Hello,
I recently extracted a set of data for 1% density for 2012 ACS, 2013 ACS, 2014 ACS and 2015 ACS samples with HHWT and REPWT1…REPWT80.
My goal is to create a weighted mean of all housing units that lack a complete plumbing for these years–I cannot use a 5 year survey due to geographies changing (I am using PUMAs that have been redrawn boundaries).
If I add all the HHWT of houses without complete plumbing across all 4 years and divide by the total HHWT of all houses, would my weighted mean be accurate, or is each HHWT specific to year and would not be accurate if combined?
I have the same question if I was to combine REPWT1-80 accross the years to get my standard error.
Thanks in advance!
The sampling weights in the IPUMS USA represent weights that calculate representative statistics within each sample. In other words, for each household, the HHWT reflects the number of households that individual household represents in the total U.S. population for a given year. Thus, if you are pooling multiple years of data together, the sampling weight needs to be adjusted. In your case you have four years of data, so you want to make an adjustment so that the total sample represents the US population on average over your 4 year period. The simplest adjustment method is to simply divide the weight (HHWT) by the number of years in your pooled data (4). Although our documentation on replicate weights doesn’t mention this, I believe the same method applies to replicate weights if you are looking to calculate average statistics of the US population over a multi-year period.
I hope this helps. Let me know if you have any additional questions.