ACS Data Users Group

Looking for clarification on the difference between the poverty estimates provided in tables S0201 and B17001, specifically for adults ages 65+. Calculating by summing across sex and ages within B17001 yields pretty different estimates compared with the percentages given in S0201 at the state level. Is this a 'householder' vs member of household difference?

I'm comparing S0201's "Estimate!!POVERTY RATES FOR FAMILIES AND PEOPLE FOR WHOM POVERTY STATUS IS DETERMINED!!All people!!65 years and over" vs summing "Estimate!!Total!!Income in the past 12 months below poverty level!!Male!!65 to 74 years" variables across age and sex to get the total number of adults ages 65+ below poverty, over the total number of adults ages 65+ for whom poverty was calculated.

I've never seen any documentation for the subject tables (or the "selected population profiles", which technically, is what S0201 is, despite having a "table code" that makes it look like a subject table)

However, just looking at the numbers for the percentages under "all people", one has to guess that the denominator is roughly "of all people of the same age", not "of all people" or "of all people in poverty". (I roughed that out comparing B17001 and B01001 )

I'd love to see systematic metadata documentation, and it seems like it could be integrated into the API system with the "variables" metadata, like https://api.census.gov/data/2019/acs/acs1/spp/variables/S0201_262E.json

An update for anyone else who is interested in these two tables in the future: the estimates DO match and it was my own calculation that was off. The confidence intervals I calculated using the B table are slightly wider than the S table, which is probably to be expected from aggregating the MOEs. And I also discovered C17001, which is preferable to the B table as it provides numbers for 65+ and also has race tables available; and all 3 tables are giving me the same estimates!