Counting on the Census? Race, Group Identity, and the Evasion of Politics.
In a year that included a presidential election that was too close to call and a census marred by statistical controversy, questions of how to count each and every American preoccupied national politics. The questions were often phrased as ones pitting humans against machines, with charges of political manipulation launched by both sides in hopes of defending their favored method as the most complete and accurate of various alternatives. In the presidential election, hand counts versus machine counts translated into the legal conflict Bush v. Gore.(1) In the context of the census, "headcounts" versus sampling" emerged in court as a conflict between different branches and levels of government(2) and threatened to spawn lawsuits between any number of racial and political interest groups and their opponents.
In both the election and census cases, however, the legal disputes over different "counting" methods represented a deeper political competition between opposing partisan and factional forces. As such, consistency of argument between the different contexts of ballots and census forms usually took a back seat to the logic needed to win a particular legal or public relations battle. Those who argued against a manual recount of ballots usually demanded that only manual methods (census form mailback and personal follow-up procedures) be used for the census. Their opponents were equally inconsistent, demanding the use of machines (that is, statistical sampling) when it came to the census, but urging manual counting when it came to presidential ballots. Important differences between elections and censuses and their related procedures for counting votes and people are not difficult to unearth. The analogies do not match neatly: In one case the measurement controversy involves how to count the paper filled out by the respondent (i.e., the voter); in another it involves whether or not to rely solely on paper forms filled in by people. "Technology" appears in the election context to prevent human bias, but in the census to correct incompleteness resulting from undercoverage. As imperfect as the analogies may appear, however, they draw our attention to common problems of completeness and bias that plague any counting system that seeks to represent accurately millions of ballots or hundreds of millions of people.
Now that both controversies appear to have subsided, we realize that the choice of counting methods appears to have been much less significant than either side in either fight believed. The manual recount of presidential ballots subsequently conducted by newspapers indicates Bush probably would have won the presidential election even if the Supreme Court had not stopped the recount in its tracks.(3) And because the census appears to have been successful in reducing the undercount to its lowest level in American history, any need for statistical adjustment has diminished dramatically.(4) One cannot help but ask whether the sound and fury of either controversy, in the end, signified anything of lasting political importance.
In Counting on the Census?: Race, Group Identity, and the Evasion of Politics, Professor Peter Skerry answers that the census adjustment controversy signifies a harmful trend of "symbolic politics" devoid of any real substantive content. The book, which attacks the original plan for the 2000 Census, is the best researched and written on the subject. Skerry examines how the process of taking a national census evolved into a rhetorical and public policy competition between conflicting demands of "science," "rights" and "politics." He makes a persuasive argument that appeals to expertise based on unique technical competence or to civil rights of representation in the census obscure the politics of collecting census data. This effective critique of the rhetoric of rights and science surrounding the census may be lost in his passionate argument against statistical adjustment to correct the census undercount, but as a piece of social science, the book could not provide more in the way of useful references, rich history and a complete description of the ongoing debate. Lawyers might find the book disturbing, however. Indeed, because we fight atop the terrain of rights that Skerry finds so objectionable with regard to the census, the book is as much an assault on lawyers (particularly from the civil rights community) as it is an argument against particular census policies. Skerry derides the census component of the civil rights agenda as emblematic of a symbolic politics guided by organizations without members, closeted in courts and bureaucracies away from democratically accountable institutions, and providing few tangible political benefits for the supposedly affected communities. Because of the gauntlet it lays down to all those involved in the recent and future census battles as well as the helpful historical and social scientific analysis it provides, Counting on the Census is a must-read for anyone wishing to understand the issues underlying the adjustment controversy.
This review proceeds in four parts. Part I summarizes the exhaustive history, causes and consequences of the census undercount that Skerry takes as his point of departure. He argues that such an undercount is, to a certain extent, unavoidable and that cures will cause more harm than the disease. Part II describes the argument against statistical adjustment of the census to correct for the pervasive undercount of racial minorities: the so-called differential undercount. Because he believes the differential undercount does not by itself constitute a serious public policy problem (even from the point of view of racial minorities), Skerry considers almost any attempt to use statistical methods to adjust the headcount as doing more harm than good. This strong opposition to adjustment is in tension with Skerry's larger argument that a discourse of science or rights obscures the baser instincts of politics that motivate the different sides of the adjustment controversy. If Skerry believes the census is "all about politics," then how can he be so convinced that the anti-adjustment position is objectively correct? Part III of this review criticizes Skerry for misunderstanding the place of the census adjustment controversy in the larger and hotter legal battle over the past decade for minority legislative representation. Litigation over census adjustment represented a fight for protection of civil rights because it constituted the first stage in what promises to be the most contentious decade of redistricting-related litigation thus far. Although Skerry rightly argues that census adjustment or lack thereof does not equal clear winners and losers in the redistricting process, he ignores the fact that the failure to adjust census data may prevent certain voting fights plaintiffs from even getting into court. The failure to adjust census data could place additional obstacles in the path of racial minorities that need to prove that their communities are sufficiently large to avail themselves of the protections of Sections 2 and 5 of the Voting Rights Act. Part IV concludes with a discussion of the substantive importance of what Skerry decries as the "symbolic politics" of the census. I argue that controversies involving symbols, particularly in the context of legal conflicts involving race and equal protection, derive their meaning from the precedents they will set for larger bodies of law. It is unfortunate that Skerry's diatribe against public interest politics might obscure the methodical and reasonable articulation of the anti-adjustment position. The chief value of the book is its acknowledgement that politics rather than science or fights underlies important census-related decisions. Its chief defect is a reliance on a political predisposition to attack the motivation of adjustment advocates.
I. THE CAUSES AND CONSEQUENCES OF THE DIFFERENTIAL UNDERCOUNT
Each census represents a snapshot of the nation at a particular time. The picture of the nation that develops is fuzzier in some areas than others, sometimes overexposed or underexposed, and with a tint and color scheme that often differs from the nation it attempts to capture in its lens. No country has ever completed a perfect census.(5) The best any country can hope for is an estimate of the "true" population. Each step in the census represents a stage in this search for truth even though all involved share an awareness that there is no complete and objective number against which we can measure the precise magnitude of the errors that are made.(6) Perhaps Skerry's chief contribution is his lucid argument that Americans have come to expect too much from the census.(7) A 100% accurate count of a population in excess of a quarter of a billion people is impossible.
A. Why the Census Misses Some Groups at a Greater Rate than Others
The main problem with the census has not been inaccuracy; however, it has been bias. The inaccuracies of the census have not been randomly distributed. They have fallen hardest on certain segments of the population. Thus, observers and policymakers have often been less concerned by the undercount--that is, the number of people the census has missed--than they have been by the "differential undercount"--why the census has counted some classes of people more accurately than others. Historically, the groups that have fared the worst in the counting process were children, renters, residents of large cities, and racial minorities.(8) Because of the importance of race for both the Constitution and the Voting Rights Act, let alone the symbolic, historical and political exceptionalism of race discrimination,(9) the undercount of racial minorities and the impact of that undercount on minority representation and federal funding allocations have been the chief areas of concern for those focusing on the inaccuracies in the census.(10)
As the net undercount steadily declined for most of the past half-century--from 5.4% in 1940 to 1.2% in 2000(11)--the differential undercount has often tripled the net undercount and remained more resilient over that period. Since 1940 when an unexpectedly high number of African-American men showed up for the draft, the U.S. Government has known about the differential racial undercount.(12) The undercount of African Americans declined from 8.4% in 1940 to 5.7% in 1990, while the non-black undercount rate has declined from 5.0% in 1940 to 1.3% in 1990. Thus, while undercounting of both populations decreased, the difference between the black and non-black undercount rates has increased.(13) A similar dynamic appeared with many racial groups later added as separate categories on the census. Undercount rates for the racial categories on the 1990 and 2000 Census forms appear in Tables I and II.(14)
Table I. Net Undercount Rates for 1990 Census Net Undercount Population Group (percent) American Indian and Alaska 12.22 Native (on reservation) Hispanic Origin(15) 4.99 Black of African American 4.57 Asian and Pacific Islander 2.36 White or Some Other Race (not Hispanic) 0.68 Total Population 1.61 Table II. Net Undercount Rates for 2000 Census Net Undercount Population Group (percent) American Indian and Alaska Native 4.74 (on reservation) American Indian and Alaska Native 3.28 (off reservation) Hispanic Origin (of any race) 2.85 Black or African American 2.17 (not Hispanic) Native Hawaiian or Other Pacific 4.60 Islander (not Hispanic) Asian (not Hispanic) 0.96 White or Some Other Race 0.67 (not Hispanic) Total Population in Households 1.18
Some of the causes of the differential undercount, lower literacy rates for example, are relatively obvious. Others are more elusive, and Skerry canvasses the social scientific, anthropological and ethnographic literature on the subject to explain why the Census Bureau fails to count every group at an equal rate.(16) He arrives at the following causes:
* "Unusual" Households: The postal service or census enumerators must know where to find people first to give them the census form or interview them. Groups are less likely to be counted if they tend to live in structures without addresses that are undetected and unreachable by the enumerators, live in "ad hoc households" (i.e., where non-family members move in and out with regularity), or live as multiple families in the same structure.(17)
* Mobile People: An individual must be at an address at some point during the period of the census in order to be counted. People who migrate from place to place to find work or housing are much less likely to be counted.(18)
* Negotiating the Form: An individual must recognize the census form, understand the questions, and be willing to fill it out in order to be counted. Literacy in English is a near-prerequisite for filling out the form.(19) At a minimum, the respondent must understand that an envelope addressed to "resident" refers to him/her and understand what the Census Bureau means by "household." The less able one is to understand the form and the culturally specific concepts it includes, the less likely one is tO be counted.(20)
* Deliberate Avoidance of the Census: To be counted individuals must not fear that the information on the census form will be used to hurt them, and often they must be willing to be interviewed by an enumerator. Thus, people are less likely to be counted if they fear investigations for welfare fraud, immigration law violations,(21) or rental agreement violations, or if they live in violent neighborhoods where opening the door to anyone, let alone an enumerator, is risky.(22)
Groups are less likely to be counted if their members are harder to find, less capable of participating, or more reluctant (for whatever reason) to cooperate with the census. Skerry is quick to point out that among undercounted groups the causes for the undercount vary such that any particular strategy to remedy the undercount will likely benefit some groups more than others.(23) But almost all (including Skerry) would agree that the undercount is an artifact of the way we conduct the census: Reliance on census form mailback and enumerator follow--up necessarily biases the numbers to the disadvantage of certain groups.
B. Consequences of the Differential Undercount
The consequences of the differential undercount itself are hard to specify. The two main areas of concern have been federal funding and legislative representation. In both instances, Skerry questions whether an undercount in the census necessarily "causes" losses in benefits to undercounted groups. Perhaps the critical paragraph in Skerry's book is the following:
[T]he implications of the undercount, and of adjusting it, have been misunderstood, in fact exaggerated. While the various participants in this continuing controversy argue as though their interests were clear cut and substantial, a careful review of the evidence reveals those interests to be ambiguous and even minimal. Moreover, these interests are highly contingent and thus hard to predict.(24)
Even though we know which groups the census undercounts, argues Skerry, winners and losers because of a failure to adjust are hard to predict in any systematic way. Undercounting, by itself, doesn't hurt any group in particular, and statistical adjustment will not produce the public policy benefits that the undercount supposedly takes away.
