Good afternoon!

I am having some trouble selecting the proper weight to be used in the analysis of hourly wage. I am introducing the following modifications to the data:

  1. According to the description of UHRSWORKORG, “for basic samples, 1982-1988, the outgoing rotation usual hours question wording matches that of UHRSWORK1. Thus this variable is included in UHRSWORK1 for those years, not in UHRSWORKORG”.

If I use UHRSWORK1 for the ORG rotation group prior to 1988, can I still use EARNWT weight? I only work with individuals eligible for the Outgoing Rotation Group.

If I can not use EARNWT, what weights can I use for the analysis of ORG hours worked prior to 1988?

  1. I am keeping only workers who were in the ORG in both month 4 and month 8, can I still use EARNWT weight?

  2. I am keeping only workers who did not change the 1990 occupation between month 4 and month 8. Again, is the EARNWT still appropriate?

Thank you so much for your help!

EARNWT is the appropriate variable for weighting analyses of individuals in the Outgoing Rotation Group (ORG). While the IPUMS CPS documentation makes notes about the appropriate weight to use for certain variables, this is really a proxy for the universe of whom the variable is asked. Your second two questions get at who you are including in your analytical sample rather than of whom the question is asked; you should still use EARNWT for variables that are asked of people in the ORG. However, because of the complex survey design of CPS, you will need to take additional steps to get the correct standard errors for subpopulations (e.g., people who do not change jobs between month 4 and 8). This IPUMS CPS resource on replicate weights (which are not available for the ORG) does include sample syntax for appropriate subsetting in Stata.