Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>



									THE POLITICAL CONSEQUENCES OF AN UNCONTROLLED REDISTRICTING Ronald Keith Gaddie Department of Political Science The University of Oklahoma Norman, OK 73019 Charles S. Bullock, III Department of Political Science The University of Georgia Athens, GA 30602

Prepared for presentation at the annual meeting of the Southern Political Science Association, New Orleans, LA, January 6-8 2005. Direct all correspondence to either of the authors at the addresses above.

THE POLITICAL CONSEQUENCES OF AN UNCONTROLLED REDISTRICTING Abstract Redistricting is the most nakedly partisan activity in American politics. Political science research and the courts generally acknowledge that redistricting undertaken in a partisan environment where one party completely controls the legislative process results in a perpetuation of the majority at the cost of the minority. The 2001 Georgia redistricting did precisely that, perpetuating a legislative majority that was a popular minority. The rejection of those maps by the federal courts in 2004 led to a politically-uncontrolled redistricting of the state legislature. In this paper, we examine the effect of changing legislative boundaries by court order on the political outcomes of the Georgia House of Representatives in 2004. By examining the changing partisan structure of the districts, the altered racial structure, and the treatment and placement of incumbents, we develop a multivariate analysis of the changing structure of seat possession in the Georgia House of Representatives. Our analysis indicates that there are significant political effects due to the unexpected redistricting. These changes were sufficient to change the control of the chamber, based on the assumption that the same relationships between predictor variables remained constant from 2002 to 2004. We also find, however, that there are changes in the relationship between predictor variables and the political outcomes, indicating that the remap alone does not explain the magnitude of political gains by the Georgia Republicans.

THE POLITICAL CONSEQUENCES OF AN UNCONTROLLED REDISTRICTING Redistricting is the most nakedly partisan activity in American politics. The decennial activity of allocating political power results in conflict among regional, partisan, racial, and ethnic communities of interest (c.f. Lublin, 1999; Rush, 2000; but see also Hill, 1995; and Overby and Cosgrove, 1996). Political science research and the courts generally acknowledge that redistricting undertaken in a partisan environment where one party completely controls the legislative process results in a perpetuation of the majority at the cost of the minority (Abramowitz, 1983; Cain, 1984; Butler and Cain, 1992; but see Niemi and Abramowitz, 1994). Tendencies toward political excess are usually deterred only by divided government, which forces bipartisan cooperation, or by the presence of a judicial lever to check the exacting of the will of the majority at some costs to the minority. The Georgia redistricting of 2001 was one of the most stark exercises of power by a political majority to perpetuate itself. Georgia Democrats had ceased to gain a majority of state legislative votes in the late 1990s, yet continued to gain a majority of seats in both chambers. The Democratic governor and legislature acted together to perpetuate the Democratic majority through a highly creative legislative redistricting. The redistricting led to two judicial challenges, two trips to the US Supreme Court, a redefinition of the preclearance provision of the Voting Rights Act (Georgia v. Ashcroft), and, ultimately, the overturn of the districts on the grounds of violation of the one-person, one-vote principle (Larios v. Cox; see also Bullock, 2004).1 A federal three-judge panel implemented a replacement map crafted by a special master.

This was the second straight decade that Georgia legislative districts were overturned by the

federal courts due to legal defect. The state legislative and congressional districts were rejected due to racial gerrymandering in 1995 (Abrams v. Johnson). 1

This map largely ignored political factors in deference to traditional redistricting principles and producing a map with de minimus population deviations of less than +/-1%. This politicallyuncontrolled redistricting was immediately followed by the loss of control of the Georgia House of Representatives by the Democratic Party for the first time since Reconstruction. In this paper, we describe and explore the political effects of the 2004 remap on legislative elections for the lower chamber of the Georgia General Assembly. The analysis indicates that significant political gains were made by Republicans as a consequence of the court-implemented, revised map. When statistical relationships in evidence in the 2002 legislative elections are applied to the demographic and structural changes in the new districts, a Republican majority is indicated under the new map, and a shift in political advantage of 10 to 13 districts is indicated.

WHAT ARE “FAIR” LEGISLATIVE MAPS? The controversies arising in redistricting are almost countless, but mainly relate to a primary question: what are the motives of the map-maker, and how do these motives affect the “fairness” of a map? This question is difficult to address, because the notion of fairness is arbitrary and relative. The term “gerrymander” means to craft legislative boundaries for political advantage. But, when presented to the public or to political scientists, a gerrymander brings to mind contorted, oddly-shaped districts resembling mythical beasts, windshield-splatted bugs, or elongated barbells. Districts of conventional geometric shape (circles, squares, hexagons, rectangles) are less subject to question when viewed on a map, but from a political fairness perspective, these same principles can be used to craft political bias into a map (see, for example,


Niemi, Grofman, Carlucci, and Hofeller, 1991; and also Altman, 1998). The Georgia redistricting of 2001 raised all of these questions, as legislative districts became less-compact, less contiguous, less equal by population, less respective of political subdivisions, and were objectively biased on the basis of party.

Continuity of Representation The treatment of political actors is also an area of concern when assessing fairness. Specifically, the treatment of incumbents and their constituencies is the area of concern. This fairness principle relates not just to incumbency, but explicitly to the question of political or partisan fairness (addressed further below). The treatment of incumbents usually focuses on three aspects of the alteration of their districts: (1) Continuity of representation: what proportion of the incumbent’s new constituency is from their old constituency (core retention)? (2) Political balance and continuity of the reelection constituency: how does the partisanship of the new district compare to the old district? (3) Pairings: are incumbents paired so as to have to run against each other? Are the pairings competitive? Are they party-neutral or do the pairing advantage one party over the other? The law does not recognize an incumbent’s property right to a seat in the legislature. But, in the assessment of the fairness or unfairness of maps, biases in the treatment of incumbents by region or party are important. Biases in incumbent treatment relate to evidence of a general partisan bias in map design. When changes in party competitiveness, core retention, and incumbent pairing


fall disproportionately and detrimentally on incumbents of one party, and are not a product of the pursuit of population equality or racial fairness or even other traditional redistricting principles this can constitute evidence of partisan gerrymandering.

Partisan Fairness Of all the fairness concerns in redistricting, none is more frustrating than partisan fairness. The use of representative political systems presumes that preferences will be efficiently translated into government, and, more specifically, that majority preferences will translate into majority government. The earliest successful challenges to malapportioned legislatures were in one-party states such as Tennessee and Georgia. While the motivation for these suits was not partisan, the notion that a system of fair representation did not disfranchise the majority to benefit a geographic minority entered American constitutional law (Reynolds v. Sims; Wesberry v. Sanders). The political notion that derives from the constitutional notion of one-person, onevote – that representative systems should be politically responsive to changes in popular preference – leads to the intellectual consideration of the efficiency or bias present in translating party votes into party seats. Partisan fairness has encountered difficulties in the American courts. The current Supreme Court majority holds that party gerrymanders are non-justiciable, and further than there is no reliable, objective way to measure gerrymanders.2 This follows because voter preferences,

The United States Supreme Court recently directed a federal three judge federal panel in Texas

to reconsider its decision regarding the 2003 Texas redistricting in light of the recent decision regarding partisan gerrymanders in Veith v. Jubelirer. For this direction to have been given to the lower court requires that at least one member of the concurring majority in the Veith decision 4

unlike race or gender, are not as stable and therefore more difficult to measure objectively in terms of bias. Even the previous legal regime, under Bandamer, held that gerrymanders were illegal only to the extent they precluded all hope of success and all political input into the political process, a standard so impossibly high that no redistricting product was ever invalidated by it. Partisan fairness has application in the assessment by courts of their own products. When reconciling legal defects in maps, courts have taken pains to avoid needlessly advantaging one political party, i.e. to not make court-based gerrymanders. These same courts have asserted that, because their job is to remedy legal defects, rather than correcting political defects, no proactive effort to undo political bias in previously legal maps is undertaken. Rather, the maps that are produced are typically deemed “fair” in terms of partisanship, based on the available measures (see, recently, Balderas v. Perry).3

Population Deviations Courts have found, in the Constitution’s Article I and the Equal Protection Clause of the Fourteenth Amendment, bases for requiring equipopulous districts in all collegial bodies save the US Senate. The former finds force through a series of court cases in the 1960s and early 1970s had to switch to the minority to create a majority requesting further review of the Texas map as a partisan gerrymander.

