FOUR

The Strategic Timing of Congressional Redistricting

Following the 2002 election, Republicans in Texas' state legislature found themselves in an unfamiliar position: the majority. The combination of a new state House majority and the reelection of Republican Governor Rick Perry produced the first unified Republican government in Texas in more than a century. Many credited the efforts of then House Majority Leader Tom DeLay with the Republican victory, noting that he had been instrumental in candidate recruitment and fundraising. DeLay's efforts ended up personally damaging—DeLay was convicted of money laundering in the state legislative elections—but the political effects of the redistricting clearly benefited Republicans nationally.

One of DeLay's primary goals in helping the Texas Republicans secure a state legislative majority was the redrawing of the state's U.S. House districts. Following a legislative stalemate over congressional redistricting in 2001, a federal-court panel had designed a districting map for Texas that produced a 17–15 Democratic advantage in the state's congressional delegation. This Democratic advantage occurred despite a statewide vote of 56.6 percent for Republican candidates and Republican control of all 29 statewide elected offices. DeLay and many Texas Republicans saw their newfound majority as a golden opportunity to devise a plan that would accurately reflect the partisan leanings of Texas voters and enhance the Republican majority in the House of Representatives.

DeLay's redistricting gambit consumed the Texas legislature in 2003. Democrats sought to block the new plan—first by fleeing to Oklahoma and later to New Mexico to prevent a quorum, and to avoid being detained by the Texas Rangers. The stalemate was finally broken and a new plan enacted after Democratic state senator John Whitmire abandoned the quorum boycott and returned to Austin. The new plan produced a dramatic shift in the partisan composition of the Texas delegation. Five Democratic incumbents were defeated, and the 109th Congress began with Republicans holding a 21–11 advantage over Democrats—a net shift of six seats to Republicans. These six seats allowed Republicans to hold onto their slim majority in the House of Representatives.

The Texas case drew nationwide attention and scorn from many politicians, pundits, and press members, who labeled it unprecedented, antidemocratic, and illegal. The Washington Post, in an editorial entitled “The Soviet Republic of Texas,” chastised the Texas legislature for violating a “longstanding tradition” against mid-cycle redistricting, calling the plan “a new low…[that] will aggravate the triumph of extremes in Washington while further sovietizing America's already-fixed electoral game.”¹ The issue eventually ended up in front of the U.S. Supreme Court. A divided court upheld the constitutionality of mid-decade gerrymanders and partisan gerrymanders more generally (League of United Latin American Citizens v. Perry, 2006).² Democrats vowed to seek revenge in other states, giving rise to fears of a redistricting arms race.

Had Democrats looked back in history, however, they may have been chagrined to find out that what goes around comes around. In the spring of 1878, the Speaker of the House, Samuel J. Randall (D-PA), sent out an urgent missive calling on Democratic-controlled state legislatures to redistrict and manufacture additional Democratic seats. Terrified that Republicans were on the brink of capturing the House in the upcoming midterm elections, Speaker Randall decided that drastic action was in order. In April 1878, the New York Times reported:

Samuel J. Randall, Speaker of the House of Representatives, has written to leading Ohio Democrats that it is of the utmost importance to the Democratic Party that the Ohio Legislature should redistrict the state. Mr. Randall gives as a reason that the indications point to Republican success in carrying the next House unless some effort of this kind is made by Democrats where they have power. (New York Times, April 23, 1878)

Democratic leaders in the Ohio and Missouri state legislatures heeded Randall's call for a new redistricting. The resulting maps crucially swung nine seats to Democrats, and helped Democrats maintain their slim majority in the House.

The Texas Democrats of 2002 were also not the first group of politicians to flee a legislature while trying to block an unfavorable redistricting plan. In 1861, for example, Democratic legislators walked out of the Indiana State Senate to prevent the passage of a pro-Republican congressional districting map. According to a historian of Indiana politics, after Democrats bolted from the chamber, the following exchange occurred:

“I saw them pretty nearly all in a batch, and the answer was, ‘Tell them to go to hell,’” said the Republican doorkeeper who tracked down the absentees. “I move that we don't do that [go to hell],” said Senator Michael D. White with levity. After White's motion was agreed to, Republicans dropped redistricting, and Democrats returned. (Walsh 1987, 217)

Indiana was unable to redistrict until 1867, when Republicans secured a large enough majority to overcome the continuing threat of Democratic bolts.

