Table of Contents Introduction I. Prongs and Puzzles A. The Gingles Framework B. Unanswered Questions II. Racial Segregation A. Hypotheses B. Trends III. Racial Polarization A. Hypotheses B. Trends C. Drivers IV. Descriptive Representation A. Hypotheses B. Trends C. Drivers V. Substantive Representation A. Hypotheses B. T rends C. Drivers VI. Implications A. Positive B. Negative C. Extensions Conclusion Appendix Introduction
Senator Orrin Hatch led the opposition to the 1982 amendments that transformed the Voting Rights Act--and with it, minority representation in America. (1) The amendments converted what had been a conventional discriminatory intent provision into a far-reaching "results test." (2) Any practice that "results in a denial or abridgement of the right ... to vote on account of race" became unlawful, regardless of the practice's motivation. (3) Throughout the congressional debate, Hatch hammered a single point. If the results test was not meant to require proportional representation for minority groups (as its backers pledged (4)), then the test had no "ultimate core value." (5) It "provide[d] absolutely no intelligible guidance to courts in determining whether or not a ... violation ha[d] been established." (6) It was an empty shell.
The amendments' supporters were unable to counter Hatch's criticism. They could not identify an "ultimate core value" (other than proportional representation) underlying the results test. Instead, they resorted to invocations of precedent, claiming it showed that the test could be fairly applied. As the Senate Report put it, "There is ... an extensive, reliable and reassuring track record of court decisions using the very standard which the Committee bill would codify." (7) In other words, the supporters could not explain how their proposal would operate--but they were confident the courts had already figured it out.
In fact, the courts had done nothing of the kind. The pre-1982 case law on racial vote dilution (the reduction of minorities' electoral influence through means other than outright disenfranchisement) was a mess. (8) It featured a dozen or so factors that judges balanced as they saw fit, weighing each element based on their own discretionary judgment. (9) It offered no "intelligible guidance" except to consider the totality of circumstances.
In the face of this confusion, it fell to the Supreme Court to fashion the results test into a more determinate inquiry. The Court famously did so in the 1986 case of Thornburg v. Gingles, its first encounter with the revised statute. (10) First, the Court held that the law aimed to provide descriptive representation to minority voters--or more precisely, representation by minority voters' candidates of choice. "The essence of a [Voting Rights Act] claim," the Court declared, "is that a certain electoral ... practice ... interacts with social and historical conditions" to prevent minority voters from being able "to elect their preferred representatives." (11)
Second, and even more crucially, the Court clarified how much representation minority voters were due. Not maximal representation: the most an electoral system could possibly deliver to them. And not proportional representation either: a share of seats equivalent to a minority's share of the population. Instead, under the Court's new framework, a minority group was entitled to elect its preferred candidates only if it met a series of preconditions. It had to be "sufficiently large and geographically compact" to constitute a local majority. (12) It had to be "politically cohesive" in its voting preferences. (13) And it had to be confronted by consistent "bloc" voting by the "white majority." (14)
The Court's answer to Hatch, then, was this: The results test is neither a mandate for proportional representation nor a blank slate. Rather, it requires for minority groups the level of representation that corresponds to their size, segregation, and polarization. Groups that are geographically compact (that is, segregated) and different from the white majority in their voting preferences (that is, polarized) must be able to elect the candidates of their choice. But groups that are spatially integrated or electorally indistinct have no such entitlement.
This answer, it is true, supplies the "ultimate core value" sought by Hatch. (15) But it raises a host of vexing questions of its own. Some of these questions are normative, and legions of scholars have strived diligently to address them. (16) Some of the questions, though, are empirical, and as to them the academy has been remiss. Almost three decades after Gingles was decided, not enough is known about the phenomena the case recognized or the relationships between them. An entire doctrinal edifice has been erected on an uncertain factual foundation.
To start, take the two key determinants of minority representation under the Court's approach: racial segregation and racial polarization in voting. A large sociological literature has found that black-white segregation is falling at the metropolitan level. (17) But what is happening to it (and to Hispanic-white segregation) at the level that matters even more for minority clout: the level of the state as a whole? Similarly, several political science studies have determined that black-white polarization declined modestly in the 1990s. (18) But what were its trends (and those of Hispanic-white polarization) before and after this decade? And is the Court right to think that desegregation might fuel depolarization--that we might be progressing toward "a society where integration and color-blindness are ... simple facts of life"? (19)
Next consider Gingles's overarching goal: the election (if its preconditions are satisfied) of minority voters' preferred candidates. The number of black and Hispanic members of Congress surged in the 1990s, the first redistricting cycle after the enactment of the 1982 amendments. (20) But what about the presence of minority politicians in the state legislative chambers that are the building blocks of American democracy? Did it increase as well, and if so, were these gains sustained in the wake of the Court's racial gerrymandering decisions, which some feared would decimate minority representation? (21)
Still more interestingly, Gingles connected the election of minorities' candidates of choice to segregation and polarization in ways the phenomena had not previously been tied. Did this linkage make a difference? That is, did the relationship between segregation and polarization on the one hand, and minority representation on the other, change as a result of Gingles? And if it did, could the relationship be evolving once again as (according to the Court) "integration and color-blindness" increasingly become "facts of life"? (22) Put more bluntly, could desegregation or depolarization now be leading to the election of fewer minority-preferred candidates?
Lastly, while Gingles stressed descriptive representation, it also evinced concern for substantive representation: legislatures that, as bodies, promote minorities' policy interests. Under the decision, "a significant lack of responsiveness on the part of elected officials to the particularized needs of the members of the minority group" is a factor that cuts in favor of liability (23) At the federal level, it is reasonably clear that a tradeoff exists between descriptive and substantive representation, at least for blacks. When more blacks are elected to Congress, fewer Democrats win seats, and the chamber's median moves in a conservative direction. (24) But does this tradeoff apply at the state legislative level too, and for all minority groups, not just blacks? And if so, is the tradeoff unavoidable or can it be mitigated--for instance by Democratic rather than Republican control of redistricting?
There is a reason why these questions have not yet been answered. It is that the information necessary to grapple with them has been absent. To date, no datasets have been compiled of segregation or polarization by state and over time. Even longitudinal estimates of descriptive representation and party vote share have not been produced at the state legislative level. This lack of evidence explains why basic doubts about Gingles--and its "ultimate core value" for the results test--persist a generation after the case was decided.
In this Article, I exploit a series of original datasets to tackle these issues. As to segregation, I used information on the racial makeup and geographic location of all census tracts over a five-decade span to calculate what is known as the spatial index of dissimilarity. (25) This is the first time that spatial segregation scores have been computed for states. As to polarization, I relied on the results of all available general election exit polls, including more than 1.2 million respondents, to determine racial differences in vote choice and political ideology. (26) Whenever state-specific polls were not conducted, I employed a new statistical technique to derive state-level estimates from the national polling data. (27)
As to descriptive representation, I consulted a range of sources to track the number of black and Hispanic state house members by state and year. (28) Congress itself collects this information at the federal
level, (29) but its data-gathering effort has no state-level analogue. And as to substantive representation, I calculated the major parties' seat and vote shares in state house elections in earlier work. (30) In a recent project, a team of political scientists also generated ideology scores for state legislators on the basis of their roll call votes. (31)
As should be clear by now, my analysis proceeds at the state house rather than at the congressional level. There are fifty state houses (32) compared to a single House of Representatives, and more than five thousand state house districts compared to 435 congressional ones. So state houses...