Indeed, partisan unfairness is recognized as a reason for crafting constituencies that might

otherwise be seen as illegal racial gerrymanders. The Cromartie decision recognized that, when partisanship is as good an explanation as race for the shape of a legislative district, a violation of the Voting Rights Act did not exist. 5

beginning with Wesberry v. Sanders. Karcher v. Daggett set the standard for population variations in congressional districts when it held that “there are no de minimus population variations which could practicably be avoided but which nonetheless meet the standard of Article I, Section 2 without justification.” State legislative districts that have a population range of less than ten percent are presumptively legal. In the recent Georgia experience, the obligation of the state relative to these ranges was the subject of some confusion, and that confusion directly affected the final legality of the Georgia legislative maps.

Racial Fairness While equal population is the foremost concern in redistricting, the second most important factor demands racial fairness. Blatant efforts to discriminate against minorities are unacceptable. Section 5 of the Voting Rights Act of 1965 (VRA) requires that changes in election laws from jurisdictions covered by the trigger mechanism must be precleared by either the attorney general or the District Court of the District of Columbia. As initially drafted the trigger mechanism brought coverage to Alabama, Georgia, Louisiana, Mississippi, South Carolina, Virginia and about half of North Carolina. Subsequent revisions of the statute extended coverage to Alaska, Arizona, Texas, and parts of Florida and parts of five other states. Redistricting plans from the covered jurisdictions require preclearance. Early on the US Department of Justice (DOJ) established that it would not approve plans that had a retrogressive effect on minority political influence. This meant that plans that broke up an existing concentration of minorities would be rejected. Section 5 dealt only with changes. It did not provide a mechanism for attacking


discriminatory practices in place before 1964. The most recent revision of the VRA in 1982 rewrote Section 2 to eliminate the need to prove an intent to discriminate when challenging existing electoral practices. Successful claims required only a showing that the practice had a discriminatory effect. Going into the post-1990 round of redistricting, DOJ incorporated Section 2 considerations into Section 5 reviews. This allowed DOJ to reject plans that failed to draw additional majority-minority districts when that would have been possible. Georgia had to submit three sets of plans for its congressional and legislative districts before securing DOJ approval. To get DOJ approval, Georgia had to increase its number of majority-black congressional districts from one to three (Holmes 1998). DOJ actions also prompted North Carolina (Sellers et al. 1998) and Louisiana (Engstrom and Kirksey 1998) to submit second sets of congressional plans that added another majority-black district.4 DOJ’s max-black strategy, so named because it intended to maximize black representation, came under attack almost immediately. In 1993 in Shaw v. Reno the Supreme Court found a challenge to North Carolina’s often-lampooned 12th congressional district – a twisted spaghetti strand that extended from Durham via Greensboro, High Point, Winston-Salem, and Charlotte to Gastonia – to be justiciable. In Miller v. Johnson – which attacked a Georgia congressional district that stretched from the Atlanta suburbs more than 200 miles to coastal Savannah – the Supreme Court found that the Equal Protection clause precludes plans drawn

Because of DOJ demands in these states, legislators and judges elsewhere concluded that they

had best increase minority districts in order to expedite the approval process. For example, the federal judges who crafted Florida’s congressional map added three majority-black districts along with a second Hispanic district. The map was subsequently found unconstitutional. 7

when racial considerations are predominant. Later the court clarified that race was an appropriate consideration when considered in concert with other factors such as partisan goals, compactness, and existing political boundaries (Easley v. Cromartie 1999). Toward the end of the 1990s, the Supreme Court found that DOJ had overstepped its authority when applying Section 2’s effects to redistricting plans submitted for approval (Bossier Parish). It banned DOJ from that requiring states making submissions under Section 5 meet Section 2 standards. In the decision overturning Georgia’s illegal legislative districts in 1995, the federal court noted the historic use of traditional redistricting principles such as county integrity, continuity of representation, compactness, and contiguity (Abrams v. Johnson; see also Shaw v. Reno, Bush v. Vera). These principles were then applied by the court in shaping new maps for Georgia.

GEORGIA REDISTRICTING 2001: DEMOCRATS’ LAST STAND Georgia Democrats entered the 2001 redistricting facing unprecedented challenges. For the first time since the immediate aftermath of the Civil War, Democrats faced the possibility of losing control of the legislature. For the better part of a decade, Democratic support among white voters had eroded, changing Georgia from a state completely dominated by Democrats to a competitive state. The most dramatic change of the 1990s saw the Republican share of the congressional delegation explode from one to eight of eleven. Also during the decade, Republicans scored their first victories in statewide constitutional offices in more than a decade. Democrats retained majorities in both chambers of the state legislature, though that success was in part the product of the districting plan in place. While their efforts to win control


of the state legislature continued to come up short, GOP candidates consistently won majorities of state legislative votes cast statewide. In 2000, Republican candidates polled 55% of the vote statewide in Senate elections and 52% of the statewide House vote. Two years earlier Republican candidates took 51% of the statewide Senate vote and 53% of the House vote. Despite these levels of success, Republicans never exceeded 43% of the seats in either legislative chamber. Population changes during the 1990s compounded the challenges confronting the Democrats. The areas having the most rapid growth tended to vote Republican while areas with stagnant growth or population declines tended to be Democratic (see Bensen 2003; Gaddie 2003). Consequently, if the legislature addressed the need to equalize population among districts in a straight-forward manner and shifted seats from underpopulated to overpopulated parts of Georgia, Republicans would surely make gains. Past governors had taken a hands off approach to redistricting, leaving the task to the legislature. Governor Roy Barnes and his chief of staff Bobby Kahn broke with tradition and became central players in the 2001 map making. In previous decades much of the work of putting together maps, the numerous iterations needed to tweak maps so as to accommodate the concerns of powerful legislators, had taken place in the Legislative Reapportionment Office. In 2001 Senate maps were drawn under the watchful eye of an out-of-state consultant who reported to Governor Barnes. Democratic legislators were shown how the map treated their district. If they objected, they were likely to be warned that they had best sign on to the Governor’s map because if they balked, they would be given something even less palatable. The House, where Speaker Tom Murphy often clashed with governors during his quarter


century leading the chamber, insisted on making some changes to the Governor’s map for that chamber. Murphy agreed to modifications urged by some of his closest associates despite warnings from the Governor’s Office that the changes put Democratic control in jeopardy. As the minority party, Republicans had no input into the maps, but unlike the past, legislative Democrats also had minimal input. Governor Barnes, who expected to win a second term in 2002, did not want to have to learn how to function in a divided government environment and sought to have maps produced that would insure continued Democratic control of the legislature. Democrats went to tremendous lengths to maximize the impact of the dwindling Democratic House vote. To maximize the influence of Democratic votes, multi-member districts (MMD) were reinstated after having been eliminated in 1992. In the new map, MMDs accounted for just over one-third of members. Local Republican concentrations that might have elected a Republican in a single-member district (SMD) were swamped by larger, Democratic majorities in adjoining areas. For example, Henry County, one of the nation’s fastest growing counties during the 1990s, had a Republican legislator who lived in the county. The new map placed the Republican in a three-person district dominated by Democratic voters in south DeKalb County. Once the Republican understood the impossible situation into which he had been placed, he pulled out of his reelection bid. A four-person district was drawn to protect an Atlanta legislator who had switched from Republican to Democrat. Had she run for reelection in her old SMD, she might well have been turned out by the voters angry about her defection. The Republican districts were generally overwhelmingly Republican, and were routinely overpopulated by 4-5% while Democratic districts were frequently underpopulated by 4-5% (see


Figures 1 and 2). As indicated in Figure 1, 37 of 109 districts (34.6%) won by the Democrats in 2002 had popualtion deviations of at least -4%, while only 30 of the districts won by Democrats (27.5%) had populations above the ideal and only a dozen (11.0%) were+4% above the ideal population. By way of contrast, Republican legislators elected 37 of their 70 members (52.8%) in at least +4% from the ideal (Figure 2). Over three-quarters of the most populous districts elected Republicans, while only one-in-ten of the least populous districts elected Democrats. (Figure 2 here) Odd shapes, numerous split counties, and strategic allocations of black voters to maximize the Democratic strength characterized the legislative maps (see Bensen 2003). Half of the Republican caucus found itself paired up, either with other incumbent Republicans or against Democrats in heavily-Democratic districts.5 One of the few Republicans in DeKalb County, Georgia’s second most populous, found himself in a two-person district dominated by Democrats. Only nine incumbent Democrats shared districts with other incumbents (Table 1). (Table 1 here) Others found themselves running in profoundly unfriendly SMDs. For example, the youngest legislator, Danae Roberts who defeated an incumbent to win a seat less than four months after graduating from the University of Georgia, saw her district dismembered. She was the only Republican in the Columbus delegation and since that area’s slow growth cost it a seat,

