best use of occ1990 for year-by-year stats in the CPS data?


I’m looking at within-occupation changes over time in the CPS, and it seems like occ1990 is the variable to use. I’m confused, though, since I still see some noticeable noticeable breaks in trend around the year when the census occupation definitions changed.

For example, if I calculate percent change in employment for each occupation, I see a noticeable jump in 2003, when the 2000 definitions kicked in. Is there a recommended way to use occ1990 in order to get smooth within-occupation trends, or is this jump just an artifact of all the recoding?




The Census Bureau significantly changing their occupational coding scheme in 2003 is the likely cause of the breaks in trend you are seeing. OCC1990 is our best effort to maintain a consistent coding scheme across time; however, these major shifts in classification can still cause inconsistencies such as you have discovered.

I am sorry that I do not have better news to deliver.