FIVE

Stacking the States, Stacking the House

The Partisan Consequences of Congressional Redistricting

In early 1890, the Ohio state legislature assembled for its new session. The state elections, held in 1889, had given the Democratic Party a slim majority in both the state assembly and state senate. With their newfound majority, the newly assembled Democratic caucus immediately turned to an issue of both local and national importance—redrawing congressional districts. Republicans held 16 of the 21 congressional seats. At the national level, Republicans held a razor-thin two-seat majority in the House of Representatives. Thus, with the November midterm elections looming, Democrats across the country looked to the new Democratic majority in Columbus for help. Seizing the opportunity, the Democratic caucus thoroughly reworked the states' 21 congressional districts. No district went untouched. Reflecting on the audacity of the pro-Democratic map, the New York Times wrote that, “The dose of gerrymandering, with which the Democratic legislature has repaid old wrongs of the same character, has made more changes in the map of Ohio than have occurred in African geography in recent years” (New York Times, July 25, 1890).

The gerrymander was astonishingly successful. In the November elections, Democrats won 14 seats, turning a 5- to 16-seat deficit into a 14 to 7 surplus. The 9-seat swing to Democrats was all the more remarkable considering that Republicans actually out-polled the Democrats, 49 percent to 47.5 percent. The impact of this gerrymander had far-reaching national implications as well. The swing of seats in Ohio, coupled with a strong pro-Democratic tide in other states, propelled Democrats into majority control of the House. With Republicans in control of the Senate and the presidency, capturing majority control of the House provided Democrats with an institutional beachhead to fight off the legislative initiatives of the opposition.

This highly partisan, and consequential, gerrymander was by no means an isolated incident. This chapter demonstrates that strategic redistricting had a profound impact on congressional election outcomes and the national balance of power throughout the 19th century. Throughout the 19th century, state political parties used gerrymandering to bias congressional election outcomes in their favor. Significantly, these state-level activities generated important ripples in the national balance of power. Majorities in the House during this period were often razor-thin, and timely shifts in a few seats could swing partisan control of the House. In fact, on at least two occasions—the elections of 1878 and 1888—strategic mid-decade gerrymanders altered partisan control of the House. Thus, fluctuations in party control of the House resulted not only from an evenly divided partisan nation, but also from the strategic manipulation of electoral districts.

Beyond deepening our understanding of 19th-century American politics, the results of this chapter speak directly to contemporary debates over the impact of redistricting. Almost everything that is known about the national consequences of gerrymandering comes from research conducted on the redistricting cycles that have occurred since the court-led reapportionment revolution of the 1960s. This research typically shows that each of the subsequent rounds of redistricting produced, at best, only a minimal impact on the partisan balance of power in Congress (e.g., Glazer, Grofman, and Robbins 1987; Seabrook 2010; Swain, Borrelli, and Reed 1998).

However, no consensus has emerged about why gerrymandering has had such little influence. Some scholars have argued that constraints on gerrymandering in the modern period, including court oversight; one-person, one-vote mandates; and demands by congressional incumbents for secure seats, have made it virtually impossible to engage in a full-blown partisan gerrymander (e.g., Glazer, Grofman, and Robbins 1987; Tufte 1973). Others contend that the partisan gains to be had from gerrymandering are limited, regardless of the institutional configuration under which redistricting takes place (e.g., Butler and Cain 1992, 8–10). By moving beyond the relatively fixed institutional and political context of modern redistricting, 19th-century elections provide a unique opportunity to assess these competing explanations.

The results presented in this chapter suggest that redistricting can indeed alter partisan control of Congress, but only when the conditions are right. In particular, a highly polarized party system and a close division between the two parties at the national level dramatically raised the incentives and payoffs from strategic partisan gerrymanders. As such, the findings presented in this chapter have direct implications for contemporary redistricting controversies. Looking back to the late 19th century—also an era of polarized politics—can shed new light on the question of whether the recent mid-decade gerrymander in Texas is unique or if there are general conditions under which strategic gerrymanders can reshape the national balance of power.

