Hello ACS data users group! This is my first post.
I am working with some local nonprofits that serve the homeless population and who help people move into transitional housing. I'd like to get estimates of the size of these two populations broken down by age (broad categories) , gender and hopefully race.
There are very few Census tabulations (either ACE 5-year or the 2010 census -only a few 2020 census tables are out yet) that "break out", the individual Group Quarters type. The PCT39: GROUP QUARTERS POPULATION BY SEX BY AGE BY GROUP QUARTERS TYPE (from 2010 census) is available. This table goes down to at least the "CSD" "county subdivisiion" geography, which includes cities and towns typically. But I need the table for years between the decennial census and having race would be very helpful
I've contacted the people at the Census about a custom tabulation with these 3 variables: group quarters type x age x sex (and possibly race - very useful) using ACS 5-year data. They seem to be receptive that such a tabulation would pass the disclosure rules at the Core-Based Statistical Areas (CBSA) geography.. The Disclosure Review Board is looking at it. Hopefully I'll be able to spread the cost across several organizations and see if there are any grants or donations available.
Someone might suggest using a PUMS file, which has a group quarters / household flag but the group quarters type is only institutional/non-institutional so that won't work. With the "big area" tabulation I hope that I might be able to "pull down" the CBSA tabulation to the CSD level using the ACS 5 year data and some type of Small Area Estimation technique (qroup quarters subtype being the outcome in the regression.)
Any thoughts on any part of this greatly appreciated. I'll post what I am quoted for the custom tabulation. Also HUD, Housing and Urban Development, conducts a homeless census but they don't collect age and sex and they don't canvas people in transitional housing.
As an aside -- I was reading an earlier post about how a college dormitory might be classified based on if it was owned by the college or an outside developer. I contacted the people at the Census about a boarding house on my street to see if it was classified as a group quarters or if the ACS used the household survey for that "structure," I was interested because a local community organization uses the building as transitional housing for the formerly homeless. The address is in the Master Address FIle https://geocoding.geo.census.gov/geocoder/geographies/address?form but you can't get the status as a group quarters v household. That piece of information is not available to the public. I believe that the local government helps with determining the status based on the US Census guidelines . But I doubt if the guidelines are detailed enough to have a classification scheme that applies across the entire country. There has to be local input.
I use R on ubuntu
Any help on any of these topics appreciated,
Dear Tim,
Great to meet you and thanks for reviewing my first post.
I did look into the HUD "Point-in-Time" census, which is done in January every year. The reason that I am looking into ACS data is because I would like to get data for cities and towns that don't participate in the HUD census . Also the HUD data only has the "marginals" for gender only and age only but not the 2-way table with age by gender/sex.
One of the organizations that I have been talking with serves homeless women and women who are able to transition into housing in (group) homes. I did talk with the people at Boston City Hall and they are just about to submit data to HUD from their January count. I may be able to sign a Data Use Agreement with them to get access to their individual "person by person" records to be able to get the cross tabulations. Still I would only have data for Boston, which has substantially different demographics from surrounding areas. I know that this is an ACS forum, but if anyone has some creative ideas about how to approach this research let me know. I think that the ACS may be the only source of "universally collected/available" data. This is nonprofit pro bono research.
Dave Dorer
Hi Dave -
I think you're on the right track in not using the point-in-time counts because they have known undercount problems.
However, I'd add some additional cautions to using ACS group quarters data to estimate those experiencing homelessness: (1) there are some methodological reasons why data by institution type may be problematic for small areas (such as only controlling GQ by type at the state level), (2) some group quarters types are out-of-scope for the ACS. (See page 70 https://www.census.gov/content/dam/Census/library/publications/2020/acs/acs_general_handbook_2020.pdf) And (3) unsheltered populations are not reported (and are likely undercounted in decennial and ACS), and (4) counts of those living in shelters in the 2010 Census appear to be an undercount, based on HUD's point-in-time count for the same year.
Another data source you might try would be your local department of education, which has an accurate count of public school students experiencing homelessness.