Hello, I want to make sure I apply weights correctly from an extract of ACS 2019 that includes race, ethnicity, occupation and industry.
A sample row:
YEAR | PERWT | RACE | HISPAN | OCC | IND
| 1 | 2019 | black | not hispanic | 5240 | 9160
PERWT is a count of how many individuals that one entry represents, right? So there are 20 other people who are black, not hispanic, work as a 5240 in industry 9160. So if I wanted to count all the Black people in industry 9160, I would filter on RACE and IND then sum the values in the PERWT column, correct?
Thanks!
That’s basically correct, but keep in mind that will be an estimate of the number of people who fit that profile. The PERWT of 20 doesn’t literally mean that exactly 20 people have that combination of race/ethnicity/occ/ind. It means that person represents 20 individuals in their local geographic area (“sampling entity”) who weren’t sampled. On average these should have similar characteristics as the sampled person. But because the ACS is not stratified by race, ethnicity, or any other characteristics of individuals (see here for details of the sampling design), the weights won’t ensure that the estimated number of people in a given category is exact.