Become a Patron


Peter Wagner

Reprinted from:  Peter Wagner, Breaking the Census: Redistricting in an Era of Mass Incarceration, 38 William Mitchell Law Review 1241 (2012) (98 footnotes omitted)


Every decade, state and local governments redraw their legislative districts to ensure equal representation as required by U.S. Supreme Court precedent. When each district contains the same population, each member of the community is afforded equal representation. This undertaking, however, is vulnerable to any flaws in the data on which redistricting relies. A longstanding flaw in the Census counts incarcerated people as residents of the prison location, even though almost all are barred from voting and are not legal residents of the surrounding community. When district population counts include incarcerated populations, people who live close to the prison are given more of a say in government than everybody else. The practice of using prison populations to dilute the votes of residents in other districts is referred to as prison-based gerrymandering.

When the Census Bureau began counting Americans in 1790, it really didn't matter that the Bureau decided to count incarcerated people as residents of the prison. At that time, the data was only used for one purpose: to gauge the relative populations of each state to determine how many seats in Congress each received. It didn't matter where incarcerated people were counted within a state because legislative redistricting didn't yet exist. Also, until 1900, most federal prisoners were kept in state prisons, so even this miniscule number of people did not cross state lines. For more than a century, the impact of the Census Bureau's method of counting people in prison on the distribution of political power was about zero.

Further, in 1880, there was only one federal prison and sixty-one state prisons. At that time, the United States had only 61 people in prison for every 100,000 people in the population. That's just above one-twentieth of one percent--a tiny figure that reflects just how infrequent incarceration was. By 1923, the federal prison system had grown to three prisons, but the state system had the same number of facilities. The prison population had grown, but it was growing only slightly faster than the overall population. In 1923, the incarceration rate in the United States was, by Census Bureau figures, 74 per 100,000.

Drawing state and local legislative districts somewhat on the basis of population became more prevalent around this time, but there was not a clear federal requirement that governments must regularly redistrict on the basis of population equality until a series of court cases that began in 1962. Prison populations were, at worst, minimal blips in the redistricting data.

The 1990 Census was the first to show a sudden increase in the rate of incarceration, with the rate more than doubling over the previous decade to 292 people incarcerated per 100,000 residents. By 2000, the number of prisons had skyrocketed to 1,668, and the prison incarceration rate had risen to 478 per 100,000. That's almost one half of one percent of the U.S. population incarcerated in state or federal prisons.

Incarceration is of course not evenly distributed in the population, and racial disparities have been increasing. In 1923, Blacks were incarcerated at a rate four-times higher than Whites. By 2000, that disparity had almost doubled. At the time of the 2000 Census, just under 3.5% of Black men were in prison and being counted as residents not of their hometowns but of often distant prison towns.

When these prison counts are used in the redistricting context, the impact on state legislative districts is dramatic, for example:

Seven New York state senate districts drawn after the 2000 Census met minimum population requirements only because they use prison populations as padding.

In Maryland, one state house district in western Maryland drawn after the 2000 Census drew 18% of its population from a large prison complex. As a result, every four voting residents in this district were granted as much political influence as five residents elsewhere.

The policy and racial justice implications are severe as well, for example:

Virtually all--98%--of New York state's prison cells were located in state senate districts that are disproportionately White, diluting the votes of African-American and Latino voters. Similarly, in Connecticut, 75% of the state's prison cells are in state house districts that were disproportionately White.

Of the seven New York senate districts discussed above, four of the senators sat on the powerful Codes Committee where they opposed reforming the state's draconian Rockefeller drug laws that boosted the state's prison population. The inflated populations of these senators' districts gave them little incentive to consider or pursue policies that might reduce the numbers of people sent to prison or the length of time they spend there. One of them, Republican New York state Senator Dale Volker, boasted that he was glad that the almost 9,000 people confined in his district cannot vote because they would never vote for me.

