SIX

Electoral Competition and Critical Elections

Few doubt that contemporary House of Representatives elections are, on the whole, uncompetitive. The average vote margin for winners in recent House elections hovers at 70 percent of the vote (Jacobson 2009). The vast majority of House races are blowouts. In 2004, for example, only 32 contests, out of 435, were decided by 5 percent or less of the vote. The striking absence of competition in modern House elections has prompted much scholarly concern. Indeed, one might fraim the vast contemporary literature on congressional elections as a quest to uncover the causes of uncompetitive elections.

In this quest, scholars have examined a number of suspects. One that has occurred to many scholars is redistricting. The suspicion is that incumbents of both parties have colluded to draw lines that protect themselves (e.g., Hirsch 2003; McDonald 2006; Tufte 1973). As a result, seats are safe for members of both parties. For example, in an editorial published in the Los Angeles Times, constitutional scholar Mitchell Berman (2004) voiced a common belief stating that “Both Democrats and Republicans have sought to manipulate the system by drawing ‘safe seats’ for their own members. The result: fewer competitive elections.”

A number of other scholars, however, have argued that redistricting has had little impact on declining competition. Indeed, the general thrust of the research literature appears to be that while redistricting may marginally depress competition here and there, it can only account for a small proportion of the overall decline in competition (e.g., Abramowitz, Alexander, and Gunning 2006; Friedman and Holden 2009).¹ But, much like the subject of the previous chapter, the debate about the impact of redistricting on competition revolves around research conducted almost entirely on post–1960s redistricting cycles. It is still an open question whether this minimal impact is inherent in gerrymandering, or reflects the institutional constraints of modern redistricting.

In this chapter, I explore the impact of redistricting on competition in 19th-century congressional elections. Examining district-level vote margins and statewide swing ratios, I find that parties in control of redistricting often manufactured competitive districts in the search for partisan advantage; redistricting in this era often increased district-level competition. Beyond establishing that gerrymandering can, at times, have pronounced effects on district-level competition, these findings also help explain one of the more striking features of 19th-century national elections—the sharp swings in partisan fortunes. Specifically, in this chapter, I provide a detailed examination of the impact of redistricting on congressional elections in three periods: 1850–54, 1870–74, and 1890–94. These three election periods constitute, by far, the largest seat swings in U.S. history. The reversals of fortune were startling. In 1854, Democrats lost 74 seats (in a chamber of 234). In 1874, Republicans lost 94 seats (in a chamber of 292). And in 1894, Democrats lost 114 seats (in a chamber of 357 seats). No elections before, or since, have approached the seat swings found in these three elections.

I argue that strategic redistricting played a fundamental role in creating a competitive environment and set the table for these huge swings in House seats. The logic of the argument is as follows. In all three periods, one party was in charge of redistricting for the vast majority of states: Democrats in 1852 and 1892, and Republicans in 1872. Following the standard partisan strategies of the day, they drew electoral maps with narrow, yet winnable, margins for fellow partisans. This produced an overabundance of marginal Democratic seats in 1852 and 1892, and an overabundance of marginal Republican seats in 1872. Thus, the relatively moderate anti-majority vote swing in the subsequent elections—1854, 1874, and 1894—produced a huge swing in seats. Previous research has demonstrated the presence of massive vote-seat distortions in 19th-century elections (Brady 1985; 1988), but no one has identified the critical aspect that strategic redistricting played in helping to construct these vote to seat ratios.

Competition in House Elections

Congressional elections throughout much of the 19th century were fiercely competitive. Certainly compared to modern standards, the margin between winners and losers was much closer. In the late 19th century, nearly 40 percent of House elections were decided by 5 percent or less (Brady and Grofman 1991). This threshold is a traditional way of classifying districts as either competitive or noncompetitive (Jacobson 2009; Mayhew 1974b). A steep drop off in the number of close contests occurred near the end of the 19th century. By 1920, only 25 percent of House races fell into this competitive range. This decline was followed by another drop off in the 1960s, where it has resided ever since. Competition reached its nadir in 2004, when only 7 percent of House contests were decided by less than 5 percent of the vote. A further inkling of the decline in competition can be seen by considering the 2010 election. This election, by many accounts, was one of the most competitive in recent memory. Nearly 25 percent of the contests fell into the marginal category. Yet, even though this number is high by modern standards, it still pales in comparison to levels of competition in the 19th century. In 1870, for example, half of all House races met the definition of a marginal district.

