Redrawing Colorado’s legislative and congressional boundaries was already going to be a race against the clock because of a monthslong delay in receiving data from the U.S. Census Bureau. But another factor could worsen the time crunch: counting prison inmates.
Last year, Colorado lawmakers passed House Bill 1010, which requires that redistricting count the state’s 14,000 inmates at their last-known address before they were incarcerated, rather than at the correctional facility where they are housed. But the task of reallocating the state’s prison population could be time-consuming, and the nonpartisan staffers tasked with handling the data worry that, on top of the census data delay, the process could jeopardize their ability to finalize maps and avoid delays to the 2022 election.
“We’re, quite frankly, very concerned that the time that it will take to do all of that will make it impossible for the commissions and the Supreme Court to do their work by the end of the year,” Jeremiah Barry, an attorney for Colorado’s independent congressional and legislative redistricting commissions, testified Monday at a House State Affairs Committee.
Advocates for changing how prisoners are counted say the Census Bureau’s method of counting incarcerated people skews the demographic profiles of the rural, largely white districts where prisons are in Colorado. Prison populations tend to be disproportionately people of color, and counting prisoners based on where they are incarcerated effectively hands representation of minorities to areas where their demographic power is diluted.
Redistricting staff and the Colorado Independent Legislative Redistricting Commission proposed addressing the issue through legislation making its way through the Capitol, Senate Bill 247, with an amendment to allow the commissions to avoid the prisoner reallocation requirement this year if it will cause significant delays. The measure currently requires adjustments for prisoner populations within 10 days of getting final census data, which isn’t expected until late September.
The preliminary maps were supposed to be done by mid-September, with a hard deadline for final approval of mid-December.
“Our desire is to reallocate the prison population, but if we find it impossible, the last thing you want to have is legislation that effectively locks us into an impossible situation and we cannot deliver the maps by the end of the year,” Carlos Perez, chair of the Legislative Redistricting Commission, testified recently to the House State Affairs Committee.
But the bipartisan legislative leaders behind Senate Bill 247 are moving forward anyway.
Senate Bill 247 is the basis of an interrogatory submitted to the Colorado Supreme Court Wednesday seeking permission for the redistricting commission to use preliminary data for initial map drawing. The idea is to allow the commissions to begin the process and then adjust their maps after the release of final census data at the end of September.
House Majority Leader Daneya Esgar, a Pueblo Democrat and prime sponsor of the bill, said the prisoner reallocation is a logistical issue that can be addressed down the road, such as by increasing staffing.
The legislation should be aimed only at setting a foundation for the General Assembly’s request for an opinion from the Supreme Court, said House Minority Leader Hugh McKean, a Loveland Republican and another prime sponsor of the bill.
“We’re trying to ask the Supreme Court a very tight question,” McKean said.
Colorado is among nine states that have passed laws to change how prisoners are counted for redistricting, according to the National Conference of State Legislatures.
Counting inmates at their last home address is aimed at reversing what advocates call “prison gerrymandering,” in which the populations of congressional and legislative districts are inflated by prisoners, giving them more representation than they should have.
The U.S. Census Bureau counts inmates based on the correctional facility where they are incarcerated, so states with reallocation laws must adjust the numbers themselves.
On top of delays in actually getting the census data this year, the Census Bureau will also begin to anonymize the data they collect and distribute to states with a new statistical method called “differential privacy.”
The agency has long scrambled its data to protect individual privacy , particularly for people living in small or sparsely-populated communities. A single census block in rural Colorado, for example, may have far fewer people living there compared to Denver, making it easier to identify a person in census data based on demographic details. The agency has a legal mandate to keep personal identifying information confidential.
The new and complicated differential privacy method uses algorithms to inject mathematical “noise,” or random intentional errors, into the data to make it hard to trace individuals and their locations in local census data. The idea is to preserve the overall statistical accuracy of census data, but add small, random errors that make it hard for someone to reverse-engineer the data and cross-reference other data sets to single out individuals.
It’s a trade-off between privacy and accuracy, but critics say the method could amplify distortions in small communities. Some immigrant and civil rights groups argue it could result in violations of the federal Voting Rights Act, by undermining accurate reporting of racial minorities.
The methodology has resulted in “much larger errors in the smallest levels of geography,” said Colorado state demographer Elizabeth Garner.
Her staff has spotted issues like data showing only children living in a census block or blocks with occupied housing units but no population.
“We are still finding weird things, even in this last set” of data, Garner told the Independent Congressional Redistricting Commission on Wednesday. The Census Bureau is actively working on improvements to its method to reduce errors.
The scrambled data also creates a challenge for states like Colorado that want to change how they count prisoners.
For example, the state might report 500 inmates at a prison, and the Census might count 500 people, but after the algorithms insert errors into the data to protect people’s privacy, the head count that’s released could be a different number, and the demographic characteristics reported for that population may not be accurate. And redistricting staff won’t be able to identify who is who.
To determine which prisoners need to be attached to a different census tract, Colorado redistricting staff will have to compare census data with Department of Corrections reports and manually identify which prisoners need to be counted at a different address, Barry said. They have already started processing the DOC data, but it will be unclear how much work is left until the state receives a batch of census data Aug. 16.
“I suppose if the population lines up closely, we would be able to do it fairly quickly. But without knowing what we’re dealing with, I’m not sure that we will know” how long the adjustments could take, Barry said.
There are currently six nonpartisan staffers working for Colorado’s independent redistricting commissions, with two more set to join at the end of Colorado’s 2021 legislative session in mid-June.
The existing staff has also requested funding for two more positions: an expert on geographic data and a person to help with public outreach.
Lawmakers will revisit the prisoner reallocation issue and could decide to beef up staffing and funding to help the commissions, Esgar said.
The Census Bureau has defended the new data privacy method as necessary to deter data miners from identifying people in the publicly released data, especially as computing power and technology has improved.
Alabama, with the support of 16 other states, is suing to halt the application of differential privacy to the numbers that will be used for redrawing congressional and legislative districts nationwide.
At least seven states, including Colorado, are reallocating their prison populations for the first time, said Wendy Underhill, a policy analyst for the National Conference of State Legislatures. So far, none has come up with a great solution for how to reconcile the anonymized census data with the need to reallocate prison populations, she said.
“With the data coming later, it’ll take a couple of weeks to do that reallocation in the best of circumstances,” Underhill said.
If Colorado doesn’t make the prisoner adjustments this year, the state loses its chance for a decade. That’s because once the new legislative and congressional maps are approved, they’re in place for 10 years.
There is some disagreement between the two commissions about whether to forgo the prisoner reallocation this year. Although the Independent Congressional Redistricting Commission hasn’t taken an official position on the prisoner reallocation, commissioners Lisa Wilkes and Danny Moore spoke Monday at a meeting about the importance of counting prisoners where they came from instead of in the prison where they are housed.
“I don’t want to remove it at all,” Wilkes said. “It is a step that is really important to minorities being represented correctly.”
State Rep. Kerry Tipper, one of the prime sponsors of the 2020 legislation changing how prisoners are counted for redistricting, questioned whether reallocating prisoners will turn out to be as time consuming as staffers fear. A “significant” number of inmates report their last-known address at a prison anyway, Tipper said.
Absent an exemption, Barry said redistricting staffers are trying to do as much prisoner-reallocation work as possible ahead of receiving the census data, and are carefully watching to see if other states come up with a solution that could help Colorado meet its deadlines.
“I guess we will see if those other states are able to come up with other ways in which this can be done,” Barry said.