1. Federal funding.
The federal government uses census data to allocate about $200 billion each year.(25) Medicaid, Head Start, and various health, education, welfare, and transportation programs all use census data in their formulas for allocating money to states and localities.(26) City governments, particularly through the U.S. Conference of Mayors,(27) have been in the forefront of the battle to adjust census numbers, arguing that much-needed federal funds were transferred out of urban areas to the suburbs because of the undercount. The Presidential Members of the Census Monitoring Board commissioned a study from Price-Waterhouse, which found that cities stood to lose $11.1 billion under a census regime that refused to adjust the data to compensate for the undercount.(28) And since the 1980s, high-growth states have expressed their frustration with the undercount, sometimes even going to court in failed attempts to order adjustment.(29)
Skerry views the funding issue as more complex than "fewer people, fewer dollars." Adjustment, he argues, would not necessarily help undercounted states and might even hurt some of them. Skerry points to extensive research that suggests that the decision to adjust census data to remedy the undercount would reallocate about half a billion dollars of federal funding, a staggering sum but only 0.33 percent of the total.(30) He also argues that grants to localities from either the federal or state government would be left largely unchanged were the Census to fail to adjust the numbers. As Skerry puts it,
In sum, there is virtual unanimity among researchers that the fiscal impacts of the undercount, though not insignificant, are relatively minimal. Knowledgeable analysts understand this. The interesting question ... is why these well-known research findings are not reflected in the political debate.(31)
This battle over whether disparate rates of undercount lead to warped aggregate funding allocations is less interesting than Skerry's larger argument that those concerned about funding misunderstand their interests.(32) Because funding formulas are based on relative population numbers--the share of the national population that lives in a given state--rather than absolute numbers, a state's loss of a large number of people due to a failure to adjust does not mean that adjustment is in the state's interest. The proportion of federal funds a state receives depends on the state's relative population growth vis a vis other states. Therefore, in order for a state to be fiscally "harmed" by the undercount, its population must be undercounted at a greater rate than other states, not just harmed in some aggregate sense (i.e., the number of people that are missed).
Two examples--one real and another made-up--might help prove the point. In 1990, New York state suffered a net undercount of about 300,000 people, whereas the census missed about 100,000 people in Arizona.(33) Yet, in federal funding formulas, Arizona would have gained money under a regime of adjustment while New York would have lost money, because New York's rate of undercount (1.72%) was much less than Arizona's (3.3%).(34) This concept is so important to Skerry's argument that it deserves further elaboration. Consider the following hypothetical nation that needs to divvy up one million dollars among four states:
Note that the state that suffers the greatest population loss due to a failure to adjust (A) actually loses money ($20,000) if the census is adjusted. The reason is that while A's total population and population loss are largest in the country, other states with smaller populations (C & D) lost a greater share of their population due to a failure to adjust. The perverse result of adjustment in such a case (which is merely an exaggeration of what actually happened in 1990(35)) is that the state that gains the most people under adjustment also loses money. Furthermore, state B, which lost just as many people as C and D because of a failure to adjust, nevertheless loses the most money because of adjustment. This happens because B had the lowest undercount rate in the nation, despite losing as many people as the smaller states.
The puzzle of federal funding gets more complicated when one factors in the different undercount rates between regions of a state. New York City, for example, with its relatively high undercount rate of 3.0% in 1990 would appear to have a different incentive on the adjustment issue than New York state with its modest 1.7% undercount rate.(36) Because most federal funds go through the state to get to the city (i.e., such that the state's undercount rate, not the city's, matters most in the federal formula), New York City will receive more money if census data are left unadjusted and its population remains substantially undercounted. This is true regardless of whether the state itself uses adjusted data when it decides how to allocate the funds within the state. No matter which data the state uses, its subdivisions will often have an interest in whichever data set will distribute more funds to the state itself.
2. Reapportionment and redistricting.
A similar dynamic governs reapportionment of the U.S. House of Representatives. Thus, for the hypothetical in Table III, substitute 435 seats in place of one million dollars, with each state receiving a number of representatives proportional to its share of the national population. Again, the undercount rate relative to other states will determine the number of representatives that a state might receive. Things get a little tricky when it comes to apportionment, however, because only a small subset of states risk losing or gaining seats because of adjustment.(37) Whereas in the federal funding context the relationship between population share and share of distributed funds appears more direct (i.e., a 1% difference in population share between the adjusted and unadjusted numbers translates into 1% difference in funds allocated to the state), in the context of reapportionment a 1% difference in population revealed through adjustment might not lead to an additional seat for the state in Congress. Because only a handful of states are in competition with each other for the last of the 435 seats in the House of Representatives, a state that needs to gain 100,000 people to earn an additional seat is in a different position than a state that is just above or below the threshold to gain an additional seat.(38)
Table III. Impact of Adjustment on Federal Funding for Hypothetical Four-State Nation Headcount Headcount Population Adjusted State Population Share Population A 1000 50% 1200 B 700 35% 800 C 200 10% 300 D 100 5% 200 Total 2000 100% 2500 Adjusted Population # of People Under-count State Share Missed Rate A 48% 200 16.7% B 32% 100 12.5% C 12% 100 33.3% D 8% 100 50.0% Total 100% 500 20% $ received if $ received $ gained due State unadjusted if adjusted to adjustment A 500,000 480,000 -20,000 B 350,000 320,000 -30,000 C 100,000 120,000 +20,000 D 50,000 80,000 +30,000 Total 1 million 1 million 0
Much time could be spent discussing the mathematical scenarios under which certain states win and lose in the apportionment process under adjustment, but the Supreme Court's decision in Department of Commerce v. United States House of Representatives(39) ended the speculation as to which numbers could be used in that context. The Court there held in a 5-4 decision that the Census Act prevents the Bureau from adjusting data for apportionment purposes.(40) That decision is based on a statute, not the Constitution, so Congress (in theory) could change it at any time. The decision is also only limited to apportionment of the House, nothing else. In fact, the plurality opinion held: "[The Census Act] requires the Secretary to use statistical sampling in assembling the myriad demographic data that are collected in connection with the decennial census.(41) Congress "changed a provision that permitted the use of sampling for purposes other than apportionment into one that required that sampling be used for such purposes if `feasible.'"(42) Therefore, adjustment advocates argue that for other purposes, particularly intrastate redistricting, the statute requires that adjusted data be used.
Redistricting within a state, as opposed to apportionment of representatives among states, may present a different set of issues, although Skerry maintains that the adjustment controversy in this context also is misunderstood. Redistricting is described by those involved in it as "political blood sport."(43) It is perhaps the most politically charged activity in which legislators engage because it allows the representatives to choose their voters and guarantee themselves job security. Because census data serve as the building blocks for redistricting plans, every aspect of those data has political significance.
For Skerry, however, the fight over the data to be used in redistricting is unwarranted and at times, nonsensical. As with federal funding and reapportionment, Skerry views the controversies surrounding redistricting as more symbolic than real and based on flawed assumptions as to the independent significance of the census.(44) The political impact of redistricting derives from where the lines are drawn, Skerry argues, not which data redistricters use in the process. Democrats' advocacy and Republicans' opposition to adjustment arises from the belief that the undercounted population is predominantly minority and Democratic. Even if that is true (and it is hard to disagree with that proposition), it is unclear who benefits from adjustment.(45)
Whether Democrats or African Americans constitute the relevant group, the number of people in a district, by itself, will not determine the identity of who is elected. Crafty line-drawers can crack among several districts or pack into a few districts the population of the party or group they wish to underrepresent. These strategies remain unaffected by redistricting officials' choice between two data sets. In fact, the incumbents drawing the lines will pay attention to precinct data on election results and partisan affiliation more than they will census data. The political fortunes of a group--defined by race or party--in the redistricting process depend on whether that group exists as a cohesive majority in the district, regardless of the size of a district's total population.
Although Skerry does not do so, one could even hypothesize scenarios where it would be in an undercounted group's interest to use unadjusted data. If a districting official in a highly racially polarized jurisdiction were intent on drawing districts with a 48% African-American population, for example, African Americans, who suffered from a 4.4% undercount rate in 1990, would arguably be "better off" if that plan used unadjusted data. Although the district would then have more people in it, African Americans would constitute a majority in the district and then (in theory)be in a better position to elect their candidate of choice.(46) One can play with the numbers to produce any type of outcome (as will redistricting officials), but the impact of district lines on the fortunes of particular candidates depends almost entirely on the political or racial composition of the population they capture, not the type of data used. The use of unadjusted data would lead to certain districts having larger populations than others, but a group's power within a district will depend entirely on how large a share of the district it represents.
Skerry proves that concerns over the undercount may be exaggerated. He shows that the actual harms in terms of funding allocations and "representation" are difficult to estimate. What he fails to prove, as I argue at length later, is that the ambiguity in the data and the effects of the undercount justify inaction and resignation to the racial disparity that results from the failure to adjust.(47) His argument only works if the dangers of adjustment outweigh these albeit relatively minor absolute costs of inaction. Because those costs fall disproportionately on minority populations defined by race, Skerry and all those who oppose adjustment have a pretty high bar to cross if their argument should win the day. Race-based effects are different for both the law and politics. Especially when viewed in the context of the legal scheme of the Voting Rights Act and the court battles for minority legislative representation over the past decade, policies that seek to address race-based biases in the census should be presumptively favored over those that acquiesce to them.
II. THE ARGUMENT AGAINST ADJUSTMENT
A. The Sampling Controversy(48)
In order to remedy the undercount problems observed in the 1990 census, the (first) Bush administration convened a panel of experts from the National Academy of Sciences to study possible ways to remedy the undercount.(49) As originally planned, the 1990 census was to be adjusted, but then-Secretary of Commerce Robert Mosbacher pulled the plug on the program at the last minute because of his doubts as to adjustment's reliability. Skerry describes the vicious fights that raged within and outside the Bureau on the issue of adjustment for the 1990 Census.(50) Charges of bureaucratic capture by census officials and politicization by the administration were thrown around in all branches of government.(51)
In the wake of the turmoil, however, the plan to adjust the 2000 census emerged. In its original plan, the Census Bureau set out to sample the population that did not return the mailed census form, a program called Sampling for Non-Response Follow Up (SNRFU).(52) According to that plan, the "headcount" would only include those people who mailed in their forms, but that number would be statistically adjusted based on findings from a sample survey of the non-responding population. In other words, based on inferences derived from the sample of the non-responding population, the Census planned to add people to the headcount until it came as close as possible to producing a dataset that it thought represented what would have happened if 100% of the population had responded to the census.
The Supreme Court's decision banning sampling for apportionment purposes sent the Bureau back to the drawing board, however. The Census punted on SNRFU because the Court now made the headcount numbers much more important than they would have been otherwise.(53) It was insufficient to rely, for example, on a headcount of 60% of the population and then sample the rest, because then only 60% could be used for purposes of apportionment. The headcount needed to be as accurate and complete as possible. Therefore, in a revised plan to substitute for SNRFU, the Census decided to send up to two census forms to every household in the country and have census enumerators follow-up personally to every address from which a form was not received by June. All this was done before the Bureau then sent out specially trained enumerators to a sample set of 300,000 addresses drawn from the entire population.(54) To reiterate, under the plan adopted for the 2000 Census, the Bureau sampled from the entire population, not just the non-responding population, in order to correct for biases in the headcount.