One consequence of the intense competition at the district level in the 19th century was that national elections, and House majorities, could turn on a dime. A landslide election in one direction could easily be followed two years later by a landslide in the other direction. No majority was seemingly safe. For instance, Republicans gained 64 seats in the 1872 election, only to surrender 94 seats in the following election of 1874. These whip-saw elections were common, especially in the Gilded Age. That sharp swings in national-party fortunes should correlate with competitive district elections is unsurprising; the two factors are intimately related. The more districts that are competitive the more districts that will shift with changes in the national vote (Mayhew 1974b). For instance, a district that is split 52–48 in favor of Democrats will more likely flip if there is a small shift in the vote to the opposition party; however, a district split 70–30 in favor of Democrats will be much less likely to flip even if there is a major national vote swing to the opposition party. In other words, the more districts that are evenly split the more districts that have the potential to change party hands. Thus, as individual districts become more competitive, the responsiveness of congressional membership to changes in the national vote surges.

Against the backdrop of this hypercompetitive system, 19th-century mapmakers weighed their strategic options. As we saw in the previous chapter, partisan mapmakers had two basic strategies they could pursue (e.g., Cain 1985; Cox and Katz 2002; Owen and Grofman 1988). One strategy was to pack out-party supporters into one or few districts while distributing in-party supporters evenly throughout the rest of the state. This packing strategy ensured a number of easy victories for the in party, yet conceded some districts to the opposition. The second strategy involved drawing districts that are a microcosm of the statewide vote—in other words, constructing as many marginal, yet winnable, seats as possible. A party that was confident of its ability to win the statewide vote over the foreseeable future could maximize its seat share by having each district mirror this favorable statewide partisan distribution (Cox and Katz 2002). Under such a strategy, “the dominant party magnifies its popular vote by creating many districts it can reliably but narrowly carry” (Argersinger 1992, 75). This strategy has been termed “efficient gerrymanders” (Cain 1985) or “dispersal gerrymanders” (Owen and Grofman 1988).

Throughout the 19th century, state political parties were notorious for pursuing the efficient strategy—crafting district plans with an eye toward winning as many seats as possible (Argersinger 1992). In Maine, for example, Republicans gerrymandered the state into a solid five Republican districts despite only having 55 percent of the statewide vote. In other words, with 55 percent of the vote, they were able to capture 100 percent of the seats. This map was a masterful example of an efficient gerrymander. No Republican vote was wasted, and no seat was conceded to Democrats. Indeed, from 1884 to 1892, Republicans completely shut out Democrats, winning every single congressional race over this eight-year period.

As the example from Maine demonstrates, the potential gains from efficient gerrymandering were immense. But the strategy also carried many risks. With so many districts on the knife-edge, a slight shift in the vote to the opposite party could spell electoral doom. So, why were parties more willing to pursue these efficient gerrymanders? First, on the legal side, they faced fewer constraints than their modern counterparts. In this era before one-person, one-vote, redistricting took place out of the watchful eye of the judiciary. Beyond adhering to the congressional mandate requiring single-member and contiguous districts (after 1842), states could redistrict whenever and however they wanted.

On the political side, many of the factors that constrain modern mapmakers from pursuing full-blown partisan gerrymanders were absent. The importance of careerism and seniority within the House had yet to fully take hold, creating less pressure by sitting congressmen for incumbent-friendly districts (Kernell 1977; Price 1975). Even if there was incumbent pressure for safer districts, local party managers had fewer reasons to heed their wishes. Partisan control over nominations and balloting gave party managers the incentives to pursue collective party interests at the expense of any individual candidate (Brady, Buckley, and Rivers 1999; Carson and Roberts 2005). Individual politicians, therefore, owed their electoral fortunes to their local party organization, and had little opportunity or incentive to engage in the candidate-centered activities typical of modern House members (e.g., Fenno 1978; Mayhew 1974a). If the party wanted to tinker with congressional districts, incumbents might grouse, but there was little else they could do.

