Thanks for the reply. I certainly agree that I've made some assumptions in the modeling, which I don't have any evidence to validate.

I'm interested in reading this analysis by Tideman and Plassman. That said, it seems directly contradicted not just by pretty much everyone's own experience, by other sources I can point to that analyze data. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4411173, for example, contains state-by-state classifications of U.S. voters even on a 1D axis, which look decidedly non-Gaussian; typically bimodal.