Why the Census Bureau can and must start collecting the home addresses of incarcerated people
Submitted by Peter Wagner, Eric Lotke and Andrew Beveridge to the U.S. Census Bureau on February 10, 2006
in advance of the Bureau's report to the Appropriations Committee on using prisoners' homes of record in the Census
Traditionally, the Census Bureau has enumerated incarcerated people as residents of the Census block that contains the prison in which they are incarcerated. The Census Bureau has long held that a fair and equitable Census count requires not just a complete count, but a count of people in the correct place. New uses of Census Bureau data and changing demographics require the periodic examination of the residence rules.
Since the 1980 Census, the percentage of Americans incarcerated in correctional facilities has increased four-fold. Recent research has shown that correctional facilities are frequently located in areas that are geographically and demographically far removed from the communities to which most incarcerated people belong. Counting prisoners as residents of the prison location causes unexpected distortions in Census data for rural communities and creates significant burdens on state and local legislative data users who rely on the Census for redistricting purposes.
The original constitutional mandate of the Census was simple: to determine the relative population of each state for Congressional apportionment. As American society has grown more complex, so too have the residence rules used to determine where to count the population. Over the last two hundred years, the Census has changed how it counts college students, missionaries, overseas military and many other groups. Counting prisoners at the facility probably made sense at the first Census in 1790 when few people were in prison and when the data was not used for redistricting. Things are different now, and the data collection methods need to change. The means used to count people in prison now requires a similar update and change.
Increasingly over the last century, Census data down to the local level has been used to guide social and governmental research and policy. Starting in the 1960s, the Supreme Court issued a series of "One Person One Vote" decisions which required state and local governments to redraw legislative district boundaries at least once per decade in order to ensure that districts were equal in population and that each resident had an equal stake in government. These cases created in state and local government a new class of data users who needed accurate demographic data down to the block level in order to draw legislative districts. Since 1975 when Congress passed PL94-171, the Census Bureau has worked closely with local governments to make Census Bureau data useful for local redistricting by adjusting Census Bureau geography to follow local political boundaries. While the Census Bureau has been responsive to matching the shape of Census blocks and tracts to local boundaries, it has not been so responsive to counting only "local" residents within those boundaries.
Prisons present a significant distortion on local populations. Currently, there are more than 2 million people in prisons and jails. Since the 1980 Census, the percentage of Americans incarcerated in correctional facilities has increased four-fold, with more than 0.7% of Americans currently incarcerated. Certain demographic groups are disproportionately affected by these trends; for example more than 12% of the population of African-American men in their late 20s is currently incarcerated.
Recent research has shown that correctional facilities are increasingly located in areas that are geographically and demographically far removed from the communities to which most incarcerated people belong. According to Department of Agriculture Demographer Calvin Beale, although non-metro counties contain only 20% of the national population, they are the host for 60% of new prison construction. In the 1990s, an astonishing 30% of new residents of upstate New York were people being sent to prison.
The result of counting large external populations of prisoners as local residents leads to misleading conclusions about the size and growth of communities. One study of incarcerated populations in the Census found 21 counties in the United States have at least 21% of their population in prison. Counties that see prisons close their doors report that their Census populations declined when in fact their actual population did not. Conversely, population growth reported by some counties is due to the importation of prisoners to a new correctional institution. If not for the construction of new prison cells, 56 counties the Census Bureau identified as growing during the 1990s would have in fact reported declining populations.
Because Latinos and Blacks are incarcerated at three to seven times the rate of Whites, where incarcerated people are counted has tremendous implications for how Black and Latino populations are reflected in the Census. For this reason, the African-American subcommittee of the Census Bureau's Race and Ethnicity Advisory Committee recommended that the Census Bureau count prisoners as residents of their pre-incarceration addresses.
Unlike other "group quarters" populations, prisoners do not choose to be at the facility and are not considered part of the surrounding community. Contrary to popular understanding, most prisoners are serving short sentences. The average state prison term served in the United States is 34 months. Typically, individuals move to several different prisons during the course of their stay, at the discretion of the custodial agency. When a prisoner's sentence is completed, he or she will not be permitted to stay in the facility and will most likely return to the community that he or she was removed from when the incarceration began. The 784,000 people in local jails on a given day are either awaiting trial or are serving sentences that are generally under one year. Thus, counting prisoners as residents of the correctional facilities incorrectly equates an involuntary short-term stay with a permanent move to the prison community. This practice is inconsistent with other census rules which properly credit hotel guests and other temporary visitors to their home communities.
Reflecting the involuntary and short-term nature of incarceration, and a concern that including the residents of correctional facilities could distort the local democratic process, many state constitutions define residence for a prisoner as the place he or she resided prior to incarceration.
In November 2005, the Appropriations Committee directed the Census Bureau to:
"undertake a study on using prisoners' permanent homes of record, as opposed to their incarceration sites, when determining their residences. The Bureau should report back to the Committees on Appropriations on its findings within 90 days of enactment of this Act."
This document is intended to review the importance of this issue and address the legal and practical concerns that have been raised by the Bureau in the past. We review the existing evidence and conclude that is both desirable and possible to count prisoners as residents of their pre-incarceration addresses in the PL94-171 redistricting dataset. Finally, we offer some suggestions as to how the Census Bureau can study this issue, collect the data and then publish it in the next decennial Census.
II. State and local legislators have unique data needs that require incarcerated people to be counted at home.
Counting prisoners as residents of the prison towns rather than of their homes affects most data users and all of the Census Bureau's data products. While the number of uses of Census Bureau data is inconceivably large, we imagine it would be preferable for most uses that prisoners not distort the statistics of the prison-hosting community. In many cases, particularly for studying high incarceration communities, it would be preferable to have those people counted at home. But redistricting is different than this discussion of data user preference because the legislatures actually need prisoners counted differently.
For that reason this document focuses on the two major groups that use the PL94-171 redistricting data set -- state and local legislatures -- and discuss how these two different groups are negatively impacted by the current method of enumerating prisoners. To avoid the most significant distortions in districting, state legislatures need prisoners counted at home, where as local legislatures merely need to be able use a PL94-171 dataset that does not contain the prison populations.
Each decade, state legislatures use Census data to redraw congressional and state legislative districts. The current practice of crediting prisoners to the census block with the prison results in drawing legislative districts near prisons that have large numbers of non-residents. One legislative district in New York is 7% prisoners; a legislative district in Texas is 12% prisoners; and 15% of one Montana district is prisoners imported from other parts of the state. The existence of these legislative districts with artificial populations increases the voting strength of the real prison-district residents and dilutes the voting strength of residents in every other district in the state. The urban communities that have large numbers of residents incarcerated outside the district, see their voting strength diluted beyond that of the average district in a state, though it is impossible to calculate this precisely because the necessary Census data on origin has never been collected.
