Data from 2021 IPUMS USA doesn't match what is on census website


I am trying to do some analysis using 2021 ACS data on Women in Labor Force with births in the last 12 months. I am using Stata and the following code gives me the number of women in the labor force with births within the last year as 2,609,656:

tab fertyr labforce [fweight = perwt], co

However, on census data site using 2021 ACS 1 year (Explore Census Data), the corresponding number is 1,783,751 (married women with birth in labor force) + 819,357 (unmarried…) =2,603,108.

The number from IPUMS doesn’t match the 2021 ACS 5 year data on census site either (Explore Census Data) – 1,749,723 (married women with birth in labor force) + 852,369 (unmarried…) = 2,602,092 .

Is there a reason that IPUMS extract doesn’t match either of the tables on census website? Am I using the weights incorrectly?

Thank you for your time!


I would not expect to exactly replicate official US Census statistics with the public use microdata (aka the PUMS files which are available directly from the Census Bureau and are the input for IPUMS USA), though I would generally expect the estimates to be within the margin of error. Your estimate is within the margin of error in the table you linked; I would not be concerned about the discrepancy between the specific point estimates. Official statistics rely on more detailed versions of the data than the public use versions available through IPUMS USA; they include additional records and more detail in some variables. This Census Bureau FAQ and this blog post from a former colleague provide a bit more information. PERWT is a probability weight; however using frequency weights will still get you the correct point estimates; note that the Census Bureau recommends using replicate weights to derive standard errors using 2005-forward ACS data.