Under Georgia law, a candidate must reside in a district for one year prior to filing in order to

run in the district. Because Georgia redistricting traditionally takes place in the summer before the April filing date, legislators who are displaced from their constituents or paired against other incumbents must “run where they stand” in order to return to the legislature. At least one-in-four incumbent Republicans was de facto denied reelection by the Democratic redistricting. 11

her Democratic colleagues picked her to be sacrificed. An additional technique used in the state legislative, but not the congressional, maps reduced the size of the black majorities in some districts in order to have reliable Democrats who could be redistributed to tilt nearby marginal districts toward the Democrats. Democrats sought to squeeze out a handful of additional seats by pushing the notion of equal population districts to the limit. Half of the House districts had populations that deviated by at least + 4% from the ideal population of 45,980. A third of the districts had population deviations of + 4.5% and 20 seats were in districts where the population was + 4.9%. In subsequent litigation, a federal court concluded that the other major cause of the deviations in both plans was an intentional effort to allow incumbent Democrats to maintain or increase their delegation, primarily by systematically underpopulating the districts held by incumbent Democrats, by overpopulating those of Republicans, and by deliberately pairing numerous Republican incumbents against one another. (Larios v. Cox). Traditional redistricting principles were generally rejected by the Georgia Democrats in the 2001 redistricting, in deference to the efficient use of minority and Democratic voters and the creative use of population deviations to perpetuate their hold on the legislative majority. This last element of the strategy, which was seemingly watertight in the opinion of counsel to the state of Georgia, would prove to be the undoing of the Georgia partisan gerrymander.

THE FIRST CHALLENGE: ASHCROFT Despite some Democrats’ unhappiness with the districts handed down to them by the governor,


the party shoved the maps through over Republicans’ futile objections. The primary selling point was that the careful analysis of past voting patterns indicated that these maps would continue to keep Republicans at bay (a matter we will address below). Georgia is subject to Section 5 of the Voting Rights Act. Therefore, federal approval of redistricting plans is required before implementation. Rather than sending the map to the US attorney general for review, Georgia opted to go to the District Court of the District of Columbia and seek a declaratory judgment that the maps did not discriminate against minorities. The state feared that the Bush Department of Justice might react negatively to their handiwork. Georgia’s effort to avoid having to secure approval of the maps from the Bush Justice Department could not be wholly successful since when a jurisdiction turns to the district court of the District of Columbia, DOJ acts as the defendant. To the disappointment of Republicans, DOJ raised no objections to either the congressional or state House maps. DOJ did, however, find the reduction in the concentration of African-Americans in three Senate districts unsettling (Georgia v. Ashcroft, 2002). In these districts the black percentage in the voting age population (VAP) dropped to less than 51%. Before being redrawn the black voting age population in two of the districts exceeded 60% and stood at 55% in the third. The reduction in black concentrations ran afoul of the established Section 5 consideration that rejected plans guilty of retrogression. What was different in DOJ’s 2001 response was that it accepted reductions in black concentrations in other Senate districts. In two other districts the reduction brought the black voting age population down to about 51%. To justify reducing the black concentrations, Georgia offered the testimony of Columbia University political scientist David Epstein (2002), who presented probit models that estimated


the black percentage at which the candidate preferred by African-Americans had a 50 - 50 probability of success. This figure was 44.3% of the Voting age population. Following this line of argument, then districts in which black voting age population exceeded 50% should be acceptable since it was more likely than not that they would elect the black preference. The political need that inspired Georgia’s approach stemmed from the point made earlier – most Georgia white voters were supporting Republican candidates. Reducing concentrations of African-Americans, who regularly cast at least 90 percent of their votes for Democrats (Bullock and Dunn 1999), freed up reliable Democrats to offset the growing tendency of whites to vote Republican. The Democratic strategy sought to pack white Republicans into as few districts as had to be conceded while strategically adding black votes to districts where they could provide the margin of victory for white Democrats. Although Georgia got most of what it wanted from the three-judge panel sitting in the District of Columbia (only three of the 249 districts submitted for approval failed), the state appealed to the Supreme Court. The high court, in a 5 - 4 decision, approved of the reduction in black concentration in the three districts at issue.

THE SECOND CHALLENGE: ONE-PERSON, ONE-VOTE Having failed to block the new maps at the preclearance stage, Republicans went to court to raise two new challenges. First, Republicans asserted that they were victims of an illegal partisan gerrymander. The second claim asserted in Larios v. Cox was that the new maps violated the one-person, one-vote requirement established by Baker v. Carr and Wesberry v. Sanders four decades earlier.


The three-judge panel that heard this suit dismissed the partisan gerrymandering claim so that the emphasis shifted exclusively to the population deviations. Since Karcher, federal courts had displayed a preference for congressional plans that had the smallest populations deviations. A number of states sought to close off the possibility of an equal-population challenge by reducing the deviations in their plans to a single individual. Since Georgia’s plan had a total population deviation of 72, it seemed that it might be vulnerable. In 2002, a federal district court had invalidated the congressional map drawn by Republicans for Pennsylvania because it contained a range in population of 17 people while the Democratic alternative had zeroed out the population differences (Veith v. Pennsylvania (Veith I) 2002). Courts had adopted much looser standards when evaluating state and local legislative plans. The presumption was that these plans would be acceptable if the total deviation was less than a range of 10 points (traditionally expressed as +/- 5%). As noted above, Georgia had scrupulously conformed to that standard. Nonetheless almost a third of the Senate districts and more than one in ten House districts had population deviations of + 4.9% or greater with some going right to the absolute 4.99% limit. Although some courts had interpreted a 10 points range of deviation as a safe harbor for districting plans, the Larios court interpreted this as a rebuttable presumption of constitutionality. While Reynolds v. Sims has not been interpreted as requiring absolute population equality among the districts of a state legislature, deviations must be justified in terms of a legitimate state interest. The Larios opinion explored the rationale offered by the state to justify the deviations in the three maps. In something of a surprise, the court found the state’s justification for the


deviation in its congressional map to be acceptable. Linda Meggers, the respected director of Georgia’s reapportionment office, testified that it would be possible to zero out population differences and in doing so split fewer counties and precincts while having more compact districts.6 Despite Meggers’ testimony, Georgia contended that any efforts to reduce population deviations would necessitate additional precinct splits and that congressional boundaries in some of the divided precincts would not be easily recognizable. The court accepted that justification as a legitimate state interest. The court found Georgia’s explanations for the population variations in the legislative plans less convincing. Witnesses for the state acknowledged that in crafting these plans, Georgia had not considered traditional districting principles such as compactness, contiguity, adherence to county boundaries or maintenance of communities of interest. Instead, the most over- or under-populated districts were often the ones that were the least compact and which strained to achieve contiguity. Six House districts had “duck contiguity.” That is, they were not contiguous on dry land, but the disparate parts were linked across bodies of water and this linkage was not across bridges or causeways. Another five House districts could be considered to be contiguous only at a touch-point. That is, the two parts of the district were contiguous only in the sense that the diagonal black squares on a checkerboard can be thought of as contiguous. Again the rationale for stretching the concept of contiguity was not justified in terms of promoting population equality as two of the touch-point districts were at least 4.5 % off of the ideal population. Nor could it be argued that the population deviations resulted from efforts to honor

The three-judge panel in Johnson v. Miller, 864 F.Supp. 1354,1361, characterized Ms.

