SINCE TUESDAY night, news organizations spanning ideology and readership — from FiveThirtyEight (holy of holies) to the Associated Press, from Moody’s Analytics to PredictWise, from Fox News to The New York Times — have engaged in mea culpas, wringing hands and scratching their heads at how they could get the results of the 2016 presidential election so wrong.
A lot’s been made about the inaccuracy of the myriad polls that preceded the election. But we haven’t heard as much about how they got their results; very little’s been said about a factor that the best, most reliable pollsters are defenseless against.
That factor led to an informational house of cards that was destined to fall, and it did on Election Day, when everyone (including yours truly) got completely sucker-punched by an unexpected outcome.
◊ ◊ ◊
Even though they weren't.
The pollsters generally took respondent answers on faith. What those poll respondents said seemed to dovetail experientially with the mood of the country. So, pollsters took that false data and ran with it, building poll after poll after poll around it, reporting the results as a kind of presidential campaign holy writ.
Then the media reported the pollsters’ findings back to the public (or the media conducted its own polling of everyday people, beginning its own version of a process that would lead to exactly the same result).
Thus, the sense of a wave building for Hillary Clinton was utterly, tragically erroneous from the jump, a dream wrapped in statistics as insubstantial as the dream itself.
◊ ◊ ◊
FALSE DATA — answers from millions of poll respondents, answers that were exactly the opposite of what those voters intended to do on Election Day, false answers to survey questions in the waning days and weeks of the campaign, when the scales of public opinion seemed to tip for Clinton — was the X factor of the 2016 election.
At the end of the day, it’s the only thing that would account for the outcome of the election, and the realistic expectation of nothing else. There’s no other reason for the best of established polling protocols — regression analyses, demographic weighting, sample sizing, sampling errors — to have so thoroughly whiffed at the plate when it counted.
You don’t have to be Peter Hart to see how this makes sense: A poll can only be as accurate, as truthful, as presumably dispositive as the root information that makes that poll possible in the first place. Garbage in, garbage out.
Image credits: Clinton projection: New York Times. Clinton-Trump electoral vote map: Moody's Analytics.