Skerry walks the reader through a parade of horribles produced when inferences are made about biases in the headcount based on the sample and then headcount numbers are changed to remedy those biases.(55) Figuring out if someone was actually missed from the headcount turns out to be a very difficult task. People move, they answer the form differently than they do an enumerator interview, and some people cannot be reached by either an enumerator or the postal service.(56) The sampling process is fraught with these types of error and a number of other types. Mistakes made in the sample survey then magnify the error in the adjusted numbers that are produced.
After completing the sample survey, the Bureau must compare the sample with the headcount. Because the Census needs to know not only how many people it missed but also their demographic characteristics and location, comparisons between the headcount and the sample are made between small slices of the population called poststrata. Each poststratum defines a particular group of people by race, gender, renter-owner status, age range, region and whether they live in an area with a low mailback return rate.(57) The choice and number of poststrata are very important and can bias the results of the adjusted data.(58) With too many poststrata, each member of the poststratum in the sampled population then becomes very powerful in terms of adjusting the headcount. The more the census splices the 300,000 addresses of the sample, the fewer people then appear in each poststratum and the greater the effect those people will have on the adjustment factors created for the poststratum.(59) Errors involved in constructing the sample, poststrata, and adjustment factors are called sampling error.
"Nonsampling" errors also plague the adjustment process, and Skerry suggests such errors will likely be greater than sampling error.(60) These types of errors include computer coding errors, which, for example, erroneously added 1 million people to the 1990 undercount estimate and enumerator errors such as when enumerators make mistakes or get lazy and falsify forms.(61) Skerry cites some harrowing statistics and studies regarding non-sampling error in the 1990 census to show that "adjusted census numbers can be overwhelmed by non-sampling error."(62)
Despite the well-recognized errors, three panels of the National Academy of Sciences in the 1990s have urged statistical adjustment of the 2000 Census.(63) But Skerry is quick to point out that there is broad and bitter disagreement in the statistical community over whether the census should be adjusted. There is no "right answer" from the standpoint of statistics or science on the question whether adjustment produces "better" numbers. One's definition of better depends on one's political priorities. Those priorities might grow from partisan biases, the degree of importance one attributes to the differential undercount, or one's expected gains from one set of numbers over another. Skerry urges us to focus on the explicitly political choices involving contending priorities instead of resigning ourselves to abstract notions of science or expertise.(64)
B. The Symbolic Politics of Census Adjustment
Skerry does a good job in proving that important census decisions are dominated more by politics than science. He describes the political tradeoffs as ones between "Counts or Shares? Races or Places?".(65) As explained in the discussion of federal funding and apportionment, the absolute number (i.e., "count") of members of a group is less relevant for most purposes for which census data are useful than the proportion of the population such a number represents (i.e., "share"). Depending on the type and forum of argument, a group may emphasize how large the group is in absolute terms (for example, how much it has grown since the last census) or how large a share of the population it represents (for example, that the group includes more people than other groups that get more benefits or legally recognized rights). The rhetoric vacillates between the two emphases, and one's interest and advocacy for adjustment may change depending on whether one focuses on "counts" or "shares."
Skerry's second pairing--"races or places"--is in many respects more insightful and important. It is important, because as Skerry briefly recognizes, the "focus on place-oriented outcomes reflects the innate bias of a political system based on representation of geographically defined jurisdictions.(66) The adjustment controversy, like enforcement of the Voting Rights Act and protection for minority representation generally,(67) forces policy makers and advocates to prioritize between different aspects of individual level identities. The importance of geographic identity comes from our form of government based on states, regionally defined jurisdictions, and single-member legislative districts. How many people live in New York City or in the State of California is important for the number of representatives that will be apportioned to those geographic regions and over how much money the regional governing authorities will then have discretion. But for similar and different reasons, race also has political and legal significance independent of "place." The Equal Protection Clause, the Fifteenth Amendment, the Voting Rights Act, and the Civil Rights Act give political primacy to race, sometimes at the explicit expense of geographically defined interests.(68) Moreover, in the political world racial group "strength" is often measured by "representation" in the census.(69) Arguments as to distributive justice or legal recognition resonate more strongly when data suggesting group numerical strength are there to back them up.
At times, Skerry appears agnostic on the politics of census adjustment, merely pointing out the inherent tradeoffs between different options, yet we are told from the beginning that this book is a "brief against census adjustment."(70) Why he has such antipathy for adjustment is unclear until the reader gets to the point in the book where he lets escape the real target of his ire: politics based on symbols and rights. In Skerry's view, most of the alleged stakeholders in the adjustment controversy have all bought into mistaken, overly symbolic criteria for evaluating the proper way to conduct a census. Skerry views Census bureaucrats as wanting to "do the right thing,"(71) meaning to try and distinguish the census as an arena of American life where race-based disparities are remedied rather than accepted. He sees Republicans as opposing adjustment because of a generalized suspicion of social scientists and reliance on a mistaken analogy to affirmative action.(72) And finally, Democrats are either cowardly in their acquiescence to minority concerns, or opportunistic in allowing adjustment to be a major priority because it is a relatively costless civil rights struggle that they can win.(73)
Skerry's opposition to adjustment rests more on what the adjustment "movement" represents than on what adjustment will or will not do. He sees adjustment as part of a larger, harmful trend in American politics that began in the 1960s. That politics is dominated by organizations without members,(74) who emphasize formalistic representation over substantive policy. As he puts it,
What began as a movement to secure the political participation of previously excluded segments of American society has eventuated in a regime in which participation has been exchanged for representation of an extremely formalistic nature. Having relied heavily on the courts and administrative bureaucracies, erstwhile reformers and even radicals are now insiders with weak ties and little accountability to those whom they represent. In this context, census numbers and other such data are important, but more as instruments of bureaucratic administration than as levers of political mobilization.(75) Our political system has come to emphasize the representation of interests of the disadvantaged over their actual participation in the political process.... Caught up in this representation-participation trade-off, we have come to emphasize group numbers over what groups actually do with their numbers. In a fights-oriented regime where entitlements from the welfare state are based at least as much on number as on political clout, statistics are a political resource. They do represent a kind of political power.... Indeed, it may well be that in the context of the contemporary administrative state, statistics may obviate the need for mobilization and organization.(76)
Skerry thus places himself in a school of thought that views "public interest politics" as a game played by liberal elites that speak in a language of rights and seek symbolic victories and media attention over policy results. His bitterness and disappointment with this "new politics" (which is at least 30 years old, and maybe much older) comes from a deeply felt belief that such movements foster and thrive on cynicism and contentiousness. Toward the end of the book, his pent-up rage flows freely:
[T]he ongoing controversy over the census undercount and adjustment does reflect the complicated cross-currents of this new regime: the extraordinarily high expectations placed on the census; the lack of clear standards by which to evaluate results; the shift of power away from political actors representing concrete interests and toward issue networks and public interest advocates articulating vaguely defined interests of hard-to-organize constituencies; the expansive conception of rights; a politics of symbols and ideas that is contentious, media-driven, and seemingly never-ending; the reliance on experts and the appeals to science; and finally, the paradox of governmental openness and distrust.(77)
Skerry's argument is weakest when he is most angry. At times it is just plain wrong. Skerry models the current state of American politics with the following causal chain. In the 1960s and 1970s, public interest organizations developed to represent previously unorganized interests. After some significant, substantive victories,(78) their reasons for existence and membership bases (to the degree they ever had them) diminished. Racially defined interest groups or organizations are particularly susceptible to this criticism because Skerry considers race socially constructed, highly contingent and variable among groups, and diminishing in importance as a group-defining characteristic.(79) The leaders of these public interest organizations remain, however, as do their third-party funding sources to which they are beholden. (In the spirit of full disclosure: this author currently works for one such organization, albeit one Skerry does not target.) To satisfy these funders, such organizations resort to high profile litigation and symbolic politics rather than mass movements that require intense mobilization of their loosely attached rank and file. The result is a politics that thrives off uncompromising stances on issues of little practical but highly symbolic importance. The battle over census adjustment is one such symbolic battle, according to Skerry.
C. The Alleged Costs of Symbolic Census Politics
Were census politics merely symbolic Skerry might not find it so damaging to the body politic. But he finds symbolic politics generally and census politics in particular to impose three distinct and severe costs for American democracy. First, he predicts that adjustment will lead to lower cooperation, mobilization and participation in the census. Second, he thinks adjustment will lead to greater cynicism and distrust of government. Third, he asserts that somehow adjustment will hurt minorities. If Skerry could prove any of these arguments, then the argument that adjustment's benefits have been overrated could be converted into an argument that such minimal benefits are outweighed by adjustment's costs. The book fails to support this critical step in his argument, however.
1. Adjustment politics and "participation meltdown."(80)
Part of Skerry's argument has fallen victim to events subsequent to the book's publication. He warned that adjustment of the census might lead to lessened participation in the headcount process. Because people would assume that they would be added into the census through adjustment, they would not feel the need to fill out the census form or respond to census enumerators. Stakeholders, such as the "memberless" organizations representing minorities that Skerry decries, would be less likely to rally the troops as they did to cooperate with the census. As it turned out, despite the Census Bureau's plans to adjust, the response rate in the 2000 headcount process was higher than expected and 2% higher than 1990.(81)
In fairness to Skerry, the Bureau did not employ the statistical method SNRFU ("Sampling for Non-Response Follow-Up") that he found to be the most dangerous threat to census participation. Under that system, as described above,(82) the Bureau would make inferences about the nonresponding population based on a sample of that population. In place of SNRFU and subsequent to the Supreme Court opinion which made accuracy in the headcount so important, the Bureau decided instead to sample the entire population to make inferences about and to adjust the headcount numbers.(83)
Nevertheless, Skerry maintains that the general discussion and controversy over adjustment would tend to chill participation in the census.(84) The dramatic success of the 2000 Census--both in practice and in the final results--has disproved this portion of Skerry's argument.(85) The Census Bureau was relatively successful in coordinating with and mobilizing locally based organizations to advocate for census participation.(86) In fact, the politics of adjustment, I would suggest, warded off the precise impact of lessened cooperation that Skerry teared. In the heat of the adjustment controversy, Republican lawmakers sought to counter the charge that their opposition to adjustment reflected their resignation to the differential undercount and an inaccurate census, in general. Responding to the claim that adjustment was the only way to increase census accuracy, Republicans adopted a "roll up your sleeves" approach, which translated into an additional billion dollars in funding for the Census Bureau to make the headcount more accurate.(87) Knowing that adjustment was the Democrats' tool for remedying the undercount, the Republicans wanted to avoid charges of apathy and insensitivity and did so by backing a remarkable increase in funding to the Bureau. That funding went toward an unprecedented advertising campaign and outreach effort that led to the increased response reflected in the headcount data.(88)
The "participation meltdown" hypothesis was suspicious even before the Census began, however, because it assumes that the broader population is as aware of the adjustment controversy as Skerry is. Skerry points to the one study of public opinion on census adjustment, which showed that only 5% of those surveyed displayed any understanding of the issue.(89) It is all the more mystifying then that he views the adjustment controversy as so notorious that it will actually have an impact on census "turnout". It takes a remarkable level of interest, understanding, decision-making and translation into action for the average person to interpret the adjustment controversy as implying that participation is less important. The argument might be more persuasive with regard to elites, such as leaders of organizations who will make decisions whether to "push" the census as an organizational priority, but no evidence of such a chilling effect materialized. Indeed, interest groups defined by race or region appeared more active than ever in promoting census participation.(90)
2. Rising cynicism and distrust of government.
The same charge could be leveled at the second cost that Skerry sees afflicting adjustment, namely, that adjusting the census will lead to cynicism and distrust of government.(91) Skerry views census adjustment as typical of a type politics that uses courts and bureaucracies instead of legislatures and grassroots organizing to propel a cause or policy. Thus, "politics" is taken out of the organs of government that are most responsive to electoral pressures and transferred into the least democratic institutions in the American constitutional scheme. The result is a citizenry detached, suspicious, and cynical about politics. According to Skerry, "these influential actors not only lack democratic legitimacy but also tend to disagree in policy debates, causing the general public to believe that those running the government do not know what they are doing. One result among ordinary citizens is ... the cynical response that `everything causes cancer.'"(92)
Again, the ability of the census to cause such widespread disenchantment would assume a level of political sophistication in the general populace that simply does not exist.(93) For people to become disenchanted, it would seem they must know what kinds of decisions are being taken out of their hands (a paradoxical concept to begin with). But Skerry asks us to view the adjustment controversy as emblematic of closed-door politics with important, but complex, decisions being made by the least democratic institutions. Although census adjustment might not be the straw that breaks the camel's back (i.e., the policy that pushes any American over the cynical brink regarding trust in government), it is part of a dangerous trend in politics that should be opposed at every level.(94)
Of course, Skerry's argument evolves into a larger (perhaps inadvertent) attack on liberal politics, the administrative state, and government intervention into society dating even as far back as the New Deal.(95) So little of today's politics is participatory in the way that Skerry dreams it should be. Public policy regarding the environment, health care, electoral reform, defense procurement policy, and the national economy, to name just a few important issue areas, is naturally complex and beyond the comprehension of anyone besides those who dedicate a considerable amount of time to studying the issue. This unfortunate aspect of modern society is rarely considered a rallying cry for resignation in the face of complex social problems.
