The 2016 Polls Weren’t Wrong
I have many friends and family who are extremely anxious about the upcoming election. I often try to reassure them by showing them Joe Biden’s lead in FiveThirtyEight’s election forecast, but the response I typically get is “well that doesn’t help because they were wrong in 2016.”
There was a moment in the 2016 election when Hillary Clinton seemed like a sure thing, specifically after the Access Hollywood tape was released and Trump’s polls cratered. His win probability got as low as 12%, however that was still several weeks before the election and his numbers improved, then Clinton took a big hit after FBI Director James Comey sent his letter to Congress stating that he had reopened the investigation into her email server. In the end Trump had a 29% chance of winning according to FiveThirtyEight’s model, but that doesn’t mean that the model was wrong.
If I were to tell you that if you rolled a die that you probably wouldn’t roll a 1 or a 2, and then you rolled a 1, that doesn’t mean that I was wrong. You only had a 33% chance of rolling one of those numbers, but things that have a one in three chance happen all the time.
That doesn’t mean that the model was perfect, there’s no such thing as a perfect prediction. Pollsters attempt to increase their accuracy by determining whether a poll subject is likely to actually vote or not, then skew their results accordingly. In 2016 non-college-educated whites turned out in much stronger numbers than history suggested, particularly in the swing states of Pennsylvania, Michigan and Wisconsin. That having been said though, in the end those states turned in results that were within the margin of error of the polls, so while those polls weren’t perfectly precise, they weren’t actually wrong. Also the national popular vote polling averages of 48.5% for Clinton and 44.9% for Trump were very close to the actual results of 48.2% and 46.1%.
Thankfully 2020 is a different ballgame. With two weeks to go Trump has only a 12% chance of winning. Yes he also was as low as 12% four years ago, however there’s been enough daylight between the dual fiascos of his manic debate performance and contraction of Covid-19 that this low point doesn’t seem to be a knee jerk reaction to those events. Also, compared to the big swings in probability of the 2016 forecast, the 2020 forecast has simply moved steadily in Biden’s direction. Finally, pollsters realize who they undercounted in 2016 and have adjusted their likely voter predictions accordingly, I don’t know which constituency Trump has left to over-perform with.
So am I saying that there’s nothing to worry about because Biden has this in the bag? Fuck no! In the past four years Trump has proven himself to be the most incompetent, divisive and corrupt president in modern history and there’s no telling where this impetuous tyrant of a man will take our country emboldened with the mandate of a second term, with zero concern about having to face voters ever again. If you rolled a die and told me that if you rolled a 1 that some horrible shit was going to happen, of course I’d be nervous.
There’s no need to panic, but we need to keep improving the odds. Vote people.