The impact on local legislatures, such as county boards, is even more pronounced. Because the district sizes tend to be relatively small, a single prison can have a significant impact. In our research, we have discovered a number of districts where the majority of the population in some districts are prisoners from other parts of the state. Rural residents who live in the same community as a prison but not immediately adjacent to it frequently see their votes diluted on local issues quite severely. This issue and the controversy it causes in New York State is most thoroughly described in this July 2004 comment letter from three residents of rural Franklin County New York to the Census Bureau's Redistricting Data Program:
We are writing to express our concern that the Census Bureau's method of counting incarcerated people as residents of the facility makes it difficult to use Census data in local county redistricting. We are two former legislators in Franklin County, New York and a private citizen. All of us have been involved in redistricting litigation to enforce the creation of equally sized districts.
While the Census Bureau is very helpful in adapting its geographic units so that county government can readily use Census data in local redistricting, we have found that who the Census Bureau counts within our election districts creates unnecessary controversy and confusion. Left uncorrected, Census counts of prisoners leads to a significant change in how our county legislatures work.
Franklin County is in Northern New York on the Canadian border. Census 2000 reported our population at 45,622 including 5,512 state prisoners in 5 state prisons. Almost 5,000 of the prisoners are in 3 large prisons in the Village of Malone. We do not consider the prisoners to be residents of our community as they originate outside of our community, have no interaction with it and immediately leave the district when their sentences expire or the Department of Corrections chooses to transfer them elsewhere.
Franklin County has always excluded state prisoners from the base figures used to draw our legislative districts. To do otherwise would contradict how we view our community and would lead to an absurd result: creating a district near Malone that was 2/3rds disenfranchised prisoners who come from other parts of the state. Such a district would dilute the votes of every Franklin County resident outside of that area and skew the county legislature. We know of no complaints from prisoners as a result, as they no doubt look to the New York City Council for the local issues of interest to them.
While Franklin County has consistently excluded prisoners from its redistricting population, our research into the practices in other counties has revealed a diversity of approaches. This diversity stems not from a difference of opinion on whether prisoners are a part of the prison town, but from a difference of opinion as to whether it is permissible to modify Census Bureau figures. We will give the examples of Greene County (south of Albany), St. Lawrence County (directly to our west) and Wayne County (east of Rochester).
Prior to 2002, the Greene County New York Legislature used unadjusted Census figures for its county redistricting. An increase in the prison population during the 1990s meant that 6% of the county's Census population was in two state prisons in Coxsackie in the northeast corner of the county. When The Daily Mail reported that Coxsackie would be getting an extra legislator from the arrangement, there was a large protest and the county reversed course and drew district lines based on data that excluded the prison population.
St. Lawrence County, which borders Franklin, is the only county we know of to have previously excluded prisoners from its redistricting base and now includes them. The county did this on the basis of advice from their county attorney about a 1993 state case, which other counties read to require exclusion. Factually separate from the legal analysis, the legislators admit that the prisoners have no stake in issues of county taxation or policy making. Most critically, there was a large public effort in the county to attempt to repeal the most recent redistricting legislation precisely because large external populations were included making the votes of 8 residents near the prisons worth as much as 10 residents elsewhere in the county.
According to our survey, about 1/3rd of the New York counties with prisons exclude prisoners from their local redistricting population base. This may be a minority of counties, but with the exception of St. Lawrence County, we know of no local county that included the prisoners after the issue became known to the public. In fact, our research in Wayne County may express the magnitude of the problems caused by how populations are credited to our communities.
The Wayne County Board of Supervisors is apportioned with weighted votes to each town. The Board told us that they do not consider prisoners to be residents and including prisoners in the Butler Shock Camp would have a significant impact on the weighted voting scheme. However, the County was under the incorrect assumption that prisoners were not credited to the town of Butler. While the Board was aware that Butler had a small population, its actual population is much smaller than the Board -- relying on Census figures -- had been lead to believe.
We appreciate the data that the Census Bureau supplies to our communities and thank you for your efforts to make it easier to use this data in our redistricting efforts. As part of your deliberations establishing the 2010 Redistricting Data Program, we urge you to not count prisoners as residents of our communities.
We are continuing to explore how local communities respond to the problem of drawing legislative districts with significant numbers of prisoners who are not members of the community. Colorado and Mississippi require county governments to exclude the correctional populations in the census from county districts. New Jersey requires school board districts to do the same. Virginia law encourages counties that have a Census population that is more than 12% prisoners to exclude the correctional populations prior to redistricting. We have also identified individual communities in California, Florida, Massachusetts, Illinois and Louisiana that without explicit state authorization have excluded prison populations prior to redistricting. But, as will be discussed in section VIII C, the current format of PL94-171 makes this very difficult for rural communities to do, and with only a small change the Census Bureau could facilitate this type of simple data correction.
Because available data on prisoner origin does not allow an analysis of the impact on urban city council districting, this document concerns itself with the very large problem created for small prison hosting communities that must draw legislative districts on the basis of population.
III. How prisoners are counted has occasional effect on funding.
Our review of state, federal and local funding streams reveals that where prisoners are counted has an occasional impact on governmental financial distributions. The most rigorous examination of how prisoner counts affect funding is "Prisoners of the Census: Electoral and Financial Consequences of Counting Prisoners Where They Go, Not Where They Come From" in the Pace Law Review. Although Census data does play a role in the distribution of more than $1.5 trillion each decade in federal funds, the vast majority of these funds are unaffected by where prisoners are counted. The two largest programs, Medicaid and Highways, amount for 74% of the total, though as block grants to states they are unaffected as long as prisoners stay within the same state. The majority of other federal funds are in tailored programs that match the program to the need. The low incomes of prisoners do depress per-capita income figures, but they are not included in the more frequently used household income and poverty statistics. Simply put, most government programs are designed with sufficient precision that they are not affected.
In contrast, state sales taxes are often affected by the decisions of the Census Bureau. Most states distribute sales taxes with portions going directly to the state general revenues, portions returning to the point-of-sale, and portions going to municipalities or other local governments on the basis of population. Thus, communities that host a prison receive an additional, unearned portion of the state sales tax.
Another example is Virginia's method of distributing state aid for K-12 education. Unlike other states, that use formulas that focus on pupils enrolled, Virginia distributes state aid based on a complex formula that includes the county population as a factor. The end result is that Sussex County, which has a Census population that is 19% people in prison, receives $115,000 extra as a result of the incarcerated population. In contrast, Henrico County (Richmond) loses roughly $200,000 as a result of "exported" population. It is not a great deal of money, but it illustrates the kind of low-level distortion created by population counts.
Other affected fund distributions include U.S. Department of Agriculture's Appalachian Regional Commission which makes $60 million available annually to impoverished Appalachian communities based on a formula that includes total population of each county. Such formulas inadvertently reward communities that build prisons by giving them a larger share of the funds intended for their region. Typical of the distortions that the Census creates, the communities that pay the largest price are not the urban communities that most people in prison call home; rather, they are similarly situated rural communities that lack the advantage of a population artificially inflated by a prison.
