The problems Fox News had on Election Night 2020 don’t bode well for the election of 2024.
iMediaEthics (5/19/21): In Arizona, “the actual results were much closer than what VoteCast predicted.”
A little before midnight Eastern time on November 3, 2020, the Fox network, which was collaborating with the Associated Press on vote projections, predicted that Biden would win Arizona. The decision desk director at the time later acknowledged it seemed “premature.” Almost two years ago, on iMediaEthics (5/19/21), I outlined the reasons why the call should not have been made, based on Associated Press’s own post mortem assessment. More recently, Nate Cohn of the New York Times (3/13/23) made a similar argument.
On Election Day 2020, Fox also predicted Democrats to win the House, with their majority expanding by “at least five seats.” That was incorrect, also noted in my article. While Democrats retained the majority, they actually lost 13 seats. Cohn does not mention this miscall.
The predictions were based on a new system that Fox and AP had developed in conjunction with the National Opinion Research Center (NORC). Called VoteCast by AP, and Fox News Voter Analysis by the network, it was born out of Fox’s frustration with the slow pace of predicting winners in 2016.
At that time, the network was part of a consortium, the National Election Pool (NEP) run by Edison Research, which uses exit polls and related data to project election winners. With its new system, Fox would presumably be able to make quicker decisions.
It did not go well.
Neither quicker nor more accurate
The New York Times‘ Peter Baker (3/4/23) suggests that Fox‘s problem was that its polling was too good.
Yet, earlier this month, Peter Baker of the New York Times (3/4/23) seemed to embrace the notion that the Fox/AP/NORC system is superior to the NEP system used by the other networks.
He alluded to a meeting of Fox executives after the 2020 election, when they were discussing how, in the future, they could avoid calling an election for a Democrat before the other networks. Information about the meeting came from evidence in Dominion Voting Systems’ $1.6 billion defamation lawsuit.
Baker wrote:
Maybe, the Fox executives mused, they should abandon the sophisticated new election-projecting system in which Fox had invested millions of dollars and revert to the slower, less accurate model.
It seems unlikely the executives would have referred to the NEP as a “less accurate” model. Slower, perhaps; that was the catalyst for developing the new system. But accuracy was not an issue, at least as publicly stated. It seems more a projected characterization by Baker.
In any case, Baker’s own words appeared to embrace VoteCast as a superior system:
Fox reached its call earlier than other networks because of the cutting-edge system that it developed after the 2016 election, a system tested during the 2018 midterm elections with great success—Fox projected that Democrats would capture the House before its competitors.
The only evidence Baker offered for VoteCast as a “cutting-edge” system is that Fox called the 2018 House contest “before its competitors.” In 2020, Fox also called the House contest before the other networks but, as already noted, that call was a rush to judgment that forced the network to eat crow more than a week later. Hardly cutting edge. NEP made no such error.
Baker ignored altogether the lopsided competition between the two systems in the 2022 House elections. Data posted on the Edison website shows that NEP correctly called the winners before AP and VoteCast in 296 congressional districts, while AP beat NEP in just 73 districts.
A dangerous competition
Personally, I’ve long been skeptical about the competition among networks to be first in calling winners. The public has no immediate need to know who the winner will “likely” be. In most cases, a few more hours will see a completed ballot count and the actual winners announced. If the counting extends for several days, so be it.
The real utility of the election night systems is the statistical information that is collected, which allows for a more in-depth understanding of the factors that motivated voters for one candidate or another. Projecting winners on Election Night is at best an added advantage, and at worst—when miscalls are made—a danger to democracy.
Then–Florida Gov. John Ellis “Jeb” Bush and his first cousin, Fox executive John Ellis, together made the decision that Fox would call Florida for their brother/cousin George W. Bush (image: Media Matters, 2/3/15).
That was the case in the 2000 election, when—at 2:16 in the morning after Election Day—Fox was the first to project George W. Bush the winner in Florida. The head of the decision desk was John Ellis, Bush’s cousin. Bush’s brother Jeb, then governor of Florida, was on the phone with Ellis, and urged his cousin to make the call, though the data did not support it. This projection caused the other major networks to follow suit, only to rescind the call hours later. Chaos ensued.
The miscall and resulting confusion caused Roger Ailes, chair and CEO of Fox News Network, later to admit, “In my heart, I do believe that democracy was harmed by my network and others on November 7, 2000.” (See my book How to Steal an Election: The Inside Story of How George Bush’s Brother and Fox Network Miscalled the 2000 Election and Changed the Course of History for further details.)
That kind of chaos could have happened again in 2020. As Cohn argues about the early call:
There’s not much reason to believe that there was a factual basis for a projection in Arizona. It came very close to being wrong. If it had been, it could have been disastrous.
The public’s confidence in elections would have taken another big hit if Mr. Trump had ultimately taken the lead after a call in Mr. Biden’s favor. It would have fueled the Trump campaign’s argument that he could and would eventually overturn the overall result.
Misleading distinctions
AP says of its VoteCast system, “We meet voters where they are”—meaning they don’t meet voters where they vote.
The hype about the VoteCast system begins with the misleading descriptions found on each news media’s website. Each has a slightly different description of the wonders of their new system, but both emphasize the limits of exit polls as its genesis.
AP: In the 2020 general election, less than a third of voters cast a ballot at a neighborhood polling place on Election Day. That’s why we meet voters where they are, surveying them via mail, phone and online to create a comprehensive data set that empowers accurate storytelling…. AP VoteCast is the product of more than a decade of research and years of experiments aimed at moving away from traditional, in-person exit polls to an approach to election research that reflects the modern approach to voting.
