For decades, redistricting arguably has been the most powerful political act legislatures perform, shaping representative democracy for decades at a time. And the current cycle is just beginning, with most of the work still ahead in the last half of 2021 and possibly through the first part of 2022.
Some redistricting facets are the same from decade to decade. For instance, plans must adhere to the one-person, one-vote principle, which means districts must be of equal population; maps must guarantee minority rights, while not relying too heavily on race as a factor; and redistricters must adhere to state-specific rules and regulations, some of which are as old as time, while others are new to this decade.
This cycle presents new wrinkles. Not only did the pandemic wreak havoc on the collection and release of census data, but outside advocates are paying far more attention and have far more sophisticated tools for comparing maps with what they consider ideal. And one of those newly adopted state-specific criteria—how prisoners are accounted for—is proving to be a conundrum.
Census Data Unknowns
Delays. Deadlines. Data quality. Disclosure avoidance. The 2020 decennial census has redistricters asking questions—and answers are in short supply.
Here’s what we know: Counting everyone living in the United States last year proved harder than expected and led to monthslong delays. The U.S. Census Bureau has so far released just one set of data: state total populations, the numbers used to apportion seats in the U.S. House among the states.
That release is only the opening act. The main event—the release of detailed data, including information on race, ethnicity and voting age population—is coming in mid-August, more than four months later than expected. That’s the data redistricters use to draw new congressional, legislative and local electoral district boundaries once every 10 years.
One thing that won’t change is the fact that the courts will have the final say about how these issues are resolved. —Michelle Davis, co-author of NCSL’s ‘Redistricting Law 2020’
Here’s what we don’t know: how states will respond to those delays, given that many have redistricting deadlines in their constitutions. Some states have asked their state supreme courts to provide extensions. Others are using numbers besides those provided by the decennial census to draft maps, something that would have seemed laughable a year ago, given that decennial data is the gold standard (and constitutionally mandated in some states). Now, using alternative data seems a reasonable response to a hard situation. Many states are looking at fall special sessions to do their redistricting or have chosen not to gavel out of their regular sessions while awaiting the necessary data.
Here’s what else we don’t know: how accurate the census data will be. With the apportionment numbers, some states are already raising questions, given that the distribution of congressional seats proved to be further off from projections than in previous cycles. For instance, projections showed Arizona, Florida and Texas gaining one, two and three additional seats, respectively. Instead, they received zero, one and two additional seats. Is that a problem? Up north, New York’s population count was 800,000 people higher than projected, and that meant it lost just one seat, instead of a projected two seats. Rhode Island unexpectedly kept its second seat. Were earlier estimates way off, or was the count off? And how do we know?
There’s another data-quality wrinkle that’s new this cycle. While state total populations are reported as counted, all other data the census releases will be treated with “differential privacy,” a new approach to protect respondent confidentiality. Differential privacy adds “noise” to the data so it will be harder to reverse engineer any individual responses. Unfortunately, that means data won’t be reported exactly as counted. The question: Does a little change here and there to protect privacy really matter when it comes to redistricting?
Redistricters want to know the answer because they must build districts that comply with one person, one vote and provide, where appropriate, districts that allow minority groups to elect candidates of their choice as guaranteed by the Voting Rights Act. Some say the effect of differential privacy makes it difficult to assess where and when the VRA applies. Others say the noise cancels out at the district level, so little variations at the block level are meaningless. A federal three-judge panel in Alabama v. U.S. Department of Commerce will decide whether to grant the plaintiff’s request that the Census Bureau provide accurate block data.
Distribution of political power within and between states is what’s at stake in all of this. “Despite the novel issues we are having this decade with census data, the one thing that won’t change is the fact that the courts will have the final say about how these issues are resolved,” says Michelle Davis, co-author of NCSL’s “Redistricting Law 2020.”
Judging Maps With Math
Redistricting battles based on whether maps favor one party over the other are old news. But how those battles are fought is changing—and this decade, attorneys and advocates are wielding new and improved weapons: statistical analyses made possible by leaps in computing power.
In the past, the complexity of redistricting meant the process was closed to all but political insiders. But today’s computing has opened the doors for people outside the redistricting process to analyze maps for partisan tilt—either during the drafting phase or before going to court.