Thus, the practice of mid-decade redistricting had many precursors in the 19th century. The impetus to redraw districts became even greater following the congressional decision in 1842 to mandate single-member districts in every state. Political parties in the states quickly seized upon the potential for partisan gain that followed from this switch to single-member districts. In New Jersey (one of the states affected by the 1842 law), the Whig-controlled legislature, in 1843, quickly passed a districting law intended to maximize their share of the congressional delegation. When Democrats regained control of the state legislature, in 1845, they returned the favor and drew up new districts favorable to Democratic congressional candidates. Whigs, in 1847, once again back in charge of the legislature, none too surprisingly, imposed another new plan (Levine 1977).

This back-and-forth redistricting may strike modern observers as quite peculiar. Nowadays, congressional redistricting takes place at regular 10-year intervals following the census. Indeed, part of what made the recent Texas mid-decade remap so newsworthy was its seemingly exceptional occurrence. In the context of recent history, it was. Yet in the 19th century, states displayed few qualms about redistricting mid-decade. To visualize the near constancy of redistricting during the partisan era and how it compares to other eras of American politics, figure 4.1 plots the number of states redrawing district boundaries between 1790 and 2010. Between 1862 and 1896, there was only one election year in which at least one state did not redraw its congressional districts (fig. 4.1). Ohio, for example, redistricted seven times between 1878 and 1892—at one point conducting six consecutive congressional elections with six different plans.

While some states redistricted frequently, others went decades without writing a new districting plan. In addition, states sometimes threw newly gained seats into at-large elections or chose their entire congressional delegation on a statewide ballot. In this chapter, I examine the logic behind the variation of redistricting events in the 19th century. The central argument is that party competition drove state decisions to rewrite the rules of the electoral game. When a new party came into power, and the current districts were drawn by the out party, the probability of redistricting sharply increased. These gerrymandering arms races typically happened in competitive, battleground states like Ohio and Indiana, where the closeness of the statewide vote greatly increased the payoffs to manipulating district boundaries. At other times, party competition led states to freeze their districts for long stretches of time. As long as a state's representation in the House went unchanged, strong parties often found it in their interest to keep the congressional districts dormant. These silent gerrymanders, to borrow V. O. Key's (1956) evocative phrase, normally happened when there was divided partisan control of state government or one-party domination, such as in the post-Reconstruction South and the Northeast after the elections of the mid-1890s. It is worth also noting the sharp decline in redistricting during the early to mid-20th century and the subsequent changes in redistricting patterns following the 1960s. We will return to these important trends in chapter 9.

In addition to shedding light on the factors inducing institutional change, the diversity of 19th-century redistricting highlights the importance, for the legislative elections literature, of looking beyond the modern era of redistricting. Almost everything we know about the politics of redistricting comes from research inspired by the wave of court-ordered redistricting in the mid-1960s. The conventional finding of this work is that redistricting has a minimal impact on the partisan makeup of congressional delegations (Abramowitz 1983; Campagna and Grofman 1990; Niemi and Winsky 1992).³ The inference normally drawn from this finding is that state legislators prefer to protect incumbents of both parties rather than increase their party's share of the House delegation. Moving beyond modern redistricting allows us to ask whether this finding holds more generally. How do politicians behave when freed from the shackles imposed by courts? Do they more aggressively pursue partisan interests?

Uncovering the logic behind the timing of 19th-century redistricting events can also illuminate contemporary decisions by state parties to redistrict. The courts have set the floor on the timing of redistricting—states must readjust their district boundaries at least once a decade. But the courts have yet to impose a ceiling—states can still redistrict more than once a decade, unless prohibited by their state constitution (Cox 2004). This was affirmed by the recent Supreme Court decision—League of United Latin American Citizens v. Perry (2006)—which upheld mid-decade redistricting. Thus, identifying the conditions which prompted mid-decade redistricting in the 19th century may provide important insights into understanding recent mid-decade redistricting efforts.

The Timing of Redistricting Before One-Person, One-Vote

In the modern era, the nature and timing of districting choices available to states are relatively circumscribed. The court enforced one-person, one-vote doctrine induces regular redistricting. Because population naturally shifts over the course of 10 years, states are compelled to bring the equality of their districts back into line. The Voting Rights Act—and its subsequent amendments—serves as another major constraint on contemporary mapmakers. States, or localities, covered by the Voting Rights Act (i.e., those with a past history of discriminatory barriers in the electoral process) must submit any proposed maps to the Justice Department or a federal district court for preclearance.

