Weight for household-level analysis of outgoing rotation group


I’m trying to do some household-level analysis–say, calculate the median household income of all households with a member making less than $15/hour, per HOURWAGE. I can easily drop all households not in the outgoing rotation group, create an indicator for households with at least one member making less than $15/hour, and keep one observation per household … but at that point, what weight should I use to calculate median household income? EARNWT is person-level, but is it sound to use HWTFINL when I’m only looking at the outgoing rotation subgroup?


Details on the various weights in the CPS can be found in Technical Paper 77, specifically see page 79. According to this document:

The family outgoing rotation weight is analogous to the family weight computed for the full sample, except that outgoing rotation weights are used, rather than second-stage weights.

Although there is no household-level ORG weight in the data, you can construct one as follows:

  1. Assign a family ORG weight to the reference person, which for most households is the EARNWT of the reference person (PERNUM=1). The exception is cases where the reference person is a man who is married to a woman who is also present in the household. In that case the wife’s EARNWT is used. The easiest way to find the wife’s weight is to use the “attach characteristics” feature when creating your extract, and attach the EARNWT of the spouse to each record.

  2. The household ORG weight is the family ORG weight of the reference person.