I’m somewhat confused about the classification used for the OCC2010 variable. In the description it is written that : “OCC2010 is a harmonized occupation coding scheme based on the Census Bureau’s 2010 occupation classification scheme.” Hence I was working based on the assumption that codes are the 2010 Census occupation codes.
In other posts in this forum, however, it seems to be suggested that in fact the classification used is the 2002 Census occupation scheme.
Could somebody kindly clarify what scheme is used here?
Thanks for bringing the conflicting information on the forum to our attention–we will work to find and provide updates on conflicting posts.
The Census Bureau’s 2010 occupation coding scheme is the basis for OCC2010.
However, there are some cases where OCC2010 collapses detail added in the 2010 coding scheme to facilitate comparability over time (often, but not exclusively, using a 2002 code that was broken into multiple occupations under the 2010 scheme). The value labels for some OCC2010 codes may not correspond to the 2010 occupation code labels, but do reflect the occupational composition of the code. For example, the OCC2010 combines the original OCC codes of 4700 (OCC label "First-Line Supervisors of Non-Retail Sales Workers) and 4710 (OCC label "First-Line Supervisors of Retail Sales Workers); OCC2010 assigns both to the code of 4700 but the label has been updated to “First-Line Supervisors of Sales Workers” (removing the retail versus non-retail specification). Note that if you do not use the IPUMS code labels this aggregation could be problematic.
In light of the updated 2018 occupation codes, we will be reassessing if some codes currently collapsed in OCC2010 should be made available without aggregation to similar occupations. We will also update the documentation for OCC2010 to clearly note where we have made decisions to collapse codes.
I hope this helps. Please let me know if you have further questions.
Are there any crosswalks from this “hybrid” classification scheme used for OCC2010 to the official 2002, 2010 and 2018 classification schemes besides the official crosswalks provided by CPS/BLS?
We do not have a crosswalk that does this. We use the crosswalks from the Census Bureau Industry & Occupation page when creating these harmonized variables. I would encourage you to review those for how the codes translate between coding schemes. To create a crosswalk of the official 2010 coding scheme and the codes used in OCC2010, you can create an extract that includes OCC and OCC2010 for the 2011-2019 CPS data; the OCC values in these samples use the 2010 codes, which makes it easy to see where the IPUMS variable OCC2010 has made modifications to the underlying 2010 occupation classification scheme.
I’m trying to convert CPS data with 2018 Census Codes to regular 2010 Census Codes so that they can then be converted to 2010 SOC codes.
Since there are about 53 codes in “OCC2010” that map to 133 different 2010 Census Codes (in the “OCC” variable for 2012-2019) and I’m not aware of conversion weights from the IPUMS “OCC2010” to regular 2010 Census Codes, do you have any advice or perhaps a crosswalk for making the merge?
I tried the link on the crosswalk page you suggested but it seems to be down and the archived page doesn’t seem to have crosswalks to “OCC2010.”
As an addendum, I just hand-corrected a crosswalk of 2018 Census occupations to occ2010 by replacing pre-2010 values of the latter with 2010 Census codes using Census’s 2010-to-2018 occupation crosswalks while flagging consolidated 2018 occupations that IPUMS maps back to a single 2010 in-neighbor in the absence of conversion weights. Happy to share the crosswalk if it’s helpful to anyone else.
It seems you figured this out already, but I wanted to follow up on this for future reference. IPUMS does not have such a crosswalk. If your goal is to convert 2018 census codes to 2010 SOC codes, I don’t think you will gain anything by first converting to OCC2010 codes. I think it’s simpler to just use the Census Bureau crosswalk between 2010 and 2018 codes, then the 2010 census-to-SOC crosswalk (both are available here). You could also use the OCCSOC conversion tables for 2010, which nicely recode census codes into SOC-like codes. These use the ACS codes which are slightly different than the CPS codes, but mostly the same.