I’m using ACS PUMS and trying to figure out % of households below X income breakdown by race of householder and spouse. The analysis unit is household as I want total household income so household weight should to be applied here, but I also want to include household head and his/her spouse’s race data which is a person level variable and person weight seems to be needed. Should I use both weights: household weight to determine N of households below X income and then person weight to calculate racial counts and percentages? Or can I simply use HHWT for both household level HHINCOME variable and person level RACE variable in my calculation?
Since your ultimate statistic is at the household level you should be using the household-level sampling weights. Here are the steps I’d take to perform this analysis (of course, there are more than one way to do anything): First, create a variable that indicates the race of the householder. Second, create a variable that indicates if a household is below X in household income. Third, using the household-level sampling weight calculate the percent of households fall under X in household income by race of the householder. Finally, repeat this same process except in the first step identify the race of the spouse of the householder (if one exists).