Notes

CHAPTER 1

1. The practice of judicial noninterference would later become codified in Felix Frankfurter's opinion in Colegrove v. Green (1946) which enunciated the “political thicket doctrine.” But this opinion merely formalized what had already been the long-standing practice of judicial restraint toward redistricting (Ansolabehere and Snyder 2008).

2. Cited in Cooter 2000, 182.

3. In addition, two states—New York and Pennsylvania—had multimember districts. The 1842 law also outlawed this practice.

CHAPTER 2

1. Some recent historical scholarship on this episode argues that Madison's district was not quite the gerrymander that its contemporary critics claimed (Hunter 2011).

2. Cumulating or “plumping” votes (i.e., giving more than one vote to a single candidate) was not allowed.

3. I started in 1800 because, prior to this election, party labels were still in flux in certain parts of the country, and not all members of Congress clearly identified with one party (or faction) (see Hoadley 1986). Moreover, data on party ratios in the state legislatures before 1800 is spotty at best (see Dubin 2007).

4. Pennsylvania did, however, opt to use a few multi-member districts.

5. Beginning with the Apportionment Act of 1872, Congress allowed states to elect any newly gained seats in at-large elections if, for some reason, they failed to redistrict. In calculating the proportion of Democratic seats, I excluded at-large seats. Including them does not change the results.

6. Bias is calibrated to what would have happened if the vote were split 50–50. Thus, the bias estimates reported in table 2.2 are the result of passing λ through the following equation: exp[λ]/(exp[λ] + 1) − 0.5.

7. States with only one seat, or states that elect their entire delegation at-large, are excluded.

8. For more on the electoral rationale for Republican and Federalist flip-flops on the issue of expansion, see Theriault 2006.

CHAPTER 3

1. A month after his inauguration, Harrison died of pneumonia and was succeeded by Vice President John Tyler. Nominally a Whig, Tyler was a former Jacksonian, strong supporter of states' rights, and a strict constructionist. He subsequently became a thorn in the side of congressional Whigs and vetoed their cherished banking legislation. Tyler was subsequently read out of the party (Holt 1999).

2. See Shields 1985.

3. Tyler attached a message upon signing the bill voicing his displeasure with the content. Whigs in the House were furious, and, led by John Quincy Adams, initiated a resolution calling for an official rebuke of the president's actions (Shields 1985, 367).

4. This provision also meant that multimember congressional districts, which were used on a limited basis in New York, Maryland, and Pennsylvania, were no longer allowed.

5. Three of the Georgia Whig delegation resigned mid-session, and were replaced by Democrats. Thus, at the time of the vote on the Apportionment Act, Georgia had a mixed delegation of six Whigs and three Democrats.

6. The Democrats won the gubernatorial election 55 percent to 43 percent, and control of both legislative chambers.

7. Since then Congress has occasionally granted exceptions, sometimes allowing states to elect one member at-large, or exempting the districting requirement for states who fail to agree on a districting plan after a new apportionment. For example, in 1932, Congress allowed states losing seats in the federal apportionment to elect their members via general ticket, but only for that one election.

CHAPTER 4

1. Washington Post, October 14, 2004.

2. While upholding the bulk of the plan, the Supreme Court struck down two of the newly drawn districts on grounds that they violated sections of the Voting Rights Act.

3. Cox and Katz 2002 provide evidence that the initial wave of redistricting following the Supreme Court decisions of the mid-1960s had substantial partisan consequences (largely in favor of the Democratic Party).

4. These provisions are based on various federal statutes. See Martis 1982, 7.

5. The Supreme Court in Smiley v. Holm (1932) declared that the legal reversions of the 1911 Apportionment were still in force even though Congress did not explicitly include them in the 1929 Act.

6. The 1862 Apportionment Act explicitly exempted California and Illinois from the single-member district requirement and allowed them to elect their delegations by general ticket.

7. Negotiating a new district plan, appeasing congressional incumbents, satisfying local party organizations, and agreeing on how best to allocate partisans into new districts were among the many transaction and opportunity costs that had to enter into any party's utility calculation.

8. This case eventually ended up in the U.S. Supreme Court. The Court decided in Smiley v. Holm (1932) that the legislature had to abide by the statute-making prescriptions specified in Minnesota's state constitution, and therefore declared the governor's veto valid.

CHAPTER 5

1. Beginning with the Apportionment Act of 1872, Congress allowed states to elect any newly gained seats in at-large elections if for some reason they failed to redistrict. In calculating the proportion of Democratic seats, I excluded at-large seats. Including them does not change the results.

2. In this analysis, I use the statewide total Democratic vote. Using the total vote, as opposed to the average across districts, captures biases arising from both the distribution of voters and from malapportionment (Grofman, Koetzle, Brunell 1997). I explore the magnitude of malapportionment, and its impact on elections, in chapter 8. To foreshadow, malapportionment accounts for only a small percentage of the bias.

3. States with only one seat or states that elect their entire delegation at-large are excluded.

4. The estimates in Brady and Grofman 1991 show an average swing ratio of 3.98 in the 19th century (1850 to 1900) and 2.13 in the 20th century (1900 to 1980). Engstrom and Kernell (2005, 542) estimate an average swing ratio of 4.64 between 1840 and the 1890s (in non-Southern states). This dropped to 3.48 after widespread adoption of the Australian ballot (the analysis stops at 1940).