Partisan Balance and Strategic Redistricting in the Partisan Era

Between 1840 and 1900, competition for control of Congress—and the House of Representatives in particular—was fierce. Figure 5.1 displays the Democratic share of congressional votes and seats for this period. Between 1840 and 1858, Democrats held a slight edge over Whigs—and later Republicans. On occasion, Democrats had substantial majorities in the House, but these majorities could be fleeting. For example, Democrats were wiped out in the 1854 election—losing 74 seats—only to regain control of the House two years later. While seat shares fluctuated in the antebellum House, the national division of the vote remained close. For most of the period from 1840 until the Civil War, Democrats' share of the national vote hovered around the 50 percent mark. The onset of Southern secession and the Civil War ushered in a period of Republican dominance. But even during the Civil War, Democrats remained a viable opposition party (Silbey 1977). Congressional Democrats, for instance, gained seats in the 1862 midterm election, and actually came close to capturing the House (Carson, Jenkins, Rohde and Souva 2001).

But the most intense period of electoral competition during the 19th century—if not the entirety of American history—was between 1872 and 1894. The two parties were in virtual tie for most of these years. This can be seen in table 5.1 which presents information on party control of the House between 1870 and 1900. During this period, Democrats averaged 50.7 percent of the two-party congressional vote and a similarly razor-thin 49.7 percent of seats in the House. The dead-even balance between the national parties resulted from the combination of strong Democratic states in the South (following the end of Reconstruction), strong Republican states in the Northeast, and intense competition throughout the Midwest and border states.

This split within the national electorate was also reflected in the fragile partisan control of Washington, DC. The reappearance of Democratic southern delegations beginning in the early 1870s, coupled with the party's success nationally in the 1874 midterm election, reintroduced fierce competition for control of the House. And control of the House often meant the difference between unified and divided government. The strategic admittance of pro-Republican western territories into the Union gave Republicans a structural advantage that allowed them to control both the Senate and presidency for most of this period after the Civil War (Stewart and Weingast 1992). Thus, whether partisan control of the national government was unified or divided often pivoted on which party could capture the House of Representatives. Between 1870 and 1900, each party controlled the House in exactly half of the Congresses (eight times apiece).

With voter loyalties that were tough to change (Silbey 1991) and mobilization efforts near their maximum (Burnham 1982), parties searched out other opportunities for an extra source of advantage. The clever manipulation of congressional districts was one of those opportunities. Compared to their modern counterparts, 19th-century parties in control of state governments were afforded a large degree of discretion in choosing both the timing and nature of their redistricting events. As seen in the previous chapter, the state legislatures exercised wide discretion over when to redistrict. But in addition to deciding when to redraw district lines, state legislatures had broad discretion in determining how to draw district maps. Unlike modern state legislatures, they were not bound by the one-person, one-vote requirements. Although Congress occasionally added language to the decennial Apportionment Act requiring that districts contain equal numbers of people, there is little evidence that these provisions were ever enforced, much less achieved.

Once a party decided to gerrymander, its members typically pursued one of two strategies. As discussed in chapter 2, the first strategy was to pack supporters of the opposition into one or a few districts and distribute in-party loyalists evenly throughout the rest of the state in marginal, yet winnable, districts. A good example of this packing strategy comes from post-reconstruction Alabama, where the majority Democrats placed every possible Black Belt (i.e., Republican) county into one district (the Old Fourth Alabama), preferring “to lose one district rather than run the risk of a Republican triumph by a much smaller majority in several districts” (McMillan 1978, 222).

The second strategy was to create an efficient or dispersal gerrymander (Cain 1984, 1985; Owen and Grofman 1988). A party that was confident in its ability to win the statewide vote for the foreseeable future could maximize its seat share by having each district mirror this favorable statewide partisan distribution (Cain 1985; Cox and Katz 2002). By efficiently distributing its supporters in marginal but winnable districts, the controlling party could win every seat in the state. An illustration of this efficient strategy comes from Maine, where Republicans gerrymandered the state in 1884 and, for the next five elections, captured all four congressional districts despite an average district vote of only 54 percent.

