Bad Wigins wrote:
glimpse of a possible future wrote:
everyone says the polls were wrong in 2016. I don’t necessarily agree. There were likely correct at the time they were taken, but I think sentiment shifted towards Trump in the last day or two and the polls were not timely enough to capture that trend.
Why were they "likely" correct? They might have been, but without evidence of this sudden shift in public opinion, one can't call it likely.
Rasmussen was only 1 point off.
There is quite a bit of evidence of a shift towards Trump in the last days, but of course this analysis was all done post hoc. There were also a fairly large number of undecided voters compared to recent elections and exit polling shows them breaking mostly for Trump.
https://fivethirtyeight.com/features/voters-really-did-switch-to-trump-at-the-last-minute/https://fivethirtyeight.com/features/the-invisible-undecided-voter/One other interesting dynamic in 2016 is that while both candidates were disliked by many voters, those that said they didn't trust or like either candidate were more likely to vote for Trump anyway. When it came to actual voting Hillary was even more polarizing than Trump and people voting against one of the candidates were more likely to vote against Hillary. That dynamic will be different in this election.
The prevailing theories about why the polls did not correctly "predict" a Trump victory are: Trump voters were more likely to lie (anonymously), Trump voters were less likely to respond to polls in the first place, and actual voters were not correctly identified in the sampling. While any of those are possible they are very difficult to prove.
I think what many people don't really understand is that a poll is not a prediction or a forecast. It's a snapshot of views or sentiment at a point in time. The polls showing Biden ahead by 10 points are meaningless for what happens in November.