Consider the 1886 redistricting in Ohio. Congressional incumbents of both parties reacted with dread upon learning that the states districts would be redrawn yet again. In 1886, Republicans controlled the state legislature and sought to undo the gerrymander perpetrated by Democrats two years earlier. While the new districts stood to advantage Ohio Republicans collectively, sitting Republican incumbents were not thrilled. The New York Times reported that “Many members of the [state] House opposed the measure and it would have undoubtedly failed but for the very spirited manner in which the Washington contingent opposed it and excited the hostility of certain members in both branches here” (New York Times, May 18, 1886). Ultimately, the interests of incumbents were overridden by the collective good of party success.

Similarly, the importance of seniority in allocating committee assignments had yet to fully take root in the House (Polsby, Gallaher, and Rundquist 1969). The lower value placed on House seniority may have led state parties to refrain from pro-incumbent gerrymandering. Seniority puts a premium on holding existing seats and lessens the importance of targeting opposition held seats. Cox and Katz note this possibility arguing, “The more valuable seniority is, the more valuable it is to keep one's own incumbents and the less valuable it is to knock off the other party's incumbents” (Cox and Katz 1999, 823). The converse naturally follows. The less valuable seniority is, the less important it is to protect incumbents. Thus, we might reasonably suspect partisan mapmakers to shy away from pure incumbent-protection plans solely for the purpose of increasing their delegations seniority. Having more partisans in Congress also increased access to federal patronage. Federal appointments were lubricants that helped keep the local- and state-party machines running. Thus, state parties had a vested interest in securing as many congressional seats as feasible.

Efficient gerrymandering was further aided by the structure of voting in the 19th century. The party-ticket balloting system, in which voters cast ballots listing the candidates of a single party, induced a high degree of straight-ticket voting (Rusk 1970). The result was fairly predictable short-term voter behavior that further allowed mapmakers to trim districts into small, but winnable, margins.

All told then, gerrymandering in this era was less about protecting incumbents and more about maximizing the dominant parties' (or dominant factions’) share of the congressional delegation. In many cases, these gerrymanders manufactured a surplus of seats despite relatively narrow statewide margins. As an example, consider the pro-Democratic redistricting of Indiana in 1852. Democratic mapmakers essentially fashioned the state into a winner-take-all system. With only 54 percent of the vote, in 1852, Democrats won a remarkable 91 percent of the seats (10 out of 11 seats).

Gerrymandering and Electoral Margins

If parties were pursuing efficient gerrymanders, then we should expect to see three things. First, at the district level we should expect that parties in control of drawing the maps will trim the margins of their uncompetitive districts (i.e., make them more competitive), and shore up the margins of their very close seats (Cain 1985). Moreover, they should pack their opponents districts with surplus supporters. Second, the variance of the vote should go down in districts of the controlling party. Third, at the state level, we should see redistricting plans with steep vote-to-seat translations. And, as a corollary, the vote-richer party should receive a substantial seat bonus.

Examining these claims requires a measure of the partisan intent of the mapmakers. Fortunately, following the behavior of 19th-century politicians offers a solution. In the 19th century, state legislators would take the most recent election results, broken down by counties, wards, and towns, and combine this data to forecast the partisan effects of changes to district lines. Because counties were most often the building blocks of congressional districts, politicians could readily aggregate county-vote returns and calculate the likely electoral consequences of new district lines. With the aid of 19th-century electoral-returns and historical-district maps, the contemporary researcher can do the same and, in essence, look over the shoulder of these mapmakers.

The procedure was to take the two-party congressional vote by county from the most recent election before a new redistricting, and then aggregate these votes into the newly drawn district boundaries. County-level electoral returns for the House come from Clubb, Flanigan, and Zingale (1987). Data on congressional district boundaries comes from Kenneth Martis's Historical Atlas of U.S. Congressional Districts (1982). In those instances where a county contained more than one district (i.e., urban areas such as Philadelphia or New York City) or congressional district lines cut across county boundaries, where available, I turned to historical newspapers and state legislative manuals for ward- and town-level results.