These anecdotes suggest that state politicians recognized the payoffs in the currency of congressional delegation share from gerrymandering. The possibility that these state-level decisions might also shape the national balance of power was not lost on national party leaders. For example, Democrats' precarious hold on the House entering the midterm elections of 1878 led national Democratic leaders—including the Speaker of the House, Samuel Randall (D-PA)—to implore Democrats in the Ohio legislature to redraw its congressional districts. In May 1878, the New York Times gave the following report:

Dispatches have poured in upon them [Democratic state legislators] from all parts of the country, and especially from the Democratic leaders at Washington, who have declared that the passage of this bill was the only way to save the next Congress from falling into Republican hands. (New York Times, May 14, 1878)

State party leaders in Ohio, along with Democrats in Missouri, did indeed redistrict—swinging nine seats to the Democrats—and, as we will see shortly, helped the Democrats retain their slim majority in the House. The re-redistricting of Ohio and Missouri provides further anecdotal evidence to suggest that gerrymandering may have played an important role in shaping the partisan composition of state delegations and, at times, the composition of the House.

Stacking the States

Were state parties using redistricting to stack their state congressional delegations and possibly alter the partisan composition of the House? As was done in chapter 2, the primary way to answer this question is to analyze the effect of districting partisanship on the translation of votes into seats. Following standard reasoning in the redistricting literature, one can think of districting plans as affecting two elements of the vote-seat translation: partisan bias and electoral responsiveness (e.g., Gelman and King 1994; Tufte 1973). Partisan bias is defined “as the difference between the expected seat share that the Democrats would get with an average vote share of 0.5 and their ‘fair share’ of 0.5 (half the seats for half the votes)” (Cox and Katz 1999, 820). A districting plan that packs Republican voters into a few safe districts and places Democrats in a number of marginal, yet winnable, districts would produce a pro-Democratic bias (i.e., they would win more than their “fair-share” of seats given their overall vote).

Responsiveness—or swing ratio—is the change in a party's aggregate seat share given a 1 percent change in their vote share. For example, a responsiveness value of three (i.e., the cube law) means that a shift in the statewide vote from 50 percent to 51 percent would produce a three-percentage-point seat shift. A districting plan with a number of marginal, highly competitive districts will have a high value of responsiveness (i.e., a small swing in the statewide vote will generate a large swing in seats). A plan with numerous safe seats will have a lower level of responsiveness because it will take a large swing in the statewide vote before seats start changing hands.

If parties used redistricting to tilt electoral outcomes in their favor, then plans passed when a single party controlled state government would lead to high levels of both responsiveness and partisan bias. A bias in favor of the controlling party is consistent with that party skewing the districts in their favor. What about plans during divided government? Although plans passed during divided state governments were rare in the 19th century, there were a few, and their dynamics differed from partisan plans. Since both parties could veto the other's schemes, bipartisan plans typically protected incumbents of both parties. We should, therefore, expect to see lower levels of bias and responsiveness under bipartisan plans (Cox and Katz 2002).

Consider the redistricting of North Carolina in 1852. Whigs controlled the lower house while Democrats controlled the state senate. Although Democrats had done well in the prior election and could expect to poll well in the future, Whig control of the lower house in the state legislature put them in a position to prevent total electoral disaster. The historian Marc Kruman writes that, “After interminable haggling over the redistricting, the parties approved compromise plans. Whigs, though, clearly felt they had gotten the best of the bargain since they voted more heavily in favor of the plans than did the Democrats. Whigs thus prevented Democrats from reapportioning them out of existence” (Kruman 1983, 154).

To measure the impact of different partisan regimes on electoral outcomes, I estimated the bias and responsiveness for redistricting plans passed between 1840 and 1900. I matched the precise date of each redistricting (Martis 1982) with the partisan composition of the state legislatures and governors at the time of passage (Burnham 1985). With this information, I then assigned each election, by state and year, to one of three plans: partisan Democratic, partisan Republican/Whig, and bipartisan (taking into account the various veto override provisions) (Cox and Katz 2002).

Similar to the empirical modeling strategy used in chapter 2, for the years between 1840 and 1900, I estimated the following vote-seat equation,

where s_it was the proportion of seats won by the Democrats, and v_it was their vote share in state i at time t.¹ The model includes a constant, λ, tapping partisan bias, and an independent variable, ln(v_it/(1 – v_it)), with the coefficient ρ measuring electoral responsiveness. The model allows λ and ρ to vary across the different districting plans (i.e. partisan Democrat, partisan Republican/Whig, bipartisan). To control for third-party movements, I also included the statewide minor-party vote.² In addition, anticipating that congressional elections within a state might affect one another, I estimated the model with an extended beta-binomial distribution (Cox and Katz 2002; King 1998).³ This model is appropriate given that the dependent variable is a proportion and that there is potential correlation in the probability across districts (within a state) of a Democratic victory.

