Getting race and gender percentages by detailed occupation by year

I’m trying to recreate this 2016 table, except with more detailed breakdowns of race and gender:

For example, instead of separate categories for Women and Race, I’d like to create percentages for “White women,” “Black women,” “Asian women,” and “Latina women,” too.

The trouble I’m having is with weighing the data. As far as I know, in order to get the percentage of a group in weighted survey data, I need to sum the WTSUPP of the group I’m looking at and divide by the sum WTSUPP of the entire sample.

So, to make sure I was on the right track, I decided to compare the female percent of IPUMS’ data set with BLS for “Healthcare support occupations” (occupation code 3600-3655). I summed up the WTSUPP of just the female rows and divided by the sum WTSUPP of all the rows (female and male). But my result appear to be off from the BLS by about 0.3% (BLS says 87.7% women, my result was 88%). Is there any reason why this would be happening? This is on year 2016 data, so I’m not sure if it’s affected by updates to weights. I’d also love to know if I’m doing this all wrong!

It sounds like you are performing the weighting of the data correctly. It doesn’t look like your estimates are that far off of the official BLS estimates. In general, we don’t expect to exactly replicate official statistics. So it looks like your estimates are well within the margin of error of the BLS estimates, so I wouldn’t worry too much about your methodology.