Using the Census Bureau’s PL94-171 Group Quarters Population Table

What you can do with this data, what you can't do, and how you can fill in the gaps.

by Aleks Kajstura, last updated December 27, 2021

Prison gerrymandering is the result of two related, but distinct, problems that require two separate solutions. First, the Census Bureau counts incarcerated people as residents of the wrong location (at the prison). Second, the Census Bureau fails to count incarcerated people as residents of the correct location (at their homes).

This memo concerns fixing the first half of the problem: finding the prisons in the Census data. Mathematically, counting incarcerated people at the prison location has a larger vote dilutive effect than simply failing to count them at the correct home address. This half of the problem is not only most important, it is also surprisingly difficult for data users to fix without Census Bureau cooperation. (Because of how the Census is conducted and the data tabulated by block, it can be difficult — even with external data from the prison system — to deduce how much of each block is incarcerated. This was sometimes possible in individual blocks, but not always, and generally not possible state-wide.)

All of this is made more complicated by the fact that the Census Bureau publishes data not for individual addresses or institutions, but aggregated to the approximately 11 million Census “blocks.” In an urban area, a block might be a city block, but particularly in rural areas the shapes can look quite arbitrary and contain a mix of institutional and residential land.

The Census Bureau has always published block level counts of people in “group quarters,” which includes prisons, but this data has traditionally not been available until the summer of the year after the Census when a file called Summary File 1 is published. Unfortunately, Summary File 1 is published too late to be useful for most redistricting. During the 2000 redistricting cycle, most counties and municipalities that wished to avoid prison gerrymandering by removing the prison populations before redistricting had to rely on creative methods that weren’t precise enough for state-wide use.

While as advocates, we have long argued for the Census Bureau to count incarcerated people at their home addresses, the Census Bureau has so far refused to make that improvement. But the Census Bureau acknowledged government bodies drawing districts need to know where the prisons are located in Census data in order to make their own decisions about removing these populations or reallocating them elsewhere. Therefore, the Bureau made a critical technical change to the 2010 Census that it will build upon in 2020. For the 2010 cycle, the Bureau published an Advance Group Quarters Summary File on April 20, 2011. This was after the redistricting data for each state was available, but months earlier than these data tables would have been available in Summary File 1. For 2020, the Census Bureau is going even further and publishing the Group Quarters data as table P5 within the PL94-171 redistricting data made available to each state. (The PL91-171 data would normally be released in February or March 2021, but the pandemic may force Congress allow the Bureau a small extension to the end of July 2021.)

In the 2020 redistricting cycle, ten states will be using these data in combination with their own home address data to count incarcerated people at home for redistricting purposes. These states will thus solve both of the problems outlined above; incarcerated people will not be counted at the prison location, and they will be counted at home. (For a detailed account of how states implemented this population adjustment in the 2010 redistricting cycle, see Implementing Reform: How Maryland & New York Ended Prison Gerrymandering)

In addition, we estimate that over 200 counties and municipalities will use the Census Bureau’s group quarters table to solve just the first part of the problem; they will adjust redistricting data in a way that avoids padding the districts that contain prison populations, but does not count incarcerated people at their home addresses.


Data structure and new release schedule

Remember: You need to know what correctional populations were counted by the Census Bureau and where those populations were counted. This is often not the same location as the prison’s mailing address, and inversely, you can’t assume that all populations counted at the prison’s address are actually incarcerated. The Census Bureau’s Group Quarters Data is essential.

For the first time, each state’s PL 94-171 redistricting data file from the Census will include block level counts of the group quarters population.

By law, the redistricting data, often referred to by the statute that required its creation, PL94-171, must be made available to the states by April 1, 2021. In practice, the Census Bureau traditionally releases it in waves starting in February. But this decade, the pandemic may force the Census Bureau to delay publication as late as July 31, 2021. This PL94-171 data is a subset of the larger Census data, and contains only 6 tables of information needed for redistricting: total population by race, total population by ethnicity, population 18 or over by race population 18 or over by ethnicity, housing occupancy, and group quarters populations.