IV. The law does not prohibit -- and may require -- the Census Bureau to collect incarcerated people's home addresses.
The Census Bureau has frequently deferred an honest discourse on the best place to count incarcerated people by pointing to D.C. v. Commerce, 789 F. Supp. 1179 (1992) that held that the Bureau's practice was not "arbitrary and capricious." The Bureau has been less eager to discuss whether its policy is wise or whether it is permissible under equal protection voting rights jurisprudence.
IV. A. D.C. v. Commerce does not prohibit the Census Bureau from exercising its discretion to modify the "usual residence rule".
The Census Bureau's conclusion that D.C. v. Commerce, 789 F. Supp. 1179 (1992) requires it to continue to count prisoners as residents of the prison towns is incorrect. In that case the Court held merely that the current counting methodology was not "arbitrary and capricious" under the Administrative Procedure Act, that the Court was not called upon to perform a de novo review, and that the Census Bureau was not, as a matter of law, required to use the "best available" procedure. (p. 1188). This ruling is hardly a ringing endorsement of present practice; indeed, it is virtually an invitation to do better.
The Census Bureau's African-American subcommittee of the Race and Ethnicity Advisory Committee accepted the invitation. In October, 2003 it recommended that:
"... prisoners, including those housed outside their states, be counted as residents of their pre-incarceration addresses for purposes of Congressional apportionment, state redistricting, the distribution of financial aid to their home communities, and to permit those communities to have more resources to meet their needs upon their return."
The Census Bureau's African-American advisory committee is a subcommittee of the larger Race and Ethnicity Advisory Committee, a group appointed to offer guidance to the Census Bureau on policy questions of race and ethnicity. Given the reality that 2.5% of African-Americans in the United States are currently incarcerated -- in prisons and jails often far from their homes -- how incarcerated populations are counted is of particular concern to those seeking a complete, accurate, and useful count of the African-American population.
Shortly after the advisory committee meeting, the Census Bureau issued the following response, which skirts the importance of this issue to a fair and accurate count of African-Americans and misstates the applicability of D.C. v. Commerce:
Currently, we have no plans to reevaluate the methodology that we use for counting prisoners at the institution where they are incarcerated....
In 1992, the United States District Court for the District of Columbia confirmed the principle of usual residence (District of Columbia v. U.S. Department of Commerce). In that case, prisoners who were residents of the District of Columbia were incarcerated in a prison located in the State of Virginia and, consequently, counted as residents of Virginia in the 1980 census. The District of Columbia claimed that this action violated the Administrative Procedures Act, Article 1, Section 2, Clause 3 of the Constitution (the authority for taking the decennial census for the purpose of apportionment), and the "due process clause" of the fifth amendment, and that it would cost the District $60 million over the decade.
The court ruled that the application of the Census Bureau's "usual residence rules" to prison inmates was a rational decision that was neither arbitrary nor capricious, and that it did not violate the constitutional command of the census clause. In other words, the court decided the Census Bureau could count the inmates at their usual place of residence as it is defined above.
This statement indicates a misinterpretation of D.C. v. Commerce. First, while the case does not require a change in the usual residence rules, neither does it prohibit one. More importantly, the case was brought by the plaintiffs after the 1990 Census was already taken, and it alleged that the Census Bureau's enumeration of D.C. prisoners as residents of Virginia was going to cost the District federal funds. Voting rights were not made a part of this after-the-fact lawsuit filed against the Census. The court made it clear that the Census Bureau has no control over how various federal agencies use the data to distribute funds, and the judge recommended that the District of Columbia seek to change how federal funds are distributed.
Under the Administrative Procedures Act, a court may reverse the actions of an agency in matters like this only where the agency's action was "arbitrary or capricious." Showing that a defendant's actions did not meet "mere rationality" is very difficult for a plaintiff to do, as the plaintiff must show not that she would have preferred a different outcome, but that the agency failed to consider the relevant factors and made a clear error of judgment. (D.C. at 1185.) Given the chaos that would be caused by courts second-guessing Census Bureau decisions after the Census is complete, such a high legal standard makes sense.
Under the facts alleged in D.C. v Commerce, counting prisoners as residents of the prison location in the 1990 Census was not an arbitrary and capricious act. But here, the African American Advisory Committee and this submission suggest that the 2010 Census should count prisoners at home. A court case saying that a methodology wasn't clearly wrong, doesn't mean that the Bureau is required to do it again, especially when the Bureau's own appointed advisors on African-American issues recommended a change more than six years before the next census.
IV. B. The law may require a change in how prisoners are counted under a 14th Amendment equal protection and voting rights analysis.
A post-Census claim to redo the Census because of an alleged impact on federal funds may be precluded by D.C. v. Commerce. However, a lawsuit alleging a violation of the 14th Amendment's due process protections because of the impact on voting rights would likely trigger review of the Census Bureau's counting method under the strict scrutiny standard. Under this standard, the burden would be on the Bureau to defend its decision to not facilitate the drawing of proper state and local legislative districts.
In 1963, the Supreme Court decided Reynolds v. Sims, in which it interpreted the 14th Amendment's equal protection clause to require state legislative districts to be divided on an equal population basis. As the court remarked, "legislators represent people, not trees or acres. Legislators are elected by voters, not farms or cities or economic interests." The principle is that "the weight of a citizen's vote cannot be made to depend on where he lives." In violation of this principle, current Census methodology means that the weight of a citizen's vote greatly depends on whether a large prison is nearby.
Crediting prison communities with large external population of prisoners -- who are local residents in no sense other than in the Census -- turns the "One Person One Vote" principle on its head. There can be no doubt that rural prison towns do not share a "community of interest" with urban prisoners or their loved ones. Above, in section II, three residents of rural Franklin County New York argued that they do not consider prisoners to be residents of their community. Taren Stinebrickner-Kauffman surveyed the Indiana legislature and found the same thing:
I sent to all members of the lower house of the Indiana state legislature a brief survey that included the following question:
Which inmate would you feel was more truly a part of your constituency?
a) An inmate who is currently incarcerated in a prison located in your district, but has no other ties to your district.
b) An inmate who is currently incarcerated in a prison in another district, but who lived in your district before being convicted and/or whose family still lives in your district.
Every single one of the forty respondents who answered the question -- regardless of their political party or the presence or absence of a prison in their district -- chose answer (b). Had the responses been more ambiguous there might have been reason to repeat the survey with other groups of legislators. However, unless there is something highly anomalous about Indiana, it is quite clear that representatives do not consider inmates to be constituents of the districts in which they are incarcerated - unless, of course, they happen to have prior ties to those districts.
Drawing state and local legislative districts that contain large numbers of prisoners from outside the district is not just unfair, it frequently has a larger impact than the 5% population deviation limit established by the Supreme Court to define the extent of permissible population deviation for legitimate state interests such as protecting traditional community boundaries. Prisoners are external populations that are not "traditionally" rural in any sense of the word. Communities should not be forced to reapportion with population data that violates the equal protection rights of the resident voters.