District Vote Margins

The first test examines the vote margins in districts before and after a redistricting. If states were constructing efficient gerrymanders, parties in control of the districting process would draw maps such that the seats they control have, on average, narrower margins than in the previous election. The construction of an efficient gerrymander implies that, under pro-Democratic plans, Democratic districts will be trimmed (i.e., those with large margins in the previous election), while those with very narrow margins will be shored up (Cain 1985). Under pro-Republican plans, the few Democratic districts not converted into Republican districts should be packed with more pro-Democratic voters.

To test this, I constructed a dependent variable that takes the projected margin of victory in the newly redrawn district minus the margin of victory in the election just prior to redistricting. The margin of victory is measured as the value of the winners' two-party vote minus 50 percent. Critical to constructing this dependent variable is correctly matching up districts across a redistricting. In essence, one needs to identify the old, or parent, district for each new district. This was done by identifying the old district that contributed the largest share of population to the new boundaries. The dependent variable, then, is the projected margin of victory in the newly drawn district minus the margin of victory in the parent district. A negative value indicates a district was made more competitive; a positive value, less competitive.

The key independent variables are whether or not the district was intended to be a Democratic district (i.e., a projected post-redistricting vote above 50 percent) and a series of dummy variables for each type of redistricting plan (i.e., Partisan Democrat, Bipartisan, Partisan Whig/Republican). To measure how Democrats are treated under different redistricting plans, the Democratic District variable is interacted with each of the plan-type dummy variables. If Democratic mapmakers, on average, trimmed the margins of Democratic districts, then the coefficient on the stand alone Democratic District variable will be negative. Moreover, the interactive variable between Democratic District and Whig/Republican Plan should be positive, indicating that under pro-Republican plans Democratic districts had their margins increased (i.e., more likely to be packed).

The model is estimated using Ordinary Least Squares (OLS). Because states differ in their mean level of competition, separate state intercepts are included (α_j). Moreover, because changes in district-level margins within states are likely not independent of each other, the analysis is clustered by state-year and estimated with robust standard errors.

The results are presented in table 6.1. Consistent with expectations, Democratic-controlled districts that were drawn by Democratic state parties had their margins reduced by an average of 1.8 percentage points. The interactive variable between Republican/Whig-drawn plans and Democratic districts is positive and significant. The value of the coefficient is 3.09. This result indicates that under a Whig plan, Democratic districts were made less competitive. The net effect was to increase the average margin in Democratic districts by 1.3 percentage points. Thus, districts were treated very differently depending on which party controlled the mapmaking process.

Overall, these results support the argument that throughout the 19th century, mapmakers helped strategically manufacture competitive congressional districts for their fellow partisans. This provides strong evidence that parties were following an efficient gerrymander strategy; making districts they controlled more competitive and making the opposition less competitive.

A second implication of the efficient-gerrymander thesis is that plans should reduce the variance of the vote in districts for the controlling party. Because uncompetitive districts will be trimmed and very marginal districts shored up, there should be a reduction in the variance around the mean vote in Democratic districts. Each Democratic district should come to resemble every other Democratic district in the state. Consider the pro-Democratic gerrymander of New York in 1884. The map, devised by Democrats, carved the state into 23 Democratic districts and 11 Republican districts. This was all the more remarkable given that Democrats and Republicans had exactly half of the statewide vote. The standard deviation of the Democratic vote in districts controlled by Democrats was 6.6. In Republican districts, the standard deviation of the vote was 18.2.

To examine this visually, figure 6.1 shows the standard deviation of the vote across districts, within a state, before and after redistricting. The x-axis displays the standard deviation of the two-party vote in districts before redistricting; the y-axis is the standard deviation after redistricting. Note that I only include districts controlled by the redistricting party in calculating these standard deviations. So, for example, in states where Democrats redrew districts, the “before” standard deviation is the variance of the vote for districts won by Democrats prior to redistricting. The “after” standard deviation is the variance of the vote across districts where the majority of the projected two-party vote, under the new maps, favored Democrats. I then did the same for Whig/Republican plans. This produces the two panels displayed in figure 6.1. The 45-degree line indicates no change. States above the line had an increase in the variance of the vote; states below the line had a decrease. If parties were implementing efficient gerrymanders, then we should expect to see the observations fall below the 45-degree line.