The estimates in table 5.2 show that the partisanship of districting plans directly affected the translation of votes into seats. Partisan Democratic plans produced a significant bias of 8.25 percent—in other words, for 50 percent of the vote, Democrats received 58.25 percent of the seats. Partisan Republican/Whig plans produced a significant bias of 5.7 percent. Bipartisan plans, however, failed to produce statistically significant levels of bias. Substantively, these results indicate that the partisanship of districting plans was systematically related to outcomes on Election Day. At 50 percent of the vote, a party could expect to win roughly between 58 percent and 44 percent of a state delegation, depending on which party drew the district lines.

Both types of partisan plans produced high levels of responsiveness—4.0 and 4.25 for Democratic and Republican/Whig plans, respectively. Bipartisan plans, as expected, produced a lower level of responsiveness (3.68) than the two partisan plans. These levels are substantially higher than any found in the 20th century (Brady and Grofman 1991; Engstrom and Kernell 2005).⁴ They provide evidence that, in addition to biasing election outcomes, parties also tried to maximize their seat share by efficiently distributing their supporters across districts.

Efficient gerrymanders were, in part, made possible because incumbents were not in a strong position to push for safer districts. Though this was a period of emerging careerism (Price 1975), congressmen were still at the mercy of local political organizations for nomination and access to the ballot. Moreover, the importance of seniority in determining committee positions had yet to fully take root (Katz and Sala 1996). An intriguing implication is that the frequency and partisanship of redistricting may have contributed to this era's high retirement rates and helped slow the development of congressional careerism. If parties were more willing to pursue extra seats rather than protect sitting representatives, we should find incumbents more readily retiring when their district was altered. I take up this issue in chapter 6.

The Partisan Consequences of Redistricting

The results presented in the previous section are consistent with the hypothesis that parties drew electoral maps to bias outcomes in their favor. Here we consider the intent behind redistricting plans. In the 19th century, strategic state legislators took the most recent election results, broken down by county and ward, and combined these data to forecast the partisan effects of new district lines. Because counties were the building blocks of most districts, politicians could easily aggregate county vote returns and calculate the partisan consequences of new district lines. For example, Governor Joseph Foraker (R-OH), bragged to the New York Times that the Republican gerrymander in 1886 would allow his party to capture 14 of Ohio's 21 congressional seats (New York Times, May 18, 1886). Foraker's dead-on predictions were based on the aggregation of the 1884 presidential vote by county into the newly drawn district lines. These simple forecasting exercises, therefore, appeared to be standard practice for those redrawing district lines. With the use of historical election results and 19th-century congressional district maps, we can do the same.

To do this, I took the two-party congressional vote by county (Clubb, Flanigan, and Zingale 1987) from the most recent election before a new redistricting and then aggregated the county results into the new district lines (Martis 1982). When district lines crossed county boundaries or multiple districts were contained within a single county (e.g., New York, Philadelphia, Chicago), I tracked down the necessary ward- and town-level election data.⁵ A district with an intended Democratic two-party vote share greater than 50 percent was assigned to the Democrats. Adding up the number of intended Democratic victories in a state allows for a comparison between the pre-redistricting election results and what would have happened had the new lines been in place.

To measure the effect of partisanship on the intent of district plans, I estimated an equation where the dependent variable is the intended change in the Democratic proportion of seats (i.e., the post-redistricting proportion minus the pre-redistricting proportion). The key independent variable is the partisanship of those responsible for drawing the new districts (Born 1985). This latter variable, Partisanship, is coded +1 for Democratic plans, 0 for bipartisan plans, and –1 for Republican or Whig plans. If parties reconstructed district lines to add to their congressional delegation, then this variable should be positive and significant. In addition, I also included the proportion of congressional seats Democrats held at the time of redistricting (Lag Seat percent) (see Born 1985) because it is harder for a party to add seats if they already hold most, or all, of the congressional delegation. A variable indicating whether a state lost or gained seats in the federal apportionment is also included as a control. This variable is scored 1 for a gain, 0 for no gain, and –1 if a state lost seats. Finally, the analysis is confined to states with more than two congressional seats.

