CPS occupation code change in 2020 - crosswalk?

The CPS basic monthly went to updated occupation codes in 2020. Is there a crosswalk?

Generally looks fine EXCEPT they went to new physician and surgeon codes and as far as I can tell we suddenly get about half as many physicians and surgeons

In the old data (2003-2019), we had the occupation code:
3060 Physicians and surgeons

And in the new data (2020 and beyond) they break it into multiple codes:
3065 Emergency medicine physicians
3070 Radiologists
3090 Other physicians
3100 Surgeons

Near as I can tell there is no other category hiding somewhere (although this breakdown is pretty bizarre - why have most MDs in the category “other physicians”?!). We go from estimating about 1.1-1.2 million physicians each month in 2018 & 2019 to about 630k each month in 2020.

It definitely seems odd that the number of physicians/surgeons is nearly halved between 2019 and 2020. The Census Bureau provides crosswalks for changes to occupational coding schemes, including the transition from the 2010 to the 2018 Census occupation codes, on this page. I also can’t identify some other category that is hiding in the crosswalk. Regarding the bizarre breakdown, the broad Census occupation category “3090 Other physicians” does include multiple standard occupation classification (SOC) codes that do provide more detail (e.g., “Anesthesiologists,” “Cardiologists”). However, this doesn’t offer much in the way of consolation if you are using Census occupation codes instead of SOC codes.

At this point, I can confirm that the drop isn’t a result of an IPUMS coding error, and that the issue is present in the original BLS version of the data. Our IPUMS CPS team linked the December 2019 (uses the 2010 Census occupation codes) and January 2020 (uses the 2018 Census occupation codes) basic monthly data to see if that offered any hints, and found that less than half of linked persons with an OCC code of 3060 in December 2019 had occupation codes in the January 2020 that you might expect based on the crosswalk. Many of the non-matching cases have physician assistant or medical technician occupation codes in January of 2020. At this point, the best suggestion I have is that you contact BLS, as they are the original data provider. The IPUMS CPS team is also looking into this, and I will certainly post an update if we receive more information.

Thanks! Linking dec 2019 to jan 2020 was a great idea and suggests this is a real problem. I am in touch with someone at bls about this and will see what they say. I’ll post their response here if/when I hear back.
One minor thing – the ipums documentation suggests that starting in 2020 there will be codes 3065 and 3070, but I also downloaded data & documentation for january 2020 file directly from bls and their documentation suggests that codes 3065 and 3070 were rolled into 3090 (and in the data, only the 3090 and 3100 codes are present).

Thank you for letting us know! We will update the IPUMS CPS documentation accordingly.

quick update: BLS is still looking into this. But looking more closely at the dec 2019 to jan 2020 link, I noticed that nearly all of the linked persons who changed from 3060 to the PA and med tech codes were “value to value” allocations (qocc=3). These people who were physicians in December 2019 are a large fraction of all the value-to-value allocations that were done in January 2020 for all occupations (a few other, smaller, occupations seem to have this problem as well). So the bug appears to be in the allocations.

Thank you for the update. We also have a message out to contacts at Census Bureau and are waiting on more information, but will share when we have something to report.

I heard back from the BLS. They implemented a new edit (resulting in all the value-to-value allocations) that was meant to recode physicians who reported less than a professional degree to other occupations (see text of the email they sent me below). But when I looked at the value-to-value allocations they were all physicians who reported a professional degree – my guess is this was the mistake, i.e. they accidentally recoded physicians with a professional degree (e.g. MD!) to other categories. I pointed this out to them and they were going to check with the programmers to be sure the edit was implemented correctly. So hopefully this will be fixed in the next release.
email text from bls:
“The 2018 Census occupation classification system introduced in January 2020 incorporates an educational attainment edit for a number of high-education occupations, like physicians, lawyers, and scientists. Such edits were not included previously.
In this new check, if an individual who was coded as a doctor had educational attainment of less than a professional degree, they were recoded to a non-doctor occupation like physician assistant or cardiovascular technologist/technician. As a result, the new classification system shows fewer people in some of the science and health care professional occupations and more people in the related technician and technologist categories.”

Thanks for the update. My colleagues who work on IPUMS CPS have been following up on this with BLS and also observed individuals with professional degrees who were allocated away from being physicians or surgeons. Hopefully we can get a bit more clarity on this soon; it is great to hear that your contacts are following up with the programmers to ensure the edits that consider education are being implemented as intended.

The BLS corrected the occupation codes on 9/23, so they should now be correct:


Has IPUMS updated the monthly cps data for 2020 to incorporate the revised data?

The IPUMS CPS team is currently working on adding the revised data files that have the corrected occupation codes, but has not yet released them. They should be available soon. Thanks for identifying the issue (and sharing the good news about the re-release)–I am really pleased that this has been addressed!

@Doug_Staiger IPUMS CPS has released the corrected BMS 2020 occupation data from Census/BLS.