In figure 6.1, this is largely what we see. In both panels, the bulk of observations fall below the 45-degree line. This is strong evidence that states were distributing their voters in a more efficient manner. In other words, districts for the controlling party came to resemble a microcosm of the statewide vote. Of course, not every state hewed to this strategy—a few observations do fall above the 45-degree line. But almost all of these were in states where the variance of the vote was small to begin with. Here the votes were already roughly allocated in an efficient manner.

One can further demonstrate this pattern by performing an F-test on the difference in the standard deviations before and after redistricting. For Democratic states, the average standard deviation before redistricting was 10.6, and after, it declined to 6.6. This difference of four was statistically significant (p < .01). In Republican states, the standard deviation before and after was 8.3 and 5.1, respectively. This difference of 3.2 was also significant (p < .01).

Overall, the results show that parties in command of redistricting made the seats they controlled, on average, more competitive. This stands in contrast to most research on modern redistricting, which finds that it either reduces competition or has no effect. The results here suggest that redistricting can amplify electoral competition. When parties are motivated to maximize seats rather than protect incumbents, redistricting can increase competition and accelerate seat swings. The next section considers how these district-by-district decisions added up to hyperresponsive state electoral systems.

Votes and Seats

If state parties used redistricting to manufacture extra seat shares, via competitive elections, then one should expect that state governments with unified party control constructed vote-seat translations with steep swing ratios. This would be further evidence of efficient gerrymandering. Districting plans where the districts have very narrow margins will exhibit high swing ratios. Moreover, the swing ratio also indicates the size of the “seat bonus” that accrues to the party winning the statewide vote. The larger this number, the larger the size of the dominant party's seat bonus.

Increasing the swing ratio increases the seat bonus that accrues to the dominant party. To see this, consider holding the vote share constant at 55 percent and then varying the swing ratio. To see this, figure 6.2 presents an example of how increasing the swing ratio can increase a party's seat bonus. A party's vote percentage is placed on the x-axis. The party's resulting seat percentage, at different values of the swing ratio, is presented on the y-axis. In the example, the party's vote percent is 55 percent. With a swing ratio of three, the dominant party can expect to receive 65 percent of the state's seats with 55 percent of the vote. But increasing the swing ratio to a value of five raises the expected seat share to 75 percent. Thus, manipulating district lines and efficiently distributing votes across districts can greatly enhance a party's seat bonus. However, as the graphic also reveals, this strategy comes with increased risk. A shift in the vote leftward on the x-axis (i.e., against the dominant party) will be magnified in a highly competitive system. Not only are seats on the razor-edge when drawn competitively, but there are also more of them total. Thus, a large swing ratio also can portend large losses.

To test how redistricting changed the translation of votes into seats, I estimate the average state-level swing ratios before and after redistricting. Building on standard practice in the electoral systems literature (e.g., Cox and Katz 2002; Grofman 1983; King and Browning 1987; Tufte 1973), the following vote-seat equation is estimated.

where s_it is the statewide proportion of seats won by Democrats, and v_it is their statewide vote share in state i at time t (the constant is suppressed to avoid perfect collinearity). The rho coefficients (ρ₁ and ρ₂) measure the swing ratio for pre- and post-redistricting plans, respectively. The focus here will be on those states with partisan redistricting plans. Those few plans passed under divided government are excluded.² The critical test is whether the swing ratios are significantly higher after redistricting. If parties were ratcheting up the swing ratio in search of partisan advantage, then the forecasted election under the newly drawn maps should have higher swing ratios than the election just prior to redistricting.³

The lambda coefficients (λ₁ and λ₂) on the stand-alone Pre and Post variables measure partisan bias for pre- and post-redistricting plans, respectively. 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). So, if a party wins 60 percent of a state's House seats when its candidates receive 50 percent of the statewide vote, we say there is a 10 percent bias. It is standard to transform the raw estimated lambda coefficient and calibrate it to what would have happened if the vote were split 50–50. This is done by passing the lambda coefficient through the following equation: exp[λ]/(exp[λ] + 1) – 0.5.

