Occupation by MSA



I am estimating employment share by occupation at the MSA level. I have seen some other posts in this forum about calculating median wage, etc of an occupation at the MSA level and it seems that this is statistically an okay procedure as long as the occupation-MSA sample is large enough and PERWTs are applied.

This might seem obvious, but is it similarly statistically sound to use iPums with the perwt to estimate employment share of an occupation at the MSA level? I guess my concern is that the census is a stratified sample and so I want to make sure that the sample isn’t stratified by occupation.




It should be fine to calculate the employment share of occupations at the MSA level. The caveat is just what you’ve stated, if you are examining occupations with very low numbers of observations within MSAs, then your estimates will have a larger margin of error. So, as long as you don’t cut your occupations with too fine of detail, you should be okay.