Moreover, it is not so clear that census adjustment politics is as technocratic and arcane as Skerry would have us believe. Elected politicians, acting perhaps on what Skerry decries as a mistaken belief that census numbers in some way affect their political future, are actively involved in the decision whether or not to adjust and can play a large role in the process of adjustment.(96) Indeed, elected politicians and the political parties seem remarkably involved in the intricacies of the census given how arcane and technocratic a process it is. So long as census data .are used for redistricting, we should expect their involvement never to fade.
Finally, Skerry's broadside suggestion that government involvement in technical issues leads to cynicism and distrust lacks support in the relevant literature. For a book that is so well-footnoted when it comes to describing the census and adjustment's shortcomings, it is remarkable that it substitutes casual prognostications about "what the American people believe" for what could be a richer description of the dynamic of political trust. The literature on political trust is voluminous and remarkably coherent in its conclusions, none of which supports Skerry's argument.(97) Instead, Skerry points to occasional paragraphs in the literature on the growth of the administrative state to support what is more properly a contention about public opinion.(98)
From the public opinion literature one will learn that Americans' distrust of government has remained relatively constant over the past twenty-five years.(99) The percent of Americans who say they trust government "about always or most of the time" reached a high of 70% in 1964 but then dropped sharply throughout the Vietnam War and Watergate, and continued to decrease throughout the stagflation of the 1970s. Since Watergate, however, political trust (as measured by the survey question stated above) has vacillated between 20% and 40%.(100) The vacillations are explained by two main factors: scandals (such as Watergate, Iran-Contra, the Clinton Impeachment) and economic conditions.(101) Americans' trust of government has waxed and waned completely irrespective of the "level of activism" or technocracy that Skerry suggests would be behind a census adjustment-cynicism link. In fact, "public desire for government services and activism has remained nearly steady over the past 30 years'"(102) even as levels of distrust have gone up and down. Our leaders' ethical lapses, the economic conditions of the 1970s, and the recession of the early 1990s are more powerful explanations for the persistent levels of distrust and cynicism.
3. The possible costs of adjustment for minorities.
It is quite clear that Skerry views adjusted data as less accurate than unadjusted data. But as he implores the reader throughout the book, census numbers are about politics, not science, so crude claims of "accuracy" divorced from context have no meaning. To be sure, as described above, Skerry provides some real horror stories from the sordid history of the 1990 census and demonstrates that adjustment of the 1990 census would have erroneously added one million people. But abstract claims of accuracy do not have real political impact unless they can be tied to the inefficient execution of public policy--a link that Skerry appears loathe to make.(103)
Instead, Skerry, at times, suggests that minorities might be worse off under adjustment. There appear to be two reasons for this: (1) the "counts versus shares" argument presented above; and (2) the argument that adjustment will prevent other types of social action and mobilization that will actually help the disadvantaged. On the first argument, Skerry shows that the benefits of adjustment will vary based on region and among different minority groups, with some standing to benefit more than others through the process of adjustment.(104) Depending on the idiosyncrasies of the given jurisdiction (e.g., the extent of racial diversity, the reasons why people did not respond to the census, and the heterogeneity within race groups), African Americans, for example, might do relatively worse under adjustment than they would with just a headcount. This is no doubt "true" in some cases, but as a reason for rejecting adjustment it falls flat. Even Skerry appears to agree that "on average" adjusted numbers will give a more accurate count of racial minorities.(105) However, averages can obscure individual variations, and to be sure, census adjustment may introduce greater error at the smallest levels of aggregation such as the census block level (i.e., adjusted numbers will be more accurate for the nation or a state than they might be for any given neighborhood). The contingency and variation inherent in adjustment, however, do not impeach it as a remedy for the differential undercount. Indeed, the accuracy of the headcount will also vary based on any number of conditions so "contingency" can hardly be an argument well-targeted to the abandonment of adjustment to correct the systemic differential undercount. Less persuasive still is the argument that we should not even try census adjustment so we might be able to learn about who wins and loses under a system different from the one that we know undercounts racial minorities.
Second, Skerry raises the "participation meltdown" hypothesis again but in a more general form. He argues that preoccupation with adjustment will not only chill participation in the census, but it will distract attention from the real social problems that afflict minorities. It is worth quoting the argument at length lest I be accused of taking quotations out of context:
Instead of fostering the empowerment and organization of minorities, adjustment would undermine them. By reducing the incentives for minority groups to mobilize around cooperating with the census to maximize their totals, adjustment would render them less able to challenge critical decisions affecting their communities. Once the capacity to mobilize people around the actual count was lost, it would not likely be regained around the arcane issues that would ensue. Again, adjustment would not obviate the need for political clout but rather would shift census politics into a rarefied arena where the relative political strengths of the disadvantaged--community networks and grass-roots relationships--would be of little use. ... In a period where the electorate is substantially demobilized, elected officials, party leaders, journalists and foundation executives find it easier to advocate census adjustment than to address the more fundamental problems facing the disadvantaged. Whether one focuses on their social and economic problems or the obstacles to their political mobilization, remedies are far from clear.... How much safer to focus on a seemingly more tractable issue like the census undercount, whose remedy--adjustment--is couched in terms of enfranchisement and empowerment! Even for opponents of adjustment, this debate is a more comfortable one than many likely alternatives.(106)
Census adjustment thereby replaces "real" issues as a focus for organizational resources. Instead of solving the problems of the poor, organizations gravitate toward solutions of how better to represent them in official statistics.
Skerry once again views the world--this time the world of liberal policy makers and advocates--as one preoccupied with the problems he finds important. As one who has attended meetings of almost all of the organizations Skerry lambastes (NAACP, MALDEF, PRLDEF), I can say with some confidence that the adjustment issue, while getting a fair amount of time in discussion, never constitutes more than one panel in a conference of six or seven panels (and usually is taken care of by one speaker on one panel). Adjustment has not prevented the NAACP or the Lawyers Committee for Civil Rights from flying down to Florida on a moment's notice to begin litigating cases relevant to the 2000 election or continuing to fight for integrated educational institutions and voting rights. It has not prevented Latino groups from focusing on immigration issues and redistributive decisions in their communities. And when the census came up in the months surrounding the 2000 Census, the adjustment issue was usually counterbalanced by equal time devoted to issues involved with mobilizing for the census,(107) issues that Skerry would prefer as the organization's focus.
Were adjustment an accepted part of the census regime would such groups somehow focus more on the technical issues of the adjustment process at the expense of "real" census participation and mobilization for "real" issues? For those, like Skerry, who preoccupy themselves with the subsections of the civil rights community that pay attention to the census issues, it seems like too much attention is and was being paid. But the myopia that Skerry attributes to the civil rights community for focusing on the census is a charge that should be thrown back at Skerry himself. For it is his focus on one portion of the civil rights community's agenda that distracts him from the larger struggles of which census adjustment is only one small part. One of these larger struggles involves the enforcement of the Voting Rights Act and the fight to craft legislative districts where minorities can elect their candidates of choice.
III. WHEN ADJUSTMENT MATTERS: ENFORCEMENT OF THE VOTING RIGHTS ACT AND GUARANTEEING MINORITY REPRESENTATION
Once the reader shaves away the unnecessary political diatribes against public interest organizations, courts, bureaucracies, and other agents of symbolic politics, Skerry's basic contribution remains: The harms of the undercount and the benefits of adjustment to any particular group are highly contingent and uncertain. Although the costs of adjustment may also be ambiguous (e.g., the extent of participation meltdown and rising cynicism), advocacy for adjustment must depend on more than an argument that all is uncertain so therefore we should give adjustment a try. The piece of the puzzle that Skerry leaves out of his description of the adjustment controversy is the importance of the census in enforcement of the Voting Rights Act.(108) Indeed, what Skerry misunderstands as symbolic politics on the part of the civil rights community is actually a small part of quite substantive (even if sometimes controversial) politics in a forty-year struggle to achieve "effective" legislative representation. Even small undercounts in the census might affect the ability of racial minority groups to get into court to present claims relating to effective legislative representation. Adjustment of the census and enhancing group numbers could be a precondition to launching a successful voting rights suit.
There is a larger purpose to this discussion than to prove that the Voting Rights Act is the exception to the general rule that the costs of the undercount and the benefits of adjustment are exaggerated. What skeptics in this area fail to recognize is the place Of the census adjustment issue in a larger controversy over issues of minority vote dilution and representation. The gathering and distribution of census numbers represent the first stages in the process of creating and litigating over legislative district lines.(109) To a certain extent, Skerry is right that census adjustment by itself will not produce any discrete substantive benefits in terms of minority representation. However, the uncertainty of the adjustment process is compounded by the uncertainty in the state of the law when it comes to minority representation. The civil rights community's fight for adjustment at the front end of the decennial redistricting process makes sense particularly when put in the uncertain context of what the Supreme Court, the Department of Justice and state governments might do to favorable minority districts in the first couple of years of the decade. When placed in the context of the Supreme Court's repeated drubbing of voting rights advocates in the last decade and the fear that an Ashcroft Justice Department will not enforce the voting rights laws, the unwillingness of the civil rights community to sit out the fight over which numbers would be used in redistricting becomes quite understandable.
As it turns out, the undercount for the 2000 Census was relatively small and the Census Bureau will not adjust the data to be released in time for redistricting.(110) States and localities will soon receive the headcount data, and the adjusted data will probably be made available sometime in 2002 after district lines are drawn. Although it does not appear as though adjustment would alter the 2000 data in a significant way, the failure to release the adjusted data in time for redistricting may have some legal impact. More important from the standpoint of the argument presented here, the lack of any serious need to adjust the 2000 data should not lead the Census Bureau to rule out adjustment for the 2010 Census or subsequent censuses. So long as the viability of some legal claims depends on crossing a numerical population threshold, the adjustment controversy will have legal significance for the voting rights community. If courts or the Department of Justice interpret the Voting Rights Act in such a way as to give a special status to majority-minority districts--that is, districts in which the minority population exceeds 50%--the census adjustment controversy will remain part of the struggle to attain effective minority representation. Although the argument in this section demonstrates the legal significance of adjustment, one could also interpret it as a criticism of the legal standards that make adjustment important.
