I've built a simulator in excel, and have set up t-tests that test for significant difference in share of preference amongst 2 mutually exclusive subgroups. (share of preference = split vote rule, not first choice).
t = (Share_cell1 - Share_cell2)/sqrt (SE_cell1^squared + SE_cell2^squared)
Where SE = standard deviation of respondent level shares/sqrt(n)
I'm now weighting the respondent level shares by volume and reproportioning to yield an Expected Share of Volume.
I am uncertain of how to calculate SE for this problem. Would the 'n' I use be the effective base? (sum of volumes)^2 / (sum of squared volumes)?