I added up all households in Census 1990 and 2000 using hhwt and pernum=1 to cross-check the totals, and I find that for 1990 my household count is 91,746,420.00 and for 2000 it is 113,264,611
. However, I found the following published numbers, 91,993,582 and 105,480,101, for each year respectively, in page 13/48 of the following report https://www.census.gov/prod/2005pubs/censr-24.pdf . Could you please help understand how to obtain the right household count for each Census and ACS files? Thank you,
My apologies for the long delay–our internal notes indicate that we posted a response to the 2000 question, but were still looking into the 1990 discrepancy you described. However, it seems that this was never shared. I am sorry for leaving you without any answer for so long.
If I remove group quarters from the 2000 count (GQ=1 or 2 or 5), I am able to replicate the published count for the year 2000 of 105,480,101.
I was not able to replicate the 1990 counts. The IPUMS USA team re-downloaded all of the original 5% PUMS data from the Census Bureau and confirmed that they match IPUMS on person and household counts (both weighted and unweighted). However, these estimates differ from the published counts from the Census Bureau. I am not aware of any documentation on this discrepancy. My best suggestion is to reach out to the Census Bureau with this question about the discrepancy between the published counts and PUMS estimates in 1990.