A. Section 5 of the Voting Rights Act(111)
Suits brought under Section 5 of the Voting Rights Act present the first context in which the census adjustment controversy will arise in court.(112) Certain jurisdictions with a history of voting rights violations against racial and "language" minorities must "preclear" laws related to voting, including redistricting plans, with the Department of Justice or the U.S. District Court for the District of Columbia.(113) Preclearance will depend on whether the law or redistricting plan is "retrogressive" in protecting minority voting power--that is, whether the law worsens the position of minority voters in their ability to elect candidates of their choice. For redistricting plans, courts have found retrogression when the new plan reduces the number of districts where the minority forms the majority (i.e., "majority-minority districts").(114) Admittedly, the Department of Justice in the 1990 round of redistricting was aggressive in supporting the creation of majority and super-majority-minority districts, rather than avoiding mere retrogression, and DOJ might adopt a relaxed standard for retrogression in the 2000 round that does not place such primacy on numerical majorities. Nevertheless, a jurisdiction that reduces the number of such districts does so at its peril and invites close administrative scrutiny of its redistricting plan.
The number of majority-minority districts that serve as the baseline for new redistricting plans will depend in part on which census data are used to create those plans.(115) The use of data that underrepresents the size of the minority population could lead DOJ to preclear what are otherwise retrogressive plans because the number of majority-minority districts will appear to remain constant. For example, assume a situation where the unadjusted data show that the plan in effect in 2000 (that is, before the new lines are drawn) has one majority-minority district but adjusted data reveal two majority-minority districts. Does the new redistricting plan for 2002 need to have one or two majority-minority districts to gain preclearance? The baseline for retrogression may depend, in part, on which data a jurisdiction employs.
The potential impact of the choice of data is displayed in Table IV for a hypothetical four-district plan in a covered jurisdiction. As the example suggests, the failure to use adjusted data to compute the minority percentages for the baseline plan in effect for 2000 could determine whether one or two majority-minority districts must be created for 2002 in order to gain preclearance. The use of adjusted data to measure retrogression might lead to DOJ rejecting the 2002 plan because it reduced the number of majority-minority districts. Were the use of unadjusted data permitted, not only would the 2002 plan not be considered retrogressive, but as the example indicates, the failure to create a baseline of two districts might open the door to strategies that dilute the minority vote under the new plan.(116) In this example, the need to create only one majority-minority district allows the jurisdiction to obtain preclearance despite the fact that the new plan packs Blacks into District 2 and depletes them from surrounding districts.(117)
Table IV. The Impact of the Choice of Unadjusted Data for Section Five Preclearance for a Hypothetical Four-District Plan in a Covered Jurisdiction ((a) Bold Indicates Majority-Minority District) % Black Using % Black Using District Unadjusted Data Adjusted Data Number (2000 Plan) (2000 Plan) 1 47 51(a) 2 65(a) 70(a) 3 10 12 4 5 6 % Black Using % Black Using District Unadjusted Data Adjusted Data Number (2002 Plan) (2002 Plan) 1 25 29 2 80(a) 85(a) 3 16 18 4 10 12
Thus, the harm the use of unadjusted data could cause comes from the different baseline indicated by those data as opposed to the adjusted set. Regardless of whether one considers minority interests to be better furthered by concentration of voters into majority-minority districts or through more "efficient" dispersion of influence among several districts, the law tends to make a distinction. Of course, there is the possibility that the law could change, as well as the possibility that the Bush Justice Department will use a different standard in determining whether to deny preclearance.(118) Moreover, the Supreme Court's decisions in Shaw v. Reno(119) and its progeny(120) have introduced considerable uncertainty into the upcoming round of redistricting. In those cases, the Court struck down eleven majority-minority congressional districts because race was allegedly the "predominant factor" in their creation,(121) Although five members of the Court apparently believe that the creation of a district in order to comply with the Voting Rights Act could justify the predominant use of race,(122) DOJ and covered jurisdictions may run scared from these decisions and avoid the creation of majority-minority districts in the 2000 round of redistricting. Given that DOJ to date has never precleared a redistricting plan that reduces the number of majority-minority districts, however, many jurisdictions will likely use that guidepost in the creation of the new set of lines. Discrepancies between the adjusted and unadjusted data, therefore, could affect whether certain districts are drawn or precleared.
B. Racial Vote Dilution Under Section 2 of the Voting Rights Act(123)
Whereas only "covered jurisdictions" must preclear redistricting plans with the Department of Justice, all redistricting plans in any part of the country are the potential subjects of a suit under Section 2 of the Voting Rights Act. To put it a bit more accurately, Section 2, like other civil rights laws, applies to every jurisdiction, but a claim under Section 2 only lies if certain conditions are met.(124) The choice of census data can affect whether such conditions exist and thus whether a claim under Section 2 will lie.
Section 2 of the Voting Rights Act protects against dilution of minority votes. Such dilution is found when
based on the totality of the circumstances, it is shown that the political processes leading to nomination or election in the State or political subdivision are not equally open to participation by members of a class of citizens protected by [the Voting Rights Act] ... in that its members have less opportunity than other members of the electorate to participate in the political process and to elect representatives of their choice.(125)
Redistricting plans can dilute minority voting power and violate Section 2 by packing minorities into a few districts, splitting up minority voters into several districts, or submerging a minority community in a multimember at-large jurisdiction.(126) But not every diluted minority group necessarily has a claim under Section 2. The plurality opinion in Thornburgh v. Gingles(127) developed a three-pronged test to determine a violation of Section 2: A minority group must show that it is "sufficiently large and geographically compact" to constitute a majority in a single-member district, that it votes cohesively, and that the majority votes as a bloc so as to prevent the election of the minority's candidates of choice.(128) Although the Court did not suggest that the test applies in all cases,(129) the Gingles test for the requisite size of a minority community (i.e., that it be able to constitute a majority in a single-member district) has had some lasting significance.(130)
The choice of data could affect the minority group's ability to show that it is sufficiently large to have a viable Section 2 claim. The math is similar to that in the example laid out in the previous discussion of Section 5. Depending on whether the jurisdiction uses unadjusted or adjusted data, a minority group may or may not appear to be large enough to constitute a majority in a single-member district. For example, in a four-member city council that elects all members at-large (i.e., because the city is not divided up into districts, every voter casts a vote for each of the four seats on the city council) a compact and cohesive minority group that appears to constitute 10% of the population could not pass Gingles' size requirement. But suppose the adjusted data suggest that the minority actually constitutes 15% of the population. Then the minority would be able to get past the first hurdle of a Section 2 claim because were the court to break up the at-large jurisdiction into four districts, the minority group could constitute a majority in one of the districts. If the adjusted data, but not the unadjusted data, indicate that the minority group is sufficiently large to have a viable Section 2 claim, then the choice of data could have legal significance.(131)
In the context of Section 5 and Section 2 proceedings it should now be clear that the importance of the choice between adjusted and unadjusted data comes from legal requirements that allow certain claims to lie only when the minority can show it is or could be a majority in a single-member district. When it is rigidly enforced, the "majority" requirement can be a pure mathematical one meaning greater than 50% of the relevant population, rather than some conceptual notion of political control.(132) If courts were to enforce a reading of the Voting Rights Act that required something rougher and more ambiguous, such as the percentage needed to elect a candidate of choice,(133) then a small undercount might get lost in the shuffle of other evidentiary requirements and would have less of a legal impact. But if courts and the Department of Justice abide by any strict numerical requirement--whether it be 50%, 65%,(134) or 40%--then the undercount and the choice of data will have a legal impact. These percentages, in a sense, may be arbitrary given the uncertainties and contingencies in census data,(135) in general, and racial data, in particular. Because Skerry overlooks this admittedly obsessive adherence to numeric requirements that some courts have paid in enforcement of the Voting Rights Act, however, Skerry misses the point of the civil rights community's vigilance in arguing for adjustment of census data.
C. The Census Adjustment Controversy in the Context of the Larger Battle for Minority Legislative Representation
Although the concrete legal harms described above ought to lay to rest the contention that the choice of data has no identifiable costs, there is a larger, albeit more murky, argument that Skerry overlooks based on his severing of the adjustment controversy from the larger struggle to protect gains in minority legislative representation. If one focuses on the adjustment controversy alone, one undoubtedly will reach Skerry's conclusion that the time expended in litigating and analyzing the problems of the undercount and adjustment exceed their tangible harms or benefits. For voting rights lawyers, however, the present adjustment controversy (and the heat it has encumbered) flows logically from a series of fights in the 1990s where the U.S. Supreme Court has articulated new rules of decision that make it more difficult for minorities to craft districts where they can elect candidates of their choice.
The U.S. Supreme Court delivered two major blows to the voting rights community in the 1990s. As already explained, Shaw v. Reno(136) created a new cause of action that led courts to strike down districts where race was the "predominant factor" in their creation.(137) Less noticeable to those outside the voting rights bar, however, was the Supreme Court's new interpretation of Section 5 in Reno v. Bossier Parish. According to the two opinions issued in the case,(138) the Department of Justice cannot deny preclearance to a plan that violates Section 2 of the Voting Rights Act or was created with a discriminatory purpose. Only if a plan has a retrogressive purpose or effect can DOJ deny preclearance.
As severe as the setbacks may have been for the voting rights community, the new uncertainties in the law have created anxiety and confusion for jurisdictions seeking to craft legally defensible plans. The law governing the upcoming redistricting process requires states to use race data to ensure compliance with the Voting Rights Act, but they cannot rely on it predominantly otherwise they will be subject to a Shaw claim. And covered jurisdictions must make sure that they do not intentionally or unintentionally retrogress, but they can purposefully discriminate without running the risk of a denial of preclearance. Jurisdictions must abide by these requirements, of course, while crafting congressional districts that obey the strict one-person, one-vote rule requiring districts to be "as equal as practicable" in population.(139) Moreover, the high political stakes of the 2000 redistricting process will ensure that all these claims--Voting Rights Act claims, Shaw claims, and one-person, one-vote claims--will be made by political partisans seeking to destroy the redistricting plan of their opponents by any means necessary. Each of these claims can represent a lever to overturn a redistricting plan that opponents disfavor for reasons wholly unrelated to the actual contentions of race discrimination or violations of equal protection. Adding insult and injury, the Florida election controversy and the Court's decision in Bush v. Gore ensured that an environment of bitter partisanship is the only predictable feature for the next decade of voting rights litigation.
Into this jurisprudential tornado dropped the census adjustment controversy. In one sense, it is highly relevant, because all the redistricting controversies will make assumptions about the population based on whichever data set the jurisdiction uses. In another sense, Skerry is right that the importance of adjustment is highly contingent on how some of these other controversies will be worked out. In the end, however, the rational strategy for civil rights lawyers in the earliest stages of the redistricting process (i.e., before the lines were drawn) was to argue that the data should appear in a form most favorable to making their claims. Whatever may happen in the redistricting process and in the torrent of cases that public interest organizations, political parties, and state governments bring throughout the next decade, the safe approach for lawyers representing groups seeking to minimize the effect of future legal and political "contingencies" was first to maximize the group's representation in the data states will use to redistrict. True, as Skerry proves, the actual benefits from the standpoint of minority representation were difficult to ascertain ex ante. Because the choice of data could affect what types of claims might be viable down the road, however, the maximization strategy when it came to adjustment of the data made perfect sense.