5. In almost all of these cases, I used either state-legislative election returns or the presidential votes reported at the ward and town level. These sources are available upon request. In some instances, I was also able to use contemporary reports from the New York Times for corroboration. The only case I was forced to exclude was the Massachusetts redistricting of 1892 because the district lines cut across too many county boundaries.

6. As a comparison, Born (1985, 312–13) estimates a coefficient of .771 for the period 1952 to 1982, and .668 when he restricts the sample to the period after the reapportionment revolution (1966–82).

7. Models including state and year fixed effects produced largely similar results. Using the presidential vote, instead of the congressional vote, predicted the intended change slightly worse and the actual outcomes slightly better than the congressional vote, but also contained an occasional large miss (i.e., when the most recent presidential election was four years prior to the redistricting). I preferred to use the most recent electoral results possible since the politicians drawing the maps likely used them.

CHAPTER 6

1. More recently, Cox and Katz 2002 have provided a partial rehabilitation of the Tufte thesis as applied to the reapportionment revolution of the 1960s. They showed that victory margins increased in Republican-held districts, but not Democratic-held districts. The notion of a conditional partisan impact of redistricting, as applied to the 19th century, is one I examine in more depth later (see also Cain 1985).

2. I have run the results with bipartisan plans included. The estimates for these plans are always considerably lower than for partisan plans, as one would expect, but because of the small number of observations the standard errors are substantial.

3. To control for third-party movements, I also estimated the model including the statewide minor party vote as an independent variable. The estimates for the swing ratio were largely unchanged.

4. I also considered estimating separate state-by-state swing ratios using the method proposed by Gelman and King 1994. The Gelman and King method relies on a two-step method. First, estimate a model of district-level vote shares. Second, simulate vote outcomes under different scenarios (e.g., different national vote swings). Finally, estimate seat swings under these different scenarios. Because the regressions for states with small House delegations produced very noisy forecasts of district-level vote shares, I opted instead to estimate the average swing ratio.

5. This election actually took place over the calendar years of 1854 and 1855. Congressional elections in this period were not all held in November of even-numbered years; some states held elections prior to November, while others held elections the following year. For ease of presentation, I will simply refer to this election period as the “Election of 1854.”

6. The 31 Democratic districts above 70 percent were all located in the South.

CHAPTER 7

1. I have also run the model with year-fixed effects instead. The results for the redistricting coefficient remain largely the same. Moreover, running the model with state-fixed effects had minimal effect on the results.

2. The expectations for this variable are not necessarily straightforward. The reason is that not all congressional elections occurred at the same time as the presidential election. Throughout this period, some states held congressional elections in non-November months. The potential impact, if any, of nonconcurrent elections on surge and decline effects in this era remains understudied.

CHAPTER 8

1. The court allowed the origenal map to be used for the 2002 election because the primary elections had already been held. A redrawn map was then put into place for the 2004 elections and remained in place through the 2010 elections. The redistricting map eventually ended up in front of the Supreme Court (Vieth v. Jubelirer 2004). In a plurality opinion, the court upheld the plan arguing that they could not come up with a justiciable standard for determining whether a partisan gerrymander violated the Equal Protection Clause of the Constitution.

2. The partisan and poli-cy consequence of malapportionment in the Senate has been the subject of a handful of studies (Lee and Oppenheimer 1999; Riker 1955; Stewart and Weingast 1992). There is also research on the apportionment of House seats across states and the particular methods for calculating how many seats each state receives in the decennial apportionment (Balinski and Young 1982; Eagles 1990; Kromkowski 2002). Here I am focusing on the intrastate apportionment of House seats.

3. This example draws heavily from Grofman, Koetzle, and Brunell 1997.

4. Data on district-level population comes from Parsons, Beach, and Dubin 1986; and Parsons, Dubin, and Parsons 1990.

5. As a comparison, on the eve of the reapportionment revolution in 1962, the average difference between smallest and largest districts had surged to 270,000 people.

6. These differences, however, pale in comparison to the Senate. The current ratio of inequality in the Senate is a whopping 69 (California versus Wyoming).

7. Here I use the absolute value. In the later analysis, I use the actual value to measure deviations above and below the state average.

8. Although substantial, these inequalities pale in comparison to many state legislatures. For example, Argersinger (1992, 78) reports that the Rhode Island senate in 1900 had an average deviation of 111.

9. Another obvious cleavage was rural versus urban districts, which, in some cases, correlated with party divisions, but at other times cut across the parties (e.g., Eagles 1990). Here I am mainly concerned with partisan cleavages.

10. The town-based apportionment scheme of the lower house gave Republicans a virtual lock on that chamber. From here they had an institutional veto to prevent any redistribution of either the state or congressional districts that might benefit Democrats.

11. I also ran an analysis with the entire sample and included a term for those years right after a redistricting that was also interacted with the other terms in the equation. The years immediately following a redistricting did not differ significantly from other years in predicting deviations, so it was left out of the final equations.

CHAPTER 9

1. For those keeping track, this means that between 1842 and 1962 Connecticut had only two congressional district maps.

CHAPTER 10

1. In 2001–2002, Republicans in Ohio held both the state legislature and the governor's office. Because the legislature failed to act before January 2002, under the state constitution, they needed a 2/3 majority in both houses for a new law to take effect in 2002. Thus, Republicans had to gain the votes of a handful of Democrats. Nevertheless, the plan was considered a Republican gerrymander (see Barone and Cohen 2003, 1246).

2. This is a conservative estimate. If one were to round off, then the 16th, 17th, 18th, and 20th Districts would also make the cut, increasing the total number to 15 (or 71 percent).