"occ" vs "occ1990" vs "occ2010"

I am working with US 1990 Census (5%) sample and wondering what is the best way to keep those employed in military occupations.

I start off by using the variable “empstatd” and keep individuals with empstatd codes 14 and 15. However, to my surprise, the reported occupations for these individuals do not necessarily correspond to military occupations in “occ”, “occ1990” and “occ2010”. For example, there are quite a few instances where an individual has the following codes:

occ = 4 (Chief executives and general administrators, public administration)
occ1990 = 905 (Military)
occ2010 = 10 (Chief executives and legislators/public administration)

How come, despite keeping only armed forces workers using empstatd, I am getting non military codes?

occ seems to refer to one’s last or most recent occupation, not necessarily one’s current occ, while empstatd seems to refer to one’s current status. I am not an IPUMS staff so I might be wrong.

I believe the answer to your question is more related to differences between the OCC, OCC1990, and OCC2010 variables than the reference periods for OCC and EMPSTAT. As a side note, the questions for EMPSTAT and OCC are based on the same reference period of ‘last week’ (see the questionnaire text for EMPSTAT and this questionnaire form for 1990 for more information).

In order to more clearly understand what you are seeing, I created a data extract myself with the 1990 5% sample. After restricting my analysis to individuals with armed forces employment status (EMPSTAT = 14,15) the only code that comes up for OCC1990 is 905 (Military), as you stated, which is what I would expect. OCC and OCC2010, however, both return a number of codes that are not military-specific. The differences between the codes that come up for these variables is likely related to the differences in how each variable is created. Occupation classification systems have been changed by the U.S. Census Bureau in almost every census since 1850. While updating the classification system can be beneficial to maintain relevant occupation codes, it creates comparability issues for studying changes in occupation over time. The IPUMS team retains original occupation information in the OCC variable and creates other harmonized occupation variables (including OCC1990 and OCC2010) that are more comparable over time. You can read more about these variables on the Integrated Occupation and Industry Codes page.

The harmonized variables created by IPUMS are based on occupation classification systems from different years (i.e. 1990 for OCC1990 and 2010 for OCC2010), which results in classification systems that are inherently different. If you are less interested in change over time, you might prefer to use OCC instead of the harmonized variables, which will allow you to avoid the anachronisms that are part of the harmonization process. Finally, since OCC and EMPSTAT come from different questions on the census form, I wouldn’t be concerned about seeing non-military-specific codes after restricting for armed forces employment status.