Possible Error: NHIS distestflg Variable

I’ve been working with the NHIS data (samples from 2015-2017), focusing on disability and I noticed that recently you replaced the “afdcomplete” variable with a new variable called “distestflg”. I like that the new variable explicitly states that the respondents who received the FDB supplement are mutually exclusive from respondents who received the AFD supplement in years 2013-2017. This is a helpful improvement over the old flags.

But I think that there might be an error in the new variable. For example, if I’m looking at the 2017 sample, even when I use the new flag (distestflg) to restrict my sample to adults (18+) who received the AFD supplement (distestflg = 4), there are many respondents who are designated “niu” for several of the AFD questions. I’m focusing on 6 variables (hrprobamt vdifamount lamemcondif lawalkclimdif lawashdresdif lacomdifego) and they all have ~58k respondents designated as “niu” even after restricting the sample to distestflg=4. I think these may all be the same respondents.

Is this an error in the new flag or have I overlooked something important about the way the AFD questions were administered?

For those who are interested, I received a response from the IPUMS team by email, here is what they said:

The issue you are describing about unexpected NIUs is related to the questionnaire structure of the NHIS. NHIS asks a subset of questions to one randomly selected adult aged 18+ (the “sample adult”). You will note on the universe tabs for each of the variables you mentioned, that they reference “sample adults.” I think that people who are not selected as the sample adult but are in a family selected to receive AFD questions are the consistent NIU group you have identified. If you further restrict your tabs to ASTATFLG == 1 (or just crosstab these variables with ASTATFLG while restricting to DISTESTFLG == 4), the universes should look as expected.

1 Like