Calculating Median Earnings with March CPS: comparing to Census Tables


I am trying to reproduce the median earnings by sex tables published by the Census (Table P-20 (…). I am using stata and I am using the March CPS files appended over several years. My commands are:
collapse (p50) incwage [pw=wtsupp] if sex==1 & age>=25, by(year)
collapse (p50) incwage [pw=wtsupp] if sex==2 & age>=25, by(year)

The medians I get are close (for most years) to the Census figures but not exact. Can this difference be explained?




I replicated the small differences that you are seeing between the microdata and published tables. We do not have an explanation for this difference, but this is very common when comparing microdata estimates and published statistics using CPS data. You could consider contacting the Census Bureau directly with your question.

I am sorry that I could not be of more direct assistance.