IV. CONCLUSION: THE RIGHT TO BE COUNTED
When taken out of the relevant legal context, the controversy over census adjustment appears as a symbolic fight over relatively tiny and uncertain changes to a complicated, even if politically charged, data set. When placed back into the context of the erosion of voting rights protections in the 1990s and the uncertain state of the law that welcomes the 2000 redistricting process, the fight takes on added and more understandable significance. The fight to adjust census data may preserve new or future claims under Sections 2 and 5 of the Voting Rights Act and perhaps provide some insurance against the next adverse decision or change in the law that lay just beyond the horizon. To suggest that the fight over census data is purely symbolic misrepresents the strategic dynamics caused by the confusion, uncertainty and urgency spawned by recent changes in the voting rights caselaw.
Moreover, the catch phrase "symbolic politics" masks a very real and deep conflict over notions of equal treatment in law and public policy. Even if Skerry is right in his suggestion that census adjustment will not bestow any significant benefit to minorities, the rabid opposition to adjusting data to correct for missed persons of color cannot go unchallenged. Indeed, the fact that the costs and benefits of adjustment are uncertain makes the opposition even to trying out adjustment to correct for the differential racial undercount all the more disturbing. As a legal matter, it appeared until very recently that jurisdictions would be faced with a choice between two types of 2000 Census data: one with a known bias against minorities and another with other, perhaps greater, possibility for error. The right to be counted takes on special constitutional significance when the uncounted are disproportionately nonwhite. Moreover, the legal fight over whether such an undercount "matters" may affect not only voting rights law but also larger legal principles concerning the margin of race-based error that the democracy is willing to tolerate.
We return in this conclusion to the analogous situation of the Florida election controversy. The Court in Bush v. Gore(140) chose between two types of equal protection violations or undercounts. By halting the recounts because they were being conducted in a haphazard and standardless manner, the Court sought to prevent the random or politically motivated discrimination that might allow one ballot to be counted as a vote while another, identical one was disregarded. By preventing any hand recount from taking place, however, the Court froze in time (and for all time) the disparities that existed between counties based on whether they had machinery that was more or less likely to register a vote. Thus, the Court chose one type of discrimination or inequality over another.
Various state actors face a similar choice when deciding whether to adjust census data. Will courts, the Census Bureau, the Department of Justice and governments at all levels acquiesce to the known race-based inaccuracies in the unadjusted census data? Or will they choose another type of inaccuracy with uncertain political ramifications but targeted to remedying the headcount's racial bias? Perhaps most important, will they halt the progress of experimentation that might allow the next census to be more accurate or less racially biased than either the current or previous ones? Although the substantive effects of the adjustment conflict may be felt by potential voting rights plaintiffs throughout the decade, the answers to these questions will be, in part, "symbolic," meaning that they will symbolize the priorities of the relevant decisionmaker. The precedents created to resolve even symbolic conflicts, however, may prove decisive in solving more substantive conflicts right around the corner. Sitting this fight out when the issues are so clear, even if the implications are not, would be irresponsible for all those fighting for political equality at each stage of the process of representation.
(1.) 121 S. Ct. 525 (2000).
(2.) See Dep't of Commerce v. United States House of Representatives, 525 U.S. 316 (1999) (interpreting the Census Act to prevent the use of adjusted data for congressional apportionment); Virginia v. Reno, 117 F. Supp. 2d 46 (D.D.C. 2000), aff'd, 121 S. Ct. 749 (2001) (holding as unripe Virginia's declaratory judgment action seeking approval of the state's ban on the use of adjusted data in redistricting); City of Los Angeles v. Evans (CD. Cal., complaint dated Feb. 21, 2001) (suit alleging Secretary of Commerce illegally usurped Census Bureau's authority to determine whether statistical adjustment of redistricting data set was feasible and demanding the release of adjusted data).
(3.) See The Florida Count, What Went Wrong: Undervotes Support Bush Win, MIAMI HERALD, available at http://www.miami.com/herald/special/news/flacount/index.htm (visited March 3, 2001) ("If a manual recount of presidental ballots had gone forward in Miami-Dade County, George W. Bush likely would have won the presidency outright, without weeks of indecision and political warfare, a review of Miami-Dade County's `undervote' ballots shows.").
(4.) See EXECUTIVE STEERING COMMITTEE FOR ACCURACY AND COVERAGE EVALUATION POLICY: RECOMMENDATION CONCERNING THE METHODOLOGY TO BE USED IN PRODUCING THE TABULATIONS OF POPULATION REPORTED TO STATES AND LOCALITIES PURSUANT TO 13 U.S.C. 141(C) (Mar. 1, 2001), available at http://www.census.gov/dmd/www/EscapRep.html (visited Mar. 3, 2001) [hereinafter ESCAP] (recommending that headcount rather than adjusted census data be used for redistricting). Despite the Census Bureau's decision not to adjust the 2000 data, the debate over the meaning of the adjustment controversy continues. Compare Kenneth Prewitt, Up for The Count, WASH. POST, Mar. 3, 2001, at A19 (arguing that baseless charges of political manipulation of the 2000 Census have damaged the image of the census as a scientific enterprise), with Peter Skerry, Dressing It Up As Science Can't Disguise the Politics, WASH. POST, Mar. 4, 2001, at B05 (arguing that the census is inherently political and that adjustment of the 2000 data could have introduced substantial error into the census).
(5.) Skerry's history of the census goes all the way back to the Bible with censuses conducted by Moses and King David. PETER SKERRY, COUNTING ON THE CENSUS?: RACE, GROUP IDENTITY, AND THE EVASION OF POLITICS 12-13 (2000).
(6.) See Prepared Statement of Kenneth Prewitt, Director, U.S. Bureau of the Census, Before the Subcomm. on the Census of the House Comm. on Gov't Reform, May 19, 2000, at 2, at http://www.census.gov/dmd/www/051900.html (last visited Oct. 8, 2000) (describing the many steps in the compilation of census data as a search for truth).
(7.) P. 82.
(8.) See Nathaniel Persily, Color by Numbers: Race, Redistricting, and the 2000 Census, 85 MINN. L. REV. 899, 910 (2000).
(9.) See pp. 82-86. Skerry emphasizes how important, for example, the Constitution's three-fifths rule has been in framing the adjustment debate. See p. 5; Richard R. Buery Jr., COP Census Politics, THE NATION, Dec. 7, 1998 at 6, 7. That provision counted slaves as three-fifths of a person for purposes of direct taxation and apportionment of the House of Representatives. U.S. CONST., art. I, [sections] 2, cl. 3.
(10.) Skerry dedicates an entire chapter to the issue of racial and ethnic categories, see pp. 43-79, which I do not review in depth here, but which provides an incredibly rich and, at times, chilling analysis of the remarkable arbitrariness of racial categories. His basic argument is that the census presents a paradox between racial self-identification and the need for authoritative categorization for purposes of enforcing the laws for which the data are used. Skerry shows how unreliable the racial data are by telling horror stories about how enumerators make erroneous or seat of the pants conclusions about the race of respondents, how the census will recode people into different categories (e.g., if a person writes in Jewish they will be recoded as white, or if they write in Hispanic on the race question, the census will impute their race as white or black depending on the racial composition of their neighborhood), and how the question wording and order will produce dramatically different results. The inherent fuzziness of racial data fuels Skerry's argument because the differential undercount then appears subsumed by generalized census errors in measuring race. In other words, if racial data are inherently unreliable, then a 5% undercount becomes overwhelmed by much greater errors in measurement of race. For the definitive history of the census race question, see MARCO ANDERSON, THE AMERICAN CENSUS: A SOCIAL HISTORY 7-158 (1988).
(11.) These net undercount rates obscure deeper inaccuracies in the census because they include both omissions (failures to count people who exist) and erroneous inclusions (counting people who do not exist). The number of erroneous inclusions in 1990 ranged from 4,4 to 10 million and omissions ranged from 9.7 to 15.5 million. See pp. 85-86. See also ESCAP, supra note 4, at 4 (displaying undercount data from 2000 Census).
(12.) See MARCO J. ANDERSON & STEVEN FEINBERG, WHO COUNTS? THE POLITICS OF CENSUS TAKING IN AMERICA 29-30 (2000).
(13.) NATIONAL RESEARCH COUNCIL, MODERNIZING THE U.S. CENSUS 32 (1995), available at http://www.nap.edu/openbooks/0309051827/html/32.html [hereinafter MODERNIZING THE U.S. CENSUS]; pp. 82-83. Skerry also notes that most of the black undercount can be attributed to undercounted black males. Those aged 20 to 64 had an undercount of 11.2%; those aged 30 to 34 had an undercount of 14.0%. Pp. 82-83.
(14.) ESCAP, supra note 4, at 4-5, provided the data for these tables. The Census changed the racial categories for the 2000 Census so two tables are necessary.
(15.) This figure excludes Hispanics who also fell into one of the following categories: Black or African American, Asian and Pacific Islander, or American Indians on Reservations.
(16.) See pp. 86-101
(17.) Pp. 87-91.
(18.) Pp. 91-93.
(19.) Although non-English forms are available, an individual must request them.
(20.) Pp. 93-94.
(21.) Contrary to prevailing wisdom, Skerry points to ethnographic research that suggests that status as an "illegal immigrant" is "relative[ly] unimportan[t]." See p. 97.
(22.) Pp. 94-97.
(23.) P. 100.
(24.) P. 121.
(25.) Seep. 122.
(26.) See LEADERSHIP CONFERENCE EDUCATION FUND, CENSUS 2000: EVERYONE COUNTS [sections][sections] 1.4, 2.1-5.1 (2000) (available at http://www.civilrights.org) [hereinafter LCEF].
(27.) See U.S. CONFERENCE OF MAYORS, THE FISCAL IMPACT OF THE CENSUS UNDERCOUNT ON CITIES (1999) (available at http://www.usmayors.org/uscm/ wash_update/census).
(28.) See U.S. Census Monitoring Board, Presidential Members, Price Waterhouse Coopers Projects $11 Billion Loss in Federal Funds to State and Local Governments if Census 2000 Is Undercounted (Mar. 9, 2000) (available at http://www.cmbp.gov/ v3/report/bystate/pressrelease.shtml).
(29.) See ANDERSON & FEINBERG, supra note 12, at 130-67.
(30.) P. 122
(31.) P. 126.
(32.) P. 126-29.
(33.) Pp. 127-28.
(35.) See pp. 122-27.
(36.) Pp. 126-29.
(37.) See pp. 129-34.
(38.) P. 131.
(39.) 525 U.S. 316, 334 (1999).
(40.) In particular, the Court found section 195 of the Census Act to prevent the use of adjusted numbers for apportionment purposes. Section 195 provides:
Except for the determination of population for purposes of apportionment of Representatives in Congress among the several States, the Secretary shall, if he considers it feasible, authorize the use of the statistical method known as "sampling" in carrying out the provisions of this title.
13 U.S.C. [sections] 195 (1994). As originally passed in 1957 this provision had a small but significant difference: It specified that "the Secretary may, where he deems it appropriate, authorize the use of ... `sampling.'" 13 U.S.C. [sections] 195 (1970) (emphasis added). In his dissent, Justice Stevens thought this change quite significant because it implied congressional intent to make sampling mandatory in all non-apportionment settings but optional (and thus permissible) in the apportionment context. Dep't of Commerce v. United States House of Representatives, 525 U.S. 316, 358-59 (1999) (Stevens, J., dissenting). See generally Persily, supra note 8, at 8-10 (discussing the decision).
