As an extension of my world religion map and in light of the fundamental flaws in ARDA data, I'm attempting to estimate religious affiliation based on only data from the Census/ACS with an eye on finding an adequate model to project to sub-county areas. The single most useful figure available is ancestry data, to which I've attempted to assign each of ~190 ancestries an approximate religious distribution. In light of the recent Pew results, I've tried to create a model that matches their state-by-state results. I also used their results to identify a possible proxy for the variance in the nonreligious distribution: white, never-married males.
The model I've come up with isn't perfect, but I think it's pretty good so far. Here's a county map for the contiguous US:
Compare to this map based on ARDA data:
I'm aware of a few issues, such as not yet having a great proxy for the Jewish population, but if there are any other glaring issues, let me know. When it's adequate, I'll try making some sub-county maps.