The stability of evaluations of Trump 2016-2020 is esp. striking given the breathless pace of news and norm-shattering.

That disconnect points to the critical importance of studying activists/social movements in addition to general population surveys like this one. https://twitter.com/dhopkins1776/status/1349480688920227843

There's not much in my survey data that hints that 1/6/2021 is on the horizon.

28% of respondents rate Trump above an 80 on the 0-100 feeling thermometer--which is lower than but not too far from 34% for Obama in 2018.

Repeatedly interviewing a sample of engaged American adults feels a bit like asking my Philadelphia-area friends which football team they root for.

"Still Eagles? After a season like that?"

"Ah, yeah. Why do you keep asking?"

Now there are some sub-group shifts our surveys can detect.

Below are OLS models of the 2016-2020 change among the same respondents in 1) vote choice (-1=Dem,0=neither,1=Trump), 2) Trump feeling thermometer (0-100), and 3) PID (7=strong GOP).

There's been a lot of talk about shifts among the Latino vote.

Note that these English-speaking, over-30 Latino respondents from a nat'l panel do shift notably toward Trump relative to others (~8 points)--but there is absolutely no change in their net partisanship 2016-2020.

Education works similarly, in that people with higher educational attainment become more anti-Trump--but also show no shift in party identification.

In terms of party ID, though, notice the pro-GOP move among 2008 union households. I wonder how much of that is public-sector unions (including the police) vs. private-sector unions.

What about gender, race?

Here's a model of the January 2016 Trump feeling thermometer.

Those who identified as female or Black were *already* much cooler on Trump.

Those identifying as White, Protestant, or Catholic were warmer on him.

Baseline=Asian Am./other, non-female