Meggers as “probably the single most knowledgeable person available on the subject of Georgian redistricting.” 16

county boundaries. The House plan split 80 counties, eight more than in the plan that it replaced. The state interest offered to justify what had been done, as expressed by witnesses at trial and also argued by state’s attorney David Walbert to the federal court was: (1) to protect the interests of rural South Georgia which for decades had grown more slowly than the rest of the state by eliminating as few districts in that region as possible; (2) protect inner city Atlanta by reducing districts that might be lost there; (3) protect Democratic incumbents who participated in the redistricting process. The plaintiffs did not contest the state’s rationale, only the contention that such a rationale represented a general state interest. The conclusions of one plaintiff’s expert’s report, in its entirety, was: The conclusion to be drawn from this remap is a simple one, summed up in an anonymous quote regarding Georgia’s county unit system, published in 1961: “the situation is simply this: we’ve got the power and you haven’t, and we ain’t going to give it up!” [Cornelius 1961: 942) The crafting of legislative districts in Georgia has defied nearly every convention of redistricting and subverted every traditional redistricting principle. Why? Every redistricting principle – incumbent protection, compactness, contiguity, core retention, county integrity – is subverted to plans with large population deviations, and which under-populate many districts while overpopulating many others in an arbitrary fashion, based on geography and politics. The deviations are not justified by any traditional redistricting criterion (Gaddie 2003: 27). The court found that the rationales offered by Georgia for population deviations did not promote legitimate state interests. Instead, the efforts to advantage some parts of the state – rural South


Georgia and inner city Atlanta – were as unconstitutional as the efforts struck down 40 years earlier in Reynolds v. Sims. The Larios court concluded that In short, the deliberate regional favoritism built into the Georgia House and Senate Plans created more than a taint of arbitrariness and discrimination, violating Equal Protection by diluting the votes of citizens of the suburban and exurban party of northern Georgia and overweighting the votes of citizens in rural Georgia and inner-city Atlanta. While protecting incumbents might be an acceptable state interest, the Larios court noted that it “is a permissible cause of population deviations only when it is limited to the avoidance of contests between incumbents and is applied in a consistent and nondiscriminatory manner.” Georgia’s plans protected only Democratic incumbents and often placed multiple Republican legislators in overpopulated districts. Republican areas tended to be those that were already overpopulated when the 2000 census data were mapped onto the existing plan. Simple probability dictates that it should have been Republican districts that shed population in order to comply with the equal population requirement. Instead the new maps managed to combine parts of overpopulated Republican districts so that GOP incumbents ended up in the same district (Gaddie 2003). In contrast, the Democratic districts that gained population did so without bringing the homes of multiple Democrats into a single district, though some districts had to be torturously shaped to achieve this goal. An illustration of the relationship between the population deviations and the partisan nature of the districts appears in Figure 3. The horizontal axis plots the percent population deviation from ideal for each of the 180 seats in the Georgia House of Representatives in 2002. The vertical axis plots: for the diamond markers, the proportion of votes cast for Republicans for


Public Service Commission in 2004; for the circular markers, the probability of the district voting a majority Republican for PSC in 2000. The relationship between the population deviations and proportion of votes cast is positive and generally linear (adjusted-R2 = .348), while the expected probability of a district voting majority Republican is so strongly related to the size and direction of the district population deviation as to be nearly perfectly linear. The population deviations were structured on the basis of partisanship. (Figure 3 here)

CRAFTING THE NEW MAPS Democratic lawyers immediately appealed the federal three-judge panel’s decision to the US Supreme Court, which refused to intercede. Georgia had to come up with new legislative districts in time for the 2004 elections. Since filing for election in Georgia was scheduled to take place during the last week of April, the trial court gave the legislature less than three weeks, until March 1, to come up with replacement plans. Republicans, despite the Democratic gerrymander that had been designed to increase their party’s Senate contingent by five, took control of the upper chamber after the 2002 election.7 They had passed a new plan for their district in 2003 only to see it languish in a House committee. After Larios, the Senate enacted a plan. In the past, each chamber had deferred to the other when it came to districting its own chamber. That House ignored that norm in 2004. Not only did the House balk at accepting the Senate plan, it proved unable to fashion a

The Democratic gerrymander actually resulted in Democrats winning 30 of 56 Senate seats

but the newly-elected Republican governor, Sonny Perdue, prevailed on four Democrats to change parties. 19

response to the court order. The House Legislative and Congressional Reapportionment Committee never released a plan. Although Democrats had a sizable advantage, holding 107 of 180 seats, they doubted whether they could hold their ranks and enact a plan. They feared that Republicans would cut deals with enough rural, conservative Democrats to enact a GOP alternative to any plan that the Democratic leadership would offer. When the legislature failed to act, the court appointed retired federal judge Joseph Hatchett to serve as special master. The judge, assisted by Professor Nathaniel Persily, drew maps for the General Assembly that had maximum deviations of + 1%. The initial maps did not consider incumbency and resulted in 66 representatives being paired (see Table 2). Of the paired incumbents, 45 were Democrats (40.4% of the Democratic caucus) while 20 were Republicans (30% of the GOP caucus). A number of these pairings occurred while a nearby district had no incumbent. Ironically, the chair of the House Reapportionment Committee who had refused to present a map ended up sharing a district with two of the House’s most powerful members, the chairs of the Appropriations and Rules Committees. (Table 2 here) The court responded to a number of suggestions from the two political parties and when a pairing could be undone with little difficulty and would not result in another pairing, the court approved changes. Members of the Legislative Black Caucus claimed that they had been singled out for pairing even though the new maps resulted in the same number of majority-black districts as the maps being replaced (Cook 2004).8 In the end 33 representatives shared districts with

Of the 38 seats elected from majority black districts in 2002, 17 had populations of greater than

-4% while only 7 had populations above the ideal. The geographic concentration of these districts contributed to the initial pairing of black incumbents elected from these districts. 20

other incumbents – 22 Democrats, 10 Republicans, and the sole independent (see Table 2).The new maps still resulted in a large number of open seats which boosted Republican hopes of winning a majority in the House and thereby take control of two branches of Georgia government.

Judicial Finality The final decision in the Georgia redistricting trial came on June 30 2004, when the US Supreme Court affirmed the judgment of the Northern US District Court of Georgia by an 8-1 vote. Unlike most affirmations without hearing, this case did lead three of the justices to place their names on opinions. Writing a concurring opinion to the majority (joined by Justice Breyer) Justice Stevens spoke in powerful terms for the defense of the principle of one person, one-vote: The District Court’s findings disclose two reasons for the unconstitutional population deviations in the state legislative reapportionment plans. The first was a ‘deliberate and systematic policy of favoring rural and inner-city interests at the expense of suburban areas north, east, and west of Atlanta’ . . . The second was ‘an intentional effort to allow incumbent Democrats to maintain or increase their delegation, primarily by systematically underpopualting the districts held by incumbent Democrats, by overpopulating those of Republicans, and by deliberately pairing numerous Republican incumbents against one another.’ . . . Democratic incumbents ‘attempted to draw districts that would further enhance their own prospects at reelection and further their other political ends.’ . . . Although ‘the numbers speak for themselves’ the District Court found that the shapes of many of the newly created districts supplied further evidence that the


plans’ drafters ‘intended not only to aid Democratic incumbents in gtting reelected but also to oust many of their Republican counterparts.’ . . . many of the districts that paired Republicans were oddly shaped and overpopulated . . . the District Court found that the population deviations did not result from any attempt to create districts that were compact or contiguous, or to keep counties whole, or to preserve the cores of prior districts . . . . ‘the population deviations were designed to allow Democrats to maintain or increase their representation’ . . . . the district court correctly held that the drafters desire to give an electoral advantage to certain regions of the State and to certain incumbents did not justify the conceded deviations from the principle of one-person, one-vote. ” No rational state policy was served by the deviations, per Reynolds v. Sims. Stevens was not satisified to just affirm this interpretation of Karcher v. Daggett and Reynolds v. Sims. He continued, working through ground debated just weeks before in Veith v. Jubelirer: The equal population principle remains the only clear limitation on improper districting practices, and we must be careful not to dilute its strength. It bears emphasis however, that had the Court in Veith adopted a standard for adjudicating partisan gerrymandering claims, the standard would have been satisfied in this case. . . . the District Court’s detailed factual findings regarding appellees’ equal protection claim confirm that an impermissible partisan gerrymander is visible to the judicial eye and subject to judiciallymanageable standards . . . “drawing district lines that have no neutral justification in order to place two incumbents ot the opposite party in the same district is probative of the same impermissible intent as the ‘uncouth twenty-eight-sided figure’ that defined the boundary of Tuskegee,Alabama in Gomillion.


Justice Stevens expressed a notion familiar from this court, that oddly-shaped constituencies are not only aesthetically offensive, but, when combined with some political agenda, also legally offensive. Justices Stevens and Breyer effectively invited additional challenges to redistricting plans on the basis of partisanship, if the machinations of the sort used in the Georgia legislative maps were present in the challenged map.