“This cycle is going to have a much faster response to maps once they get proposed than we’ve ever seen before,” says Hannah Wheelen, senior analyst at the Electoral Innovation Lab at Princeton University.
The 2010s saw numerous cases alleging unconstitutional partisanship in redistricting, and expert witnesses in those cases both perfected existing mathematical tests for analyzing district maps and developed new tests of their own. For instance, the “efficiency gap” made its debut in an unsuccessful challenge to Wisconsin’s state legislative districts in Gill v. Whitford.
The efficiency gap and the oldie-but-goodie mean-median test weren’t designed to be one-size-fits-all. “The important thing about [statistical tests for partisanship] is they work really well for some states, but don’t say much about other states. So, you have to take a very state-specific approach,” Wheelen says.
The mean-median test measures partisanship by subtracting the median vote share of either party across all districts from the average vote share of the same party across all districts. It works better in states that are closely divided politically between the two major parties than it does in states where one party has a clear advantage, Wheelen says. The efficiency gap, which measures “wasted votes”—the number of votes over 50% each party receives in districts they win, plus all losing votes—has similar efficacy.
One way around this uses mathematical ensembles to draw hundreds of thousands or even millions of maps in a particular state. Those maps create a distribution of possible partisan outcomes in a state, or a “bell curve” of possible plans. Drafters—or, later, public mapping groups or adversaries in court—can then place any specific redistricting plan on the curve to see where it falls. This “outlier” analysis enables people to instantly know if a redistricting plan is more Democratic-leaning or more Republican-leaning than other possible plans in a state, as shown by the curve. Legislators can proactively use these tools to better understand their own maps and how they might be perceived by the public.
“It’s going to be hard for any maps to sneak through that have built-in advantages for incumbents or others,” Wheelen says.
Prisoner Reallocation for Redistricting
Home is where the heart is—unless you’re the Census Bureau. In that case, home is where a person eats and sleeps.
Applying its “eats and sleeps” definition, the bureau counts prisoners at the address where they were incarcerated on Census Day (April 1 of a census year, or April 1, 2020, this cycle). Because prisons are often built in rural areas, the bureau’s residency policy, while logical in other contexts, “creates a disconnect from where people reside versus political power,” says Aleks Kajstura, legal director of the Prison Policy Initiative.
Imagine two neighboring rural counties, each with about 25,000 people. One of the counties has a state penitentiary with 5,000 prisoners counted among its 25,000. Except in Vermont and Maine, those incarcerated people rarely vote and don’t consider the prison to be their home, and elected officials from the region don’t see the prisoners as their constituents. But in most states, they “count” at the prison address for redistricting purposes. Effectively, the elected officials in the county with the prison represent 5,000 fewer people than those in the neighboring county. Does that mean they deserve a bigger voice in the legislature than their neighbors?
“The most comprehensive solution for states to redress this issue is to count everybody at home,” says Kajstura, who considers “home” to be the pre-prison address.
To reallocate prisoners according to this concept, states can compare prisoners’ addresses from the department of corrections data file with the population at the prison as reported in the census. Then, for redistricting purposes only, the state can virtually “move” the prisoners back to their pre-incarceration addresses, and in effect move the political power associated with the head count as well.
This won’t impact the flow of federal funds tied to the census, Kajstura says, because this newly created dataset wouldn’t be used for anything other than drawing districts. The data management task is no small job and can take a couple of weeks. This year, with a monthslong delay in the release of census data needed for redistricting, every week counts.
In the 2010 cycle, New York and Maryland reallocated prisoners for redistricting. California, Colorado, Delaware, Nevada, New Jersey, Virginia and Washington all enacted legislation in the last decade to do the same; Illinois has adopted the policy for the 2030 redistricting cycle.
Another option, chosen by the rest of the states: go along with the Census Bureau’s rule on residency for prisoners, and count prisoners where they sleep.
Wendy Underhill directs NCSL’s Elections and Redistricting Program; Ben Williams is a policy specialist in the program.
This story was first published in the Summer 2021 edition of State Legislatures magazine.