By Mike Dorf
With the decennial census now under way, a familiar practice has again (and rightly) come under fire: The census counts prisoners as residing wherever the prison in which they are incarcerated happens to be, rather than in the communities from which they originate. However, in nearly all states, prisoners aren't allowed to vote. As a result, the count is depressed for disproportionately poor and minority communities from which prisoners disproportionately come, while the disproportionately non-minority, disproportionately rural communities in which prisons are located get a population bonus. Here I'll try to clear up a little bit of the confusion in the public discussion of this phenomenon.
In 2006, Congress asked the Census Bureau to explore the feasibility of counting prisoners at their prior homes rather than in prison. The Census Bureau produced a document raising a host of objections. The biggest problem would be the need to meet with and interview individual prisoners but also, the Census Bureau would have no uniform way of verifying the home addresses or even of defining home address.
With due respect, that was a misleading answer because the census data serve multiple functions. At the federal constitutional level, they are used to decide how many seats each state gets in the House (and thus also how many Electors in the Electoral College). Congress has interpreted the "actual Enumeration" language of Article I, Section 2 to forbid statistical sampling with respect to this core constitutional function. However, the same statute empowers the Census Bureau to use sampling for other purposes. Accordingly, the practical difficulties of conducting a non-sampling enumeration of prisoners should present no obstacle to the Census Bureau's providing estimates of local populations once prisoners are allocated to their prior residences.
That's significant because the numbers used for inter-state allocations of House seats need not be used--or can be supplemented by other data--in making allocations of benefits and, the focus of the current discussion, representatives within states. Here the relevant limits are the Voting Rights Act and equal protection limits embodied by the one-person-one-vote apportionment rule. To my knowledge (and I admit that I haven't researched this piece thoroughly), neither Congress nor the Supreme Court has spoken to the question of whether sampling-based data can be (or must be) used for these other voting-related purposes.
If I'm right about that, then a state could choose to allocate prisoners to their homes for purposes of allocating representatives among districts--so long as it had a reasonable basis for doing so, and sampled data would be such a basis. (According to a 2001 Harvard Law Review Student Note, a few states forbid the use of sampled data for internal purposes.) States that wanted to do so could probably generate the necessary data on their own, but they would do a lot better if the Census Bureau helped them.