V. Counting prisoners is far easier than counting overseas Americans.
Counting special populations is always controversial, but a change in how prisoners are counted would not meet the difficulties seen in recent Census test of the feasibility of including overseas American in the decennial census. In response to citizen request and a Congressional mandate, the Census Bureau conducted a test enumeration of overseas Americans in early 2004 in Kuwait, Mexico and France in order to study the feasibility of including these people in the next decennial Census. While U.S. government employees overseas have typically been counted, other groups of Americans have generally been excluded.
In a September 2004 review of the test enumeration, the Government Accounting Office concluded that the test was a failure and the effort should be halted because of low response resulting from one simple fact: overseas Americans are hard to locate. Americans overseas are not required to register their locations with the State Department or local governments, so there is no master list of overseas Americans to send forms to. The only method that could be used was a public relations campaign that resulted in less than 5,400 responses in these three countries with high numbers of Americans.
In the stateside Census, participation in the Census is mandatory and the Census Bureau starts with a master address list of every dwelling in the country. The Bureau can readily determine the completeness of their count by visiting residences that do not return a form. In correctional facilities, finding the population is even simpler: it is the only population that is by definition counted by the government several times a day.
In fact, in many cases the data the Census Bureau wants already exists in administrative records.
VI. Changing how prisoners are counted would not mean that how students are counted would need to be changed.
Some Census Bureau officials have suggested that changing how prisoners are counted might also require that the method of counting college students also be changed. We do not believe this to be the case. Unlike college students, prisoners are not a part of the local community. While some colleges are partially self-reliant communities, prisons are entirely so. Students are welcome and encouraged to come in to town to purchase goods and services and to rent apartments. Students can register to vote in the local community and may decide to stay permanently after graduation. While the majority of students will eventually leave the college town, they will not necessarily return to their parents' home community. College is a time of transition in a person's life. But for the duration of their time at the college, the college is the place they willingly intend to be for the time being: that's the very definition of residence.
People in prison are different. Prisoners are barred from leaving the facility without official permission. Forty-eight states bar prisoners from voting, and those that allow prisoners to vote require the voting to be done via absentee ballot back home. Finally, parole regulations frequently require prisoners to be returned to their original county of conviction upon release.
While students are a part of the community that hosts the college, prisoners are not.
VII. It is feasible for the Census Bureau to collect and process prisoner's home address information as part of Census 2010.
At the request of the National Academy of Sciences Residence Rules Panel, we examined which sources of home residence data existed and made a number of suggestions in our written submissions concerning the best way to gather this information. We summarize this information below, and then add a new piece of evidence from Census 2000 showing that the Census Bureau can collect and process home address information for incarcerated people.
VII. A. Direct enumeration of prisoners is possible and would provide additional benefits to Census data beyond home residence.
The most direct solution to the problem is to collect information from incarcerated people directly rather than from administrative records. The National Academy of Sciences has twice recommended that the Census Bureau rely less on administrative records for enumerating group quarters populations. Direct enumeration of prisoners is safe and appropriate. Although prison officials often warn of safety problems, social workers, volunteers and other civilians enter prisons quite frequently to no ill effect. The journalist accounts of the prisons where Census 2000 forms were distributed directly to prisoners reported that the prisoners were cooperative, suggesting that this direct approach could be used more widely.
Distributing forms directly to prisoners would have at least 5 collateral benefits to overall Census accuracy:
1. Increasing precision. Addresses provided to the Census by people in prison at the time of the census will be more precise than addresses provided to prison administrators at time of intake, simply because they are more current. For example, people who were living with their families and expect to return to their families can provide the current family address. Addresses so provided will be at least as accurate as the unstable addresses used for homeless people, migrant workers and even the highly mobile urban youth who sometimes end up behind bars for a while, and is certainly more accurate than counting them as residents of entirely different congressional districts.
2. Increasing respondent candor and accuracy. The Census Bureau has considerable experience working with sensitive populations. In its work, the Census Bureau is clear about why it collects the data it does, how it will be used and how its privacy will be protected. It would seem reasonable to expect prisoners to be more candid with Census Bureau officials than they would be with the state Department of Corrections regarding the last address and other information.
3. Creating uniformity in data collection procedures.In Census 2000, the method of counting prisoners varied from prison to prison. In some prisons, forms were distributed to and collected directly from prisoners. In other facilities, administrators used official records and submitted a report to the Census Bureau.
4. Reducing the imputation rate for data collected about prison populations. Census 2000 suffered from an unexpectedly high reliance on administrative records for group quarters populations. The missing data rates "reached as high as 50 percent for all group quarters residents and as high as 75 percent for prison inmates." As described further in this section, prisoners are likely to fully participate in the Census at very high levels and make large scale imputation unnecessary.
5. Aiding the Census Bureau in spotting duplicates in the response process. Collecting only facility addresses for group quarters populations makes it difficult to distinguish processed from unprocessed Census forms, resulting in problems with electronic duplication of paper records. An earlier National Academies panel suggested in Counting Under Adversity and another panel formally recommended in Reengineering the 2010 Census that home address information be collected to facilitate unduplication.
If prison and Census officials encourage prisoners to fill out Census forms, we expect the great majority will do so. About.com reported the enumeration of prisoners at Folsom State Prison in California and favorably compared their cooperation to that of the public at large:
At 4:30 pm on April 3, 2000, officials of California's Folsom State Prison handed out 4,000 Census 2000 questionnaires to prisoners. Three hours later, 3,300 -- 82 percent -- of the census forms had been completed and returned. Not bad compared to a response rate of only 67 percent from California's non-inmate population.
This report from the Salt Lake Tribune indicates similar findings even with the 10 prisoners on death row, and shows that the prisoners willingly participated in the Census. Unfortunately, many Census officials in other regions did not "prefer personal interviews with residents to ensure accuracy":
Rather than asking administrators for a nose count, census takers prefer personal interviews with residents to ensure accuracy, particularly for the wider-ranging long forms given to every sixth American.
As a crew chief for the special-places operation, [Census Bureau employee Robert] Comfort covered the state prison in Draper. Prison guards feared some prisoners might smear excrement, blood or other substances on their forms, but the documents came back clean and, almost without exception, completed.
"The prisoners in maximum security were actually pretty helpful," Comfort said. "We also counted about 10 people from death row, and they didn't give us a fight. . . . The guards may have told them they would lose privileges if they didn't help out."
"Not so, said Jack Ford, prison spokesman, who speculated the prisoners just chose to behave. Still, he admitted he worried when the census team went through the prison's maximum-security facility.
"We have some prisoners in super-max that have proven to be security problems in the past," Ford said. "Rather than pass forms back and forth," Ford said, "We had [census takers] ask them the questions through the cell."
We will discuss the quality of the data collected at the Utah State Prison in section VII C.
VII. B. Much of the home address information the Census Bureau would require already exists in administrative records
The National Academy of Sciences asked us if state Departments of Correction maintain electronic records of home addresses and what the research says about the accuracy of this data. The majority of states do collect home residence information at the time of intake into the correctional system and maintain these records in an electronic format. It should therefore be possible to use these records to assign most incarcerated people to a new address outside of the facility. However, the process may be labor intensive and it is not available in every state or many local jails. Although it seems reasonable to expect state correctional datasets to grow more complex and complete over time, we do not expect that administrative records can answer all questions the Census Bureau needs in 2010.