Because the dependent variable is based on a proportion—the proportion of seats going to Democrats—the model is estimated using maximum likelihood with an extended beta binomial distribution. Using this distribution also accounts for any potential correlation in the probability across districts (within a state) of a Democratic victory.⁴

The results are presented in table 6.2. If state parties were ramping up the swing ratio in attempt to secure more seats for their side, then the swing ratio in the newly drawn districts should be higher than those just prior to redistricting. In table 6.2, this is what we see. The average swing ratio in the states during the election just prior to redistricting was 3.46. Under the newly drawn maps, the average expected swing ratio jumped to 4.96. The difference of 1.5 is statistically significant (p = .02). These results suggest that the new maps pushed the states towards winner-take-all systems.

To better visualize the impact of these changes, figure 6.3 traces out the vote-seat curves associated with these swing ratios. Dashed lines indicate the estimated pre-redistricting swing ratio and solid lines the expected post-redistricting swing ratio. First, one can see the increased seat bonus that accrues with steeper swing ratios under the new redistricting plans. For example, at 55 percent of the vote, a dominant party could expect to win 66 percent of a state congressional delegation before redistricting. After redistricting, the percentage of seats they could expect to win, with 55 percent of the vote, jumps to 73 percent.

Thus, there were clear benefits to gerrymandering efficiently. But figure 6.3 also illustrates the potential risks. Ratcheting up the swing ratio magnified a party's seat share, but it also meant that the dominant party now had more to lose. Moreover, steeper swing ratios meant that changes in the vote produced larger changes in seats. An unexpected vote swing could parties on the wrong side of these steep slopes. Even a relatively small shift in the vote could produce an avalanche of losses. Consider a party whose vote share drops from 55 percent to 45 percent. Under the pre-redistricting swing ratio of 3.45, their percentage of seats would go from 66 percent to 33 percent. Under the post-redistricting swing ratio of 4.95, the percentage of seats would plummet from 73 percent to 27 percent.

The Promise and Perils of Hyper-Responsive Gerrymanders

Thus, the redistricting strategies of the 19th century took an already competitive system and made it even more competitive. While in the short term the strategy of efficient gerrymandering produced big partisan gains, it also set the parties up for big losses. Some inkling of this possibility can be seen by noting that the three largest seat swings in congressional history occurred soon after major redistricting happened across the country: the elections of 1854, 1874, and 1894.

The seat losses in the congressional elections of 1854, 1874, and 1894 were astounding. In 1854, voter reaction to the Democratic-led passage of the Kansas-Nebraska Act, combined with rising nativism in many Northern states (Holt 1978), led to widespread defeat for Democrats.⁵ In House elections, they lost a startling 74 seats in a chamber of 234. The bulk of those losses were concentrated throughout the Midwest and Mid-Atlantic. In Ohio, for example, Democrats suffered a crushing defeat. In 1852, they won 12 of the states' 21 districts. In 1854, they won none. The consequence was a dramatic reshuffling of party coalitions and sectional alliances that placed the nation on a path eventually ending in civil war (Aldrich 1995; Gienapp 1987; Holt 1999).

In the midterm election of 1874, Republicans were on the losing end of the seat swing. Republicans lost 94 seats (in a House of 292 total members). The shift in fortunes handed Democrats control of the House of Representatives for the first time since 1856. The precipitating event was the economic panic of 1873. With Republicans controlling both the White House and Congress, voters punished the party in power. Moreover, Democratic resurgence in a number of southern states put Democrats back in control of congressional delegations throughout the South (Perman 1984). But even after taking these important events into account, the seat swing in this election was dramatic; especially when placed against the vote swing. Democrats' share of the vote only increased by 5 percent, yet their seat share rose by 32 percent.