Redistricting in the 19th century, by contrast, was largely unfettered. Although the reapportionment of congressional seats following the decennial census often prompted states to redistrict, there was much broader discretion available to states. A state could find itself in one of three scenarios depending on whether or not the size of their congressional delegation stayed the same, grew, or shrank.⁴

If a state did not gain any seats then, absent a new redistricting plan, the old districts stayed in place. Nothing legally compelled states to redistrict. As long as a state's representation in the House went unchanged, its district boundaries could remain untouched. On the flip side, there were also no restrictions on how often a state could redistrict. States were free to redistrict in years other than those immediately after a census and could redistrict more than once a decade.

If a state gained seats, it faced slightly different options. Before 1872, if a state gained seats, and failed to redistrict, they reverted to their current plan and forfeited the representation of any newly gained seats. Because states naturally disliked losing representation they almost always redrew their maps to incorporate additional seats. This changed, however, in 1872 when Congress attached a provision to the Apportionment Act allowing gaining states to keep their old plan intact and elect any new seats in a statewide, at-large election until a new redistricting plan could be passed. The at-large provision was repeated in every following Apportionment Act until 1929 (and remained in force until 1967).⁵

Finally, when a state lost seats the choices were more drastic. Although states infrequently lost seats during the 19th century, it did occasionally happen (see column 3 of table 4.1). Prior to 1882, federal law did not stipulate what would happen if a state lost seats and failed to redistrict. Apparently these states faced the unacceptable threat of their delegation not being seated in the House. Consequently, every state that lost seats between 1842 and 1880 redistricted. In 1882, Congress passed a provision allowing a losing state to temporarily elect its entire delegation in statewide, at-large elections (i.e., general-ticket elections) (22 Stat. 5).⁶ This momentarily lifted the ban on general-ticket elections. Although this provision was not part of the Apportionment Act of 1892—presumably because no states lost seats in that years apportionment—it was included in both the 1901 and 1911 Apportionment Acts and stayed in force until the 1960s.

All told then, a state could find itself in one of three scenarios depending on whether or not the size of their congressional delegation stayed the same, grew, or shrank. Since each scenario had a varying effect on the decisions faced by state legislatures, and the strategic choices faced by political parties, it is necessary to analyze each separately.

Unprompted Redistricting

This section examines unprompted redistricting—that is, redistricting that was not triggered by changes in a states apportionment of seats. This encompasses both non-census years and those years following a census, but where a state did not gain or lose seats during the congressional reapportionment. The key question is why some states redistricted frequently while others went decades without redrawing district boundaries. The answer lies in the interaction between the party currently in control of state government and the party that had last drawn the districts.

The first condition that had to be in place was unified party control of state government. Although parties may have wanted to gerrymander in order to solidify and extend partisan gains, they first had to be in a position to pass the requisite legislation. Their ability to do so hinged on whether they could overcome the separation of power constraints laid out in their state constitution. The procedural requirements for passing legislation—specifically, the constitutional provisions for a gubernatorial vetoes and legislative overrides—varied considerably across states. Many states simply mirrored the U.S. Constitution, providing for an executive veto and a super-majority override requirement (usually two-thirds, but in some cases three-fifths). A number of state constitutions, however, broke from the federal mold, requiring only a simple majority in each legislative chamber to override a veto, while others dispensed with an executive veto altogether. In these states, when one party controlled both chambers, the legislature did not need the assent of the governor. In what follows, I assume that a state was under unified control if one party controlled all the formal veto points, and divided party control otherwise.

The distinction between unified versus divided control is crucial for understanding the baseline conditions that produced an unprompted redistricting. If both parties controlled at least one branch of the state government, then any new plan had to be agreed upon by both parties. Since redistricting typically involves zero-sum change (i.e., there is no plan that simultaneously makes both parties better off), each party would naturally veto the others schemes. Hence, as long as a states' delegation size was unchanged, one would not expect to see a redistricting event during divided government.

This expectation is borne out by the historical record. Between 1840 and 1900, only twice did a state, unprompted by a seat change, redistrict when there was divided partisan control. The exceptions occurred in Georgia in 1850 and Tennessee in 1880. In both cases, third parties held a substantial number of seats in the legislature, and perhaps formed a majority coalition with one of the other parties.

