why do privown, privdep, and privout not return the # of people with privately-purchased health insurance?

I understand your question as asking why the number of cases in PRIVOWN, PRIVDEP, and PRIVOUT are different. Please correct me if I’m wrong. The answer is due to the different universe statements of each variable. PRIVOWN includes persons 15 years and older. PRIVDEP includes all persons. And PRIVOUT includes persons 15 years and older who have privately purchased insurance.

I hope this is helpful.

Hi Jeff, thanks for your answer. Let me clarify my question: the weighted sum of these three variables yields a number close to the published Census estimate of the number of people with privately-purchased insurance, but that weighted sum is outside of the margin of error for the Census estimate. Should I be using different variables? Thanks!

Thanks for the additonal clarification. I’m assuming you are looking at Table 1 in the sourced document. If so, here are a couple notes to be aware of. First, while the table reports figures for 2013, 2014, and 2015, these figures are calculated using the 2014, 2015, and 2016 CPS ASEC samples. This is becasue the question asks about the previous year, so the 2014 sample reports on 2013 insurance coverage, and so on. Second, note that the 2014 ASEC supplement used an experimental redesign, which requrires a bit of an addjustment when using wieghts. See this page for more details. I think once these details are taken into account your method for relplicating Table 1 should work out.