THE ELECTORAL CONSEQUENCES OF THE 2004 REMAP The political consequences of the remap were far-reaching. Twenty-six Democratic incumbents did not stand in the general election, and another six were beaten in the general election. Republicans dominated the open seats and reelected 63 of their 65 incumbents. When the dust of the November election settled, Democrats had lost a net of 24 seats, compared to 25 gained by the Republicans (plus the defeat of one independent), leaving Republicans with a 95-85 legislative majority.9 In 2002, 136 incumbents stood for reelection in the general election: 87 of 107 Democrats (81.4%) and 49 of 72 Republicans (68.1%). In 2004, 143 incumbents stood in the general election: 81 of 109 Democrats (74.3%) and 62 of 70 Republicans (88.5%). The Republican gain is in part a product of the nature of the redistricting pairings in the federal court map. Of the 180 new single-member districts, 15 were true open seats, with no incumbent in the district; 65 hosted one Republican incumbent; 85 hosted one Democratic incumbent; four districts hosted two Republican incumbents; six districts hosted two or more Democratic incumbents; and 5 districts hosted an incumbent from each of the major parties. By

The majority would grow in the coming weeks as south Georgia Democrats would cross party

lines to join the GOP caucus, boosting its ranks to 99 of 180 seats. 23

definition four Republican incumbents and seven Democratic incumbents were precluded from renomination.

The Two-Incumbent Pairings The five bipartisan pairings were in districts 34, 93, 146, 165, and 166. Based on the descriptions of the districts, they would have appeared to favored the Democrats. Three of the districts (93, 165, 166) had given at least 55% of their votes to Democrats for the PSC in 2000, and the three were between 32% and 61% black voting age population. The other two districts (34 and 146) voted a majority Republican for the PSC in 2000 and had black voting age populations below 22%. Republicans actually carried three of the five districts, including the two Republican-tending districts and also district 166, which had voted 55% for the Democrats for PSC in 2000 (see Table 3). (Table 3 here) Democrats retained all six of the districts where two or more Democrats were paired (57, 84, 92, 99, 149, 161). None of these districts had voted less than 63% for the Democrats for the PSC in 2000, and four of the six had a sufficiently large black population to guarantee the election of a black candidate of choice, based on David Epstein’s (2002) analysis. In the districts with two Republican incumbents, all were retained by the GOP (7, 17, 18, 32). All of these districts had cast a majority of votes for Republicans for the PSC in 2000, including three that voted over 60% Republican. None of the districts had a black voting age population of over 17%, and one had a scant 0.19% black voting age population.


One-Incumbent Districts In the 65 districts where one GOP incumbent was placed, Republicans retained 63 (96.9%). Incumbents sought reelection in 53 of these districts. Overall these were very safe GOP constituencies. The average vote for the PSC in 2000 was 61.3%, and the average Voting age population was only 11.4% black. These were typically very white, very Republican districts (see Table 3). The 85 districts where lone Democratic incumbents were located showed more variety and political volatility. In 73 of the districts the incumbent did seek reelection. But, Republicans won 13 of the 85 districts where Democratic incumbents were placed (15.2%) . Of the seats that changed party, seven came from the dozen open seats created by the retirement of Democratic incumbents. The other six came from the defeat of Democratic incumbents by Republican challengers. On average, these districts were located in good Democratic territory. The average Democratic vote for PSC in 2000 was over 61%, and the average black voting age population was just over 40%.

The Open Seats Of the fifteen seats created by redistricting, a dozen (80%) were captured by the Republicans. On average these were very Republican districts, with a mean PSC2000 vote of 56.9% Republican and an average black voting age population of just 16.9%. Seven of these seats were taken by Republicans without opposition in the general election. Nine of the open seats had a GOP vote for PSC in 2000 of over 55%, and the three other districts captured by the GOP had a PSC 2000 vote of at least 46% Republican.


In the absence of incumbents, Republicans were dominant in the 2004 Georgia house elections, taking 28 of 38 open seats (73.7%). Seventeen of those seats were taken without opposition from the Democrats, while just three open seats (all with majority black voting age population) were conceded by the Republicans. In the contested open seats, Republicans won eleven of eighteen instances.

Descriptive Change: Black Population Concentrations The revision of the Georgia maps by the courts did affect the distribution and concentration of black voters. As indicated in Table 4, the map rejected by the federal court included 38 Voting age population majority-black districts and another 10 districts with more than 40% black voting age population that are assumed to be safely Democratic if not generally accessible to black populations to elect candidates of choice (see Black and Black 2002). Among the remaining districts, 82 were less than 20% black and 50 were less than 10% black by population. (Table 4 here) The strong structural role of race in determining the composition of the general assembly is evident from the data in Table 4. Republicans won none of the 48 districts that were over 40% black voting age population and carried 90% of the 50 districts that were less than 10% black by Voting age population. Republicans only won 25 of the 82 districts (30.5%) between 10% and 40% black voting age population, including just a bare majority (17 of 32 seats) with a black voting age population between 10% and 20%. While Republicans came to dominate congressional districts in the Deep South of less than 35% black voting age population (Black and Black 2002), it was only in overwhelmingly-white areas that Republicans dominated


Georgia state house elections in 2002. The change in state legislative boundaries scattered the selective packing and placement of black voters in the state legislative map. As indicated in part B of Table 4, the number of districts less than 20% black voting age population increased from 82 to 89, and the number of black-majority Voting age population districts increased by +1 to 39, while the number of districts over 60% black voting age population changed by +3, and the number of districts between 40% and 50% black voting age population increased by a net of 2. The number of districts between 20% and 30% black voting age population decreased by 6, while the number of districts with black voting age population between 30% and 40% black decreased by 4. The political impact of the shifts exceeds the demographic changes. In 2002, Republicans won 17 of 32 districts between 10% and 20% black voting age population. In 2004 the number increased to 31 of 37 such districts (83.8% of seats). Among districts between 10% and 40% black voting age population, Republicans carried 50 of 77 seats (64.9%), and they even carried 8 of 18 districts with black voting age population between 30% and 40%. Republicans consolidated their domination over nearly all of the most-heavily white constituencies and also ran even in competing for seats with sizeable black minorities. Redistricting created more constituencies that were favorable, on the basis of racial demographics, to Republicans than existed under the old map, yet Republican gains exceeded the expectations given previous GOP success in state legislative districts on the basis of race.

Descriptive Change: Underlying Republican Partisanship The changes in legislative boundaries also shifted the underlying partisan makeup of the districts.


Of the 180 state house districts created in 2001, 78 cast a majority of ballots for Republican for the Public Service Commission in 2000, and 102 gave majorities to the Democrats. Only 18 of 180 districts had cast between 50% and 55% of ballots for Republicans, and just 24 cast between 50% and 55% of ballots for Democrats (see Table 5). (Table 5 here) Among the safest Republican districts (less than 45% Democratic for the PSC in 2000), 57 of 60 seats were won by a Republican in 2002. None of the 47 safest Democratic districts (over 60% Democratic for PSC in 2000) went to the GOP. This ategory also includes nearly all of the majority-black seats in the 2001 map. Only 5 of 102 seats (4.8%) with a PSC 2000 vote of over 50% Democratic voted elected a Republican to the legislature in 2002. The 2004 map substantially altered the distribution of partisans and the nuber of districts dominated by either party. Of 180 districts created, 89 had a GOP majority for the PSC in 2000 and 74 had voted at least 55% Republican (changes of +11and +14 respectively over the 2001 map). Republicans dominated those newly-Republican districts, picking up a total of 16 seats in the majority-Republican PSC districts to take 81 of 89 such districts (91.0%, compared to just 70.5% of the 78 such seats in 2002). Republicans also took 14 of the majority-Democratic PSC 2000 districts, compared to just 5 such districts in 2002.