The impact of prison-based gerrymandering on state legislative districting gets the most attention from policymakers, but the problem is even more significant in rural counties and cities that contain prisons. Their county board districts and city council districts are smaller than state legislative districts, so a single prison can have a massive effect. The most well-known example is in Anamosa, Iowa, where the state's largest prison constituted 96% of the city's second ward. In 2005, there were no second ward candidates for city election, and the winner won with two write-in votes, one cast by his wife and another by a neighbor. Citizen outcry about the unfairness of granting some residents twenty-five times as much political influence as other voters led Anamosa to change its form of city government.

The extreme example of Anamosa is far from unique. Other examples include:

Lake County, Tennessee drew a district after the 2000 Census where 88% of the population in County Commissioner District 1 was not local residents, but incarcerated people in the Northwest Correctional Complex. The result was that every group of 3 residents in District 1 [had] as much say in county affairs as 25 residents in other districts.

Half of one city ward in Rome, New York, drawn after the 2000 Census, was incarcerated, and the majority of the clout given to the Chair of the Livingston County New York Board of Supervisors came from claiming incarcerated people as residents of his town.

Wisconsin has a number of county and municipal districts where prisons constitute the majority of individual districts. The Waupun City Council drew a district after the 2000 Census that was 79% incarcerated, and Juneau County drew a district after the 2010 Census that was 80% incarcerated.

The most troubling example may be from Somerset County, Maryland where prison-based gerrymandering made it impossible to elect an African-American.

Somerset County, which until 2010 had never elected an African-American to county government, settled a voting rights act lawsuit in the 1980s by agreeing to create one district where African-Americans could elect the candidate of their choice. Unfortunately, a prison was built and the 1990 Census was taken shortly after the first election, leaving a small African-American vote-eligible population in the district. This made it difficult for residents of the district to field strong candidates and for voters to elect an African-American Commissioner. An effective African-American district could have been drawn if the prison population had not been included in the population count.

II. Federal Law Does Not Require Reliance on the Census Bureau

While state and local governments are required by federal law to redistrict each decade and typically use the decennial census to do so, federal law does not require that choice. Most governments rely on the U.S. Census for redistricting because the data is high quality and free. But the Supreme Court has said that states, and local governments by extension, are free to use other sources of data.

One Supreme Court case, Burns v. Richardson, implicitly approved the type of adjustments for prison populations discussed here:

Neither in Reynolds v. Sims nor in any other decision has this Court suggested that the States are required to include . . . persons denied the vote for conviction of crime in the apportionment base by which their legislators are distributed and against which compliance with the Equal Protection Clause is to be measured. The decision to include or exclude any such group involves choices about the nature of representation with which we have been shown no constitutionally founded reason to interfere.

When states draw their congressional districts, they are required to use the best population data available, and, although as a general matter, Census data will be the best available, states are not required to use data they know to be flawed, just because it is Census data.

Similarly, the Supreme Court in Mahan v. Howell rejected Virginia's argument that it was compelled to use Census Bureau assignments of residences of military personnel in its state legislative redistricting and suggested that a state may not use Census data it knows to be incorrect.

State and local governments are therefore free, at least under federal law, to create their own censuses from scratch, or to simply correct how the federal Census counts people in prison. Recently, a federal three-judge panel specifically rejected claims that adjusting Census data to count incarcerated people as residents of their legal home addresses for redistricting purposes is unconstitutional.

Not only does the federal government not require state and local governments to count incarcerated people at the prison location, but the Census Bureau has recently begun helping these governments do the opposite. The Census Bureau recently changed how it publishes its data to make it easier for states and municipalities to draw districts without including the prison population. Although this is not commonly understood, there is no one monolithic decennial Census. That is, there are actually at least three different decennial data products provided by the U.S. Census Bureau that are relevant to redistricting, each with its own purpose, methodology, and numbers. The first is the state-level counts, which include military and federal employees overseas and is used for Congressional apportionment. The second is the PL94-171 redistricting data file, produced since 1980, that is most commonly used for redistricting. The third is the Advance Group Quarters Summary file, produced for the first time after the 2010 Census, for the explicit purpose of helping governments with the problem of prison-based gerrymandering. As the Census Bureau explains:

This early release of data on the group quarters population may be beneficial to many data users including those in the redistricting community who must consider whether to include or exclude certain populations in redrawing boundaries . . . . It will permit state and local redistricting officials to overlay this file with the 2010 Census Redistricting Data (Public Law 94-171) Summary File data.