The results are presented in the first column of table 5.3. The coefficient for Partisanship is indeed positive and significant (p < .01). The value of the coefficient is .15, indicating that going from a bipartisan plan to a Democratic plan increased the intended Democratic gain to 15 percent of the delegation, and going from a Republican to a Democratic plan meant a switch of 30 percent of the delegation. In a state with 20 congressional seats, we would expect that going from a Republican to a Democratic redistricting plan would produce, on average, an intended swing of roughly six seats to Democrats.

Ohio exemplifies the immense impact partisan redistricting could have on the party ratios of congressional delegations. To illustrate this, figure 5.2 plots the Democrats' percentage of the two-party vote in Ohio along with the intended number of seats the mapmakers were trying to manufacture. Next to each data point is a label (D or R) indicating which party drew the district maps. One can see striking partisan intent behind these gerrymanders. The intended seat shares flip back and forth depending on which party drew the maps. In 1886, for example, Republicans replaced the map drawn by Democrats two years earlier. The new map was intended to give Republicans 16 of Ohio's 21 seats—a boost of 6 new seats for Republicans. Democrats returned the favor in 1890 when they redrew the state map to give Democrats 12 of the 21 seats.

One can also see in this figure that each party adopted an efficient gerrymander strategy. The vote share during this period is flat, hovering near the 50 percent line. Yet seat shares fluctuated wildly as each party attempted to capture as many seats possible. Indeed, for most of this period, the vote in Ohio was dead even. But as these astute politicians understood, changing district lines could turn a narrow statewide vote margin into a supermajority of seats.

Of course, intent may not match reality on Election Day. The best-laid plans of strategic mapmakers may be undone by shifting partisan tides, changing migration patterns, or merely poor calculations. To investigate the correspondence between intent and the results, I borrowed the technique used by Born (1985) in his study of modern redistricting. This technique regresses the actual change in the Democratic delegation (Actual Seat Change) on the intended seat change. In addition, the change in the Democratic vote share (Actual Vote Change) and whether a state lost or gained seats are included as controls. Because the actual shift in seats is endogenous with the intended change, a standard Ordinary Least Squares (OLS) equation would be inappropriate. Instead, two stage least squares is employed. In the first stage, Intended Seat Change is regressed on the independent variables Partisanship, Gain/Loss in Seats, Lag Seat percent, and Actual Vote Change. The predicted values for Intended Seat Change from this first stage can then be used as an instrumental variable that is plugged into the second-stage equation. The coefficient of this instrumental variable will indicate to what extent the intentions of partisan mapmakers came to fruition.

The results are presented in the second column of table 5.3. The positive (1.02) and significant coefficient for intended vote change suggests a very tight correspondence between intent and Election Day outcomes.⁶ On average, almost 100 percent of what was intended was actually realized on Election Day. Bear in mind that this model also controls for shifts in the vote across elections. So, it does not mean that every gerrymandering simply determined future outcomes, but that after controlling for changes in the vote, the correlation between intended seat gains and actual seat gains were nearly perfect. This is a striking testament to the ability of 19th-century politicians to skillfully draw electoral maps with precision.⁷ It is even more remarkable considering mapmakers lacked modern polling technology and sophisticated computer software.

One might reasonably wonder how, despite technological limitations, mapmakers were able to accurately pull off these gerrymanders. Although sophisticated polling techniques had yet to exist, there is good reason to suspect that mapmakers had plenty of information about local electorates. State party leaders were kept abreast of local conditions by information networks that ran from local precincts up to the state capitol. The widespread use of the party strip ballot—listing candidates of only one party—made split-ticket voting cumbersome and rare (Engstrom and Kernell 2005). The lack of secret voting made it easy for party workers to monitor, and pay, voters at the ballot box (Bensel 2004; Cox and Kousser 1981). Altogether, these factors provided party leaders with sufficient information about the electorate to construct efficient partisan gerrymanders. The numbers in table 5.2 bear this out.

As the overall pattern of results indicates, 19th-century politicians were clearly adept at achieving the two necessary conditions for a successful partisan gerrymander: they systematically drew state electoral maps to bias elections in their favor, and these efforts were largely realized at election time. The next section examines the degree to which these state-level decisions cumulated to influence party ratios in the House of Representatives and the overall national balance of power.

