ind1990 for IPUMS-USA: crosswalk recode between 1990 and 2000 Census classifications


Hi, I have a question about variable ind1990 in IPUMS-USA. More specifically about how Census Industrial Classification (CIC) 2000 is recoded into CIC 1990. I have looked at the documentation
for ind1990. This points to a crosswalk between CIC 1990 and CIC 2000. This is an Excel file which is also available on the Census website (Technical paper #65, October 2003; see also Table 1 here:…).

This Excel file actually shows what fraction of the employment of a CIC1990 industry belongs to a specific CIC2000 industry. However, converting a CIC2000 industry into a CIC1990 industry requires some extra work. As the documentation for ind1990 explains:

“To take one example from the technical paper produced after the 2000 census: of persons coded in “Footwear” in 2000 (2000 code 177), the Census Bureau determined that 11% would have been coded as " Other rubber products, and plastics footwear and belting” in 1990 (1990 code 211), while 89% would have been coded as " Footwear, except rubber and plastic" (1990 code 221). Thus, IND1990 assigns a code of 221 to the cases in the 2000 IPUMS sample having an original 2000 IND value of 177."

I am trying to reproduce this recode myself to make sure I understand how it works and some questions came up. Take the footwear example above.

  1. Suppose instead that only 60% of Footwear in 2000 would have been coded to 1990 code 221. Would the IPUMS still assign 221 to the cases in the 2000 IPUMS sample having an original 2000 IND value of 177? In other words, does the IPUMS use a threshold for the reallocation?

  2. If not, what if the highest fraction of a CIC2000 belonging to a single CIC1990 is less than 50%: does IPUMS assign the CIC2000 to the CIC1990 with the highest fraction of CIC2000 employment?

Obviously, if you can share the code files which performs the recode, that would be fine with me as well!



There is no threshold for reallocation. Thus, if the largest proportion is less than 50%, we still assign all of the cases to the 1990 code with the largest proportion. For example, the 2000 code of 298 is split into several 1990 industries, the largest of which is code 300 at about 46%. We still map all persons with an original industry code of 298 in 2000 to a code of 300 for IND1990.

We perform the recode with a translation table, which you can find here.