I learned from an earlier post that CPS is representative at the state level as long as appropriate weights are applied.

However, from this post, I understand that ASEC is no longer representative when zooming in on specific groups even with proper weighting.

Is there a general rule (or a best practice/experienced rated recommendation) when disaggregating is ‘too much’, i.e. state representativeness fails?

For example, suppose I am interested in computing mean incomes and federal taxes (FEDTAXAC) of households with a working age head. Can I expect that using ASECWTH as a weight will deliver results which are representative at the state level?

Thanks a lot for guidance on this.

(PS: Following @Matthew_Bombyk’s recommendation, I worked through section 10-3 of this report but I have not been able to find what I was looking for.)