Second follow-up on health insurance coverage edits question

Hi Brandon,

The SHADAC document you provided has been very helpful. Thank you for that! I still had doubts about the edits taking place, so I spoke to a SHADAC person. The edits do seem to take place in the raw ACS data, nevertheless, she confirmed that in the IPUMS data some of these edits were not reflected. Using the example of the VA case, directly from her email to me:

“Using IPUMS, if restrict your data to hinsva=2 (has VA) and vetstat=1 (not a veteran) and the conditions above (other coverage yes and as reported) I find two unweighted cases that look like they should have been changed to “no” for VA coverage. I didn’t see this in the raw data from Census, so I assume it as something to do with IPUMS editing, possibly of vetstat.”

Replicating what she did and using the appropriate weights I get this:

These 54 people (really 1 unweighted observation) should technically not have VA coverage. My theory would be that these people are people who you reclassified from veterans to “no military service” for the variable vetstat, but who still have the VA coverage from the ACS.

I just thought I’d let you know about this, in case your team hadn’t noticed it yet.



It looks like the problem actually is in three of our allocation flags: QHINSCAR, QHINSCAI, and QHINSTRI. We’re accidentally coding cases with a code of ‘11’ in the original data as ‘0’ in the IPUMS data. They should instead receive an IPUMS code of ‘3’. When adjusting for this error the tabulation you cited is consistent with what is in the original data. We’ll get this fixed as soon as possible.