I am working on building a synthetic population that includes group quarters and am stumped by a pretty large inconsistency between the ACS 5-year and decennial census reported group quarter populations in the block group that contains the Cook County Jail in Chicago (block group 170318435001). The ACS 5-year estimates for the group quarter population in this block group hover around 9k in 2019, 2020, 2021, and 2022. However, the decennial 2020 estimate is only around 5k. The current Cook County Jail population is listed around 6k (https://www.cookcountysheriffil.gov/wp-content/uploads/2024/03/CCSO_BIU_CommunicationsCCDOC_v1_2024_03_08.pdf).

Wondering if anyone has any idea why this discrepancy could be so large? I am looking at table B09019 for ACS and the 2020 Census Redistricting Data (P.L. 94-171) Summary Files (P5) (https://www.census.gov/programs-surveys/decennial-census/about/rdo/summary-files.html#P1).

Appreciate any input!

I have worked with the group quarters population and the group quarter facilities counts in several different situations. West Feliciana LA (a Parish) Houses the Angola Prison, one of the largest maximum security priosns in the US. In 2010 the blocks encompassing it reported about 5500 individuals of which about 75% were black, in 2020 the report showed that 15% were black and the rest had become white. The Bureau refused to say how this occurred. The numbers from the ACS mirror the 2010 Census. DP could be a factor, but no one is saying. For DP they left invariant the number of housing units and GQ facilities. They eventually published a file by block that gave the facility type for GQ by block along with the housing units. This is not officially a part of the redistricting file, though it should be. For DP they required at least one resident in every GQ facility. As youmust know they severed the relationship between housing units and GQ and population. Put another way, all of the occupans are made up. This has lead to Occupied Units with no People,, and people with no units. See https://www.nytimes.com/2022/04/21/us/census-data-privacy-concerns.html

Technology advances forced the Census Bureau to use sweeping measures to ensure privacy for respondents. The ensuing debate goes to the heart of what a census is.

Garbage In, Garbage Out. Use the ACS!!!!

Appreciate the input!

The most likely explanation is one of timing. Many prisons populations fell dramatically during the COVID-19 pandemic. Certainly the 2019 and 2020 ACS data--which would have been benchmarked to 2010 Census--would not have caught up with that change.

(For nationwide trends on the magnitude of the dip in 2020, you can refer to BJS statistics here: https://bjs.ojp.gov/document/p22st.pdf)

For how the decennial count compares with 2021 and 2022 ACS... those ACS years use a "blended base" approach--incorporating data from several sources. So there are a number of possibilities for why the numbers still don't match (could be that blended base hasn't caught up with the dip, could be that the prison population rebounded by 2021, could be something else entirely).

2020 Census disclosure avoidance did add noise, but the magnitude of the change you're describing would be incredibly unlikely. The disclosure-related changes are usually single-digit changes, sometimes double-digit changes--not thousands.

Beth with all due respect, the Angola Prison did not send anyone home during COVID, according to the department of correction in LA it had the same count in 2020 as it did in 2010. No the problem is gross error, and the fact that the Bureau in this case refused to dabd o anything to correct a gorss error. The DP on GQ's was very strange since it required at least one person in each GQ, but of course, GQ's generally have more than one person. In short, the Census Bureau turned Angola prison from 75 percent black to 15 percent black, or reported it that way.

Such gross errors when they are found, cannot be corrected, in fact the Bureau has a policy of not changiing characteristics. As to DP errors, according to the Urban Institute study, 12 percent of blocks cannot be true. Connie Citro and Ron Prevost did a major study and there is map up at the Urban Institue that shows which blocks are impossible. In Westchester where I live, there are people supposedly living in the Football Statdium of New Rochelle High School, yet there is no housing there.

The Census has had lots of trouble with Group Quarerts in the 2010 Census there were many mislocated Colleges (SUNY New Paltz was one) and Brison, the Prison in Beacon was another.

The only thing that was reported as found were GQ facilities and Housing Units, but not the number of people living in either.

The advent of DP has meant that for small areas, the Census is much less useful than before.

Andy

Andy, with all due respect, the April 2020 count *at the prison in question* is consistent with the 2020 Census results. I found the data here: https://www.cookcountysheriffil.gov/wp-content/uploads/2020/04/CCSO_BIU_DailyCCDOC_v9_2020_04_01.v1.pdf

Not every quirk in the 2020 Census is a disclosure avoidance error. In fact, not all of the quirks are errors at all.

I'm not trying to be a Pollyana. As you well know, there are some very serious concerns with 2020 Census data quality (age heaping, college and nursing home GQ, young child and tribal area undercounts, I could go on...) But lumping all of the quality issues into "it's all just the noise" risks missing important structural problems that can be fixed for 2030.

If you're interested, the CNSTAT report has a very thorough chapter on all of the fielding challenges with the GQ count https://nap.nationalacademies.org/catalog/27150/assessing-the-2020-census-final-report

Beth--I am not writing about Cook County, I am wrting about Angola Prison in West Feliciana Louisiana, which I mentioned in my email. I brought this up with James Whitehorne, and I will send tyou the material in our correspondence. But the discrepancy started when the % Black in West Feliciana plunged from 46 percednt black to 23 percent balck. I will send you the memo I sent to Jamew Whitehorne, as well as his response from the Bureau. I also discussed this with a lawyer for Louisana and a lawyer for the NAACP LDF, as well as the local head of the NAACP. The question in West Feliciana is not what happened, but how it happened. There is no question about the count of the population, the question is about the composition of the prison population which numbers 5,500 more or less and did so in 2010 and still no one is willing to say why the gross error happened.

Maybe you can get a better response, I will send you the meos off li line. But as the Chicago example showed the ACS does not confirm the massive change in composition, no one in West Feliciana believes it and it does not conform to the state collected data. However, since it is an error that may affect state wide redistricting, Louisiana did not want to bring an official query. The bar to questioning the count in the Census is very high and requires a public body to inquire. It seems to me that checking out obvious errors like this should be possible, but apparently is not.

But to mischaracterize the composition of one of the largest maximum security prisons in the United States seems to me to be an error worth fixing. It is thigs like this that erode the Censu's credibility.

Andy