Polling was very inaccurate in 2016. Results were outside stated margins of error in several key states. For instance, the bad polling had Hillary up 7 points in Wisconsin. It was enough to convince her poll-gullible staff she did not need to campaign there. Bad polling also led her to mistakenly believe she was close in Arizona.
The polls have a big problem: people who answer their phones are an increasingly rare and unrepresentative subset of the population. They also have a major problem estimating the weights to give each population segment in the final weighted average.
- Younger voters will not show up for Biden. Nominally Democrats, they have zero enthusiasm for the candidate and are worried he will impose national lockdowns and mask mandates.
- Trump will get over 15% of the Black vote, up from 7% in 2016. Trump will also do better with Hispanics in Florida, Texas, Pennsylvania, and Arizona than he did in 2016.
- Suburban women in Minnesota, Wisconsin, New Hampshire will not turn out for Biden. Concern for safety, security, and fear of higher taxes will overwhelm dislike of the Trump personna.
So despite predictions otherwise, I forsee a Trump electoral college victory in a relatively low turnout election.