Among third-plus-generation non-Hispanic white adults, the share raised Catholic rose from 27.0% (2000-09) to 29.4% (2021-24), +2.40 points [+0.73, +4.37]. Splitting it by birth cohort, cohort replacement is +1.07 [+0.64, +1.54] and the within-cohort part is +1.33 [-0.27, +3.23].
I predicted, before running, that nearly all of it would be replacement, because religion at 16 cannot change. That is only partly supported: replacement is about 45%, and the within-cohort share is too uncertain to call. The 1945-59 cohort rises clearly (+1.05 [+0.20, +1.88]); the 1975-and-later cohort falls (-0.94 [-1.84, +0.08]).
Limits: General Social Survey (GSS) recall at 16, no tests yet for recall error or mortality selection, and no fertility measure, so the immigration-versus-fertility question stays open. Log: research/catholic-demography/LOG.md.