Fox: With more voters than ever voting early or by mail, the new method overcomes the limitations of in-person exit polls and captures the views of all Americans.
The descriptions imply the old, unnamed system, the one they used to belong to (NEP), relies solely on “in-person exit polls.” It does not. And the people at Fox and AP know that.
NEP has been much more than an exit poll operation, ever since its incarnation in 2004. I was with the previous media consortium, called Voter News Service (VNS), on Election Nights 1996 and 2000, and even then, the consortium supplemented exit polls with pre-election polls to measure the preferences of early and by-mail voters.
These days, NEP supplements Election Day exit polls with exit polls at early voting locations around the country, plus multi-mode pre-election polls of absentee voters, including interviews conducted by phone and web.
You can see a comparison of the two methodologies as outlined by NEP and AP Votecast. The comparison reveals two major differences:
- The Fox/AP system relies solely on surveys of voters done before polls close, while NEP uses pre-election polls to measure preferences of absentee voters and exit polls to measure preferences of voters as they have just finished voting.
- All voter preferences gathered by NEP are based on probability samples, the “gold standard of survey research.” Less than a third of VoteCast respondents are selected using probability methods.
To be fair, given the low response rates of phone surveys, or even of multi-mode surveys (those which include, as NEP does, phone and web), it’s not clear that probability samples continue to be superior to non-probability surveys (Pew, 5/2/16; 538, 8/11/14; 3Streams, 3/18/21).
VoteCast or NEP?
Journalists should welcome the addition of a statistically based Election Day coverage system like VoteCast to compete with NEP. Until 1990, the three major broadcast networks—ABC, CBS and NBC—each conducted their own exit poll operation, providing somewhat different takes on the electorate. But in 1990, because of prohibitive costs, they formed a consortium (and added CNN), originally called Voter Research and Surveys. The consortium expanded to include Fox and AP, but still there was only one take on the electorate.
Just as it’s useful to have more than one poll on any given race, it’s useful to have more than one election night operation. But there’s nothing to substantiate the idea that this new operation is especially “cutting edge” or superior to the one that already existed.
In fact, VoteCast is not so much cutting edge as duller edge. Its main advantage in avoiding any in-person exit polling and using mostly non-probability samples of voters is lower cost, rather than any increase in quality.
VoteCast cuts costs dramatically by getting rid of the whole exit poll operation, both in early voting states, and especially on Election Day, which (for NEP) includes 734 exit poll stations across the country, along with recruiting and training interviewers and establishing a live call-in reporting process for the results.
Fox and AP are not alone in trying to find cheaper methods of polling voters. There is an industry-wide effort to cut polling costs, because of abysmally low response rates. As Pew (5/2/16) noted several years ago:
For decades the gold standard for public opinion surveys has been the probability poll, where a sample of randomly selected adults is polled and results are used to measure public opinion across an entire population. But the cost of these traditional polls is growing each year, leading many pollsters to turn to online nonprobability surveys, which do not rely on random sampling and instead recruit through ads, pop-up solicitations and other approaches.
By 2020, most election polls had in fact turned to non-probability samples. As one article noted, from September 1 to November 1, 2020, only 23% of the reported election polls on 538 were based on strict probability samples. The rest were based either totally (61%) or partially (16%) on non-probability samples. Pew observed:
The advantages of these online surveys are obvious—they are fast and relatively inexpensive, and the technology for them is pervasive. But are they accurate?
That is the question that faces the industry overall. The advent of VoteCast, which mostly relies on non-probability samples, is yet another effort to develop more cost-effective ways of measuring public opinion. As such, it should provide useful information for other pollsters as the industry morphs away from the very expensive probability standard.
But the key test should not be which system is quicker in projecting winners, though it is naïve to assume the networks won’t continue to compete in this area. Instead, an evaluation of the two systems should rely on how accurate and plausible are the data each system provides about the nature of the electorate, and the factors that influenced the election.
What about Election Night?
Fox News CEO Suzanne Scott was worried about “the impact to the brand” of calling Arizona for Joe Biden (New York Times, 3/4/23).
Among the media partners in each system, it appears that only one media organization can’t be trusted to make projections in a timely manner based on the statistical findings. Baker makes clear in his New York Times article (3/4/23) that following the 2020 election, Fox executives’ primary concern about the Arizona call was not that it was right, but that coming before any other network, it infuriated Trump and his aides, and angered their own viewers.
Discussions followed, even by their two main news anchors, who, according to Baker,
suggested it was not enough to call a state based on numerical calculations, the standard by which networks have made such determinations for generations, but that viewer reaction should be considered.
As Baker points out, that had already happened. When its decision desk decided to call Nevada for Biden on Friday night, November 6, Fox president Jay Wallace refused to air it. By VoteCast’s models, Arizona would have given Biden the electoral votes he needed to be declared president. Wallace didn’t want his network to be the first. He waited until all the other networks had made the call the next day, and then allowed Fox to follow suit.
Once a network has decided it’s more important to tell viewers what they want to hear, rather than what the data provide, it doesn’t matter how good the election night system might be. The calls can’t be trusted.
The post What Fox’s Bad Calls on Election Night 2020 Say About 2024 appeared first on FAIR.
This content originally appeared on FAIR and was authored by David W. Moore.
David W. Moore | Radio Free (2023-03-24T22:12:57+00:00) What Fox’s Bad Calls on Election Night 2020 Say About 2024. Retrieved from https://www.radiofree.org/2023/03/24/what-foxs-bad-calls-on-election-night-2020-say-about-2024/
Please log in to upload a file.
There are no updates yet.
Click the Upload button above to add an update.