Family Income is is a Household Variable on the Census PUMS Site, but is a Person Variable on IPUMS

In the PUMS data files on the US Census website (which have indefinitely been unavailable for downloading due to “system maintenance”–but that’s another topic for another day), family income is a household variable. However, here on the IPUMS website, family income is a person variable. Why is that? Is there any way I can use family income as a household variable?

The harmonized IPUMS variable FTOTINC is based on the IPUMS variable FAMUNIT; the families identified by FAMUNIT are not necessarily synonymous with household and a household may include more than one family. By offering FTOTINC as a person-level variable, household members may have different values for family income. I will note that for some family types (specifically those with unmarried partners), FTOTINC does not necessarily sum the income of all family members (see this forum post for an example and discussion of the related issues).

There are a few options for getting family income as a household-level variable (including some that do not use the IPUMS variable FTOTINC if the way it summarizes income is not appropriate for your research application):

  1. Leverage the source variables option available through IPUMS USA (see screenshot below); this includes the unharmonized original data in your extract alongside harmonized IPUMS variables. I suspect that the variable FINCP is what you are looking at in the original census data. Using the 2018 5-year ACS as an example, you could include the source variable US2018C_FINCP–which is a household-level variable in your IPUMS extract. You may also be interested in US2018C_HINCP (another household-level variable that reports income for the household rather than family). Please note that there is less documentation of source variables than harmonized variables created by IPUMS.
  2. Assign the FTOTINC value of the first person in the household (PERNUM = 1) to all members of the household. Note that this approach will eliminate any within household variation for households that contain more than one FAMUNIIT (see above for caveat about families with unmarried partners).
  3. Create your own household-level measure by summing the person-level variable INCTOT across all members of the household (noting that this would be a value of household income rather than family income). You could also modify this approach to include only certain members of the household (e.g., persons meeting certain criteria for defining families).

Ok, this is good to know. Thank you for your help.

1 Like