.
Edmonds: Is there a short version of how you will do the work on election night and in overtime if necessary? I am envisioning you behind platoons of editors and a giant control board, but that may not be right.
Buzbee: Here are the basics:
- We have a race caller for every state. Most of them work from the states, or this year because of the pandemic, from their homes spread out across the U.S. They are all political junkies who know their state cold, and they are all strong at math and analysis. They go through extensive training and do extensive “studying” throughout the year. Even the most experienced race callers do extensive refresher training. Circumstances in the state may have changed, for example. There could be more expected mail-in ballots. How does that change the pace of the night? That type of thing.
- Each of them works with an analyst based in Washington, D.C. These analysts are highly skilled and trained, too — often they are veteran, cream-of-the-crop race callers. The race caller is poring over the data as it comes in, deciding when the race can be called, talking with the analyst, back and forth. When they think a race can be called, they make a recommendation.
- One of our two Decision Desk editors based in Washington must look the data over, ask any questions, probe, and sign off on each race call. Then we move an AP NewsAlert to our customers and the world, saying the race has been called. Then we write journalism about it and produce video pieces about it.
- In the presidential race, a candidate must gain 270 Electoral College votes to win the presidency. So the Decision Desk editors are also tallying the Electoral College votes. When one candidate gets more than 270 electoral votes, they have won the election. They will be the next president. In 2016, at 2:29 a.m. Eastern, we declared that Donald Trump had won the state of Wisconsin. Wisconsin’s electoral votes gave Trump more than 270 electoral votes, so we then immediately moved a flash that said Trump had been elected president.
- Ultimately, we work as a team and we are a good, collaborative team. I am responsible for ensuring we get this right. This year of course we have a pandemic. Washington bureau chief Julie Pace and I and our Decision Desk editors have been working from home since March. But we made a decision to bring a core team into the Washington bureau on election night. Many of our journalists will still work from home, but a core team of us will be in the bureau. We think that will allow us to work more accurately, efficiently and quickly, because so much of this work is collaborative.
- We keep our race callers in something of a bubble: looking just at the facts, just at the data. We don’t want them looking at the competition or anything else. We don’t want a reporter nagging them to call a race. They sit in one part of the bureau, and then the writers/video producers will quickly work off those race calls, to update our stories and spread the news.
- This year, we’ll all be wearing masks!!
Edmonds: Does there remain a tension between getting it right and getting it fast? Obviously accuracy comes first, but can you — or would you even want to — factor out speed.
Buzbee: We definitely want to be as speedy as possible. The world wants to know who won this election, and we understand that.
But, accuracy absolutely comes first for AP, and we stress that every day. We feel a heavy responsibility. The world depends on us for the answer: Who won? We are not going to get that wrong. It is our mission, our job — to get it right, and we embrace that responsibility.
Declaring a winner does not involve magic. It’s based on math, and facts. We will go over the math, and we will look at every fact in front of us. We will examine “what might we be missing?” We will question the situation, and poke at it, and probe it. And when we are sure, doing all of that as quickly as we can, and only when we are sure, then we will call the race.
Edmonds: Do some of your clients — TV especially — draw on your work plus their own when they make calls? And am I right that AP stays out of the probability game as exemplified by FiveThirtyEight and the now-discontinued New York Times needle?
Buzbee: We deeply respect the race calling of the networks in the United States: NBC, Fox, CNN, ABC and CBS. Each of us calls races independently. … Personally I have the highest regard and admiration for each of their operations. Yes, you are right: AP does not make pre-election predictions.
As Buzbee alludes to, a consortium, known as the National Election Pool, was for many years a shared effort among the major networks and AP. After the 2016 election, the wire service decided it wanted out over the issue of exit polling. So did Fox News. The two are collaborating on VoteCast (which Fox brands as the Fox News Voter Analysis).
NPR, PBS NewsHour, Univision News, USA Today Network, and The Wall Street Journal will also use AP VoteCast (and call it that) on election night.
In a webinar on Oct. 14, sponsored by PEN America, Sam Feist, Washington bureau chief of CNN, said that his network and others remaining have made adjustments that will allow a version of exit polling for those who vote early in person or by mail.
Also, The New York Times no longer shows an election probability estimate with the graphic of a needle. However, the Times, FiveThirtyEight, The Washington Post and, this cycle, The Economist all are including probability estimates in reporting the likely outcome of the presidential vote.