correct. you could do some amount areal weighting (this zip code is 60% CBG A and 40% CBG B), but you’d be assuming the population is spread evenly across those CBGs and zips, you have to decide what a zipcode polygon actually is (probably ZCTAs, but they don’t cover the whole country like CBGs), and I’m not sure what that would mean for the margins of error in the census estimates.
Eugene you are the god of the census
Well actually it’s Gina Raimondo