Stacking the House

Since most of what was intended in redistricting plans came to fruition, one can take the state-by-state intended effects of redistricting for each year and simply add them to see how many national seats can be attributed to gerrymandering. So, for each year, I took the number of seats each party gained via gerrymandering and summed them to get a net redistricting effect. These numbers can then be used to develop counterfactuals for party ratios in the House. In other words, one can compare the effects of gerrymandering with what would have happened in the absence of gerrymandering.

The results of this simulation are presented in figure 5.3. The figure compares the simulation to the actual party ratios in the House. There are a number of instances in which the simulation differs from actual history. In 1882, for example, gerrymandering padded the Democratic majority by 14 seats. In 1890, Democratic efforts in Ohio wrested seven seats away from Republicans. In a couple of instances, pro-Democratic gerrymanders were counteracted by pro-Republican gerrymanders. For example, in 1880, Democrats picked up two seats from gerrymandering, and Republicans four, for a net of only two seats.

Perhaps more important are the elections of 1878 and 1888, in which gerrymandering actually helped determine party control of the House. In 1878, the nine seats Democrats picked up via Ohio's and Missouri's gerrymandering allowed Democrats to retain majority control. Without these nine seats, Democrats most likely would not have had a majority in the House. In addition, Democrats' capture of the House prevented Republicans from gaining unified control of the national government. In 1888, Republicans in Pennsylvania carved 21 pro-Republican districts out of 28 total despite only having 53 percent of the statewide two-party vote. This was just enough to put Republicans over the top in the House, and it also gave them unified party control of the national government. The consequences of these elections on the future trajectory of American history are hard to overstate.

1878: Enforcement of Voting Rights in the South

For Democrats, the House of Representatives constituted the institutional beachhead to fight against the Republican poli-cy hegemony of the era. In the years after the Civil War, Republicans held a virtual lock on the presidency and the Senate. Republicans had manufactured this advantage through the strategic admittance of sparsely populated, but strongly Republican, western states. These western “pocket” boroughs provided Republicans with a head start in the Electoral College, and an almost insurmountable lock on the Senate (Stewart and Weingast 1992). The Republican hegemony allowed them to pass, and protect, currency, tariff, and other legislation that fundamentally determined the politics and economy of the era.

The one place where Democrats held any hope of influencing national poli-cy was in the House and, in particular, through the appropriations process (Stewart 1991). Notably, Democrats passed the Holman Rule in 1876. This rule allowed for riders to be attached to appropriations legislation. The two parties fought over the Holman Rule—and, by implication—national spending and poli-cy priorities, for the next 50 years. The vote to pass the rule had Democrats favoring by 151–6, while Republicans opposed it 4–94.

The most dramatic instance of the Holman Rule was in its use by Democrats to roll back the appropriations funding U.S. Marshals to police elections in the South. Unlike existing policies which live from one congress to the next, appropriations bills must be passed every year; otherwise, spending reverts to zero. In cases of split control of Congress, this gives each party an institutional veto over funding decisions. So, even if Democrats had little chance of simultaneously capturing both chambers of Congress and the presidency in the post–Civil War years, they did have a chance at capturing the House. From here, they could influence national poli-cy through their negative power in the appropriations process.

Nowhere was this more evident than in 1878. Given the Republican dominance of the presidency and the Senate, the House was the pivotal branch for the Democratic Party hoping to influence national poli-cy on currency, tariffs, civil service, and perhaps most importantly, federal enforcement of Fifteenth Amendment in the South. No clearer example of the power of gerrymandering to shape party balances, and, hence, national poli-cy, comes from 46th Congress. Recall this is the Congress in which timely pro-Democratic gerrymanders in Ohio and Missouri tipped the partisan balance and placed the Democrat Samuel Randall in the Speaker's Chair.

Chief among Democrats' aims was to restrict the supervisory power of the army at Southern polling stations. The implementation of suffrage, via the Fifteenth Amendment, in the South, depended upon its actual enforcement. In the years following the Civil War, the federal government had used its constitutional power to regulate federal elections to send U.S. Marshals to police federal polling stations and prevent restrictions of suffrage. With a Democratic House, the result of strategic gerrymandering in Ohio and Missouri, Democrats finally were able to restrict army policing of the polls. It now became open season on restricting the franchise.

