Why does the data change so much on the boundaries of states? I would expect the data to be locally correlated, and there wouldn't be discrepancies between neighboring rural counties that are in different states.
"State of residence" was a predictor in their linear model. A person's state (and its government) will probably affect the outcome variable to some extend, so it's not really surprising that it might be retained as a significant term. However, I find it hard to believe it's so important. I wonder if that predictor is simply confounded with some unobserved variable relating to methods of surveying that differed by state.
Even just looking at counties I'm familiar with it doesn't pass the sniff test.
You've got the "hicks live here" county that's the same color as the "yuppies buy houses in these suburbs" county that's the same color as the "not much English gets spoken here" county and so on. And yet it's all different than the adjacent counties that happen to be across state lines that have similar demographics.
Methodology here.
https://data.cdc.gov/stories/s/Vaccine-Hesitancy-for-COVID-1...
There are also effects here around metro areas.
often, population density affects culture/attitudes toward these sorts of things. as with many such maps, adding an additional analysis layer denoting population density would yield interesting results. (though if you intuitively know the population densities of various "outlier" states, and you know where major cities are in the lesser-populated states, you can already intuitively draw some correlations yourself.)