This redistricting cycle, Maryland and New York both relied on the Advanced Group Quarters Summary file in conjunction with the state's corrections departments' data to adjust the data used for redistricting. Courts in both states have approved the laws requiring these adjustments in order to count incarcerated people at their home addresses for redistricting purposes.

III. Legislative and Other Reforms

Four states have recently passed legislation that will count incarcerated people at their homes for redistricting purposes. Maryland and New York passed legislation effective for the 2010 round of redistricting. Delaware passed similar legislation although the state subsequently postponed implementation until 2020. Similarly, California passed a bill that will take effect in 2020.

The New York legislation applies to state legislative, county, and municipal governmental redistricting. The Maryland legislation is similar, but also applies to Congressional redistricting. The Delaware legislation applies only to state legislative districting. The 2008 amendments to the California Constitution transfer redistricting authority to the Citizens Redistricting Commission, so the California law asks the Citizens Redistricting Commission to deem each incarcerated person as residing at his or her last known place of residence, rather than at the institution of his or her incarceration when drawing future districts, and then mandates that the Department of Corrections and Rehabilitation provide the necessary data to the Citizens Redistricting Commission.

As discussed above, the greatest impact of prison-based gerrymandering is on county and municipal governments where a single prison population can easily make up the majority of a local government district. More than 100 counties and municipalities independently chose to reject the Census Bureau's prison miscount after the 2000 Census and drew districts without including the prison populations. In addition, in several states, including Michigan, Colorado, and New Jersey, exclusion of incarcerated populations is mandatory according to state statutes, and in Mississippi the Attorney General instructs counties to exclude the prison populations when redistricting:

[I]nmates under the jurisdiction of the Mississippi Department of Corrections as well as inmates of local jurisdictions in local jails . . . are not deemed residents of that county or locality, as incarceration cannot be viewed as a voluntary abandonment of residency in one locale in favor of residency in the facility or jail. . . . Such inmates should not be used . . . for redistricting purposes by virtue of their temporary presence in a detention facility or jail in the county, unless their actual place of residence is also in the county.

IV. Impediments to Reform

The biggest impediments to prison-based gerrymandering reform are policy inertia and basic misunderstandings about the mechanics of the criminal justice system and how federal and state funding formulas operate.

In a nutshell, the Census Bureau counts people in prison as residents of the prison location because that is where they have always been counted. Thankfully, Census Bureau policies are not fixed in stone, and as the country, its population, and its needs have evolved, so too has Census Bureau methodology. The problem of the Census Bureau's prison miscount and the prison-based gerrymandering that results is a problem new to the era of mass incarceration, and the Bureau's methodology needs to catch up with modern America in this regard.

Former Census Bureau Director Kenneth Prewitt neatly summarized the problem with the outdated methodology: Current census residency rules ignore the reality of prison life. Incarcerated people have virtually no contact with the community surrounding the prison. Upon release the vast majority return to the community in which they lived prior to incarceration.

The prison reality that Director Prewitt spoke of is the fact that, while the prison buildings themselves may exude permanence, the people inside are in fact quite transient. Indeed, in New York state, for example, for those incarcerated on January 1, 2008, the median length of stay for people at their current prison was only 7.1 months.

In no practical sense are people in prison residents of the prison location, and the one common context where states have contemplated the question--voting--they have concluded that a prison cell is not a legal residence. Most states have explicit constitutional provisions or statutes that declare that a prison cell is not a residence. As discussed above, people in prison are generally barred from voting, and in the rare cases where people in prison can vote, they must always vote absentee at home, not in the district in which the prison is contained.

As mentioned, another impediment to reform is often an unsupported fear of negative changes to federal or state funding. Our research has found that both sides of the debate often rely on an oversimplification of how Census data is used to distribute funds and therefore overstate the impact of prison counts on funding formulas. In general, prison populations have very little impact on the distribution of federal and state funds, and the state and municipal solutions discussed in this article would have no impact whatsoever because there are no formulas tied to redistricting data.