(41.) 525 U.S. at 339 (emphasis added).
(42.) Id. at 34.1.
(43.) See Raja Mishra, State races see big money: Parties looking ahead to 2001 redistricting, BOSTON GLOBE, 11/6/2000 (available at http://www.boston.com/news/politics/ campaign2000/news/State_races_see_big_money+.shtm) (quoting Tim Storey, a redistricting specialist for the National Conference of State Legislatures).
(44.) Seep. 134.
(45.) See pp. 136-37.
(46.) I say "in theory" because I am assuming a 100% politically cohesive African American population that turns out to vote at the same rate as the non-African American population.
(47.) Toward the end of the book Skerry provides some alternatives to adjustment that I do not discuss in depth here because they only represent two pages of the entire book. Pp. 197-99. To be fair though and not to accuse him of resignation in the face of the undercount, here are the alternatives he lists (some of which, such as the first two, he admits are infeasible or undesirable and many of which such as 4, 5, 6, and 7 were integrated into the 2000 Census):
1. Requiring people to remain in their house on Census Day;
2. Compiling a national register and requiring people to notify the government when they move;
3. Census holiday;
4. Shortened questionnaire (post card census);
5. Curtailment of INS enforcement in the period during the census headcount (already performed for the 2000 census);
6. Hiring local enumerators who are integrated into the undercounted communities;
7. Making census forms more readily available in public places;
8. Greater reliance on administrative records;
9. Using continuous measurement to supplement or replace a once-a-decade census;
10. Using administrative records (instead of a post-enumeration survey) to adjust the census numbers;
11. Each state could also conduct its own census or Congress could use other than census data for its funding formulas.
(48.) A good sample of articles describing the sampling controversy itself can be found in the December 2000 issue of the journal PS: Political Science & Politics. See Lynn Billard, The Census Count: Who Counts? How Do We Count? When Do We Count, 33 PS: POL. SCI. & POL. 767 (2000); Thomas L. Brunell, Using Statistical Sampling to Estimate the U.S. Population: The Methodological and Political Debate over Census 2000, 33 PS: POL. SCl. & POL. 775 (2000); Margo Anderson & Stephen E. Fienberg, Partisan Politics at Work: Sampling and the 2000 Census, 33 PS: POL. Sci. & POL. 795 (2000); Thomas L. Brunell, Making Sense of the Census: It's Political, 33 PS: POL. SCI. & POE. 801 (2000).
(49.) See Persily, supra note 8, at 4.
(50.) Pp. 22-42.
(51.) See Pp. 22-23; ANDERSON & FEINBERG, supra note 12, at 70-190 (describing extensive legislative and political history of the 1990 Census controversy).
(52.) Pp. 101-02.
(53.) P. 102; Persily, supra note 8, at 8-9.
(54.) P. 103; Persily, supra note 8, at 5.
(55.) Pp. 101-20.
(56.) Pp. 103-04.
(57.) 1,392 poststrata existed for the 1990 census. For example, one poststratum would include Hispanic male renters in their twenties who live in Western cities with a low mailback return rate. In the process of adjusting the census, the Bureau compares the number of poststratum members in the headcount with the number found in the sample. Thus, if 1% of the headcount fits into the category described above, but 2% of the sample has those characteristics, the census assumes that the undercount for that category is 50%. In the process of adjustment, the Bureau then creates an adjustment factor, which is basically a number by which to multiply the headcount number to get the sample number. So for the example above, the adjustment factor would be 2.0 because the sample indicated that twice the number of poststratum members exists in the population than were picked up by the headcount. In the final step of the adjustment process for each census block with the above characteristics (i.e., Western cities with low mailback return rates), the census would double the number of Hispanic male renters in their twenties. P. 105. Skerry does not include the mailback return rate as defining poststrata because the Bureau decided to include that characteristic after the book went to press. See Persily, supra note 8, at 6-8.
(58.) P. 105.
(59.) Pp. 106-07.
(60.) P. 107.
(61.) Pp. 107-11. The nature of this error and the scope of its unrealized impact is a subject of vigorous debate. Compare Margo Anderson & Stephen E. Fienberg, History, Myth Making, attd Statistics: A Short Story about the Reapportionment of Congress and the 1990 Census, 33 PS: POL. SCI. & POL. 783 (2000) (suggesting the accusations of error in the 1990 adjustment process overlook the complex interaction of several decisions), with Thomas L. Brunell, Rejoinder to Anderson and Feinberg, 33 PS: POL. SCI. & POL. 793 (2000) (suggesting that both computer coding and other errors in the adjustment process would have led to disastrous misrepresentations of state population totals).
(62.) P. 108.
(63.) P. 115 ("[A] consensus in favor of adjustment has developed among policy experts working oil census issues."); Persily, supra note 8, at 4; LCEF, supra note 26, at [sections] 15.1.
(64.) Pp. 32-36.
(65). Pp. 118-20.
(66.) P. 119 (referring to Secretary Mosbacher's focus).
(67.) See generally LANI GUINIER, TYRANNY OF THE MAJORITY (1996).
(68.) See text accompanying notes 120-26 infra (discussing the conflict between the requirements of the Voting Rights Act and the Supreme Court's decision in Shaw v. Reno).
(69.) See pp. 148-50, 159.
(70.) P. 1.
(71.) P. 156.
(72.) Pp. 157-59.
(73.) Pp. 159-60.
(74.) See pp. 180-81 (describing the Mexican American Legal Defense and Education Fund); id, at 187 ("[U]nderlying these dynamics is a politics not only more and more moved by symbols and ideas, but also increasingly untethered from material interests.... [T]he cutting edge of American politics today, particularly with regard to social policy, is characterized by vaguely defined interests, interpreted by elites whose ties to the constituencies they seek to represent are often weaker than to third-party funding sources.").
(75.) P 178.
(76.) P. 179.
(77.) P. 190 (emphasis in original).
(78.) It is unclear whether Skerry views public interest organizations as serving any beneficent purpose. He writes, "[t]he virtue here ... is that previously excluded interests and groups are now represented in the political process. The vice is that the process is generally more contentious and less civil." P. 187. I assume Skerry views the Voting Rights Act and Civil Rights Act as important pieces of legislation, but maybe he believes that those were the last real, substantive victories for a civil fights movement that is now irrelevant and memberless. He casts a broad net for this "new politics," I think, to include all the major pieces of legislation of the Great Society and even Nixon-era policies, such as environmental protection and disability rights laws. Census adjustment is just a prism through which to view a harmful, liberal trend in American politics.
(79.) His attack on public interest organizations, and particularly racial interest group organizations such as the NAACP, as organizations without members is unfair even if it could be proven. Groups defined, in part, by their lower socio-economic status do not have the resources (money, education, time, etc.) to create mass organizations with membership lists like the AARP or NRA. A fair criticism might be made that such organizations may have different interests than the constituencies they represent. But that empirical question, which would require some analysis of the disjuncture between elite and mass attitudes on given policies, is one Skerry does not investigate.
(80.) P. 162 (defining "participation meltdown" as "a dramatic decline in cooperation with the census due to the perception that adjustment would obviate the need to fill out census forms").
(81.) See U.S. Census Bureau, Census 2000 Final Response Rates, available at http://rates.census.gov/rates.php3? (visited on Dec. 22, 2000).
(82.) See text accompanying notes 52-64 supra.
(83.) See text accompanying notes 49-64 supra; Persily, supra note 8, at 5.
(84.) See p. 165 (citing RICHARD A. KULKA ET AL., THE POTENTIAL IMPACT OF ADJUSTING OR NOT ADJUSTING THE 1990 CENSUS: EVIDENCE FROM A TELEPHONE REINTERVIEW TO THE SURVEY OF 1990 CENSUS PARTICIPATION 10-11 (1991)).
(85.) See U.S. Census Bureau, Census 2000 Final Response Rates, available at http://rates.census.gov/rates.php?3 (visited on Dec. 22, 2000).
(86.) U.S. Census Bureau, Overview: Partnership Program, at http://www.census.gov/ dmd/www/partner.htm (visited Dec. 21, 2000).
(87.) See generally Census Monitoring Board, Congressional Members, at htpp://www.cmbc.gov (visited Dec. 22, 2000) (compiling Republican proposals against adjustment).
(88.) See U.S. Census Bureau, Overview: Advertising and Promotion, at http://www.census.gov/dmd/www/advoverview.htm (visited Dec. 22, 2000).
(89.) P. 165 (citing RICHARD A. KULKA ET AL., THE POTENTIAL IMPACT OF ADJUSTING OR NOT ADJUSTING THE 1990 CENSUS: EVIDENCE FROM A TELEPHONE REINTERVIEW TO THE SURVEY OF 1990 CENSUS PARTICIPATION 10-11 (1991)).
(90.) See, e.g., LCEF, supra note 26, at [sections] 23.1.
(91.) P. 188.
(92.) P. 189 (quoting Hugh Heclo, Issue Networks and the Executive Establishment, in ANTHONY KING, ED., THE NEW AMERICAN POLITICAL SYSTEM (1978)).
(93.) The literature on political belief systems, information, and attitude formation is vast. See, e.g., WARREN E. MILLER & MERRYL SHANKS, THE NEW AMERICAN VOTER (1996); Phillip E. Converse & Gregory B. Markus, Plus Ca Change ...: The New CPS Election Study Panel, 73 AMER. POL. SCI. REV. 32 (1979); Phillip E. Converse, The Nature of Belief Systems in Mass Publics, in DAVID E. APTER ED., IDEOLOGY AND DISCONTENT 206-61 (1964); Phillip E. Converse, Information Flow and the Stability of Partisan Attitudes, 26 PUB. OP. Q. 578 (1962).
(94.) P. 189.
(95.) Id. ("Cynicism is also fostered by a hyperactive government that inevitably and repeatedly promises more than it can deliver.... When government attempts such ambitious goals, it raises expectations whose (sic) disappointment undermines faith in its ability to do anything.").
(96.) See ANDERSON & FEINBERG, supra note 12, at 205-13.
(97.) See, e.g., Jack Citrin, Comment: The Political Relevance of Trust in Government, 68 AMER. POL. SCI. REV. 973-88 (1974); JOSEPH S. NYE, PHILIP D. ZELIKOW, DAVID C. KING EDS., WHY PEOPLE DON'T TRUST GOVERNMENT (1997); JOHN P. ROBINSON ED., MEASURES OF POLITICAL ATTITUDES (1999); PIPPA NORRIS ED., CRITICAL CITIZENS: GLOBAL SUPPORT FOR DEMOCRATIC GOVERNMENT (1999); Jack Citrin & Donald Green, Presidential Leadership and the Resurgence of Political Trust, 16 BRIT. J. POL. SCI. 431-53 (1986); Jack Citrin & Samantha Luks, Revisiting Cynicism in an Angry Age? (unpublished manuscript, on file with the Stanford Law Review).
(98.) See pp. 188-90 (citing various works by Hugh Heclo and R. Shep Melnick).
(99.) See Pew Research Center for the People and the Press, Deconstructing Distrust: How Americans View Government (1997), available at http://www.peoplepress.org/ trustrpt.htm.
(101.) Id.; Citrin, supra note 97.
(102.) Pew, supra note 99.
