Dear All,
I have been working through some analysis using the methods in
Glassman
https://www.census.gov/library/working-papers/2025/demo/sehsd-wp2025-09.html
The Relationship Between the Population Experiencing
Homelessness and Living in Shelters and Poverty
SEHSD Working Paper FY-2025-09
July 2025
Brian Glassman
Poverty Statistics Branch
Social, Economic, and Housing Statistics Division
U.S. Census Bureau
In this paper Glassman looks at the Non-institutional Group Quarters population.
and examines variables such as poverty age race and sex in Education and Employment this population and compares the results to the overall Households population.
As noted the non-institutional group quarters population contains students in college dormitories.
Non-institutional does not include nursing homes.
The question is can you "subset down" to the emergency homeless shelters using other ACS/PUMS variables.
I took a shot at this by using the PUMS variable SCHG which gives the grade level for people in schools. This should remove school dormitories.
One way to "tease out" the homeless shelter population is to subset on the ESR == 6 "not in labor force"
I've also looked at using PINCP (total person's income) with a cutoff of 5000
Being able to create a variable that define a population that is close to the population in homeless shelters would be very useful.
One thing that I looked at was to use MIL to select veterans from this population.
Also tried health insurance HICOV HINS3 and HINS1-7 to select people medicare/medicaid
I know that many of you have looked at this problem
Any thoughts ?
Dave Dorer