Household income by race & household structure

Hey folks,

In the past, I’ve calculated median household income with HHINCOME & HHWT, with “Household” as my structure.

Now, however, I’m trying to figure out the HHINCOME of different racial demographics. But when I add the RACE variable, I can only download it with “Rectangular (Person)” as my structure.

Can I still run the same calculations with HHINCOME & HHWT with that structure? Or should I use PERWT instead? I’m a little lost as to what’s the best approach.

HHINCOME is a household-level variable; it is appropriate to use HHWT for analyses focused on this variable. RACE is a person-level variable. Data rectangulized to the person-level include household-level variables on each person record; if you would like to include household records separately from person records, you can instead choose a hierarchical data extract (see image below). To include person-level data as part of household-level analysis, I recommend that you create household-level measures that summarize the person-level characteristics of interest on the first person record (e.g., iterate through the RACE values for all household members and then classify households based on their racial composition) and then restrict your analytical sample to one person per household by using only the first person in each household (PERNUM == 1) and apply household weights.

Hi Kari, thanks for the reply!

I ended up doing this, let me know if it works or not at all. I used the Rectangular data structure and also added the RELATE variable. From there, I filtered to only head of households, and used their HHINCOME and HHWT (along with RACE) to calculate the HHINCOME percentiles that I was looking for.

It is worth noting that the RACE value for the householder may not be the same as all other household members, but leave it to you to define the best way to measure RACE at a household level for your specific research application. Your approach sounds like an appropriate way to filter rectangularized data at the person-level to include a single observation per household and apply HHWT.

Hi there. So as I understood we cant use the same calculations with HHINCOME & HHWT when we have different value for the householder and for other household members?

Household-level variables (e.g., HHINCOME, HHWT) should be identical for all household members. Person-level variables (e.g., RACE) may differ for each person in a household. If you are interested in the full detail of all values for a person-level variable within a household, then using only the householder is problematic. For example, if you wanted to compare completely white households to completely Black households, you cannot rely on the RACE value for the householder only. If, however, you are interested in comparing households by the race of the householder, using the RACE value for the householder only without consideration of this person-level variable for other household members is fine.