According to the leading geo-spatial analyst of prisoner origin, policy analyst Eric Cadora, about 30 state Departments of Correction maintain electronic administrative records of the home addresses of prisoners.
The states that maintain this information most often do so by asking prisoners to self-report a last address during intake, but often the addresses are not updated until the time of release. For example, the state of Kansas contains a complete dataset of the last known address for its state prisoner population. Based on data published to the county level, the North Carolina Department of Corrections appears to have a virtually complete dataset of home residences for its population. According to an analysis of the prison population on October 31, 2004 published in the Department's online research query tool, in only 0.17% of the records (63 out of 36,136) was the home residence county "unknown". This is only slightly higher than the number of records of county of conviction that were reported as "missing" and "other". The North Carolina data appears to be complete, though the precision below the county level is unclear.
There has been one published study that includes a discussion of the geo-spatial accuracy of this type of data: Political Punishment: The Consequences of Felon Disenfranchisement for Rhode Island Communities, released by the Rhode Island Family Life Center in September 2004. The study found that the Rhode Island Department of Corrections collects a last known home address from persons entering the correctional system and maintains this list for all persons incarcerated under sentence, on probation or on parole. Over 88% of the relevant data in that study was successfully located to the tract level.
The Rhode Island Family Life Center excluded from its analysis those who reported out-of-state addresses. In only 4.4% of the remaining records, research partner Providence Plan was unable to geo-code the address back to the town level. The address records were essentially blank. Because their analysis also required a tract level analysis of urban Providence, the Providence Plan attempted to geo-code all addresses reported within that city. They were able to do so in 92.6% of the Providence cases, which the analysts reported in an interview that this was a higher success rate than they had experienced using other datasets.
VII. C. Census 2000 data for the question "Where did this person live 5 years ago" demonstrates that the Census can in fact collect and process home information for the incarcerated population.
The Census Bureau has considerable experience gathering and geo-coding addresses provided by respondents, and in at least some cases successfully did this with prisoners. While most Census forms are processed based on the address the form was mailed to, on the long form the Census also collects work addresses down to the street level for commuting statistics and collects migration data by asking the respondent for the town of residence 5 years prior. The migration data is geocoded to the tract level, and the previous residence information is published at the county level.
As discussed above, in 2000 the Census Bureau's method of counting prisoners varied from region to region and from facility to facility. In many places, the Bureau relied on administrative records and received much criticism from the National Academy of Sciences for this. We know from press reports and our interviews that in a small number of states and prisons, the Bureau distributed short forms to all prisoners. We have been able to identify a small number of prisons were the Census Bureau also distributed long forms to prisoners and we can evaluate the quality of this data by looking at the published imputation rates for particular questions.
When the respondents fail to answer all of the questions on the form, the Census Bureau frequently fills in the blanks with the statistical process of "imputation". For example, if one household skips a few questions, the Census Bureau might use the answers from the neighbor's forms to estimate race, ethnicity or other data. Low imputation rates mean that the Census Bureau received many complete and valid responses, high imputation rates means that the Bureau was unable to collect the data it sought and had to fill it in by statistical means. The frequency of imputations for various questions are published in separate tables in the Summary Files, so it is possible to determine how effectively prisoners can provide -- and the Census Bureau process -- an address outside of the facility for their residence 5 years ago.
According to press reports, long forms were distributed to 1/6th of the prisoners at the Utah State Prison and the California State Prison at Lancaster. Lancaster has its own Census Block group, and the prisoners in the Utah State Prison account for 66% of the population in its block group. By way of comparison, Rikers Island (the New York City jail) appears to have been enumerated by administrative records. Not surprisingly, the 5 year migration data was imputed for 100% of Rikers Island's block group, but only 51% of the California State Prison and 5.6% of the block group that contains the Utah State Prison.
This analysis of Census 2000 data shows that the distributing census forms directly to prisoners results in higher quality data than relying on administrative records. Further, it shows that it is possible for the Bureau to enter prisons, distribute forms, collect home address information and then process that data. Our analysis is based upon press accounts of Census Bureau procedures and an analysis of block groups that are predominantly prisons. The publicly accessible Census data does not describe which methods of enumeration were used in which prisons, nor does it allow us to separate incarcerated responses from non-incarcerated responses in the same block group. For these reasons, we can not be more precise about low imputation rates of prisons where prisoners filled out long forms; but the internal Census data will contain this information.
The Census Bureau should conduct a review of its records as to how individual prisons where enumerated and compare that with the imputation rate generally and the migration imputation rate specifically, in order to determine which of the enumeration methods used in 2000 was the most successful at accurately capturing complete data that they didn't need to impute. With almost 1% of the population in prison -- and 12% of young Black men -- it is critical that the Census Bureau count this population fairly and accurately.
VIII. What reform would look like -- 3 options.
Although this document expresses concern about the impact of prisoner-misenumeration on demographic analysis, its primary concern is with the impact on the political districting process. As the Census Bureau is currently planning to replace the long form in 2010 with the American Community Survey, we here address only potential changes to data sets based on the short form: PL94-171 and Summary Files 1 and 2.
We discuss three different proposals that would each meet the needs of redistricting data users while potentially also benefiting, but not harming, other data users. The first proposal counts prisoners only at home in all data products based on the short form. The second proposal creates alternative tabulations for only the PL94-171 redistricting data. The third proposal is a subset of the second, and would allow users of PL94-171 to remove incarcerated populations from the census counts but would not provide the necessary data to credit prisoners back to their home communities. None of these proposals affects the residence rules for prisoners in the American Community Survey.
VIII. A. Enumerate incarcerated people as residents of their home Census block as part of a new group quarters housing type: "temporarily incarcerated".
The Census Bureau could change the usual residence rule to count prisoners only as residents of their home Census block in all data products based on the short form: PL94-171, Summary File 1 and Summary File 2. This new residence rule would be modeled on that designed for non-institutional group quarters : Count prisoners at the facility only if they do not report a usual and valid address elsewhere.
The Census Bureau should create a new group quarters population type for prisoners: "temporarily incarcerated". As a result, prisoners would continue to be excluded from the tables based on the "household" universe, such as Table P16 in Summary File 1 "Population in households", or Table P17 "Average household size". However, prisoner's home addresses would be indistinguishably represented in the PL94-171 redistricting data and in all Summary File 1 and 2 tables based on the total population.
VIII. B. Allow redistricting data users to choose whether they wish to count prisoners at home, at the facility, or neither, by providing alternative counts for census blocks that contain correctional facilities or the home addresses of incarcerated people.