The new table will officially be designated as P5 and called “Group Quarters Population by Group Quarters Type”.

What you will find in the PL 94-171 Group Quarters Table 5:

  • For each of the ~11 million census blocks, the number of people in:
    • Correctional facilities for adults (P0050003)
    • Juvenile facilities (P0050004)
    • Nursing/skilled care facilities (P0050005)
    • Other institutionalized facilities (P0050006)
    • College/university student housing (P0050008)
    • Military quarters (P0050009)
    • Other non-institutional facilities (P0050010)

Race and ethnicity data for correctional populations

The 2020 Census will have limited options for determining the race and ethnicity of correctional populations. We recommend several workarounds.

As we noted, the data published from the 2020 Census will have fewer tables from which the race and ethnicity data for correctional populations can be obtained. We recommend several workarounds, in this order:

  • If 100% of the block population is incarcerated — that is if the correctional population for the block reported in table P0050003 matched the total population reported in table P0010001 — you can assume the populations and demographics reported in tables P1, P2, P3 and P4 are that of the correctional facility in the block. (We expect that this approach will work for many relevant blocks. In 2010, 47 % of the Census blocks that contained correctional facilities did not mix any non-correctional populations into the block.)
  • For blocks where only a portion of the population is incarcerated, you may be able to make some estimates of the race/ethnicity breakdown of the correctional population from the race and ethnicity populations reported for the block itself. (In 2010, correctional populations in Census blocks with very large correctional facilities tended to constitute all or almost all of that block's total population. Specifically, 92% of blocks with a correctional population of at least 1,000 people had a block population that was at least 80% incarcerated.)
  • If the above is not helpful in your situation, and there is only one correctional block in the county, you can use tables DHC PCO29A-I to determine the number of people for 9 types of race and ethnicity for correctional facilities in the county, which would logically also be the race and ethnicity of the correctional facility in your block in question. However, because the biggest impact of prison gerrymandering is created by state prisons, and most counties have at least one local jail facility, the prerequisite conditions for this workaround are unlikely to exist very often.
  • You may need to make some assumptions and calculate your districts under each separate assumption. For example, if the block has 1,000 people of whom 900 are incarcerated and the block is 50% African-American and 50% white, you could make sure that your districts meet your goals regardless of whether the non-incarcerated population is 100 white people or 100 African-American people or somewhere in between the two. In many cases, you don't need complete precision.
  • You can ask the facility to provide you with race and ethnicity data, and they may tell you. You can ask for the data as of April 2020, but if that's not available then data from any recent year will be better than the next alternative. If you can't get data for a specific state facility, you could also consider using available statistics for the race and ethnicity of the entire state's correctional system, where available, and impute that demographic breakdown to the population counted in each facility.
  • You can also look at the race/ethnicity of the facility as it was counted in the 2010 Census and use that to impute the demographic breakdown of the 2020 population on the basis that it is the most precise and recent data available. We've made this data available for almost every correctional block in the country, along with detailed sourcing information, in our 2010 Correctional Facility Locator.

Lastly, in addition to table P3, which will include the VAP (voting age population) in the PL data files, the Census Bureau is planning to publish CVAP (citizen voting age population) for each block. The Bureau, however, has yet to develop a methodology for creating such data and has not yet identified a release timeline for the data, should they succeed in producing it.

The Group Quarters Table won’t have everything:

  • No home addresses of people counted in group quarters.
  • No names of facilities. (See below for a partial solution.)
  • No indication whether each block contains one correctional facility or multiple facilities. (See below for a partial solution.)
  • No indication of the types (prison, jail, private, etc.) of facilities. (See below for a partial solution.)
  • No data on race/ethnicity/age (Although it may be possible to generate close estimates of race/ethnicity/age immediately if the correctional population is similar to the total population of the block, and possibly other estimates if you can wait until your states’ Demographic and Housing Characteristic File is available (see below).)
  • No other demographic information.

