It’s mid-May 2020, which means we are in the heart of the next Presidential election campaign. You have not heard much about it in the media, because the media is playing politics with the COVID-19 virus, since that is where their best opportunity lies, or so they think. As the hysteria over the virus gradually wanes and fades, the media will turn to its common weapon of pre-general campaign politics political opinion polls.
You may recall that the polls in 2016 fared, well, poorly. OK, that happens. A smart guy looks at what went wrong, is honest about his mistakes, makes corrections and moves on. What will happen this year, from what I see, is basically the opposite – the poll agencies and analysts will deny their mistakes, pretend what happened was a one-off and not their fault anyway, and go back to business as usual. This is an important thing to remember when you read polls during the Spring and Summer, and the attendant cheerleading for Biden.
An unfortunate example of this can be seen in the case of Nate Silver. Nate made his name as a baseball stats guru, then turned his attention to politics and got lucky in some predictions. Lucky not because his math was off, but he grabbed the attention of the New York Times, who sponsored his website FiveThirtyEight.com. giving him fame and fortune … and an ego which led him to blunder in 2016.
After the election, Nate wrote a series of articles about what went wrong in the partisan expectations, but while he did a few honorable things, like admitting he completely screwed up the RNC nomination of Trump, he stopped short of admitting his assumptions were wrong, and while Nate is honest enough that all his old articles are still up on his website, the plain fact is that he did a number of things he denied. Nate, for example, has repeatedly claimed that the polls were correct in 2016 and the only mistake were media pundits using poll numbers to project a Clinton win, he did exactly the same thing himself in giving Clinton a “71.4%” chance of winning the election on November 8, 2016.
To understand why this was a total face-plant, I am writing a series of articles on political polling, how they work, how to read their reports, how political narratives play into polls, who does polls and who reports on them, and other important factors in polling.
Today I start with the train wreck of 2016, and the excuses made by the media, as presented by Nate Silver.
In January of 2017, just before President Trump was inaugurated, Nate Silver realized that ignoring the blunders in his web site would lead to further damage to his reputation as a polling guru, so he wrote a long article for his website, called “The Real Story of 2016”. Like Hillary Clinton’s explanations of what happened, Mr. Silver’s version of causes and facts are subjective and self-serving. I must admit he put a lot of work into the article: Twenty-one paragraphs and three thousand, two hundred and ninety-eight words. I printed out the article and it came to eleven single-spaced pages. But not once did Nate Silver admit the core mistake of 2016 – that he and others believed Hillary would win, and wrote to support that belief.
Mr. Silver worked hard to claim that what happened was not his fault, rather than honestly evaluate the matter and dig into his starting assumptions. The 2017 article shows that he embraced his original assumptions and is simply selling them in a new package. I say “is”, because a reading of Silver’s subsequent writings demonstrates that the assumptions he made when he wrote the article in 2017 are the same ones he bases his thinking on now. A review of statements Silver made in the article are useful in understanding the continuing bias and blunders of the media, including analysts like Nate Silver.
(paragraph 2) “While data geeks and traditional journalists each made their share of mistakes when assessing Trump’s chances during the campaign, their behavior since the election has been different.”
False. Not one major polling group has admitted they got the call wrong, or made any changes to their methodology to do better in the future. The assessments published by polls and media all worked to basically cover their butts with excuses.
(paragraph 4) “Another myth is that Trump’s victory represented some sort of catastrophic failure for the polls.”
“Trump …beat his polls by only 2 to 3 percentage points in the average swing state.”
Now these are telling statements, and very much dishonest. First, if there is a job one for a poll report, especially for analysts, it is to get the main point correct. Not one poll ever predicted a Trump win.
That, given the election’s outcome, cannot honestly be described as anything but a ‘catastrophic failure’.
Also, anyone familiar with polling would catch the significant admission in the second quote. If Trump’s results beat the ‘average swing state average’ by 2 or 3 points across the board, as Silver admitted, that is a direct admission that Trump’s support was consistently better than the polls reported. That, by definition, is an admission that the polls failed to correctly measure support for candidate Trump.
(paragraph 9) “how much the polls shifted toward Trump in the immediate aftermath of Comey’s letter”
… except that is not what happened. Comey’s letter was released October 28, 2016. On that day, Clinton led Trump in the national polls 44.9% to 41.0%. On election day those polls had shifted to 45.5% Clinton, 42.2% Trump.
Silver’s comment here demonstrates the excuse machine in action. This particular one was trotted out by a lot of different polls to excuse their screw-up. But frankly saying Trump enjoyed a major jump after the Comey letter is just plain lying.
Starting in paragraph 11, Nate started offering what he considered important “elements” in how Trump won. They are as follows:
First, the background conditions were pretty good for Trump.
Second, demographics gave Trump a big advantage in the Electoral College.
Third, voter preferences varied substantially based on news events, and the news cycle ended on a downturn for Clinton.
Those statements are just revisionism. First, if Trump had really enjoyed good background conditions, you’d figure he would have led in the national polls for a decent part of the general campaign, but there was absolutely no time where the polls showed Trump ahead. That means either the conditions were not in Trump’s favor, making this a false claim, or Silver is admitting the media was indeed consistently biased against Trump.
Second, I recall the Democrats telling us for years how Republicans were doomed by demographics. How minorities would never support a Republican, how women voted more than men, making it impossible for a Republican to win .., and so on. To claim now that demographics favored Trump is – again – to either make a deliberately false statement or again admit the poll analysts manipulated the results so they could favor Clinton.
The third statement I have already addressed. There simply was no end-of-campaign jump in support for Trump, so making that claim – and seriously, this one keeps coming up over and over – is just lying.
So maybe by now Nate has seen the light and is aware that he did a really bad job in 2016? According to Newsweek, no he has not.
The polling agencies put together their own panel to look into 2016, but like Silver the AAPOR report also ignored the problem, in favor of assumptions.
As a starting point for 2020, therefore, it’s important to understand that polls are reported using a narrative, based on assumptions, and this is as true in 2020 as it was in 2016.