Wednesday, March 31, 2010

The Law of the Census

Nate Persily

In anticipation of Census Day (tomorrow), here are a few thoughts on the law of the census (one of my thoroughly unmarketable areas of expertise for nine out of every ten years). The 2010 Census presents all kinds of interesting legal issues concerning how to count, what to count, whom to count and where to count them. These comments focus on the relationship between the census and reapportionment/redistricting, but the problems affect federal funding as well.

How to Count

When one thinks of controversies surrounding the census, the "sampling" debate ordinarily comes first to mind. The Supreme Court has ruled that the Census Act prohibits the use of "sampling" for apportionment purposes, but allows the use of other statistical methods such as "hot deck imputation." The Act allows sampling for other purposes, such as redistricting. The Bush administration scuttled plans for statistical adjustment in the 2000 Census, which ended up probably being the right thing to do given the success of the 2000 Census as measured by the shrinking of the differential undercount (that is, the relative rate at which census undercounts member of certain demographic groups, particularly racial groups).

For the 2010 Census, sampling is not even being considered as an option, but it is somewhat difficult to predict the character of the likely undercounted population. The high rate of foreclosures in the last year and subsequent shifts in population in some cases to nontraditional housing alter some of the assumptions that had guided the original operational plan for the census. The anti-census rhetoric from Michelle Bachmann and Glenn Beck and the general suspicion of the federal government among tea party sympathizers might deter participation from populations (i.e., whites) that historically have been more accurately counted. (Given that Minnesota may be on the cusp of losing a congressional district, Bachmann has altered her criticism to focus more on the race question on the census.)

What to Count
About half of the decennial census form now deals with questions concerning race. For the second time, the 2010 census will allow respondents to check off more than one race. For those of us who use the data for redistricting purposes, this means the census will publish data for a mind-bending 126 different racial and ethnic combinations for all levels of geography. The OMB has issued guidelines that instruct the federal government for purposes of civil rights enforcement to reallocate multiracial individuals to the non-white race. In a much overlooked footnote in Georgia v. Ashcroft, Justice O'Connor's opinion for the Court seemed to adopt this approach as well in the context of a voting rights case. It will be interesting to see whether the Roberts Court has an opportunity to confront the OMB guidelines directly in a coming case.

Many jurisdictions feared in 2000 that the multiracial checkoff would wreak havoc on the redistricting process. That turned out not to be true and likely will not be true this time either. This is due to the fact that most redistricting cases involve African Americans or Hispanics. African Americans have a very low rate of multiracial identification and Hispanics (considered an ethnicity, not a race, for the census) have their own separate question on the census and are therefore unaffected by the multiracial checkoff. In the recent case of Bartlett v. Strickland, the Court has required plaintiffs in section 2 VRA suits to demonstrate that they could constitute a majority (i.e., over 50%) in a single member district. Perhaps some groups hovering close to that number would have their claims affected by the multiracial checkoff, but that should be a rare case.

The more interesting controversy surrounding "what to count" concerns citizenship. The decennial census does not contain a question about the respondents citizenship, for fear that to do so would chill participation. In 2000, one out of every six households received the "long form" of the census, which did have a citizenship question. The long form of the census has been abandoned in 2010. It has been replaced by the American Community Survey (ACS), which is a survey sent to 2.5% of households every year throughout the decade. The census promises to publish citizenship data from five year averages from the ACS. While reliable for large areas of geography, such as counties, the data become less usable and accurate for smaller areas of geography, such as census blocks, which form the building blocks of redistricting plans.

This is important because the 5th Circuit, among others, requires voting rights plaintiffs to demonstrate that they would constitute a majority of the citizen voting age population (CVAP) in a district to bring a section 2 VRA claim. The less reliable the citizenship estimates, the greater the tension between the available data and the legal hurdles the courts place in the way of VRA plaintiffs. (Bartlett v. Strickland seemed to suggest that voting age population, rather than CVAP, is the appropriate basis for a VRA claim, but the 5th Circuit in a recent case declared that Bartlett did not overturn or undermine its earlier precedent on this issue.)

Also, a lawsuit has been filed recently in Irving, Texas, arguing that the city council districts there should be drawn on the basis of equal numbers of citizens, rather than equal numbers of people. One of the districts there, with a population that is majority Hispanic, has half the number of citizens as adjoining districts. As with the question regarding the use of CVAP in VRA lawsuits, the Supreme Court has never directly dealt with the question of citizenship in the context of a one person one vote case. However, in Burns v. Richardson (1966), the Court expressed considerable deference to the states' choice among different population bases for redistricting, when it allowed Hawaii to draw its districts based on equal numbers of registered voters per district. In a dissent from a denial of cert in Chen v. City of Houston, 532 U.S. 1046, 1046 (2001), Justice Thomas recently has urged the Court to consider and clarify whether the one-person, one-vote rule requires equal numbers of citizens.

Although the Irving, Texas case is probably dead on arrival anyway, the absence of block level citizenship data should make the notion of requiring redistricting according to equal numbers of citizens particularly untenable.

Whom to Count and Where to Count Them

The 2000 Census also saw controversies over whether (and where) certain populations should be counted for purposes of reapportionment. In Utah v. Evans, for example, Utah argued that it was deprived of a congressional seat because the census allocated overseas federal employees to their respective states but failed to do the same for LDS missionaries.

For the 2010 Census, the main point of controversy concerns the counting of the prison population. For the first time, the census will make available in time for redistricting the number of prisoners by census block. This will allow jurisdictions to subtract out prisoners for purposes of redistricting in order to prevent the possibility that some districts become padded with large numbers of ineligible voters. In some state legislative districts over ten percent of the population is incarcerated. See here. And in some states, such as New York, this means that a heavily minority prison population ends up padding rural, largely white legislative districts. (In other states, the disparate racial impact is less clear given the racial composition of the census blocks from which prisoners come and to which they are transferred.)

Several state legislatures are considering bills that would require alteration of of census data in this manner for purposes of redistricting. We should expect litigation under both the VRA and one person, one vote as a result either of states' decisions to subtract the prison population or their decisions not to do so.

Those are just a few of the many legal issues we should expect to flow from the 2010 census in the coming year. For now, just send in your form!