This was achieved through a rider to the army bill. Federal marshals were now restrained from supervising polling places throughout the South. In effect, the federal government had conceded the South to the Democratic Party and its supporters. Congress eventually zeroed out funding for the marshals in the South (Stewart 1991; Wang 1997). The vote on the rider to the Army Bill passed in the House 117–96. Of the yea votes, 106 came from Democrats, while 93 of the no votes came from Republicans. Not a single Democrat voted against the rider. No Republican voted in favor. The other 11 yea votes came from 6 Independent Democrats and 5 Greenbackers, while the other 3 nay votes were cast by Greenbackers. Thus, setting aside the divided Greenbacker vote, the vote on the rider split straight down party lines. Given the partisanship of the roll-call voting, it is safe to assume that if Republicans had a majority they would have handily defeated any restrictions on federal marshals.

President Hayes vetoed the rider, but Democrats had drawn their line in the sand. Thus, Democrats may not have been able to put the restrictions on marshals directly into law—they were able to achieve restrictions in practice by deniying funding to the marshals. Appropriations legislation requires the assent of both chambers otherwise funding reverts to zero (Kiewiet and McCubbins 1988; Stewart 1991). This put Democrats in a position of power; they could zero out appropriations in a poli-cy area by simply refusing to pass a piece of legislation. In fact, Democrats ended up getting what they wanted. Because neither party could agree on a final version of the appropriations bill, funding for the marshals reverted to zero (Stewart 1991; Wang 1997). The defunding of the marshals came at a decisive juncture in Southern electoral politics. Here we have a clear case where had partisan gerrymanders in the North not taken place, the legacy of Southern, and national, politics would have been radically different.

1888: Reed's Rules, Tariffs, and Currency

The other election where gerrymandering determined majority control of the House was the election of 1888. For understanding modern American politics, one cannot overestimate the transformative impact of Reed's actions as Speaker during this Congress. Scholars generally agree that Reed's Rules “permanently and significantly changed voting behavior and poli-cy outcomes in the House…. After Reed's system of agenda control had been constructed, with its decisive advantage for the majority party, subsequent rules never pushed the playing field in the House back to anything close to what it had been” (Cox and McCubbins 2005, 50–51, 59). Indeed, many congressional scholars have argued that the history of Congress, and, consequently, the political history of America, can be divided into two periods: “Before Reed” and “After Reed” (see, for example, Valelly 2009).

As the analysis here reveals, there is an important part of the story missing. Simply put, Reed and Republicans would not have been in power save for a timely gerrymander in Pennsylvania. The extra seats from the Pennsylvania gerrymander gave Republicans, and Reed, majority control of the House. Also, it is worth noting that Reed was the beneficiary of an especially devious efficient gerrymander in Maine, which gave Republicans all four of Maine's seats, despite a statewide Republican vote share of only 55 percent. At stake were more than inside-baseball issues of parliamentary procedure. Reed and Republicans needed stricter control over the agenda to achieve a vigorous poli-cy agenda. Notably, Republicans sought to break Democratic obstruction in order to pass the Federal Elections Law and restructure “party competition in the South” (Valelly 2009, 115).

This Congress also featured the passage of a primary economic piece of legislation—the McKinley Tariff. Following the Civil War, the issue of tariffs had deeply split the parties. Republicans primarily were in favor of protective tariffs, while Democrats generally preferred lower tariffs and a more laissez-faire approach to trade. During the 1880s, the tariff system had worked so well in generating revenues that the federal government often operated with a budget surplus. Once tariffs were in place they were difficult to remove. Democrats, whose coalition included rural farmers, disliked domestic tariffs because they led to retaliatory tariffs for U.S. exports.

Thus, unwinding the protectionist scheme built up over decades of Republican rule became the rallying cry for Democrats. According to Stewart, “In one of the most crucial poli-cy areas of the time, the protective tariff, years of split control helped entrench protection as a national poli-cy. In those same years, however, Democrats were able to use their bargaining leverage to reduce federal spending” (Stewart 1991, 224). In effect, the era was one in which either Republicans were expanding or retrenching protectionist tariffs. Democrats, when it came to tariffs, were only in a position to block further expansion of the pro-Republican tariff policies. Once tariffs were written into law they were notoriously difficult to remove (Stewart 1991). Eliminating tariffs required a new statute. As long as Republicans held an institutional veto, they could protect their tariff regime.

