Something I've been thinking about recently is how reliant a lot of us are on county-level results specifically to parse the political geography of different states. This can can be misleading--as
Geographer Martin Lewis pointed out in his Demic Atlas project:
The biggest distortion that results from using states or quasi-states as all-encompassing spatial containers for socio-economic comparison is that lightly populated areas might receive precise scrutiny, while some of the world’s most populous places are subjected to extraordinarily crude aggregation.
If you replace this with counties you can see how this can apply. A rather famous example of this around these parts is Georgia 2020, where only 1 county flipped (to Trump) yet margins in the state's most populous counties shifted in such a way to give Biden the state.
But there are other good examples. One is the Mid-Cities area of Dallas-Fort Worth. The City of Arlington, which is a massive suburb of both cities, began to trend leftward as early as 2008--McCain got 52% in the Mid Cities to Obama's 47%. However, the swings in greater Dallas only came to view later.
Another example, at a different level of aggregation, comes from Tennessee. We often think of Tennessee as titanium R with a few isolated specks of Blue. However, if you break down the state at a different level--grand divisions instead of counties--you find that the Western Grand Division is actually a swing region, with a narrow Republican lead. Memphis is an immensely powerful metro area, making Western Tennessee purple on a regional level.
