Are the polls twice as inaccurate?

Rewind four years ago to 2016. FiveThirtyEight national polls showed Hilary Clinton with a 5 to 6 percentage point advantage over then-candidate Donald Trump. Although this narrowed to about 4 points the day before the election, most were confident this would ensure an easy Clinton victory. It turned out the polls were missing something. I surmise part of the miscalibration was attributable to a large tranche of voters that preferred not to broadcast their support for Trump.

Fast forward to 2020, FiveThirtyEight is showing Biden with a 10 to 11 percentage point advantage over Trump in the national polls. Is it reasonable to think that the polls are twice as inaccurate this time around? Considering the polarization that has occurred since 2016, maybe so. If it was already quite taboo to vocalize one’s support for Trump in 2016, in 2020 it has become anathema.

Let’s say the national polls don’t matter much. It’s really about polls in the battleground states, for they do more to determine the election outcome than the total popular vote.

In 2016, with 13 days before the election, Clinton enjoyed a 4.1% advantage over Trump in battleground states (Pennsylvania, Michigan, Wisconsin, Florida, North Carolina, Arizona), according to Real Clear Politics. Fast forward to today, and Real Clear Politics has Biden with a 4.0% advantage in the same states. While Trump still trails Biden, net-net it represents a +0.1% shift for Trump.

This would seem to suggest that despite his unpopularity, Trump finds himself in a comparable position in 2020, if not stronger, than he was in 2016.