Family incomes ACS B19113

IPUMS – I’m trying to get as close as I can through IPUMS to the median family incomes HUD uses for income limits for a given metro area (show below as # xxxxx). They state that they use ACS table B19113.

To get close I’ve used the following variables but I’m hoping you can guide me/correct me on this.

Dependent variable: ftotinc (* - )
Weight: hhwt
filter1: met2013(XXXXX)
filter2: famsize (2-

It looks like you are using the SDA online analysis tool; please let me know if that is not the case.

Unfortunately, it may not be realistic to replicate this table using SDA. Family income as measured by FTOTINC is an IPUMS-created variable that sums individual income for family members based on families defined as FAMUNIT. IPUMS USA only offers the Census Bureau measure of family income as an unharmonized or source variable (US2018C_FINCP for the 2014-2018 5-year ACS data); source variables are not available for SDA , but you may include them as part of a data extract for analysis in a statistical package. I encourage you to review the documentation in FTOTINC and this information on families and income in ACS data for some context about the different definitions of families.

If you want to use FTOTINC and SDA to create a table of the median family income as measured by FTOTINC, remember to exclude any missing/out of universe values for FTOTINC. If you weight with HHWT, you should restrict your analytical sample to a single person per household (I recommend doing this by specifying pernum(1-1) in the selection filters field). Note that if a household contains more than one family, this approach will not capture the second family’s unique value for FTOTINC. You may be interested in including the IPUMS USA variable HHTYPE in your selection filters to try and mimic the family households that Census uses in its version of the table; I suspect that this is what you are getting at with your selection filter using FAMSIZE. It seems like you are using the “Means” tab and selecting median as the percentile to display under the “Median/Percentile Options” drop down menu; I don’t know of another way to use SDA to calculate the median for continuous variables with as many values as FTOTINC.

Note also that the MET2013 variable doesn’t correspond exactly to the metro areas identified in ACS summary tables, for two reasons:

  1. MET2013 corresponds to the 2013 delineations of metro areas. The ACS summary tables use those delineations only in the 2013, 2014 and 2015 ACS releases. For later years, the ACS tables use later metro area delineations, which deviate somewhat from the 2013 delineations.

  2. More significantly, most metro areas can’t be exactly identified in public-use microdata. MET2013 identifies metro areas as closely as possible but there may be as much as 15% mismatch between a MET2013 code and the corresponding 2013 metro area’s population, and there are numerous metro areas that MET2013 doesn’t identify because a mismatch below 15% isn’t possible. See the MET2013 description for more info.

If by chance you’d be able to use ACS summary data for your work, note also that IPUMS provides ACS summary tables (including B19113) through IPUMS NHGIS. These tables identify all metro areas exactly as they’re identified in the ACS source data.