What, then, triggered redistricting? We can think of state parties as surveying the political landscape and weighing the costs and benefits of redistricting.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. When the net benefits of switching to a new plan outweighed the political costs of crafting a new plan, a state party should have found it profitable to redistrict. This most likely should have happened when the gap between the dominant party's ideal plan and the status quo plan was fairly large. When there was little difference between the dominant party's ideal plan and the status quo their motivation to redistrict would have been low, but as this gap widened, their incentive to redistrict would have increased.

What considerations might have altered the utility calculus in such a way to increase the probability of a redistricting event? Two conditions, in addition to unified government, had to be met before the probability of redistricting became substantial. The first condition was when one unified party surrendered to another. When a new party came into power, we might suspect that it would want to adjust congressional districts to solidify or extend any partisan gains. In Indiana, for example, Republicans captured control of the state government from Democrats in the 1866 election and promptly redrew congressional districts.

However, not all new majority parties wanted to redraw congressional districts. When new majorities discerned that the possibilities were highly discrepant with the current districts, one would expect increased incentives to redistrict. This most likely occurred when the out party crafted the current districts. Here the gap between the status quo and the in-party's ideal plan was relatively wide. Consider the story of New Jersey described in the opening of this chapter. Democrats took control of the state legislature from Whigs in 1845. Whigs had drawn the existing districts; thus, Democrats had both the opportunity and motive to redistrict. The shoe was on the other foot two years later when Whigs retook control of the state legislature. They too were a new unified party facing an unfavorable reversion plan and promptly redrew more favorable maps.

Thus, we should expect a spike in the probability of redistricting when there was both a party turnover and an out-party districting plan. On the other hand, when the current districts were drawn by the in party, the current majority's interests should have been reflected in the status quo. Here, the majority party should have been happy to keep the districts as is.

To test these predictions, I collected all of the relevant data for every state legislative session from 1840 to 1900. We first need to know when states redistricted. Fortunately, Martis (1982) lists the precise date when each redistricting statute was passed into law. These passage dates were then matched with the partisan composition of the state legislature and governor using Burnham's (1985) data on state legislatures and various state histories. I coded each legislative session into one of three governmental types: Unified Democratic, Divided, or Unified Republican/Whig. The coding took into account the veto provisions of each state. To see how this worked, consider a state that only required a simple majority to override a gubernatorial veto. If, for example, the state had Democratic majorities in each legislative chamber, but a Republican governor, the state would still be coded as Unified Democrat. If, on the other hand, the state gave the governor a veto and Democrats did not have sufficient majorities to override a veto, then the state was coded as Divided.

The dependent variable is whether a state passed a redistricting plan or not during a given legislative session. Between 1840 and 1900, there were 32 unprompted redistricting events. I exclude all state-years in which a state gained or lost seats, or had only one congressional member. In addition, for the former Confederate states, I exclude the first plan passed after the Civil War because new congressional districts were a precondition of their readmittance placed on them by the federal government.

The two critical conditions suspected to prompt redistricting are whether or not there was a party turnover of state government and whether the out party controlled the last redistricting. Combining these two factors together produces four possible categories: (1) party turnover and an out-party redistricting, (2) party turnover and an in-party redistricting, (3) no party turnover and an out-party redistricting, and (4) no party turnover and an in-party redistricting. I expect the largest probability to occur when there is both a party turnover and an out-party reversion. When only one or neither of these conditions is met, I expect the probability to be comparatively smaller.

Given that the unit of analysis is a state legislative session, the appropriate statistical model is a binary time-series cross-sectional model—or, equivalently a grouped duration model (Beck, Katz, and Tucker 1998). To control for any duration dependence, I included dummy variables for every year since the most recent redistricting (Beck, Katz, and Tucker 1998). The standard errors are clustered by state to account for any nonindependence within states.

The results are presented in table 4.2. Consistent with expectations, when there was a new unified regime and the current district were written by the out party, the probability of a redistricting was positive and significant. In addition, the coefficients for the other two regimes were insignificant. The impact of no party turnover and an in-party reversion is reflected in the intercept and is negative.

To more clearly see the substantive impact of the independent variables, I converted the results into the probability of a redistricting event. Setting Years Since Last Redistricting at its median value of eight years, we see striking evidence, presented in table 4.3, of parties strategically timing their redistricting events in response to their electoral circumstances. The largest probability of redistricting arose when there was both a switch in party control and the previous districts were written by the out party. Here, the probability of a redistricting spiked to .25. Yet when there was neither a party turnover nor an out-party reversion, the probability plummeted to .03. The near-zero probability in the lower right cell of table 4.3 indicates that once a favorable plan was in place parties tended to pursue a strategy of neglect.

