“When you come to a fork in the road, take it.”
— Attributed to Yogi Berra and, unlike some Yogi-isms, believed to be authentic, because his house in Montclair, New Jersey, could be reached by bearing left or right at a fork
After the 2016 presidential elections, it was clear the established way of calling the result state-by-state needed fixing. More and more of the electorate was voting by mail or early in person. That shift accelerated in 2020 during the pandemic. Day-of-election exit polling no longer made sense as the key to a definitive projection by the “decision desks” of the data providers and their clients.
At that point, the members of the 15-year-old media collective known as the National Election Pool split into two competing groups.
The Associated Press developed a new methodology called AP VoteCast and sold it to big news organizations, including Fox News and The Wall Street Journal. With complex screening checks to identify those who will vote, VoteCast relies — to a large degree — on self-reporting in a survey.
The other big national networks in the pool (CNN, ABC, CBS, and NBC) along with The Washington Post and BBC instead decided to stick with an established vendor, Edison Research, whose choice was to keep exit polling at the core of calling races but to modernize it.
I recently spoke with Joe Lenski, executive vice president of Edison, who directs the firm’s election research. He offered both an explanation and a defense of the approach.
Through Edison’s advance work, Lenski said, an informed estimate is that 50% of American voters will cast a ballot the old-fashioned way, in person, on Election Day this year. Another 15 to 20% will vote early in person and the remainder will vote by mail ballot.
“So we have two-thirds of voters who we know are voters — we speak to them as they leave the polls,” Lenski said. That, he believes, is a big advantage in reliability.
For the mail ballot portion, there is no escaping self-reporting, but Lenski said a series of calls and screening questions yields a representative and reliable sample — tested in 2022 races and presidential primaries, where it has performed well.
I wondered if Edison’s estimates of the percentages that vote in those three different ways might be off. If so, wouldn’t that throw off a forecast of the result?
Maybe, Lenski said, but counts of how many vote early and how many absentee ballots are ordered become available well before Election Day, so models can be adjusted.
Whichever vendor a network or newspaper outlet chooses, they combine that input with election analysis of their own. The result is what you see on air, with analysts like NBC News’ Steve Kornacki or CNN’s John King. Behind the scenes, political directors also have input.
Under bright lights, in what could be a very close election, both Edison and The Associated Press want an extra degree of certainty before calling a given state or the overall winner in electoral votes.
AP executive editor Julie Pace told me that the wire service’s standard remains that a race can be called when analysts believe that the trailing candidate has no path to victory. But, of course, that point can be adjusted election cycle-by-cycle.
Similarly, Lenski told me that “if we were looking for 99.5% certainty in 2020, this time it needs to be even closer to 100%.”
Pace and Lenski also contrasted the contentious present with determining results with the way it was done back in the 20th century.
“It used to be that when AP called a race, that was it,” Pace said.
Lenski recalled that he started as a 23-year-old low-level associate for CBS’s election desk in 1988. “We mistakenly called a couple of states for Dukakis — but it was no big deal.” The error didn’t matter as Democratic presidential candidate Michael Dukakis lost in a landslide to Republican George H.W. Bush.
Bottom line, with the rosters of clients remaining the same and the differing methodologies still in place, the potential exists again that states will be called differently or, at least at different times.
That was dramatically the case in 2020. Fox News has its own extensive data and analysis operation led by Arnon Mishkin, a superstar in the field. The conservative network called Arizona for Joe Biden on election night, as did The Wall Street Journal and NPR. That brought howls of protest from candidate Donald Trump and other Republicans.
It was nine days later before the National Election Pool group followed suit.
While the competition between the units is mutually respectful and low key, each side still claims they called Arizona right.
Fox and the AP said that Biden won, as they said he would, albeit by a thin margin.
Lenski told me Edison’s models showed that mail-in votes that would be counted later over the following week would break hard for Trump. They did, he said, ”and in the end we were justified” in holding off making a call.
I am not enough of a student of the relevant math to say just how rare the kind of discrepancy that played out in Arizona is. My guess is that the odds are long that there will be a repeat, with the election in the balance.
As with Yogi’s house, 2016’s fork in the road between the two players in decision desk data could just as easily lead the competing units to the same place next month.
Comments