High family incomes among families below poverty line in ACS?

Hello. Does anyone know why there might be very high family incomes even among families below the poverty line? For example, in the ACS 2019 data, among those ages 25 to 64 not living in group quarters, the maximum family income (FTOTINC) among those below the poverty level (POVERTY<100) is 1,112,000.

This mismatch between FTOTINC and POVERTY values actually reflects an error in the calculation of FTOTINC, and is related to the different ways that the Census Bureau and IPUMS identify those in primary families. I will explain how the each organization defines families below:

  • The Census Bureau assigns all individuals related to the householder by blood or marriage to the primary family. Census Bureau primary families can be identified by looking at the variable RELATE. Everyone with a RELATE value of 10 or less will be considered part of the primary family.
  • The IPUMS USA definition includes all individuals who are in Census Bureau designated primary families, and also includes any unmarried partners of the household head. People in the primary family of a household (using the IPUMS definition) can be identified by having FAMUNIT=1.

The mismatch that you are seeing arises because FTOTINC is calculated by IPUMS by summing the values of INCTOT for each individual in the Census-identified primary family, but then that value is assigned to everyone in the IPUMS-identified primary family (those with FAMUNIT=1). Since the family definitions do not align in cases where the householder has an unmarried partner in the household, the value of FTOTINC is sometimes incorrect in those cases.

On the other hand, the value of POVERTY is taken directly from the original data, which is based on the Census Bureau definition of primary family. This means that unmarried partners of the householder sometimes appear to have a high family income (from FTOTINC), but a low poverty value, because their poverty value is treating them as an unrelated individual, and thus only their individual income (INCTOT) is being used to calculate their poverty value. You can clearly see this dynamic in action by looking at the household of the person with FTOTINC=1,112,000 and POVERTY=8 (see screenshot).

Thank you for pointing this out. We like to show our appreciation for our data users who report errors such as this by sending them an IPUMS mug. If you send us your mailing address to ipums@umn.edu, we’ll send you one in the mail. The IPUMS USA team is looking into the best way to fix the error, but in the meantime you should be fine using just POVERTY. If your analysis requires a consistent family income and poverty measure, you can recalculate FTOTINC by summing INCTOT within each family unit (identified by SERIAL and FAMUNIT), and then recalculating POVERTY by first converting your new family income value to 1999 dollars by multiplying it by CPI99, and then comparing it to the poverty thresholds found on this page.

1 Like

Thank you for the explanation. I do need consistent family income and poverty measures, so I am recalculating those variables as advised. In the poverty thresholds table on this page, what does “RP” refer to in the row label “Two people, RP under 65 years”?

RP is the “reference person,” also sometimes referred to as the householder. In the IPUMS USA data this is the person with RELATE=1.

1 Like

Thank you. In order to count the number of related children under age 18 in each family, I am using RELATED to first identify children. Are the correct values of RELATED to include 301 (child), 302 (adopted child), and 303 (stepchild), but not 401 (child-in-law), 402 (step child-in-law), or 1242 (foster children)?

In the Census family definition, the primary family includes all household members related to the householder by birth, marriage, or adoption. By this definition, the primary family will consist of everyone in the household with a RELATED value less than 1100. This includes all the codes you listed except for Foster children.

For counting up children under 18 in the primary family, I suggest first using RELATED to identify all family members, and then count up the number of these family members who are under age 18 (regardless of their RELATED value). This will capture not only the child, stepchild, etc. relationships that you listed but also cases where, e.g., one of the children is the grandchild of the householder.

1 Like