The confusion about funding arises in part because the Census Bureau encourages participation in the Census by appealing to the important use of Census data in funding formulas. This leads to a misunderstanding about how the population data is actually used. Most large federal and state funding formulas, particularly those targeted to individual municipalities or school districts, do not use total population for their population component. Instead, they use more targeted factors, like people in poverty (which does not include people in prison or other people not in households), the number of school-age children, or non-Census data like the number of children enrolled in school. As a result, the impact of prison populations on funding formulas tends to be quite small.

Furthermore, each funding formula is a complicated effort to match the program's resources with the need being addressed. Each formula has its own specific data sources and methodology, none of which rely on state or local redistricting data. Thus, any changes to the redistricting data, be it for state redistricting or municipal redistricting, will have no effect whatsoever.

This logical conclusion is confirmed by the experience of the more than 100 rural counties and cities that removed the prison populations when redistricting after the 2000 Census to no ill financial effect.

Not only is there nothing to lose in ending prison-based gerrymandering, but everyone who does not live immediately adjacent to the state's largest prison complex benefits in at least one way if the practice is ended. At the state level, everyone living outside the state legislative district with the largest prison benefits; and the majority of the people living within that state district benefit at the local level from fairer county and municipal redistricting. The successful reform efforts to date have all been structured to maintain a broad coalition of all the interest groups that would benefit from reform.

In New York, where a bill to end prison-based gerrymandering ultimately passed on a narrow partisan vote, a Quinnipiac University poll shows it was supported by the majority of the state, urban and rural, Democrat and Republican, and the bill received editorial support from urban and rural upstate papers. Once the laws passed, rural supporters fought to keep them in place, resisting pushback from New York legislators responsible for drawing the new district lines. When a lawsuit was filed in New York challenging the new law, residents from every area of the state intervened to defend the law.

In Delaware, a bipartisan reform bill passed unanimously in the House. In Maryland, too, the legislation passed with bipartisan, urban and rural support. One White Republican state senator spoke from the floor about why he was voting for the bill, and both lead sponsors had massive prisons in their districts. As Delegate Joseline Peña-Melnyk explained, It doesn't matter, . . . . To me, it is just a fair way to count. Senator Catherine Pugh agreed, It was the right thing to do.

V. Solutions and Best Practices

The ideal solution is for the Census Bureau to count incarcerated people as residents of their home communities rather than of the prisons where they are incarcerated. This federal fix would solve all of the problems that lead to prison-based gerrymandering, but state and local governments that don't want to wait for the federal government to act have to find their own solutions.

In most states, the state constitution is silent on the data source to be used for redistricting, leaving states free to pass legislation to improve the federal Census data as Maryland and New York have done. A notable exception is Massachusetts, where restrictive language mandates the use of the federal Census for redistricting, leaving the state with the larger task of amending the state constitution or a more speculative effort to lobby the Census Bureau for a change.

Model legislation --prepared by a coalition of civil rights, voting rights, and criminal justice reform organizations working to end prison-based gerrymandering--offers a basis for ending prison-based gerrymandering in states that have the constitutional ability to pass a law reallocating incarcerated people to their home addresses. Various political and practical realities may dictate other choices, but the model bill recommends three, somewhat subtle, best practices:

(1) The legislation should grant a specific non-legislative agency the task of receiving the Department of Corrections' data and performing the reallocation procedure and there should be specific deadlines for this work to be completed. In many states, the Secretary of State is an ideal choice, particularly where the Secretary plays an active non-partisan role in elections administration and has the necessary technical skills. Experience has shown that the question of who is responsible for the reallocation can have a major impact on the process. In Maryland, for example, the statute did not specify who would do the reallocation, but the Maryland Department of Planning took the initiative and did an impressive job. New York's bill gave this task to the Legislative Taskforce on Demographic Research and Reapportionment (LATFOR), the partisan redistricting taskforce, but did not specify a deadline. Partisan wrangling over technical implementation delayed the completion of the process, leading to advocates' concerns and a federal lawsuit alleging that the legislature was unable or unwilling to implement the law. Tasking an independent agency with implementation would not remove the risk of a legislature repealing the law, but it would separate the minor technical issues from the larger policy ones.

