Matching Census Family Status Estimates

CPS
1

Hello,

I’m currently attempting to match the family status estimates in Table H-01 of the Census’ tables on health insurance coverage and would love some insight into what I am potentially getting wrong!

Specifically, I’m interested in the values in column B of the Excel workbook (the “Total” column). The specific categories I’m attempting to match under the “Family Status” bolded header are “In families”, “Householder”, “In unrelated subfamilies”, and “Unrelated individuals”.

I was able to get a spot-on match for the “Householder” and “In unrelated subfamilies” group using IPUMS microdata, so those groups are not an issue. However, I’m having difficulties matching the “In families” and “Unrelated individuals” groups.

For “In families”, I’m filtering my extract using FAMREL (specifically FAMREL != 0) and then summing the ASECWT. For “Unrelated individuals”, I’m filtering my extract using FAMREL (specifically FAMREL == 0) and then summing the ASECWT.

While the estimates I obtain are close to the Census estimates, they are still off by a large enough amount that I’m not 100% comfortable proceeding with my additional analyses. Any insight into how I could better filter the data to try to match the Census estimates would be greatly appreciated!

2

I believe you should be defining families using FTYPE=1,3 and unrelated individuals as FTYPE=2,5. FTYPE gives the family or subfamily type, while FAMREL gives the status of individuals within the family (useful for identifying householders). Here’s what I calculated:

Table H-01 IPUMS % diff
In families 262,991 263,696 0.27%
…Householder 83,698 83,698 0.00%
In unrelated subfamilies 941 941 0.00%
Unrelated individuals 60,618 60,631 0.02%

Is this the same as you found? It’s not clear to me where the discrepancies are coming from, but I think they’re close enough not to worry about the definitions.

3

Thank you so much for the feedback! My number for families was over 264 million so your approach is definitely an improvement. I’ll use that approach. My number for unrelated individuals was the same. It sounds like the estimates you produced are as close as I can get, so I will proceed with that. Many thanks!