Advocates warn Census Bureau of impending test failure, inaccurate counts of incarcerated people
Without testing improved ways of counting people in correctional facilities, the Bureau is destined to repeat the errors of past decades again in the 2030 Census.
by Aleks Kajstura, March 5, 2026
The Census Bureau is gearing up to start one of its main tests leading to the 2030 Census, and just announced changes that will make their counts of incarcerated people even less accurate than the last Census. The Prison Policy Initiative submitted a comment letter explaining how the proposed test changes not only fail to count incarcerated people at home, but guarantee failed counts at correctional facilities, too. This Census Bureau failure will make efforts to address prison gerrymandering even harder.
Prison gerrymandering is a problem created because the Census Bureau incorrectly counts incarcerated people as residents of their prison cells rather than their home communities. As a result, when states use Census data to draw new state or local districts, they inadvertently give residents of districts with prisons greater political clout than all other state residents. While the Census Bureau drags its feet, states are working to fix prison gerrymandering on their own; in fact, nearly half of the U.S. population now lives in a place that has addressed it. Now the Census Bureau is about to make their work harder.
The Bureau has a decades-long pattern of failing to accurately count people in “correctional group quarters” — the Census term for prisons and jails — and this test was an opportunity to correct course. The Bureau was actually planning to use this test — officially called the “2026 Operational Test in Support of the 2030 Census” — to try out new procedures for gathering counts from correctional facilities in hopes of improving the process. Instead, it suddenly reversed course and is cutting all testing in prisons and jails.
The 2026 test was originally set to be carried out across six test sites scattered across the U.S., several of which were picked specifically to test counting methods for “group quarters” — a type of housing that includes prisons and jails. These plans have now been drastically slashed: limited to just two sites, and only counting people in private residences (among a slew of other cuts).
This disastrous policy change will ensure that the Bureau fails to get the correct counts for the number of people in each facility, which the states rely on when they adjust that data to count people at home. Therefore, the choice to exclude group quarters from the test will affect the count for the 2030 Census, and make it more difficult for states to implement their successful prison gerrymandering reforms.
Our comment explains how the Bureau’s failure to prepare for the 2030 Census will result in a repeat of past failures and an inaccurate count of incarcerated people. We make three main points:
The Bureau’s mistakes create false race and ethnicity data
The concentrated populations in group quarters, particularly the correctional facilities, magnify otherwise-insignificant errors. A single input error can flip the race characteristics of thousands of people, as the Bureau did in reporting 2020 Census populations for the blocks containing Angola Prison in Louisiana.
The Bureau routinely misplaces correctional facilities
During the 2020 count, the Census Bureau counted at least one prison or jail in every state in the wrong place. Most of the populations counted erroneously were small, but occasionally, the census reported hundreds or thousands of incarcerated people in the wrong place.
Two factors combine for an inaccurate count of incarcerated people
Apart from placing prisons in the wrong place, the Census Bureau counts incarcerated people in the wrong place at a more basic level — in a prison or jail cell — instead of in their actual communities […]
States and local governments have spent over two decades trying to fix this problem, but that work is undermined when the Bureau routinely mistakenly and unknowingly places prisons and jails in the wrong location in its data or dramatically miscounts them.
For more details, check out our full letter here (PDF).