Multivariate Analysis The redistricting strategy pursued by the Democrats in 2001 was to strategically place reliable black voters, other Democrats, and incumbents in such a fashion as to maximize Democratic opportunities while minimizing Republican incumbency effects and packing as many Republican


voters into as few districts as possible. The 2004 court-ordered redistricting was incumbent neutral in its effects, but it also redistributed Democratic voters in such a fashion as to create several opportunities for Republicans to make legislative gains. When we consider the Georgia redistricting from a structural standpoint, there are three factors that were considered important by Democratic mapmakers: the party potential of the district; the placement of minority voters; and the placement of incumbents. These three components can be represented by a set of five variables: Party potential is measured by the vote in 2000 for public service commission. The PSC is a traditionally low-profile, low cost office that is representative of the GOP party baseline in Georgia entering this redistricting round (CITE). Two PSC seats were contested in 2000, and when the votes of the two contests are aggregated the statewide average is 5_% Republican. The effect of minority voter concentrations is measured by the% Black voting age population and the % Latino Voting age population (respectively) in each district. Incumbency is measured using a pair of dichotomous variables, one each to indicate the presence or absence of a Republican or Democratic incumbent.10 Table 6 shows two estimates of the Republican share of the vote for legislature in the contested districts in Georgia. In the first estimate, the 2000 Public Service Commission vote and controls for the presence of Democratic and Republican incumbent candidates are

The data for this analysis were collected from the Georgia Secretary of State’s Election Board

website; from various issues of the Atlanta Journal-Constitution; and from data collected and analyzed by the authors on behalf of the state of Georgia and the Georgia Republican Party, respectively, for litigation related to the Georgia redistricting in Georgia v. Aschcroft and Larios v. Cox. 29

introduced. In the second estimate, the incumbency controls are retained along with controls for the black and Latino Voting age populations. The racial controls and the PSC 2000 vote are not included in the same model because the black voting age population in particular is highly correlated with the PSC 2000 vote at the district level (adjusted-R2 = .85). The analyses reveal the strong relationship between partisanship in the closely contested, low-profile PSC contests and the vote for the legislature in 2002. In both analyses we also see that, while incumbency acts in the direction expected for incumbents of both parties, the effect of the presence of a Democratic incumbent is roughly twice that of a Republican incumbent. (Table 6 here) In 2002 both the strategic placement of Democratic incumbents and the elimination from contention of Republican incumbents alone could affect the expected vote distribution by four to seven points, controlling for base partisanship. The same models are estimated for 2004. It is worth noting that, in the model controlling for partisanship through the PSC 2000 vote, the intercept is 14 points higher than for the districts in 2002. The slope coefficient for PSC 2000 is somewhat attenuated ( b = .833 in 2004 as opposed to b = 1.106 in 2002). The model is similarly robust in 2004 as in 2002. The most notable change is in the effect of incumbency. In 2002, the slope for Democratic incumbency was roughly twice that for Republican incumbency in both regression estimates. The Republican and Democratic slopes are of roughly the same magnitude though in opposite directions. In the racial analysis the slope for black voting age population is somewhat attenuated but still highly significant. The relationship between the predictor variables and Republican success in winning


districts is tested on the 2002 and 2004 house elections. For each election, the political, racial, and incumbency variables are initially tested separately on the outcome of interest – Republican wins a seat – and then the five variables are combined into a single estimate. Then, a reduced, combined model is presented that eliminates insignificant predictors from the initial, combined estimation. The 2002 election outcomes were strongly structured by the underlying partisanship of the district and by the presence or absence of a Democratic incumbent running in a district. The analysis in Table 7 reveals that the PSC 2000 vote alone reduced 80% of the predictive error in determining which party won a district. The racial/ethnic variables alone reduced the predictive error by 60%, though only black voting age population was a statistically-significant predictor (Latino Voting age population failed significance). The incumbency controls reduced the predictive error by just over 47%. When the five variables are combined in a single estimate, only the PSC vote and the presence of a Democratic incumbent significantly affected the party winning the seat, though the predictive error reduction is less than for the PSC variable alone. The reduced, combined two-variable model reduced the predictive error by 80%. (Table 7 here) The 2004 elections are also strongly structured by the underlying partisanship of the district. The PSC vote alone reduced the predictive error in who won a state house seat by 79% (see Table 8). The racial/ethnic variables reduced the predictive error in who won by about 65%, though again black voting age population is the only significant predictor. The incumbency variables reduced predictive error by 80%, but in 2004 both Democratic and Republican incumbency attain significance. In a combined model the incumbency variables and partisanship


attain significance at conventional levels, and reduce the predictive error by 89%; eliminating the Latino Voting age population control results in black population attaining significance in the reduced, combined model. (Table 8 here)

CONCLUSIONS The Georgia redistricting of 2001 has been noted as an example of political creativity that clearly pushed the limits of redistricting technology and the law in pursuit of political advantage. In order to successfully implement their initial redistricting proposal, Georgia Democrats had to take an unprecedented avenue in the obtaining preclearance under the Voting Rights Act. The US Supreme Court had to redefine election law with regard to the act to uphold Georgia’s actions. Nearly every convention of redistricting was set aside in pursuit of political advantage. And, ultimately, the map was undone by the inability to justify seemingly legal population deviations to the satisfaction of the courts. The defects of the Georgia legislative maps were only justified by the pursuit of political advantage by the party in power. When the state’s districts were crafted to comply with a de minimus standard for population deviations, while pursuing traditional redistricting criteria, the carefully-crafted political advantages of the gerrymander came completely undone. The tripod of the gerrymander – strategic placement of Democratic incumbents and the strategic dislocation of Republican incumbents, together with the strategic placement of black and Democratic voters, respectively – was knocked out from under the Democratic majority in the Georgia General Assembly. How much of the Democrats’ undoing is a product of the court-ordered remap?


As a check on the impact of the redistricting, we applied the results of the regression analyses for 2002 to the data for 2004, in an effort to construct expected GOP vote shares for each district. We first estimated GOP vote shares for all 180 seats in 2002, based on the racial or partisan data for those districts and the circumstance of incumbency in the election.11 Using the estimates from Table 6(A) – the PSC2000 vote with incumbency controls – indicated an expected result in 2002 of 74 GOP seats and 106 Democratic seats (Republicans won 70 seats, Democrats 109 seats, and one independent was elected). The estimates from Table 6(B) – the racial demographics with incumbency controls – indicated 83 GOP seats and 97 Democratic seats for 2002. Applying the regression equation results for 2002 to the data and incumbency circumstances for 2004 lays bare the substantial impact of the redistricting on the political balance in the Georgia General Assembly. Of 180 districts, 92 were expected to vote a majority for Republicans according to the PSC2000/incumbency model; the racial/incumbency model expected 102 GOP seats and 78 Democratic seats. Of the 92 seats expected to vote majorityRepublican by the PSC2000/incumbency model for 2002, 86 did in fact elect a Republican; of the 102 seats expected to vote majority-Republican in 2004 based on the racial/incumbency model for 2002, 91 did so. Overall, if we were to assume that any district that one or the other regression equation expected to vote majority Republican would do so, we would correctly call 91 cases, and miss the call on 16 cases (4 that were expected to vote Democratic and went Republican, and 12 that were expected to vote Republican and did not). Now, if we take the analysis a step further, and eliminate the incumbency influence, we

See Gelman and King, 1994; Kouser, 1996; Gaddie and Bullock, 2000, for examples of variants

on the application of this technique. 33

can get a feel for the role of incumbency in assuring Democratic hegemony in the Georgia General Assembly. By suppressing the effects of incumbency to 0 when applying the equations from Table 6(A) and 6(B), we realize the precarious nature of the Democratic majority. In 2002 Democrats retained the lower chamber based largely on the creative placement of core Democrats (especially minority voters) plus the power of their incumbents. According to estimated vote shares from the equation in Table 6 (A) and holding incumbency to 0 (assuming all contested, open seats), in 2002 Republicans and Democrats should have each garnered 90 of 180 seats in the lower chamber. The estimates from the racial model – Table 6(B)– indicate that Republicans would have been advantaged in 101 districts. Things are far worse for the Democrats under the new map. When the incumbents are removed from the equation as applied to the map for 2004, OLS estimates indicate that Republicans would be expected to garner a majority in 103 of 180 districts, based on the PSC2000 equation in Table 6(A), and a majority in 113 districts based on the racial equation in Table 6 (B). This result indicates to us that a shift of ten to thirteen seats to the advantage of the Republicans occurred due to the effects of redistricting, laying aside the secondary effects of the placement or displacement of incumbents of each party under the new map. The Georgia redistricting is politically and legally significant. The political significance is self-evident, in that it substantially disrupted a political strategy designed to continue the control of a party that had lost its popular majority. The product of the court, which sought to minimize the detrimental effects on incumbents to the extent possible while crafting a neutral-principles map, nonetheless undid the Democrats by


denying to them sufficient incumbents and sufficient attractive districts to run in. This analysis also indicates that the bottom has probably not yet been reached by the Georgia Democrats in the state house. They could conceivably fall to just over one-third of seats based on the alignment of current districts. From a legal standpoint, the Georgia remap can serve as a potential roadmap for the courts to assess and rectify partisan gerrymanders. Two justices of the US Supreme Court have explicitly stated that the evidence of partisan bias evident in the Georgia maps was sufficient to demonstrate that a gerrymander is “visible to the judicial eye” and the Georgia map would meet the standards of an illegal gerrymander according to the minority’s standard in the Veith case. The three-judge panel of the northern circuit has shown the propriety of a neutral principles map, at least in this case, as a solution to a partisan gerrymander. The compact, basically incumbent-neutral map (which nonetheless took pains to retroactively uncouple many paired incumbents) translated a majority of votes into a majority of seats, as Republican state house candidates garnered 56.5% of the votes for the lower chamber in 2004 and won 52.8% of seats.