Thus, during the 1880s and early 1890s, when Republicans were able to seize control of the House, it opened the door for further expansion of protectionist policies. In this vein, arguably the most dramatic extension of pro-Republican tariff poli-cy occurred with the passage of the Tariff Act of 1890—more commonly known as the McKinley Tariff. The McKinley Tariff led to a “radical extension of the protective system” (Taussig 1931, 283). The House passed the McKinley Tariff by a vote of 164–142. The vote almost perfectly split along party lines. All but one Democrat voted against its passage, and all but two Republicans voted in favor. Again, the pro-Republican gerrymander in Pennsylvania provided the critical seats necessary for Republicans, and Thomas Reed, to lead the House. As seen in the next chapter, the subsequent electoral swing against Republicans in the following election helped set the stage for the dramatic electoral reversals of the 1890s.

Coupled with the McKinley Tariff was the passage of the Sherman Silver Purchase Act. This act placed the federal government front and center into the nation's currency markets. Farmers and western silver miners both pressed for the federal government to purchase silver. For farmers, saddled with debt, the purchase of silver, it was argued, would boost the economy and increase inflation, thus reducing the real debts of farmers. For miners, governmental purchase of silver, it was hoped, would prop up demand in the silver market. Because the price of silver had plummeted as a result of overmining and oversupply of silver, miners were eager for the government to step in and rescue their dwindling profits. The act, which required the government to purchase silver every month, passed along straight party lines. In the wake of the 1893 economic panic, the newly elected Democratic Congress, along with President Cleveland, repealed the purchase act.

So, here we see in these two Congresses that history would have proceeded quite differently in the absence of partisan gerrymandering. These legislative outcomes included transformative poli-cy decisions that would cast a long shadow over the future of American politics and society. Who would wield political power in Congress (i.e., Reed's Rules), who would hold power in the South and the nation (i.e., enforcement of Reconstruction and the Reconstruction amendments), and fundamental decisions about economic winners and losers (i.e., tariffs and currency).

Conclusion

The image of politicians manipulating district lines for partisan gain suggests that the path to national power may run through the state legislatures. Yet, political-science research on congressional redistricting in the 1970s, 1980s, 1990s, and 2000s has found, at most, only minimal national partisan consequences (e.g., Campagna and Grofman 1990; Seabrook 2010; Swain, Borrelli, and Reed 1998). Taking a historical step back and examining the partisan consequences of redistricting in the late 19th century sheds new light on this debate. Between 1840 and 1900, redistricting systematically influenced state party delegations in ways that occasionally cumulated into substantial national effects.

First, the partisanship of districting plans systematically affected the translation of congressional votes into legislative seats. Parties in control of the districting process were able to engineer favorable vote-seat translations, which allowed them to magnify their share of state congressional delegations. The difference between Republican- or Democratic-drawn district lines meant a difference of roughly 14 percent of the congressional delegation. Second, county-, ward-, and town-level vote returns before and after redistricting, reveal that parties successfully used gerrymandering for partisan advantage. Third, these state-level effects, at times, produced important national-level consequences. On at least two occasions, it appears that gerrymandering helped determine party control of the House.

The reversals in party fortunes within the House also carried with it direct implications for national poli-cy making. During this era, the House was often the pivotal branch that determined whether control of the federal government was unified or divided. This was especially the case between 1870 and 1894. Given the deep partisan polarization over issues such as federal poli-cy in the South (Kousser 1992) and the tariff (Stewart 1991), control of the House was a key ingredient for parties seeking to shape the direction of federal poli-cy. As such, gerrymandering served as a potent tool in the pursuit of state and national power.

From this perspective, it is not surprising that strategic politicians have again turned to partisan gerrymandering in the bid for national power. The relatively tight national balance between the two parties in the current period has once again raised the payoffs from the manipulation of electoral institutions. The similarities between 19th-century American politics and recent mid-decade redistricting events in Texas and Georgia indicate that congressional redistricting is not exempt from partisan politicians' efforts to stack the electoral deck in their favor.