Alternatively, the Census Bureau could follow the advice of redistricting and Census Bureau expert Nathaniel Persily at the University of Pennsylvania Law School and not decide where to count prisoners. The Bureau can simply collect both home and facility addresses as described above in section VIII A and publish alternative versions of only the PL94-171 redistricting dataset. As Professor Persily wrote, in a paper submitted to a Census Bureau symposium in 2004: "In doing so, the Bureau would remain true to its mission as information provider without having to take a stand on exactly where states and localities should place prisoners." States and localities could decide what dataset they wish to use.
This proposal is even simpler than it sounds because PL94-171 is a very small dataset with only 4 tables of summary statistics on total population, for the population 18 years and over, for different races and for Hispanic/Latino origin. By contrast, Summary File 1 contains 286 detailed tables plus more than 100 maps and other products. Summary File 2 contains 47 detailed tables.
It would be possible for the Census Bureau to create a special tabulation of the PL94-171 data that would identify the prison populations at the prison and allow jurisdictions to remove prisoners from the census data. The almost 2 million prisoners counted in Census 2000 were counted within only 5,692 Census blocks, so only that small number of blocks would need to be reported.
Allowing jurisdictions to add the prisoners back to their home addresses would require a second special tabulation of PL94-171 tables, which is well within the Bureau's abilities. Although there are more than 8 million census blocks, we estimate that perhaps no more than 200,000 census blocks would include the homes of incarcerated individuals. Because about 3 million census blocks contain no population, the American population lives in only 5 million census blocks. As the two million prisoners disproportionately come from a small number of communities, it would seem reasonable to expect that only a tiny portion of census blocks would be affected by this special tabulation.
VIII. C. Allow data users to locate and remove correctional facilities from the PL94-171 data, which would be simple to implement with minimal impact on other operations.
Currently, many jurisdictions that might like to remove correctional facilities from the Census cannot do so because the law requires the completion of redistricting before or only shortly after the only Census product to identify correctional facilities -- Summary File 1 -- is released. Identifying the Census blocks that contain correctional facilities and providing their correctional populations in the PL94-171 data is the simplest of the reforms suggested above. If the Census Bureau feels that insufficient time remains before the 2010 Census to fully implement either proposal A or B (above) in full, it could allow PL94-171 data users to identify and remove correctional facilities without waiting for the later release of Summary File 1. There is no legitimate reason not to implement this reform.
IX. State and local proposals to adjust census data do not absolve the Census Bureau of the responsibility to change how prisoners are counted.
In previous articles and submissions we have highlighted the efforts of jurisdictions to adjust the federal census data so that it meets their needs and people are counted in the right place. These efforts are vivid evidence that people would prefer that the Census counted incarcerated people differently. However, the fact that these jurisdictions can fix the Census Bureau's mistakes does not mean the Bureau should continue to make them. In order to fairly serve its data users in state and local government, the Census Bureau must change how it counts prisoners. As we will show, it is not only inconvenient for data users to fix the Census Bureau's mistakes, for some communities it would be impossible for practical or legal reasons.
IX. A. Local adjustments of Census counts of prisoners
The New York constitution declares that incarceration does not change a residence. In rural Franklin County, where almost 11% of the population reported in the Census is incarcerated prisoners from elsewhere in the state, the chair of the county legislature called ignoring the prisoners during redistricting a "no-brainer". He was right. It made no sense to draw a district around the prisons in the town of Malone that was would be 2/3rd prisoners who did not live in the county. But there is another relevant issue: It was possible for Franklin County to take the 2000 Census data, and two years later, take the prisoners out and be left with just the actual residents of Franklin County.
The New York City Council -- in to whose districts most of the prisoners counted in Franklin County actually belong-- had no such choice. When they drew their districts in 2002, it was too late to put the prisoners back in to their Census data because the Census Bureau did not collect this information. In the 2002 Importing Constituents: Prisoners and Political Clout in New York report, the Prison Policy Initiative was able to use correctional data to calculate that the Census Bureau credited 43,740 New York City residents to upstate prison towns, but those records didn't say exactly where in New York City those prisoners reside. Taking the prisoners out is easy, but putting them back in after the fact is a different matter altogether when the Census Bureau didn't ask the right questions about residence.
IX. B. State adjustments of Census counts of prisoners
A different, legal, problem exists with some state level adjustments. Kansas currently adjusts the federal census to count students and military at their home addresses for purposes of redistricting. The state starts the planning for this about a year before the federal census is undertaken and then there is a relatively simple process of fixing the federal census data when it arrives. Illinois and New York State currently have bills pending that would create similar processes in those states for prisoner counting.
Unfortunately, not every state that wants to can follow suit with a simple change in their statutes. In Massachusetts, for example, such a statute would violate a 1990 amendment to the state constitution that the "federal census shall be the basis for determining the representative districts". Massachusetts, which was the last state to abolish its state census to rely on the federal census precisely because its top court viewed the U.S. Census Bureau's definition of usual residence to be incompatible with the state constitution's definition of "inhabitant", is now stuck. Unless it changes the state constitution yet again, Massachusetts is forced to use Census data regardless of how illogically the Census Bureau may count the population. While the state constitutions in New York and Illinois have some flexibility on what data is used for redistricting, many other states are in the same situation as Massachusetts and must use whatever the federal Census provides. It is therefore critical that the Census Bureau change how it counts prisoners if every state is to be able to fairly draw legislative districts to contain equal numbers of actual residents.
IX. C. Proposals to adjust census do not release the Census Bureau from its obligation to count the population fairly and accurately.
Counting prisoners at the facility probably made sense at the first Census in 1790 when few people were in prison and when the data was not used for redistricting. Things are different now, and the data collection methods need to change. Jurisdictions that are impatient waiting for the Census to update its methodology should be commended for taking matters in to their own hands. But the Census Bureau should not assume that every jurisdiction has the practical ability or even the legal option to fix the Bureau's mistakes after the fact.
If we want an efficient national fix to the problem of where prisoners are counted in the Census, that solution needs to come from the Census Bureau itself.
The decennial census is an extraordinary tool and the effort that goes into it is truly impressive. We believe that solving this problem of people in correctional facilities is critical - they presently number roughly 0.7% of the entire population and in excess of 12% of certain segments of the population (African American men in their late twenties). Numbers this large need to be enumerated with the greatest possible precision.
In 1999, then Census Bureau director Kenneth Prewitt opposed an 11th hour attempt to change how prisoners were counted in the Census. But more well conceived proposals presented during the 2010 Census planning process have since gained his support. In 2004, he wrote that accuracy requires not that just the population be counted, but that they are counted in the correct place. Fairness is also at stake:
Changes in the criminal justice system over the last three decades call into question the fairness of counting persons where they are imprisoned rather than where they were living when arrested, and to which they return on release. 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. (In these, and in additional ways, prisoners differ from college students, the other sizable group living, though in their case voluntarily so, away from 'home.') With over 1.4 million people in prison, and 650,000 people returning home from prison annually, where to count the incarcerated population is no small matter. Counting people in prison as residents of their home communities offers a more accurate picture of the size, demographics, and needs of our nation's communities...
