CPSUPDATE, UHRSWORKT and EARNWEEK

Hey IPUMS time use users,

If I understand correctly, the variable CPSUPDATE in ATUS (I’m using 2003-2019) should indicate whether the information regarding earnings (e.g., EARNWEEK) and hours worked (e.g., UHRSWORKT) was carried over from the CPS sample completed 2-5 months prior to the ATUS survey. However, when comparing, for example, UHRSWORKT and UHRSWORKT_CPS8 only within CPSUPDATE=0 respondents I find that the information in these two variables does not match in a substantial proportion of the cases.
Am I missing something? Why don’t these variables match in cases in which the information was carried over?

Your answers will be much apprenticed and thank you very much for your help!

Based on the original BLS ATUS data dictionaries and data user’s guide, I think CPSUPDATE reflects updates to ATUS earnings variables, not hours worked variables. The IPUMS ATUS value labels for CPSUPDATE clearly reflect that this variable reports updates to earnings, but it is confusing that the variable description briefly mentions hours worked. The IPUMS ATUS team will review and update this documentation. In 2003-2019, I only see 47 cases different EARNWEEK and EARNWEEK_CPS8 values for persons with CPSUPDATE values of 0; this seems more reasonable than the (as you noted) sizeable discrepancies between UHRSWORKT and UHRSWORKT_CPS8.

Whether or not values should match between usual hours worked at the time of ATUS and at the time of CPS for persons whose earnings information was carried forward from CPS is a good question to ask the ATUS folks at BLS who collect these data. I am not aware of any documentation that clearly answers this (and would appreciate it if you were willing to share what you learn).

Thank you for the reply Kari,

I will ask the BLS staff and update if/when I get an answer.
I would just like to add to the discussion that the description of UHRSWORKT does more than briefly mention the connection between hours worked and CPSUPDATA. Rather, it clearly states that “…information on hours worked was only collected at [the ATUS interview] for respondents who changed jobs or employers since the final CPS interview, or whose earnings were allocated in the final CPS interview. For other respondents, information on hours worked was carried forward from the final CPS interview.”

Thank you very much for your reply!

Thank you for pointing out the UHRSWORKT discrepancy! I had only been reviewing the information about hours worked on the CPSUPDATE page when responding to your query. I will ask the IPUMS ATUS team to follow up with BLS as well to confirm that the information in UHRSWORKT is indeed in error, and provide an update here when I hear back.

After further exploration and conversation with the IPUMS ATUS team, I can confirm that CPSUPDATE does not report if hours worked was updated or not; this variable only reports if earnings information was updated at the time of the ATUS interview. Please note, however, that variables about employment other than earnings are eligible to be carried forward from CPS; CPSUPDATE just explicitly flags updates to earnings. The phrasing on UHRSWORKT is ambiguous, but not technically incorrect. The statement “CPSUPDATE may be useful as an indication that this information may be out of date.” is intended to convey that if the earnings values were not updated at the time of ATUS, a researcher may want to be cautious about using employment variables for that respondent. We are planning to update the documentation to remove this comment; the information about when UHRWORKT is or is not updated between the two surveys conveys the relevant information. Removing the comment about CPSUPDATE allows researchers to make their own assessments about if the information is current enough for their research application.