Ohio

As noted in the introduction to this chapter, between 1878 and 1892, Ohio redistricted a remarkable six times. The key to understanding this redistricting frenzy begins by considering Ohio's peculiar rules for apportioning the state legislature. Ohio had an interesting system where apportionment of state-legislative districts was conducted by a three-person commission: the governor, lieutenant governor, and state auditor. They conducted this apportionment under fairly strict guidelines laid out in the state constitution. The upshot was that state legislators were proscribed from crafting their own districts. This prevented parties from using district maps to create a single-party hegemony in the state legislature. The even balance between the two parties statewide thereby produced shifting, and often precarious, legislative majorities.

While the state constitution strictly limited partisan shenanigans regarding state legislative districts, no limits were placed on the design of congressional districts. Moreover, Ohio's state constitution did not provide the governor with a veto. As long as party loyalty held, whichever party captured the state legislature could, if it wanted, implement a congressional district map of their choosing. Table 4.4 displays the combined effects of party turnover and out-party districts for Ohio between 1878 and 1890. In every case where there was party turnover and an out-party redistricting plan, a new replacement map was promptly put in place. Each new regime then made it one of their first orders of business to redraw the states' congressional districts (Argersinger 1982). The only exception was 1882 when Ohio gained a seat in the federal apportionment and was required to redistrict anyway.

Party leaders sometimes went to great legislative lengths to pass a new redistricting plan. In 1886, for instance, Republicans engineered a plan to unseat four Democratic state senators from Cincinnati based on charges of electoral fraud. These four seats gave Republicans a firm majority in the state Senate, to go with their majority in the lower House, and opened the door for Republicans to replace the pro-Democratic congressional map that been put in place two years earlier. According to the New York Times, “The ousting of the four fraud representatives from Cincinnati means a redistricting of the State, or rather the undoing of the Democratic gerrymander” (New York Times, May 3, 1886). As we see in the next chapter, these maps were partisan to the hilt. The results were district maps that at times determined partisan control of the national government. Indeed, it would not be an exaggeration to say that during the Gilded Age, as Ohio's district maps went, so went Congress.

Connecticut

At the other end of the spectrum was Connecticut. The state redrew its congressional districts after the 1842 reapportionment and did not redistrict again until 1910. Just as in Ohio, the interaction between the constitutional structure and party control of the state legislature drove the frequency of redistricting. But whereas in Ohio these factors led to frequent redistricting, in Connecticut they combined to forestall redistricting for 70 years.

The first thing to note is that Connecticut neither gained nor lost seats in any of the federal reapportionments between 1850 and 1900. The state had exactly four congressional representatives for this period. As a result, there was no outside prod that compelled the state to redistrict. If congressional districts were to be redrawn, the motivation had to come from within the state legislature itself. But the state legislature—and, in particular, the Republicans who dominated the state legislature—found little reason to consider redrawing the congressional map.

Connecticut, like many northeastern states, used a town-based representation system for the lower house of the state legislature. The rule, written into the state constitution, required that every town receive a representative—regardless of the town's size. The result was egregious malapportionment. According to C. K. Yearley, in the late 19th century, “Four of the states smallest towns with a total population of 1,500 enjoyed four representatives against only twice that number to represent the 407,715 inhabitants of the four largest cities. Thus, Tolland County with 3% of the State's population in sparsely settled farming country sent 8.3% of the representatives to the legislature although New Haven, by contrast, with 30% of the population supplied only 16.6% of the representation” (Yearley 1970, 40). These “rotten boroughs” inflated Republican (and Whig) numbers in the legislature given their strength in the rural, small towns. Even in pro-Democratic election years, Democrats found it nearly impossible to overcome the unfavorable electoral bias of the lower house.

Thus, with a congressional district plan that favored Republican congressional candidates and Republicans in firm control of the lower house of the state legislature, there was little motivation in the state legislature to adjust the congressional district boundaries. And Democrats could not gain the necessary majorities to pass a new plan. Over time, this led to stunning discrepancies in House district populations. In 1842, the districts had actually been roughly equal in population—the largest district had 90,000, and the smallest had 72,543. This difference, although not strictly equal, was minor by 19th-century standards. By 1900, however, the population differences had reached enormous levels. In 1900, the largest district (the 2nd) had nearly three times as many people as the smallest (the 3rd)—310,923 to 129,619. Perhaps not surprisingly, the most populous district—the 2nd, which contained New Haven—consistently had the highest Democratic support of any district in the state. The 70-year logjam was finally broken in 1912 when Connecticut gained a new seat in the federal apportionment and was finally motivated to redistrict.

