At the start of Fox News’s election night coverage, commentator Monica Crowley assessed the state of polling on the 2016 presidential campaign: A “reliable predictive model,” said Crowley, might not be “operative.”
Only, Crowley nailed it hard. The seven hours that followed her statement showed, precisely, that a reliable predictive model wasn’t operative. Evidence to that effect was everywhere, especially on the graphs of outlets that had attempted to predict the outcome of the presidential election:
Earlier on Election Day, another common face on Fox News said similar things. “I do think a lot of the polls are purposely wrong,” said Donald Trump in a Tuesday morning interview on “Fox & Friends.” “The media’s very dishonest, extremely dishonest, and I think a lot of the polls are phony. I don’t even think they interview people. I think they just put out phony numbers.” This was laughable coming from Trump, a guy who touted his polling awesomeness throughout the primary race. Back in January, for example, he said in an Iowa rally, “There’s actually this theory that even though I’m leading in all the polls, I’ll actually do better than the polls.”
As with many proclamations from Trump, yesterday’s dig at polling practices had its share of hyperbole/falsehood: We have no evidence that pollsters fabricated their work — just that they messed it up on an historical level. Larry Sabato of the University of Virginia said earlier today on Fox News: “We were wrong, the entire punditry industry, the entire polling industry and the entire analyst industry,” said Sabato, adding that he didn’t have egg on his face. “I have omelet on my face.” GOP strategist and NBC analyst Mike Murphy had this to tweet:
The subject line on this morning’s Politico Playbook includes this disclaimer: “WE WERE ALL WRONG.”
All around the dial last night, it was the same: major media figures admitting that their polling numbers stunk. The meae culpae were uncharacteristically groveling and absolute coming from media types, who generally like to equivocate and make excuses before issuing corrections. Yet here, the data was just overwhelming: It was the media’s precious random-sampling polls vs. actual voting results from duly empowered vote-counters. There was no conceptual hiding place.
And just how wrong were those polls? Have a look at this tweet from Natalie Jackson, the Huffington Post’s senior polling editor:
Trump ended up squeaking through in Wisconsin, 48 to 47 percent. Speaking of the Huffington Post, it predicted a comfy Clinton win in the electoral college. “Our final forecast is that she’ll win 323 electoral votes, far more than the 270 she needs to win. According to our model, she’ll win Florida, North Carolina and New Hampshire, and that’ll be that. Early night,” wrote Huffington Post Washington Bureau Chief Ryan Grim in a mass email that was proved wrong within about five hours. Cockiness built upon cockiness at Huffington Post, as Grim last weekend taunted Nate Silver of FiveThirtyEight for his methodology in arriving at election forecasts. “The short version is that Nate is changing the results of polls to fit where he thinks the polls truly are, rather than simply entering the poll numbers into his model and crunching them,” wrote Grim.
As it turned out, Silver should have been doing even more of that skewing stuff. His final forecast for the election gave Clinton a 72 percent shot. The Huffington Post settled somewhere near 100 percent for Clinton, leading Grim to ditch his bumptiousness on Twitter last night:
Apologies of that sort are flying up and down the Acela corridor today: I’m so sorry. It is as if the entire coastal elite just spilled hot coffee on someone.
Balance requires referencing a supporter of the quantitative efforts. In an interview on Fox News this morning, Tom Bevan of Real Clear Politics said, “Everyone’s running to say that the polls were terribly, terribly wrong. That’s not actually true. If you look, Hillary Clinton’s going to win the popular vote. That’s what the polls said. In the final Real Clear Politics four-way average, she was ahead by 3.2 percent. She’s going to probably end up with roughly a 1 percentage point popular-vote victory. That’s a variation of only 2.2 percent, that’s actually less than what we had in 2012.”
Pressed by host Martha MacCallum on how the polls suggested there was no way Trump could win, Bevan fired back: “Part of the reason for that is that’s how people frame the race.” He noted that there were polls showing Trump ahead in Florida, North Carolina, New Hampshire.
The Washington Post’s Chris Cillizza argued on Twitter:
Important thing to remember:— Chris Cillizza (@ChrisCillizza) November 9, 2016
1. It's ok to be wrong about elections
2. The key is to figure out why and then explain it to people
That approach understates the stakes here. Trump predicated his campaign on bashing the “dishonest” media. Ninety-nine percent of the time he brandished his media critiques, he was either lying or distorting the record to the point of blurriness. Like when he said camera operators weren’t showing his crowds; like when he protested that he opposed the Iraq War in spite of numerous fact-checks finding the exact opposite; like the way he attempted to discredit the New York Times. With each attack, Trump ratcheted up the media’s duty to make absolutely no errors in covering the Trump phenomenon. And yet here we are with this turd.
Get ready for it: For the next four years, Trump will retreat to these unfair polls every time the media catches him in a lie or a mistake.
And misfires in pre-election polls wouldn’t matter so much if media outlets didn’t rely on them so darned much. But they do, especially on cable news. “NEW POLL” is verily a perma-chyron on the major cable networks. It’s how they start out their coverage, how they finish their coverage and how they pad their coverage in between. Polls are newsy, easy to understand, predictable for planning purposes and fun to display on the screen. That’s why they’re ubiquitous, and also why last night is far more of an embarrassment than it ever should have been.