Changes to OCC2010 Across Samples

I am using the harmonized occ2010 variable for an occupation-level analysis that combines the 2017 to 2019 ACS 1-year PUMS samples. I’m noticing in the code listing for occ2010 that some occupations are available in 2017 but not for later years. This confused me because I was under the impression that occ2010 was intended to be constant across samples. For example, gaming managers (330) is only available for 2017 but not 2018 and 2019. Another example is Counter, Cafeteria, Food Concession, and Coffee Shop Attendants (4060), available in 2017 but not later.

My central question: when pooling 2017-2019 together, how should I recode gaming managers, counter attendants, and others so that the categories persist across my three-year study span? I saw the integrated_ind_occ_crosswalk but it doesn’t seem to answer this question. Many thanks!

Thanks for your post and apologies for the slow response.

When we use “harmonized,” what we mean is applying a consistent coding scheme to different input data. OCC2010 in this case, applies the 2010 coding scheme to all years of data. However, this variable cannot infer detail not available in the original data. The Census Bureau periodically updates their occupation codes to accurately reflect the types of occupations people have, adding new codes and detail for emerging occupations and removing or condensing occupations that are declining. For example, there is much more detail in IT occupations in more recent coding schemes; in this case, previously general occupations are split into more nuanced occupation titles.

We perform occupation harmonization via a series of crosswalks. The crosswalks describe how adjacent coding schemes relate to one another. These comparisons are sometimes straightforward (e.g., a new occupation in the latest scheme is simply one of several new occupation titles from a previously singular category; this is a many-to-one relationship between new and old codes). However, some new occupation codes will be associated with more than one occupation under the previous scheme (i.e., a one-to-many relationship between the new and previous coding schemes).

Gaming managers are an example of a many-to-one relationship between the 2018 and 2010 schemes. Under the 2018 occupation coding scheme, there is no longer a “330: Gaming Managers” code; instead there is “335: Entertainment and Recreation Managers”. This new code includes all those who would have been coded as “Gaming Managers” under the 2010 scheme. However, it also includes a subset of people who would have been coded as “430, Managers, All Other” in 2010; the “All others” code was divided between four codes between the 2010 and the 2018 schemes: 335, 426, 440, and 705. Our protocol is to compare the size of the two groups (e.g., 100% of the “Gaming Managers” compared to the subset of “Managers, All Other” who are assigned to 335 under the new scheme) and assign all cases of the new “contested” code (335) to whichever code under the previous system contributes the most cases (330 or 430). In the situation of the 2018 code “335: Entertainment and Recreation Managers”, more cases are contributed by those who would have been classified as “Managers, All Other” under the 2010 scheme (weighted estimates of 71,585 versus 21,506 Gaming Managers). Accordingly, all cases of 335 in the 2018-forward data, are assigned to 430 in OCC2010.

What is happening with each specific occupation will vary (e.g., the 2010 code for counter attendants was merged with another for combined food preparation workers into a new singular category in 2018: “Fast Food and Counter Workers”; I suspect the latter group contributed more cases and is why you don’t see counter attendants in 2018 forward but I haven’t confirmed that). When harmonizing across occupation schemes, some detail will be lost and some occupation categories will be omitted entirely as you have found. The easiest option is to aggregate occupations as recommended on the OCC2010 description tab. However, you may want to review the proposed harmonization (see this blog post about creating your own crosswalk between unharmonized and harmonized versions of occupation and industry variables) along with the Census Bureau crosswalks that IPUMS uses when making harmonized occupation and industry variables to make occupation-specific determinations. For more details on the process for assigning new occupations into previous coding schemes, you may be interested in this working paper on adding the 2010 occupation codes to the harmonized variable OCC1990–while the focal years are different, the process is the same.