CASES Abrams v. Johnson 521 US 74 (1997) Baker v. Carr, 369 US 186 (1962) Balderas v. Texas, Civil Action No. 6:01 CV 158 (E.D. TX. Tyler Division, 2001). Bush v. Vera, 517 U.S. 952 (1996). Cox v. Larios 542 US ____ (2004) Easley v. Cromartie, 532 U.S. 234 (2001). Georgia v. Ashcroft, 123 Sct. 2498 (2003). Johnson v. Miller (CV 194-008) 922 F. Supp. 1556, 1995 U.S. Dist. Lexis 19774. Karcher v. Daggett, 462 U.S. 725 (1983). Larios v. Cox, Civil Action No. 1:03-CV-693-CAP (N.D. GA 2004). Miller v. Johnson, 515 U.S. 9000 (1995). Reynolds v. Sims, 377 U.S. 579 (1964). Veith v. Jubelirer (CV 02_1580, 2004) Wesberry v. Sanders, 376 U.S. 1 (1964)


REFERENCES Abramowitz, Alan I. 1983.Partisan Redistricting and the 1982 Congressional Elections. The Journal of Politics 45: 767-770. Altman, Micah. 1998. Traditional Redistricting Principles: Judicial Myths v. Reality. Social Science History 22: 159-200. Bensen, Clark. 2003. Expert Report of Clark Bensen, J.D. Prepared for submission in Larios, et al., v. Perdue, et al., 1:03-CV-0693, United States District Court for the Northern District of Georgia, Atlanta Division (October 20). Black, Earl, and Merle Black. 2002. The Rise of Southern Republicans. Cambridge: Harvard University Press. Bullock, Charles S., III, 2004. Two Generations of Redistricting: An Overview. Extensions Fall 2004: 9-13 Bullock, Charles S., and Richard E. Dunn. 1999. The Demise of Racial Districting and the Future of Black Representation. Emory Law Journal 48 (Fall): 12091253. Butler, David, and Bruce Cain. 1992. Congressional Redistricting: Comparative and Theoretical Perspectives. New York: MacMillan. Butler, Katharine Inglis . 2001. REDISTRICTING IN A POST-SHAW ERA: A Small Treatise Accompanied by Districting Guidelines for Legislators, Litigants, and Courts. Typescript, University of South Carolina Law School. Cain, Bruce. 1984. The Reapportionment Puzzle. Berkeley, CA: University of California Press.


Cook, Rhonda. 2004. Democrats Stall Map Decision. Atlanta Journal Constitution (March 18): D5. Cornelius, William G. 1961. The County Unit System of Georgia: Facts and Prospects. The Western Political Quarterly 14: 942_960. Engstrom, Richard L. and Jason E. Kirksey. 1998. Race and Representational Districting in Louisiana. In Race and Redistricting in the 1990s, edited by Bernard Grofman. New York: Agathon Press. Epstein, David. 2002. Expert Report of David Epstein Ph.D. Prepared for submission in Georgia v Ashcroft. Gaddie, Ronald Keith. 2003. Expert Report of Ronald Keith Gaddie, Ph.D. Prepared for submission in Larios, et al., v. Perdue, et al., 1:03-CV-0693, United States District Court for the Northern District of Georgia, Atlanta Division (October 20). Gaddie, Ronald Keith, and Charles S. Bullock III. 2000. Elections to Open Seats in the US House: Where the Action Is. Boulder: Rowman and Littelfield. Gelman, Andrew, and Gary King. 1994. Enhancing Democracy Through Legislative Redistricting. American Political Science Review 88: 541-559. Glazer, Amahai, Bernard Grofman, and Marc Robbins. 1987. Partisan and Incumbency effects of the 1970s Congressional Redistricting. American Journal of Political Science 31: 680-707. Hill, Kevin A. 1995. Does the Creation of Majority Black Districts Aid Republicans? An Analysis of the 1992 Congressional Elections in Eight Southern States The Journal of Politics 57: 384_401.


Holmes, Robert A. 1998. Reapportionment Strategies in the 1990s: The Case of Georgia. In Race and Redistricting in the 1990s, edited by Bernard Grofman. New York: Agathon Press. King, Gary. 1989. Representation Through Legislative Redistricting: A Stochastic model. American Journal of Political Science 33: 463-477. Kousser, J. Morgan. 1996. Estimating the Partisan Consequences of Redistricting Plans_Simply. Legislative Studies Quarterly 21: 521_541 Niemi, Richard G., and Alan I. Abramowitz. 1994. Partisan Redistricting and the 1992 Congressional Elections. The Journal of Politics 56: 811_817. Niemi, Richard G., Bernard Grofman, Carl Carluucci, and Thomas Hofeller. 1991. Measuring Compactness and the Role of Compactness Standard in a test for partisan and Racial Gerrymandering. Journal of Politics 52: 1155-1181. Overby, L. Marvin, and Kenneth M. Cosgrove. 1996. Unintended Consequences? Racial Redistricting and the Representation of Minority Interests. The Journal of Politics 58: 540_550. Rush, Mark E. 2000. Does Redistricting Make a Difference? Partisan Representation and Electoral Behavior. Lanham, MD: Lexington Books Sellers, Patrick J., David T. Canon, and Matthew M. Schousen. 1998. Congressional Redistricting in North Carolina. In Race and Redistricting in the 1990s, edited by Bernard Grofman. New York: Agathon Press.


Table 1: Georgia State House Incumbent Pairings 2002

Paired Incumbents District 3-2Post 14-2Post 17 30 35 44 46 52 56-2Post 61-3Post 67-2Post 76 85-2Post 97 106 110 113 126 127 137 138 Hammontree (R), Williams (R), Forster (R) Pinholster (R), C. Smith (R), Knox (R) Scheid (R), Franklin (R) Cooper (R), Kaye (R) Wiles (R), Hines (R) McKinney (D), Collins (I) Snelling (R), Hembree (R) Millar (R), Davis (R) Jennings (R), Stuckey (D)* Ragas (D), Sailor (D), J. Williams (R)* Mills (R), Coan (R), Reese (R) Hudgens (R), B. Smith (R) Cox (R), Yates (R), Lunsford (R) Burmeister (R), Allen (D) Graves (R), Reichert (D) V. Smith (R), Roberts (R) Hugley (R), Taylor (D) Mueller (R), Day (R) Lanier (R), DeLoach (I) Everett (R), Bullock (R) Holland (D), Scott (R)

Pop.Dev. (%) 4.740 3.990 4.820 4.570 4.990 -.680 4.240 2.000 -4.960 -.740 4.950 -1.460 4.300 -4.290 4.470 .570 -3.680 4.790 4.800 4.450 3.100

’00 PSC Average 61.80 51.04 71.25 70.47 64.13 36.26 60.46 61.62 40.82 31.56 69.58 58.96 69.28 34.11 60.32 62.19 24.76 68.57 51.79 61.31 46.55

*All of these pairing are fatal except in districts 56 and 67, where each incumbent could run in a separate post. Source: Gaddie 2003a