The next census is still more than 4 years away, but if the Census Bureau begins planning now, there is still time to test and implement even the largest of the proposals suggested here. The Census Bureau has selected San Joaquin County in California as one of the sites for its 2008 dress rehearsal. This county contains more than 7,000 prisoners in a number of different facilities and would be an excellent place for the Census Bureau to test and fine tune the optimal way to gather, process and publish the home address information of our nation's incarcerated population.
We encourage the Census Bureau to make the necessary adjustments and to fulfill its tradition of providing information of the highest quality. We are available to assist if we can be useful.
About the authors
Peter Wagner, J.D., co-founder and executive director of the Prison Policy Initiative, teaches, lectures, and writes about the negative impact of mass incarceration in the United States. His current focus is on working to demonstrate-- through graphics, legal research, and state-by-state analyses-- the distortion of the democratic process that results from the U.S. Census Bureau's practice of counting the nation's mostly urban prisoners as residents of the often remote communities in which they are incarcerated. The New York Times editorial board has three times written editorials supported his efforts to change the way prisoners are counted, and the Boston Globe identified him as the "leading public critic" of the prisoner miscount. He has presented his research at national and international conferences and meetings, including a Census Bureau Symposium, a meeting of the National Academy of Sciences, and a keynote address to a conference at Harvard University. Mr. Wagner's publications include Importing Constituents: Prisoners and Political Clout in New York (2002); The Prison Index: Taking the Pulse of the Crime Control Industry (2003); and, with Eric Lotke, Prisoners of the Census: Electoral and Financial Consequences of Counting Prisoners Where They Go, Not Where They Come From, 24 Pace L. Rev. 587 (2004).
Eric Lotke is a consultant in private practice in criminal justice administration and policy. Formerly he was Director of Policy and Research at the Justice Policy Institute, an organization that provided research and technical assistance to solve persistent problems in the criminal and juvenile justice systems. He was also a Soros Senior Justice fellow and the Senior Policy Analyst in the national office of the Open Society Institute. As a Justice Fellow, he performed original research on the political and financial consequences of the U.S. Census Bureau counting people in prison where they are confined rather than their original homes. Mr. Lotke was Executive Director of D.C. Prisoners' Legal Services Project, a private non-profit law firm that provides legal services to people incarcerated in the District of Columbia. He represented the inmates in a groundbreaking class action lawsuit against the private Corrections Corporation of America regarding a prison in Youngstown, Ohio; he pioneered litigation over excessive, non-competitive price of collect phone calls from people in prison. He also acted as Chief Operating Officer of the Alliance of Concerned Men, a direct service organization in the District of Columbia that runs programs for troubled youths and returning prisoners, and that intervenes at moments of intense conflict among rival gangs. Mr. Lotke has published widely on issues affecting the criminal justice system. He authored pathbreaking research on patterns of juvenile homicide and on the demographics of involvement in the justice system. He has addressed issues ranging from the Federal Sentencing Guidelines to the treatment of sex offenders and the origins of criminal responsibility. He was associate editor of The Real War on Crime published by HarperCollins (1996). He has been an adjunct professor at the Georgetown Law Center, the George Washington Law School, and the University of the District of Columbia School of Law. He was formerly a judicial clerk on the Supreme Court of Connecticut. Mr. Lotke's formal education includes a Bachelor of Arts from Wesleyan University and a J.D. and a Master's in Philosophy from the University of Wisconsin.
Dr. Andrew Beveridge, Professor of Sociology at Queens College received his Ph.D. from Yale in 1973. He is an authority on recent social and demographic trends affecting the New York Metropolitan Area and the United States, as well on the application of computer based demographic mapping and analysis methods. He is a consultant to the New York Times, which has published many news reports based upon his analysis of Census data. He is the co-author of African Businessmen and Development in Zambia, published by Princeton University Press, and numerous articles, papers and reports. He has consulted in a number of civil rights cases with the Connecticut Civil Liberties Union, the ACLU of Maryland, the Open Housing Center of New York City, Westchester Legal Services, Center for Constitutional Rights, the Capital Defenders Office, Davis Polk and Sullivan and Cromwell. His research work has received grant and fellowship support from the American Council of Learned Societies, the National Science Foundation, the National Endowment for the Humanities, the American Philosophical Society, the Department of Housing and Urban Development, the Putnam Foundation, the Robert Wood Johnson Foundation and other agencies.
The initial research behind this report was supported by grants from the Soros Justice Fellowship Program of the Open Society Institute. We are grateful to William Cooper for analyzing block level census data and to Barbara Fedders, Aleks Kajstura, Daniel Jenkins, and Jon Yount for helping to improve this report.
 By mid-2004, there were 1.3 million prisoners under state jurisdiction, and 167,000 prisoners in federal prisons throughout the country. An additional 784,000 people were confined under the jurisdiction of local jails. Jail inmates are often incarcerated in their own county, although there are growing trends towards the construction of multi-county regional jails or the use of local jails to house state and federal prisons on a per-diem basis.
 Ibid, Table 14.
 Rose Heyer and Peter Wagner, Too big to ignore: How counting people in prisons distorted Census 2000, April 2004, Prison Policy Initiative.
 See Peter Wagner, Actual Constituents: Students and Political Clout in New York, Prison Policy Initiative, October 2004, and the letter from Daniel Jenkins, Mark Flack Wells, and Norman Gervais to the Census Bureau published as Rural citizens call for change in how Census counts prisoners.
 See Bureau of Justice Statistics, Sourcebook of Criminal Justice Statistics, 2002, Table 6.38. Felony jail sentences and misdemeanor jail sentences are significantly shorter.
 See for example, Arizona State Const. Art. VII S 3, Nevada State Const. Art II S, and New York State Const. Art, 2, S 4.
 Peter Wagner, Importing Constituents: Prisoners and Political Clout in New York, Prison Policy Initiative, April 2002.
 Peter Wagner and Rose Heyer, Importing Constituents: Prisoners and Political Clout in Texas, Prison Policy Initiative, November 2004.
 Peter Wagner, Importing Constituents: Prisoners and Political Clout in Montana, Prison Policy Initiative, December 2004
 Letter from Daniel Jenkins, Mark Flack Wells, and Norman Gervais to the Director of the Census Bureau on July 9, 2004 in regards to the Establishment of the 2010 Census Redistricting Data Program.
 Peter Wagner, "State officials tell counties to exclude prison populations from county level redistricting", Prisoners of the Census, November 1, 2004.
 Eric Lotke and Peter Wagner, "Prisoners of the Census: Electoral and Financial Consequences of Counting Prisoners Where They Go, Not Where They Come From"24 Pace L Rev 587 (Spring 2004) available at http://www.prisonpolicy.org/reports/pace.pdf
 It is unearned because other costs associated with hosting a prison such as burden on roads and sewerage is typically accommodated in other parts of the budget.
 Lotke and Wagner, op. cit. at 601-602
 The population enters via the per capita income. Ibid at 601-602, 604.
