Here's a map I made of Atlas predictions by county. I set each poll to have a voting deadline the day before election day, so the results should be a pure representation of forum consensus at the time.
Not visible is Kalawao, which is colored dark blue. Let me know if I made any errors!
Unsurprisingly, there are a lot of counties which voted Trump which the forum predicted would vote for Harris, given the election results. Some of these are classic swing counties which were close and getting wrong makes sense for given the polls, like Northampton, Nash, Tippecanoe, or Marion.
However, there were quite a few counties where Atlas consensus was fairly strong and wrong. For example, 88% of voters thought Harris would win Ada County, 84% thought Harris would win Oklahoma County, and 81% thought Harris would win Fayette County. If there's any pattern to counties which Atlas was especially incorrect about, it was rapidly growing, sunbelt esque suburban areas. Meanwhile, the only incorrectly predicted Harris county was Montgomery, Ohio.
This suggests imo that we are overly reliant on putative trends in education polarization and population growth, and don't think enough about persuasion and material conditions in our analysis.