(2) The legislation should apply to county, municipal, and other local districts as well as to state legislative districts. While most county and municipal governments already avoid prison-based gerrymandering when redistricting, the exceptions are dramatically negative, and the entire process is inconsistent and cumbersome. Politically speaking, proposing one consistent data set for state and local redistricting is a proven way to build urban and rural coalitions to improve democracy for everyone.

(3) The legislation should specify that when the proper residential address of an incarcerated person is unknown or in another state, the redistricting data should reflect that person as being counted at an unknown geographic location within the State. People at unknown geographic locations should not be included in the calculations for ideal district size or population deviations. The method is similar to the way overseas military are counted as at-large residents of a state for congressional apportionment but not included in specific districts. Notably, this specific aspect of the model bill was explicitly endorsed by the NAACP in a 2010 Convention resolution.

In addition to the concerns addressed by the above best practices, at the county, municipal, and other local levels of government, there is tremendous variation in how governments adjust the Census and in the level of detail given to documenting the rationale. I'd like to identify some best practices in this regard.

Some municipalities and counties adjust the Census figures, some cut a hole in their map where the prison is, and some overpopulate the district that contains the prison by the exact size of the prison population. To the line drawer, these methods are very different, but the outcome of each is identical, and the redistricting professional's convenience should dictate the methodology.

The justifications and documentation of the redistricting process are more important. In some cases, municipalities and counties note the adjustment on their redistricting map, but the best practice is illustrated in New York's Essex County, where the county explained its rationale for excluding the prison population in Local Law Number 1 of 2003:

Persons incarcerated in state and federal correctional institutions live in a separate environment, do not participate in the life of Essex County, and do not affect the social and economic character of the towns in which . . . the correctional facilities where they are incarcerated are located.

The inclusion of these federal and state correctional facility inmates unfairly dilutes the votes or voting weight of persons residing in other towns within Essex County. This is particularly so if the 1,898 inmates in the town of North Elba are included in its population total of 8,661 since those inmates would then represent 21.914% of the town of North Elba's population.

The Board of Supervisors finds that the population base to be utilized in and by the plan apportioning the Essex County Board of Supervisors should exclude state and federal inmates. Putting this rationale into the public record would show the basis for an adjustment to any court looking at the districts, and would make it more likely that the legislature in a decade's time would recall and repeat the previous decision. In our research, we've found many examples where municipalities and counties were unaware of the basis of their previous maps until we performed a population analysis for them.

Finally, the ideal solution is for the Census Bureau to count incarcerated people as residents of their home communities, not the prisons where they are incarcerated. The Census Bureau has the legal discretion to determine where to count people in prison. Fortunately, the Census Bureau director has given reason to be hopeful, writing in his blog: Counting members of all group quarters is complicated; we re-evaluate our residence rules' after each census, to keep pace with changes in the society. We'll do that again after the 2010 Census.

The challenge for advocates seeking change at the Census Bureau is that the public and policymakers alike tend to pay attention to Census issues only in years ending in 9, 0, and 1 when the data is being gathered and published. The critical policy decisions--and the scientific research to support those policy decisions--naturally take place in the middle of the decade, though, when interest is less intense. We must make sure the Census Bureau asks the right questions to inform its decision about improving where incarcerated people are counted. If this mid-decade opportunity is missed, states and local governments will be forced to continue to develop solutions on their own.

The problem of prison-based gerrymandering is a historical accident. The Census Bureau never intended for your right to vote to depend in large part on whether or not you live next to a large prison. The combination of an old methodology, an unprecedented change in incarceration patterns, and a modern constitutional mandate to draw districts on the basis of equal populations has created an undeniable problem for our democracy.

Of course, if mass incarceration ended tomorrow, the need for the Census Bureau to update its prison counting methodology would evaporate. Until then, we need to make sure criminal justice policy--and all policy--decisions are made by the willing majority and are not the result of the Census Bureau counting 2 million incarcerated people in the wrong place.



. Peter Wagner is Executive Director of the Massachusetts-based Prison Policy Initiative.