(103.) See pp. 115-18; id. at 149 ("numbers by themselves mean little").
(104.) Pp. 141-43.
(105.) See pp. 117-18.
(106.) Pp. 191-92.
(107.) See, e.g., LCEF, supra note 26, at [sections] 23.1.
(108.) Skerry mentions the Voting Rights Act and its enforcement, in passing, as part of a larger argument that constructing districts to maximize the concentrations of racial minorities could hurt both the minorities and Democrats. See p. 136.
(109.) See NATHANIEL PERSILY ED., THE REAL Y2K PROBLEM: CENSUS 2000 DATA AND REDISTRICTING TECHNOLOGY 2 (2000).
(110.) See ESCAP, supra note 4.
(111.) 42 U.S.C. [sections] 1973c (2000).
(112.) In fact, the state of Virginia sued for preclearance of its antisampling law even before the census data were collected. See Virginia v. Reno, 117 F. Supp. 2d 46 (D.D.C. 2000), aff'd, 121 S. Ct. 749 (2001). Because the data had not yet been used in a redistricting plan and thus one could not tell if the choice of data worsened the prospects for African American representation, the U.S. District Court for the District of Columbia held the case to be unripe and the Supreme Court summarily affirmed. See id.
(113.) See 42 U.S.C. [sections] 1973c (2000); Beer v. United States, 425 U.S. 130, 133 (1976); Georgia v. United States, 411 U.S. 526, 531 (1973).
(114.) See Beer, 425 U.S. at 141-42.
(115.) Retrogression is determined by comparing the new plan to the last legally enforceable plan. The new census data are used to determine the racial composition for both plans, however. In other words, DOJ overlays the old districts lines onto the new census data and then compares the minority percentages in the former districts with the percentages that will be created by redrawing the lines according to the new plan. See 28 C.F.R. 51.54(b)(2); City of Rome v. United States, 446 U.S. 156, 186 (1980).
(116.) As stated in the text, DOJ required the drawing of even more majority-minority districts in 1992 than existed in the baseline plan, and in the upcoming round they may adopt a more flexible approach to minority influence that downplays whether the community surmounts the 50% threshold. See Guidance Concerning Redistricting and Retrogression Under Section 5 of the Voting Rights Act, 66 Fed. Reg. 5412 (2001) ("Although comparison of the census population of districts in the benchmark and proposed plans is the important starting point of any retrogression analysis, our review and analysis will be greatly facilitated by inclusion of additional demographic and election data in the submission.... [E]lection history and voting patterns within the jurisdiction, voter registration and turnout information, and other similar information.... is used to compare minority voting strength in the benchmark plan as a whole with minority voting strength in the proposed plan as a whole.").
(117.) I do not want to foster the misimpression that such packing will necessarily survive preclearance even under a "no reduction in majority-minority districts" standard. Other factors, as the preceding footnote explains, will enter into DOJ's decision as to whether a 2002 districting plan causes retrogression in minority voting power.
(118.) See supra note 116.
(119.) 509 U.S. 630 (1993).
(120.) See, e.g., Bush v. Vera, 517 U.S. 952, 976, 981, 982-83, 985-86 (1996) (plurality opinion); Miller v. Johnson, 515 U.S. 900, 905 (1995); United States v. Hays, 515 U.S. 737, 738-39 (1995).
(121.) See PENDA D. HAIR & PAMELA S. KARLAN, REDISTRICTING FOR INCLUSIVE DEMOCRACY: A SURVEY OF THE VOTING RIGHTS LANDSCAPE AND STRATEGIES FOR POST-2000 REDISTRICTING at App. 5 (2000). Hunt v. Cromartie, 120 S. Ct. 2715 (2000) (noting probable jurisdiction and setting oral argument), currently before the Court, may decide the question whether Shaw's test of racial predominance could invalidate a district where the minority community constitutes even less than 50% of the population.
(122.) See Bush v. Vera, 517 U.S. 952, 990 (1996) (O'Connor, J. concurring).
(123.) 42 U.S.C. [sections] 1973 (2000).
(124.) See generally SAMUEL ISSACHAROFF, PAMELA S. KARLAN, & RICHARD H. PILDES, THE LAW OF DEMOCRACY 441-545 (1998).
(126.) See generally Pamela S. Karlan, All Over the Map: The Supreme Court's Voting Rights Trilogy, 1993 SuP. Cz. REV. 2,$5, 250 (describing "cracking," "packing" and "stacking" as strategies to dilute minority voting power); Persily, supra note 8, at 11-12 (same).
(127.) See 478 U.S. 30 (1986).
(128.) Id. at 50-51.
(129.) See HAIR & KARLAN, supra note 121, at App. 4 ("This question--whether minority voters can seek under the Voting Rights Act to create a minority opportunity district with less than 50% minority population--remains open today; the Supreme Court has several times pointed to the issue, but has not yet resolved it."). The authors note that Justice O'Connor's opinion concurring in the judgment in Gingles specifically "express[ed] no view as to whether the ability of a minority group to constitute a majority in a single-member district should constitute a threshold requirement for a claim that the use of multimember districts impairs the ability of minority voters to participate in the political processes and to elect: representatives of their choice." Gingles, 478 U.S. at 89 n.1 (O'Connor, J., concurring).
(130.) See HAIR & KARLAN, supra note 121, at App. 4 ("[I]n cases where the Gingles preconditions are applied rigidly ... the undercount will cause a minority community to fail to meet the threshold that it is large enough and compact enough to constitute a majority in a single-member district. The group may actually meet the requirement, but the census data may not support that claim because of the undercount."). Moreover, the authors note that in Johnson v. DeGrandy, 512 U.S. 997, 1014 & n.11 (1994), the court also pointed out that the number of majority-minority districts in a jurisdiction would be one factor in proving whether a districting plan denied minority voters "equal political opportunity."
(131.) One study has concluded that adjustment of the 1990 census data "would have had the potential to affect minority voter opportunities in more than forty state legislative seats." ALAN LICHTMAN, REPORT ON THE IMPLICATIONS FOR MINORITY VOTER OPPORTUNITIES IF CORRECTED CENSUS DATA HAD BEEN USED FOR THE POST-1990 REDISTRICTING: STATES WITH THE LARGEST NUMERICAL UNDERCOUNT (2001), available at http://www.cmbp.gov/reports/Lichtman/default.asp. The study is unclear in its definition of "minority voter opportunities," so it is difficult to understand why the use of one set of numbers actually affected the political fortunes of the group. Skerry would rightly have a field day with this study because it deftly ignores the problem of contingency central to the argument that we cannot know how lines would have been drawn if adjusted data would have been used. The counterfactuals Lichtman constructs exaggerate the importance for redistricting of the failure to adjust the 1990 data. (Indeed, it is hard to imagine many more majority-minority districts being created during the last round). Nevertheless, he does add some empirical support to the argument I make here that for states covered by Section 5 the use of adjusted data could change the baseline number of majority-minority districts.
(132.) Compare Perez v. Pasadena Indep. Sch. Dist., 165 F.3d 368, 372-74 (5th Cir. 1999) (applying "bright-line" rule for Section 2 claims that would bar claims by any group that constituted less than a majority of the citizen voting age population of a potential single-member district), with Puerto Rican Legal Def. and Educ. Fund, Inc. v. Gantt, 796 F. Supp. 681, 689 (E.D.N.Y. 1992) (three-judge court) (assembling cases applying varying standards for Gingles' first prong).
(133.) Several political scientists have tried to operationalize a more flexible approach to measuring whether minorities have an equal chance to elect their candidates of choice. See, e.g., Bernard Grofman, Lisa Handley & David Lublin, Drawing Effective Minority Districts: A Conceptual Framework and Some Empirical Evidence, 79 N.C.L. REV. (forthcoming June 2001) (manuscript on file with the North Carolina Law Review). The authors also point to language in cases questioning whether "majority-minority" really means over 50% minority.
(134.) The Department of Justice was accused of enforcing a 65% rule for Section 5 preclearance in the 1990 round of redistricting. A raw majority was insufficient to place minorities on an equal playing field, the argument went, because low rates of voter turnout ensured that the minority group would still fail to elect its candidate of choice. See ISSACHAROFF ET AL., supra note 124, at 473; cf. United Jewish Org. of Williamsburgh v. Carey, 430 U.S. 144, 164, 177 (1977); Barnett v. City of Chicago, 141 F.3d 699, 702-03 (7th Cir, 1998) ("[I]t is a rule of thumb that blacks must be at least 65 percent of the total population of a district in order to be able to elect a black.... Likewise, because of both age and the percentage of noncitizens, Latinos must be 65 to 70 percent of the total population in order to be confident of electing a Latino.").
(135.) Grofman et al., supra note 133, point out, in particular, that even when a rigid numerical threshold is applied the courts disagree as to what such a number should measure. Does a majority-minority district, for example, mean a majority of the population, a majority of the voting age population, or a majority of the citizen voting age population?
(136.) 509 U.S. 630 (1993). The Shaw line of cases includes Hunt v. Cromartie, 526 U.S. 541 (1999); Abrams v. Johnson, 521 U.S. 74 (1997); Shaw v. Hunt, 517 U.S. 899 (1996); Bush v. Vera, 517 U.S. 952 (1996), Miller v. Johnson, 515 U.S. 900 (1995); United States v Hays, 515 U.S. 737 (1995).
(137.) These cases, upholding equal protection claims that rely on "expressive harms" caused by districts where race predominates, run directly counter to the guarantees of the Voting Rights Act that require states to focus on race to ensure that they are not retrogressing or diluting minority votes. See NATHANIEL PERSILY, THE REAL Y2K PROBLEM: CENSUS 2000 DATA AND REDISTRICTING TECHNOLOGY i (2000) ("[S]tates find themselves caught between the rock of the Voting Rights Act, which requires them to take race into account in their redistricting plans to ensure that minorities have an equal opportunity to elect their candidates of choice, and the hard (incomprehensible) place of the Supreme Court's counterposing set of decisions beginning with Shaw v. Reno, which order states not to allow race to become the `predominant factor' in a redistricting plan.").
(138.) See Reno v. Bossier Parish, 520 U.S. 471 (1997); Reno v. Bossier Parish, 120 S. Ct. 866 (2000).
(139.) See Wesberry v. Sanders, 376 U.S. 1 (1964). I have not discussed here, but discuss at length elsewhere, see Persily, supra note 8, at 13-14, the most likely context in which the sampling issue will arise: challenges alleging violation of the one-person, one-vote rule. Both sides of the adjustment controversy might make such claims because they will argue that using the data they oppose will create districts of wildly varying population ranges.
(140.) 121 S. Ct. 525 (2000).
Nathaniel Persily, Staff Attorney, Brennan Center for Justice at NYU School of Law; Adjunct Professor, Columbia Law School. Thanks go to Elizabeth Daniel, Sam Hirsch, Praveen Krishna, and Luke McLoughlin for comments on earlier drafts, and to Kirby Faciane and David Weiner for their editing and patience with a review that needed frequent updating to keep pace with current events.
([dagger]) Professor of Political Science, Claremont McKenna College; Nonresident Senior Fellow, Brookings Institution.
|Printer friendly Cite/link Email Feedback|
|Publication:||Stanford Law Review|
|Article Type:||Book Review|
|Date:||Apr 1, 2001|
|Previous Article:||Of kitsch and kachinas: a critical analysis of the Indian Arts and Crafts Act of 1990.|
|Next Article:||Foreword: does the Solicitor General matter?|