Democratic presidential nominee Joe Biden opened up a 21 percent national lead over President Donald Trump in the three days following news of Trump's positive COVID-19 test result, according to a new poll.
But I hate to underestimate the dark side of the force.
Polls are fine. They weren’t even that far off in 2016 - nationally they missed by like a point and a half, which is pretty normal. They missed by more in a few key states in the rust belt (due to not weighting for education), which sucks, but isn’t evidence that polling is fundamentally broken. They were actually off by more in 2012, but nobody remembers that because it didn’t flip the race.
The problem was that the media narrative didn’t adjust as the race tightened in the final weeks of the 2016 election. Clinton had a healthy but not insurmountable lead in early October that almost entirely evaporated by the end of the month. She was still ahead, but her lead now in the margin of error. In other words, we got to a point where the polls showed a race close enough that a very normal, routine polling error would be enough to flip the race. For some reason, MSM chugged along as if either A) she still had the strong lead she’d had weeks prior, or B) polling errors didn’t exist.
538 (who many in lib media accused of putting their thumb on the scale to inflate Trump’s chances for clicks) had an article like two days before the election that was basically like “hey y’all, we’re really just a very ordinary polling error (1-2 points) from Trump winning, we think he’s now got about a 1/3 chance of winning, maybe we should wait to see what happens.”
Thus when Trump won, people who were really deep in the polls were only mildly surprised. The people who were good and truly SHOCKED were duped by the stats-illiterate MSM.
(Also, 99% of people who “called it” for Trump were guys like Bill Mitchell who had no actual prognostic heuristic, they’re just biased and have been jerking themselves off to their own hindsight bias ever since).
Stop blaming the poor numbers. They’re innocent. Blame the morons that still gave Hillary a 95%+ chance even when the numbers clearly showed a tighter race.
Polls are fine. They weren’t even that far off in 2016 - nationally they missed by like a point and a half, which is pretty normal. They missed by more in a few key states in the rust belt (due to not weighting for education), which sucks, but isn’t evidence that polling is fundamentally broken. They were actually off by more in 2012, but nobody remembers that because it didn’t flip the race.
The problem was that the media narrative didn’t adjust as the race tightened in the final weeks of the 2016 election. Clinton had a healthy but not insurmountable lead in early October that almost entirely evaporated by the end of the month. She was still ahead, but her lead now in the margin of error. In other words, we got to a point where the polls showed a race close enough that a very normal, routine polling error would be enough to flip the race. For some reason, MSM chugged along as if either A) she still had the strong lead she’d had weeks prior, or B) polling errors didn’t exist.
538 (who many in lib media accused of putting their thumb on the scale to inflate Trump’s chances for clicks) had an article like two days before the election that was basically like “hey y’all, we’re really just a very ordinary polling error (1-2 points) from Trump winning, we think he’s now got about a 1/3 chance of winning, maybe we should wait to see what happens.”
Thus when Trump won, people who were really deep in the polls were only mildly surprised. The people who were good and truly SHOCKED were duped by the stats-illiterate MSM.
(Also, 99% of people who “called it” for Trump were guys like Bill Mitchell who had no actual prognostic heuristic, they’re just biased and have been jerking themselves off to their own hindsight bias ever since).
Stop blaming the poor numbers. They’re innocent. Blame the morons that still gave Hillary a 95%+ chance even when the numbers clearly showed a tighter race.