RE: The Spatial Migration of Political Support
Really nice article. I would imagine that this visualization style could be used to show the effects of gerrymandering as well. One could create an algorithm to identify gerrymandered districts (based on their shape), and then only mark those counties with color to identify trends. Theoretically, a gerrymandered county would have no variation (once red, it stays red, and once blue, it stays blue).
This approach could also be interesting to see the effect of local news stations. The Sinclair Broadcast Group has a reputation of inserting right-wing, pro Trump viewpoints into local news programs. It would be interesting to identify the counties covered by these broadcasts and then look at their blue/red coloring over time. For more on Sinclair, see https://www.vox.com/2017/5/15/15598270/sinclair-broadcast-imminent-conservative-takeover-of-local-tv-news-explained.
It would be interesting to try to look at gerrymandering with spatial data. The algorithm would only need to look at the ratio of polygon nodes to polygon area...the gerrymandered districts should show up with a higher number of nodes for a smaller area. I'll have to give it some thought, but I might eventually take a look at the data and see what pops up. Thanks for the thoughts!
You're welcome. There are other algorithms fully covered in computer graphics textbooks (like Graphics Gems) that could be used to identify convex versus concave polygons. Another approach would be to treat it as a physics problem and calculate the center of mass, and then determine how much mass exists in the polygon far from the center. The mean and standard deviation for all county data could then be calculated.
Canada has an independent, non-political commission that makes determinations about gerrymandering, and I remember reading somewhere that they have developed algorithms to do this. If you intend to pursue, it might be useful to track them down and see what techniques they've made publicly available.