Following the 2000 Census, Georgia redrew its fifty-six state senate districts to comply with the one-person-one-vote rule. At the time, the assembly and senate both had Democratic majorities. The governor, Roy Barnes, was a Democrat as well, and led the charge to construct a districting plan that could stem the expected Republican tide in the upcoming 2002 elections. The key to this plan was to "unpack" many of the heavily Democratic districts and distribute loyal Democratic voters to surrounding districts. In particular, black voters were reallocated away from districts with either especially high or low levels of black voting age population (BVAP) to create more districts in the 25 to 40 percent BVAP range, so-called influence districts. This meant that some districts with black populations above 55 percent or 60 percent were brought down close to the 50 percent mark. This strategy is illustrated graphically in figure 4.1. To construct this graph, the BVAPs in each district for both the existing baseline plan and the proposed plan were sorted from highest to lowest, and the difference between corresponding entries was calculated. The figure clearly shows the reallocation of black voters away from the extremes of the distribution and toward the center. The state submitted its plan directly to the D.C. District Court for preclear-ance, and the Justice Department (DOJ) indicated its intention to interpose objections to Senate districts 2, 12, and 26, whose BVAPs were slated to fall from 60.6 percent, 55.4 percent, and 62.5 percent to 50.3 percent, 50.7 percent, and 50.8 percent, respectively. The state submitted evidence showing that the point of equal opportunity - the level of BVAP at which a minority-preferred candidate has a 50 percent probability of winning - was 44.3 percent, and argued that the redrawn districts should still offer black candidates a healthy chance of gaining office. The DOJ disagreed, arguing that the state had not met its burden of proving the proposed plan non-retrogressive. The district court agreed with the DOJ and refused to preclear the plan. The Supreme Court eventually overruled in the case Georgia v. Ashcroft (539 U.S. 461), ruling that the district court had not taken sufficiently into account the state's avowed objective of increasing substantive representation, even at a possible cost (Figure presented) to descriptive representation. The Court relied heavily on the testimony of black state legislators, including civil rights leader and U.S. Representative John Lewis, who supported the plan as an attempt to expand Democratic control of state government.1 The reaction to Ashcroft was swift and heated. Karlan (2004) denounced the decision as a first step towards "gutting" section 5 preclearance. Others claimed that it "greatly weakened the enforcement provisions of Section 5" (Benson 2004). An ACLU official's reaction was that "The danger⋯ is that it may allow states to turn black and other minority voters into second-class voters, who can influence the election of white candidates but cannot elect candidates of their own race" (House Committee on the Judiciary 2005, 6). Others viewed the decision more favorably. Henry Louis Gates, for example, wrote in the New York Times that descriptive representation "came at the cost of substantive representation - the likelihood that lawmakers, taken as a whole, would represent the group's substantive interests. Blacks were winning battles but losing the war as conservative Republicans beat white moderate Democrats" (September 23, 2004, A23). At the heart of the latter set of arguments lies the notion of a tradeoff between substantive and descriptive representation: getting more of one necessarily entails accepting less of the other. This topic has been the subject of previous research: Morgan Kousser (1993) discusses the possibility of such a tradeoff in the context of influence districts, as does Richard Pildes (1995) in his review of social science research on the changing political landscape in the South. More quantitative studies of the tradeoff include Charles Cameron, David Epstein, and Sharyn O'Halloran (1996) and David Lublin (1997), both of which studies detected the possibility of an emerging tradeoff in their analyses of election and roll call data through the early 1990s. One of our goals is to expand and update this work, studying patterns of minority electability and congressional voting on minority-supported legislation for the years 1974 to 2000, and the districting plans that maximize substantive and descriptive representation. We also, for the first time, include Hispanic as well as black voters in these calculations, both to see if the patterns emerging for Hispanic representation differ significantly from black representation and to examine the degree of cross-minority support between blacks and Hispanics in elections and in the legislature. Our results show that up until roughly the mid-1980s few nonminority legislators in the South championed black causes in Congress. As a result, districting schemes that maximized black descriptive representation were identical to those that maximized substantive representation. Over the past two decades, though, these objectives diverged, so that recent increases in the number of blacks elected to Congress have come at the cost of substantive representation. For Latinos, the story is rather different. It has always been relatively difficult to elect Hispanic representatives, quite possibly due to the gap between total voting age population and citizen voting age population in concentrated Latino districts. On the other hand, white Democratic representatives have been generally supportive of Latino positions on roll call votes, much more so than Republicans. Thus there has always been a tradeoff between Latino descriptive and substantive representation, with the dimensions of this tradeoff changing little over time. The situation is much the same for blacks outside the South: supportive votes from nonminority Democrats have led to tradeoffs between the two types of representation throughout the period. Finally, we find interesting relationships between issues of black and Latino representation. There remains an almost perfect electoral separation: districts with more black than Hispanic voters rarely elect a Hispanic representative and vice-versa. Further, adding Hispanic voters to a heavily black district increases the probability that a black representative is elected, but adding blacks to a Hispanic district does little to help Hispanic candidates win office.