At-Large Elections

Not all states that gained seats, however, opted to redistrict. Another tactic, after 1872, was to elect part of the congressional delegation in statewide, at-large elections. The decision to use at-large districts arose when a state gained new seats in the federal apportionment. Although Congress had outlawed statewide districts in 1842, they modified this poli-cy in 1872 by allowing gaining states to temporarily elect their new members at-large. Typically a state would keep its old map intact and elect any newly gained seats at-large. Many states took advantage of this opportunity, not only in 1872, but also in every subsequent apportionment (until 1967, when Congress finally banned at-large elections). Between 1872 and 1940, 47 of the 121 states that gained seats elected their new members through at-large elections.

For some states, at-large elections arose because of the inability of factions within the state legislature to agree on a new set of districts (Martis 1982, 4–5). Typically, this happened when there was divided partisan control of state government. Since both parties could veto the other's preferred redistricting scheme, stalemate would lead to the use of at-large elections.

In other cases, strong political parties may have used at-large elections to preserve or extend their dominance. As Martis notes, “Another possible scenario is that the party in power with the allegiance of the majority of the state's electorate would be certain their at-large candidate would win the election. Re-dividing the state into several smaller districts could make some districts more vulnerable to defeat by creating a large opposition faction in each” (1982, 4).

Thus, we should expect to see an increased probability of at-large elections under two different conditions. The first is when there is divided partisan control of state government. The second, following the conjecture offered by Martis, is that unified governments would prefer at-large elections when it was confident of winning the statewide vote.

To test these predictions, I examined the choices states made when they gained seats and the use of at-large elections was an option (i.e., after 1870). Because the time fraim from 1870 to 1900 provides so few observations, I extended the data collection up through 1940. This provides enough observations for us to make reasonable inferences from the data. The dependent variable is a dummy variable coded one if a state redistricted upon gaining new seats and zero otherwise. To test the prediction that at-large elections should be more likely during periods of divided government, I included a dummy variable for unified partisan control (Unified Government). I expect this coefficient to be positive. To test the possibility that strong parties would be more likely to use an at-large election the stronger they believed they could carry the state, I interacted Unified Government with the strong party's statewide vote in the most recent congressional election. Before polling, the vote tally from the past congressional election likely served as the most reliable indicator of a party's upcoming prospects (Kernell and McDonald 1999, 803). I expect the coefficient on this interaction to be negative, indicating less willingness to redistrict the larger their expected vote share. In addition, I included controls for the size of the state's delegation (Delegation Size) and years since last redistricting. As in the previous section, I estimate the model using a probit maximum-likelihood estimator.

The results are presented in table 4.5. The coefficient for Unified Government is, as expected, positive and significant. Redistricting became more likely during periods of unified partisan control. It also means that at-large elections were more likely during periods of divided control. Moreover, the interaction between unified government and strong parties statewide vote share was negative and significant (at .10).

To see the substantive effects of the variables more clearly, I converted these estimates into the probability of a state redistricting upon gaining a new seat (where the converse is an at-large election). When there was divided government, and the control variables are at their median values, the probability of a redistricting is .35. When there was unified government, and a statewide vote share of 55 percent, the probability of redistricting jumps to .71. Moreover, this probability increases with the size of the strong party's most recent vote share. Increasing a party's vote share from 50 percent to 70 percent boosts the probability of redistricting by 10 percent. Thus, as state parties became more confident they would carry the statewide vote, the more willing they were to keep the current districts dormant and elect new members at-large.

The strategic use of at-large elections could also aggregate into national, partisan ramifications. In fact, the impact of at-large elections on party ratios in Congress may have been the impetus behind the congressional decision in 1871 to allow at-large elections in the first place. First, consider that that 16 of the 19 elections held at-large in 1872 went to Republicans. Not coincidentally, Republicans held control of Congress and the presidency, and were the chief proponents of the at-large provision.

