I’m currently working with the ACS data series. We are working on a project to match an estimation methodology that uses the IPUMS data using the ACS data files directly from the Census site.
We have noticed that the ACS 5 year file from IPUM is significantly different compared to the files located from the Census website in terms of the class of worker variable (COW) in the Census ACS data dictionary for the 5 year data there is a 0 category which represents-“.N/A (less than 16 years old/NILF who last worked more than 5 years ago or never worked)”.
However, the IPUM doesn’t have this category and my team suspects that IPUM allocates these 0 category class of worker observations to other classes.
Would someone be able to share with me the IPUM procedure of allocating the 0 category class of worker observations to other classes?
Thank you for your patience and apologies for the delayed response. The original Census class of worker (COW) variable is the input for the harmonized IPUMS variable CLASSWKR. The IPUMS harmonization process is designed to facilitate comparability across time while preserving detail that is only available in some samples. The codes assigned to harmonized variables often differ from those used in the original variables. As you noted, CLASSWKR uses a 0 code to denote out of universe cases (for persons who were not asked the question associated with the variable), whereas the original census data have a blank or “.” for these cases. The IPUMS out of universe category (0) includes cases that are blank in the original COW variable (N/A - less than 16 years old/NILF who last worked more than 5 years ago or never worked) as well as persons who indicated they were “unemployed and last worked 5 years ago or earlier or never worked”. We do not, however, allocate the original blank/out of universe to valid CLASSWKR cases.
IPUMS USA also offers the unharmonized source variables (see how to view these below) to allow users to recode the variables differently, and to increase transparency about the relationship between harmonized and unharmonized versions of variables. Below I have included a crosstab of both the detailed version of CLASSWKR (called CLASSWKRD) and the general version of CLASSWKR (which simply denotes out of universe, self-employed, or works for wages) crossed with the original COW variable from the 2019 5-year ACS.
You can toggle between source variables and harmonized variables from the main data browsing page:
I hope this information is helpful and answers your question.