 The Bureau relied on this case to make that argument as recently as April 19, 2005, in Director Kincannon's testimony to the House Subcommittee On Federalism And The Census.
 D.C. v. Commerce at 1190.
 5 U.S.C. S 706(2).
 See Rosanna M. Taormina, Defying One-Person, One-Vote: Prisoners and the "Usual Residence" Principle, 152 U. PA. L. REV. 431 (2003) and Peter Wagner, Importing Constituents: Prisoners and Political Clout in New York, Prison Policy Initiative (2002) fn 5,
 Reynolds v. Sims 377 U.S. 533 (1963) The equal protection clause "requires that the seats in both houses of a bicameral state legislature must be apportioned on a population basis... An individual's right to vote for state legislators is unconstitutionally impaired when its weight is in a substantial fashion diluted when compared with the votes of citizens in other parts of the State."
 Reynolds at 562.
 Reynolds at 567.
 Counting Matters: Prison inmates, population bases, and "one person, one vote", by Taren Stinebrickner-Kauffman 11 Va. J. Soc. Pol'y & L. 229, 303. (Winter 2004)
 See Peter Wagner, "Counting prisoners at home would not share problems seen in Overseas Americans test enumeration", Prisoners of the Census, September 13, 2004 and "Counting Americans Overseas as Part of the Decennial Census Would Not be Cost-Effective", General Accounting Office, August 2004.
 See Peter Wagner, "If prisoners could vote, they would vote at home, not in the prison town" Prisoners of the Census, December 15, 2003.
 See Constance F. Citro, Daniel L. Cork, and Janet L. Norwood, eds., The 2000 Census: Counting Under Adversity, National Research Council of the National Academies (Washington, 2004) p. 156 and Daniel L. Cork, Michael L. Cohen, and Benjamin F. King, eds., Reengineering the 2010 Census: Risks and Challenges, National Research Council of the National Academies (Washington, 2004) p. 153.
 See "Headcount comes to prisons" by John Chrisofferson, Associated Press, April 4, 2000 (reporting that "Connecticut counted more than 17,000 prisoners at 20 prisons on Monday only a handful of Connecticut inmates refused to fill out the form, census and state officials said."), Kevin Cantera, "Prison County: Census Takers Put Gloves On", Salt Lake Tribune, May 16, 2000, p. A1 and Annette Kondo, "At prison: census reactions of all stripes". Los Angeles Times, April 20, 2000, p B1.
 See Recommendation 2, "Enforce a Uniform Prison Enumeration" in the Brennan Center for Justice report: Accuracy Counts: Incarcerated People & the Census [PDF] by Patricia Allard and Kirsten D. Levingston, April 8 2004, p. 16.
 Constance F. Citro, Daniel L. Cork, and Janet L. Norwood, eds., The 2000 Census: Counting Under Adversity, National Research Council of the National Academies (Washington, 2004) p. 297.
 Ibid., p. 156.
 Daniel L. Cork, Michael L. Cohen, and Benjamin F. King, eds., Reengineering the 2010 Census: Risks and Challenges, National Research Council of the National Academies (Washington, 2004) p. 153.
 Kevin Cantera, "Prison County: Census Takers Put Gloves On", Salt Lake Tribune, May 16, 2000, p. A1.
 Mr. Cadora works with Departments of Corrections, court systems and state legislatures to study where prisoners come from and to develop strategies that might lower the rates of incarceration in particular neighborhoods. His research in New Jersey, Louisiana, Connecticut, and Kentucky all relied upon state Department of Corrections data. We interviewed Mr. Cadora in the fall of 2004.
 Interview with William Cummings of the Kansas Department of Corrections, November 20, 2004.
 See Peter Wagner, "County of conviction and prisoner residence" Prisoners of the Census, October 20, 2003. Custom reports of North Carolina Department of Corrections statistics can be generated from the "Corrections Statistics" link available at http://www.doc.state.nc.us/offenders/
 Marshall Clement and Nina Keough, Rhode Island Family Life Center, Political Punishment: The Consequences of Felon Disenfranchisement for Rhode Island Communities, September, 2004.
 Kevin Cantera, "Prison County: Census Takers Put Gloves On", Salt Lake Tribune, May 16, 2000, p. A1 and Annette Kondo, "At prison: census reactions of all stripes". Los Angeles Times, April 20, 2000, p B1.
 Rikers Island is Tract 1 in Bronx County New York, Utah State Prison is Block Group 2, Tract 1128.07 in Salt Lake County Utah, and Lancaster State Prison is Block Group 1, Tract 9010.03, Los Angeles County, California. The imputation of population items for the population in group quarters is in Summary File 3, table 95, and the imputation of mobility status for the population 5 years and over is in Summary File 3, table P114.
 See U.S. Census, Residence Rules in the U.S. Census, Rule 12, People in Noninstitutional group quarters.
 Quoted with the author's permission from Nathaniel Persily, "The Law of the Census: How to Count, What to Count, Whom to Count and Where to Count Them" draft paper presented at the U.S. Census Bureau symposium "America's Scorecard: The Historic Role of the Census in an Ever-Changing Nation" March 4, 2004.
 This analysis was performed by William Cooper, on December 23, 2005 making use of Summary File 1, Table P37. Age detail is given in P38, and P17A through P17H give data at the tract level by race and Hispanic status. P17A through P17H cannot be used for redistricting, however, since they only go to the tract level and are not in the same format as the four tables released in the PL94-171 data. Mr. Cooper is a GIS, Demographic and Redistricting Consultant based in Richmond, VA. His work was cited in the Supreme Court Opinion Reno v. Bossier Parish. Since the release of the 2000 census, Cooper has prepared election plans for minority citizens in 150 localities in 20 states. In August 2005, a federal court ordered the State of South Dakota to remedy a Section 2 voting rights violation and adopt a state legislative plan he developed (Bone Shirt v. Nelson). According to Cooper, "the prison population miscount issue has significantly undermined the one-person, one-vote principle in many areas where I have been asked to draft election plans over the past 20 years. It is a much more serious problem than the undercount, insofar as redistricting is concerned."
 This analysis was also performed by William Cooper and is based upon extrapolating the results from 12 states.
 The June primary schedule in a number of states, including Virginia, and Mississippi, requires counties and states to complete redistricting shortly after the PL94-171 data is released in March of the year after the Census rather than waiting until the Summary File 1 data is released in the summer of the year following the Census.
 See the "Proposed finding and recommendation submitted to the Residence Rules in the Decennial Census Panel at the National Academy of Sciences" [PDF] by Peter Wagner, Eric Lotke and Andrew Beveridge on June 2, 2005. See also "Local governments that exclude prison populations" at http://prisonersofthecensus.org/local/ which lists articles about cities, towns and counties that have removed prisoners from the census data prior to redistricting.
 Opinion of the Justices, 365 Mass 661, 663-664 (1974).
 Hearing before the Subcomm. on the Census of the Comm. on Government Reform, 106th Cong. 106-39 (1999) (statement of Kenneth Prewitt, Director, Bureau of the Census)