According to the IPUMS website, the ind90 indicators are supposed to be stable over time. However, starting in 2003, industry codes started having 4-digit codes, which makes me wonder what they mean by having ind90 as consistent codes if they eventually change. How can I keep track of which codes changed to four digits? Am I misunderstanding this? Thanks!
The ACS (and the decennial census prior to the introduction of the ACS) reports industry values using the Census Industry Codes. The Census Industry Codes are derived from the North American Industry Classification System (NAICS) for 1997-forward; prior to 1997 they were derived from the Standard Industrial Classification (SIC). Because these coding systems are intended to provide comprehensive coverage of economic activities, they are updated regularly to capture changes in the economy, often at very granular detail. While these changes are important to accurately capture the economy, they pose a challenge for working with the data across time.
Our harmonized industry variable IND1990 translates year-specific, unharmonized industry codes (available in the IND variable) into the 1990 classification scheme. We do this using crosswalks that show how industry codes relate to one another across adjacent schemes. The codes reported in IND1990 should always use the three-digit 1990 Census industry codes, though the inputs that we harmonize into this variable may use the four-digit classification (i.e., in the 2003-forward data).
In the 1990 scheme, the Census industry code for crop production is 010. Under the 2000 scheme (used for the 2000 decennial and 2000-2002 ACS), the Census Industry Code for crop production is 017. The scheme based on the 2002 NAICS (used in 2003-2007) uses a code of 0170 for crop production. As of this writing (July 2025), the code for crop production remains 0170 for all 2003-forward ACS samples (i.e., under the schemes based on the 2007, 2012, 2017, and 2022 NAICS codes). IND will report these year-specific codes, but IND1990 will report any case with one of these contemporaneous IND codes as 010 in IND1990. If you are interested in creating a crosswalk that shows how the unharmonized industry variables map onto the harmonized IPUMS versions, I encourage you to review this blog post.
Note that this harmonization across time has limitations (particularly for a variable like industry) – a singular industry may split into multiple new codes or some industries may disappear altogether. Depending on your application (e.g., if you are interested in highly specific occupations or you are interested in major industries), it may be preferable to work with the unharmonized variables to ensure you are including exactly the industries of interest, or to aggregate multiple IND1990 codes together.
I hope this helps. Please follow up with any further questions.