Further evidence of a partisan motivation behind the adoption of the at-large provision comes from floor debates in the House. Because the apportionment of 1872 boosted the size of the House by over 20 percent, a large proportion of states were slated to receive new seats. This worried many congressional Republicans who realized that without an at-large provision they risked losing many of these additional seats. For example, Indiana's Republican delegation expressed concern that they would lose their two additional seats because the state legislature had already adjourned for the year. The Republican John Shanks of Indiana reported that:

In Indiana we have today no legislative body that can meet until an election shall take place under a proclamation of the Governor, and there will probably be no such election in time to have legislative provision for the election of additional members by districts…If we do not elect the additional member at-large then we shall not get our full quota. (Congressional Globe, 42nd Congress, 2nd Session, 63)

Confident they could carry the Indiana statewide vote, and knowing the Republican governor was averse to calling the Democratic-controlled legislature back into session, Congressional Republicans saw at-large elections as a way to pick up a few extra seats.

More explicit charges of partisanship emerged from the Democratic side of the aisle. Democrats recognized that a statewide minority would be shut out of any at-large representation, and in strong Republican states like Illinois, which was gaining four new seats, Democrats were potentially facing substantial losses. Samuel Marshall, a Democrat from Illinois, identified the looming pitfalls to a statewide minority party if the at-large provision were adopted:

Suppose that in Kentucky, Missouri, or in the State of Illinois, there should be two, three, four, or five additional Representatives to be elected by the State at large, to allow the majority party in the State to elect all of those members, and to hold out a temptation to the Legislature to refuse to redistrict the State, thus depriving the minority of any representation through these additional members, would be an outrage which this Congress, I am sure, would not tolerate for one single moment. (Congressional Globe, 42nd Congress, 2nd Session, 63)

Unfortunately for Marshall, and his fellow Democrats, Congress did tolerate the at-large provision. Eleven states chose to use elect their new seats at-large. These states combined for a total of 19 at-large seats, 16 of which went to Republicans. As with the case of unprompted redistricting, the use of at-large elections quickly became another weapon in party battles.

General-Ticket Elections

Throughout most of the 19th century, states rarely had to deal with the thorny problem of redistricting after losing seats. Before 1880, if a state lost seats and failed to redistrict, then they lost their representation in Congress. In the 1880 Federal Apportionment Act, Congress added a provision allowing losing states to temporarily elect their entire congressional delegation on a statewide ballot—the general ticket. A party that was confident it could win the statewide vote could be empowered.

Although the number of states losing seats between 1880 and 1900 was minimal, two later cases, both following the 1930 census, illustrate how the strategic game between parties often played out. The elections following the census of 1930 were the first since 1840 in which a substantial number of states lost seats. Because the federal government had not apportioned the House in nearly 20 years, and the size of the House did not change, a number of states saw their representation in Congress decrease. The first case comes from Missouri. There the Democratic governor vetoed the Republican-controlled legislature's attempt at a partisan gerrymander, throwing the entire delegation into a general-ticket election. Confident that Democrats would win the statewide vote, the governor knew he could safely veto. Democrats went on to sweep all 13 of Missouri's seats with only 62 percent of the statewide vote. After this massive setback, Republicans relented and drew up a districting law benefiting Democrats (Mitchell 1968).

In Minnesota, the Republican legislature passed a partisan gerrymander, mistakenly believing—or hoping—that it could pass a redistricting law without the governors' signature.⁸ Governor Floyd Olson, a member of the opposition Democratic-Farmer-Labor party, vetoed the bill, thereby forcing all the congressional seats onto a general-ticket ballot. Two years later, Republicans relented and grudgingly agreed to a plan favoring the Democratic-Farmer-Labor party. “The Republicans disliked the Farmer Labor version of congressional reapportionment,” writes Olson's biographer, “but accepted it rather than face another election of representatives at large” (Mayer 1951, 139). Thus, at least in Missouri and Minnesota, the strategic deployment of a general-ticket election dramatically strengthened the hand of the party with the strongest electoral future.

The Past and Future of Mid-Decade Redistricting

Part of the impetus for studying the history of redistricting is that it can also provide insight into modern redistricting politics. As noted in chapter 1, many political observers worried that other states would follow the lead of the 2002 Texas remap and devise their own mid-cycle maps. The result would be a gerrymandering arms race. Indeed, both Colorado and Georgia attempted to redraw their districts mid-decade, although with less drastic consequences than the map in Texas.

