2014 ACS and CPS-ASEC population estimates does not match after weighting


#1

Hi,

I estimate the total population size using 2014 ACS data (weight= PERWT) as 318,857,056 and using CPS-ASEC (selecting either HFLAG=0 or HFLAG= 1, weight=WTSUPP) as 318,857,056. I was wondering why they are not equal or at least similar? Do they have different reference dates?

Do not they need to sum up to Census Bureau population estimates for 2014?

Finally, is there a weight that I can use all sample from the CP0S-ASEC that sums up to total population? I am interested in migration not in income questions?

Thanks a lot,

Dilek


#2

The 2014 CPS-ASEC sample is a bit different due to an experimental redesign. 3/8ths of the total sample were selected to receive a redesigned survey while 5/8ths received the typical survey. The weighting variable WTSUPP is designed so that, in the 2014 ASEC sample, either sample is representative of the entire US population. So, if both samples are analyzed together and weighted with WTSUPP then the population estimates will be twice the US population. If you want to analyze both samples, then the WTSUPP variable should be adjusted by dividing it by two.


#3

Even if I choose only one sample the total population estimate is 5 million people less than ACS estimate. What is the reason for this difference?


#4

Ah, yes. This is because the CPS is only a sample of only the non-institutionalized population, whereas the ACS is a sample of the total US population. See this page for more information on the sample of the CPS. Sorry I didn’t carify this earlier, I misunderstood your initial question.


#5

Thanks. I accidentaly copied the same number for both surveys in the question.