The Census Bureau may also be publishing other relevant tables in mid or late 2021 that may be relevant for addressing prison gerrymandering if they are published in time. The Bureau has not made a final announcement of which tables will be included in the Demographic and Housing Characteristics File (previously called Summary File 1), and the publication of this file may be delayed beyond the expected dates of June to August 2021 because of the pandemic. We’ll update this documentation when we know more, but we’ll be looking to see the Bureau’s plans for potentially helpful tables like:

  • Sex by age (<18, 18-64, 65+) for the adult correctional population may be in a table tentatively numbered P15 Group Quarters Population By Sex By Age By Major Group Quarters Type. (The equivalent 2010 table was Summary File 1, P43.)

There are also some tables produced for 2010 Census Summary File 1 that we expect will not be available in 2020. Data users who used these tables will need to develop other solutions:

  • 2010 Summary File 1 PCT21 Group Quarters Population By Sex By Age By Group Quarters Type which contained group quarters subtype information (ie federal prison, state prison, local jail) is not expected to be published in 2020 except at the state level in table PST1, and therefore will be not useful for making tract-level redistricting decisions. This change should not be critical because the Prison Policy Initiative will again be annotating most of the census blocks that contain correctional facilities with the type of the facility, although of course it was helpful to have this data from the Census Bureau where it was available.
  • 2010 Summary File 1 PCT20A to PCT20I Group Quarters Population By Major Group Quarters Type for nine combinations of race and ethnicity is not expected to be published in 2020 except at the county level in tables PCO29A through PCO29I, and will therefore not be useful for making tract-level redistricting decisions.
  • 2010 Summary File 1 P29A-I Household Type by Relationship for nine combinations of race and ethnicity is not expected to be published in 2020 except at the county level in tables PCO28A-I.
  • 2010 Summary File 1 PCT20 Group Quarters Population by Group Quarters Type which had counts by group quarters subtype (ie federal prison, state prison, local jail) is not expected to be published in 2020.
  • 2010 Summary File 1 P42 Group Quarters Population by Major Group Quarters Type is not expected to be published in 2020. But this data will be available at the block level in P5, which will be published in the PL data release.

What the Prison Policy Initiative will be providing to further support data users:

Avoiding prison gerrymandering means having access to the data to know whether the proposed districts contain large correctional facilities, whether those correctional facilities are likely to contain people from outside of that district, and whether the presence of the correctional institutions will change how opportunity districts perform. The Prison Policy Initiative is publishing resources to help:

  • A short primer on determining which types of correctional facilities are most relevant to correcting prison gerrymandering.(Originally prepared as a sidebar to a report about Oklahoma, this explanation of who is in different types of facilities and why states may want to treat different facility types differently should helpful in most states.)
  • Annotations of the Census Bureau’s prison count data to help line drawers identify facilities of different types. We’ll be making these — and other — annotations available in a variety of tables and digital mapping formats within days of the data’s release by the Census Bureau. See our data page for a full list and schedule.
  • A primer with some suggestions on different ways to determine whether correctional facilities are impacting the published race and ethnicity data for your districts. (Coming in winter 2020, depending on when the Bureau publishes the final list of tables in the Demographic and Housing Characteristic File.)

Conclusion

The Census Bureau’s decision to include the Group Quarters population table in the PL94-171 file is a huge step forward that will give communities more choices about how to draw their districts.

As this article describes, the Group Quarters table has limitations, but with advance planning line drawers can minimize the impact of those limitations, and draw better districts to minimize the effect of prison gerrymandering in this round of redistricting. We continue to work to convince the Census Bureau to eliminate prison gerrymandering by counting incarcerated people at their home addresses.



Stay Informed


Get the latest updates:



Share on 𝕏 Donate