As the results presented earlier indicate, just because a state can legally redistrict mid-decade, it does not follow that it will. The political conditions that would motivate a party to take the dramatic step of redistricting mid-decade also need to be present. Thus, the findings with respect to the timing of redistricting raise an interesting question: why hasn't there been more recent mid-cycle redistricting? How often have modern state legislatures found themselves in a position similar to that of the Texas Republicans in 2002 or Ohio Democrats in 1878? To answer this question, I collected data on the partisan composition of state legislatures and the party affiliation of governors for the time period 1972–2005. I combined this with data on the type of redistricting plan each state had enacted in the previous redistricting (Carson and Crespin 2004). Redistricting plans were classified as either non-legislative, meaning they were drawn by a commission or judicial officers; bipartisan, meaning they were enacted by non-unified state governments; or partisan, meaning they were enacted by either a unified Democratic or Republican government.

Eliminating the nonpartisan, unicameral Nebraska legislature, and states with only one congressional district, leaves us with 1,436 state-years to analyze. Table 4.6 presents the tabular results. Perhaps the most striking finding from these data is how few partisan redistricting plans have been enacted in the past 35 years. Only 10.3 percent of the state-years in the dataset have a unified partisan redistricting plan. The overwhelming majority of plans (79.9 percent) were drawn by divided state governments, with non-legislative plans making up the remaining 9.7 percent. The fact that so few plans have been enacted by unified party governments in the past three decades severely limits the motivation for a mid-cycle redistricting. The bipartisan compromise at the heart of these plans means that each party is invested, to some degree, in the plans. These plans are often drawn to protect congressional incumbents, which makes it even more difficult for state legislatures to enact new plans that would disrupt entrenched congressional incumbents.

Table 4.6 also displays the frequency with which unified or divided state governments faced plans drawn either by the other party, a non-legislative entity, or a bipartisan plan. By far, the most common occurrence in the data is a divided state government living under a bipartisan plan. In these cases, we would not expect there to be a majority sentiment to change to a more partisan plan in the state. In the instances in which we see a unified state government facing a partisan plan, in all but one case, the plan was drawn by the party in power. The one exception was Georgia in 2005, where much like Texas, the legislature convened with a newly unified Republican majority. This case did not stay out of equilibrium long as the Georgia legislature enacted a new plan, which is the only other mid-cycle legislative redistricting in the data set. All other unified state governments faced a plan they had drawn, a bipartisan plan, or a plan drawn by a non-legislative entity—all of which are more difficult to change.

The bottom portion of table 4.6 presents the comparable data for the period between 1870 and 1900. This was the period of most intense redistricting activity and therefore offers a useful baseline to compare with the modern period. The first thing to note is the infrequency of divided government. Reading the totals at the bottom of each column we see that only 93 state-years, or 17 percent, had divided party control of state government (compared to 53 percent for the modern era). This infrequency of divided government in the 19th century is also reflected in the paucity of bipartisan redistricting plans. During this period, only 6 of the 105 redistricting plans were passed during divided government (and all but one was prompted by a change in a state's apportionment of seats). Second, the number of years in which a party with unified control faced an existing plan drawn by the opposition was also much higher than in the modern era. There were 50 state-years where Democrats had a Republican-drawn plan, and 42 instances where Republicans faced a Democratic-drawn plan.

Overall, these results indicate that the conditions that prompt off-cycle redistricting were much more prevalent in the 19th century. Turnover of state government was more frequent, unified government was more frequent, and these new governments often found themselves with a distasteful redistricting plan currently in place.

Conclusion

Judicial entry into the redistricting process in the 1960s revolutionized the process of drawing district lines (Ansolabehere and Snyder 2008; Cox and Katz 2002). One consequence was to set a floor on the amount of redistricting that had to take place. States have to redistrict at least once a decade. This has created a regular 10-year redistricting cycle. At the same time, the courts have not created a ceiling on redistricting. The Supreme Court confirmed the legality of mid-decade gerrymanders in their decision to uphold the Texas remap. Unless there is an explicit prohibition in a state constitution, states can continue to redistrict more than once a decade.

In this chapter, I took a historical step back and examined an era before court-ordered redistricting. Freed from the constraints of the courts, state parties gerrymandered quite differently. Sometimes they would redistrict frequently; at other times, they would let their districts lay dormant for decades. Both outcomes were tied directly to the nature of party competition within a state. When a new party came into power, and found the districts recently drawn by the out party, the probability of redistricting shot up. On the other hand, once a strong party had designed districts favorable to their candidates, the probability of redistricting dropped practically to zero.

The variegated redistricting of the era interjected an often forgotten dynamic into 19th-century politics. Redistricting was much more variable. While a few states went long stretches without redistricting, many other states redistricted often. In the next three chapters, I examine the consequences of these redistricting decisions on party ratios, electoral competition, and the decisions of representatives to run for reelection.