Table 2: Georgia State House Incumbent Pairings 2004 District** Paired Incumbents, 1st Map 149 G. Green (D) and Hugh Broome (D) B. Mitchell (D) and M. Henson (D) P. Smith (D) and B. Childers (D) D. Wix (D) and A. Thomas (D) 57 D. Teper (D) and P. Gardner (D) S. Benefield (D), J. McClinton (D) and 84 G. Maddox (D) R. Teilet (D) and J. Noel (D) T. Brooks (D) and B. Holmes (D) K. Ashe (D) and D. Dean (D) G. Sinkfield (D) and R. Dodson (D) P. Stephenson (D), S. Watson (D), and 92 T Greene-Johnson (D) V. Hill (D) and D. Jordan (D) J. Skipper (D) and D. Buckner (D) T. Buck (D), C. Smyre (D), and C. Hugley (D) 161 L. Jackson (D) and M. Stephens (D) P. Houston (D) and J. Shaw (D) 34 D. Stoner (D) and R. Golick (D) 93 R. Sailor (D) and B. Bunn (D) D. Casas (R), H. Floyd (D), and C. Bannister (D) 99 L. Walker (D)* and L. O’Neal (R) 146 T. Barnard (R) and B. Oliver (D) 166 165 B. DeLoach (I) and A. Williams (D) B. Hanner (D) and E. Rynders (R) R. Forster (R) and R. Williams (R) J. White (R) and D. Ralston (R) 7 T. Knox (R) and J. Murphy (R) M. Butler (R) and C. Harper (R) 18 M. Coan (R) and D. Sheldon (R) B. Heath (R) and R. Maxwell (R) 17 D. Parsons (R) and M. Dollar (R) M. Burkhalter (R) and T. Rice (R) Paired Incumbents, 2d Map G. Green (D) and Hugh Broome (D)

D. Teper (D) and P. Gardner (D) B. Mobley (D), J. McClinton (D) and G. Maddox (D)*

P. Stephenson (D) and T GreeneJohnson (D)*

L. Jackson (D) and M. Stephens (D) P. Houston (D) and J. Shaw (D) D. Stoner (D) and R. Golick (D) R. Sailor (D) and B. Bunn (D) C. Thompson (D) and H. Floyd (D)* L. Walker (D) and L. O’Neal (R)* T. Barnard (R) and B. Oliver (D) B. DeLoach (I) and A. Williams (D) R. Forster (R) and R. Williams (R) J. White (R) and D. Ralston (R)* M. Butler (R) and C. Harper (R) B. Heath (R) and R. Maxwell (R)

*At least one incumbent had previously announced plans to retire or seek other office. **Note: District numbers are included only for final pairings. The numbering of the initial, proposed districts by the federal court’s special master did not conform to the final district numbering, and included districts numbered only with letters or letters and numbers.

Table 3: District Profiles where Incumbents were Paired
A. Bipartisan Pairings (n = 5) DISTRICT Black VAP 34 21.26 93 61.54 146 18.13 165 45.59 166 32.55 PSC2000 50.36 32.73 58.89 33.04 44.17 GOPWIN? YES NO YES NO* YES

B. Democrats-Only Pairings (n = 6) DISTRICT Black VAP PSC2000 57 8.52 34.65 84 71.97 12.40 92 58.92 31.81 99 25.51 36.54 149 43.22 35.69 161 56.39 31.87


C. Republicans-Only pairings (n = 4) DISTRICT Black VAP PSC2000 7 .19 61.38 17 5.93 60.79 18 16.92 51.48 32 8.08 68.42


*Pairing included independent Buddy DeLoach, who lost to incumbent Democrat Al Williams in the general election. No Republican was included in this pairing.

Table 4: Minority Black VAP, # seats, 2002 and 2004

A. 2002

Winning Party: N* Dem GOP 50 32 28 22 10 20 18 5 15 21 20 10 20 18 45 17 6*** 2 0 0 0

% GOP 90.0 53.1 21.4 9.1 0.0 0.0 0.0 Net  from 2002 +2 total, +0 GOP +5 total, +14 GOP -6 total, +5 GOP -4 total, +6 GOP +2 total, +0 GOP -2 total, +0 GOP +3 total, +0 GOP

< 10% 10-20% 20-30%** 30-40% 40-50% 50-60% 60%+

B. 2004 N < 10% 10-20% 20-30% 30-40%**** 40-50% 50-60% 60%+ 52 37 22 18 12 18 21

Winning Party: Dem GOP %GOP 7 6 11 10 12 18 21 45 31 11 8 0 0 0 86.5 83.8 50.0 44.4 0.0 0.0 0.0

*In 2002 180 seats were elected in 147 districts, including 66 seats elected in multimember districts. The “N” for each category represents the number of seats elected from a category, rather than the number of districts (an MMD with three seats counts as three districts with identical racial demo graphics). ** One seat was carried by independent Buddy DeLoach. ***Highest %black won by a Republican was 34.4%. ****Highest %black won by a Republican was 37.2%

Table 5: Democratic Tendency of Districts, PSC2000 Vote, 2002 and 2004

A. 2002

Winning Party: N Dem GOP 51 0 9 24 31 47 3 9 8* 23 27 47 48 100.0 44.4 1 4 0

%GOP 94.1

< 40% 40-45% 9 45-50% 18 50-55% 55-60% >60%

4.2 12.9 0.0 Net  from 2002 +1 total, +3 GOP +13 total, +12 GOP -3 total, +1 GOP -1 total, +8 GOP -15 total, +0 GOP -6 total, +1 GOP

B. 2004

Winning Party: N Dem GOP 52 22 15 23 16 53 1 1 6 14 12 52 51 21 9 9 4 1

%GOP 98.1 95.5 60.0 39.1 25.0 18.9

< 40% 40-45% 45-50% 50-55% 55-60% >60%

*One district was carried by independent Buddy DeLoach.


Contested seats in 2002 A. B Intercept -2.007 PSC 2000 1.106 Dem. Incumbent -7.048 Rep. Incumbent 3.978 Adjusted R N B. Intercept Dem. Incumbent Rep. Incumbent % Latino VAP % Black VAP Adjusted R N
2 2

s.e.b 3.726 .072 1.617 2.099

t -.539 15.277*** -4.358*** 1.895*

.835 71

B 68.466 -10.230 4.996 -.365 -.557 .668 71

s.e.b 3.048 2.378 3.013 .218 .061

t 22.463 -4.302*** 1.658* -1.680* -9.103***

Contested Seats in 2004 C. B Intercept 12.089 PSC 2000 Dem. Incumbent Rep. Incumbent Adjusted R N D. Intercept Dem. Incumbent Rep. Incumbent % Black VAP % Latino VAP Adjusted R N ***p < .001
2 2

s.e.b 5.336 .100 2.334 2.566

t 2.265* 8.348*** -3.233** 2.613**

.833 -7.546 6.704 .785 71

B 65.166 -9.863 10.971 -.409 -.315 .702 71

s.e.b 2.818 2.703 2.971 .076 .129

t 23.127 -3.649*** 3.693*** -5.380*** -2.443**

**p < .01

*p < .05

Table 7: Logistic Regression Estimates of Party Winning Seat, 2002 Georgia House Racial/ Ethnic Model 2.991 Reduced, Combined Model -12.149 .243***

Political Model Constant PSC 2000 Black % VAP Latino % VAP Inc. Democrat Inc. Republican log-likelihood null prediction % Correct Pred. PRE n = 180 85.59 61.11% 92.22% .800 ***p < .001 -14.812 .279***

Incumbency Model -.000

Combined Model -8.721 .189**

-.157*** -.064 -2.436*** 1.634** 121.655 61.11% 84.44% .600 **p < .01 87.257 61.11% 79.44% .471 *p < .05

.045 .001 -1.538* .544 77.354 61.11% 91.67% .654 +p < .10 79.361 61.11% 92.22% .894 -1.596**

Table 8: Logistic Regression Estimates of Party Winning Seat, 2004 Georgia House Racial/ Ethnic Model 3.162 Reduced, Combined Model -21.445 .413*** .088*

Political Model Constant PSC 2000 Black % VAP Latino % VAP Inc. Democrat Inc. Republican log-likelihood null prediction % Correct Pred. PRE n = 180 88.24 52.78% 90.00% .788 ***p < .001 -13.424*** .276***

Incumbency Model 1.123

Combined Model -21.765 .417***

-.116*** -.031 -3.764*** 3.370* 137.29 52.78% 83.33% .647 **p < .01 90.177 52.78% 90.56% .800 *p < .05

.090+ .010 -2.283* 4.901** 48.541 52.78% 95.00% .894 +p < .10

-2.313** 4.953** 48.560 52.78% 95.00% .894





14 12 10 8 6 3 3 4


Std. Dev = 3.32 Mean = -1.56 N = 110.00

0 -4.50 -2.50 -.50

3.50 1.50

Population Deviation




9 4 4 2 4 4 2

Std. Dev = 3.04 Mean = 2.48 N = 70.00

0 -4.50


-.50 -2.50 1.50


Population Deviation


1.0 .9 .8 .7 .6 .5 .4 .3 .2 .1 0.0 -6 -4 -2 0 2 4 6
% GOP PSC2000 Rsq = 0.3478 Prob. GOP Majority Rsq = 0.9953

Population Deviation

To top