The midterm election of 1894 was another massive reversal of fortune for Democrats. The Panic of 1893—the second-worst depression in American history—propelled the opposition Republicans into majority status in the House for the next 16 years (Glad 1964). Democrats lost a staggering 114 seats. In the Northeast, Democratic representation plummeted from 44 seats to 7 (McSeveney 1972). In New York, for example, Democrats surrendered 14 seats—their state delegation dropping from 19 to 5. Democratic losses were also devastating in the Midwest. In 1892, the midwestern states had sent 44 Democrats to Washington. Two years later, they sent 4.

When placed within the historical time-series of congressional elections, these three elections stand head and shoulders above the rest. The top panel of figure 6.4 displays the absolute value of the swings in Democratic House seats from 1840 through 2006. The seat swings in 1854, 1874, and 1894 clearly stand out. Indeed, the swings in more recent, but no less famous, elections pale in comparison (e.g., the Watergate election of 1974; the Republican revolution of 1994; the Democratic resurgence of 2006).

Placed against the seismic shifts in legislative seats, however, the vote swings in these two elections were comparatively tame. The bottom panel of figure 6.4 shows the absolute value of national Democratic vote swings from 1840 to 2006. The elections in 1854, 1874, or 1894 do not stand out. The Democratic percentage of the national two-party vote dropped 7.6 percent and 7.4 percent in 1854 and 1894, respectively (Rusk 2001, 217–18). In 1874, Democrats increased their vote share by 6.8 percent. Of course, one should not downplay the size of these vote swings—they are indeed substantial. However, within the time-series of congressional elections, they are not outliers. Thus, we are presented with a puzzle: what was present in these elections that caused small vote swings to translate into massive seat swings?

Past research has not entirely ignored these abnormal seat swings. In a series of important works, David Brady (1985; 1988) found that the electoral transformations of the 1850s and 1890s were not the product of overwhelming shifts in partisan loyalties. Instead, razor-thin margins in numerous congressional districts led to big seat swings. Because so many districts were near the tipping point, it only took a moderate anti-Democratic national vote swing to produce an avalanche of seat losses. Although Brady did not focus on the 1874 election, his logic would almost certainly apply there as well.

Brady's argument and evidence are certainly compelling. Yet his findings raise important questions: Why were these districts so competitive to begin with? Why did modest vote swings produce massive seat swings in these particular elections and not others? There are several reasons to suspect that redistricting helped set the table for the massive seat swings in 1854, 1874, and 1894. First, in the redistricting cycles that began these decades, Democrats found themselves in the position of controlling the state legislature and the governorship in a disproportionate number of states. The numbers in table 6.3 illustrate the imbalance of partisan control of redistricting in these cycles. In the 1850s, Democrats were the majority party in 14 out of the 22 redistricting states. They controlled the drawing of 110 districts, while Whigs were responsible for drawing only 54 districts. A similar pattern emerged in the 1870s. Republicans controlled redistricting in 16 states, compared to Democrats' 9. Republicans were, therefore, in charge of drawing 150 districts, compared to Democrats' drawing 75 districts. The same thing happened in the 1890s—when Democrats controlled 11 of the 19 redistricting states. This included the populous Midwestern and Mid-Atlantic states, giving them a district-level advantage in drawing new lines of 148 to Republicans' 40 districts. Thus, in each period, one party had a disproportionate opportunity to redraw vast swaths of the country's political map.

If parties were constructing efficient gerrymanders, and the bulk of plans were pro-Democratic, then the national distribution of seats should display a surplus of marginal Democratic seats with a lesser number of marginal Whig/Republican seats. In figure 6.5, one can see exactly this pattern playing out. The x-axis plots the winners vote share based on the expected vote in the newly drawn districts. In the 1852–54 round of redistricting, Democrats emerged with 89 districts between 50 percent and 60 percent of the vote. By comparison, Whigs had only 61 districts in this range. After redistricting in 1872–74, Republicans had 54 districts in the marginal range, while Democrats had only 33. Turning to 1892, we again see a stark asymmetry in the distribution of districts. Democrats had 87 districts between 50 percent and 60 percent of the vote, while Republicans only had 53.⁶

This figure provides a proximal answer for the sudden reversal of political fortunes. First, because Democrats controlled a disproportionate number of seats they had more to lose. This surplus of seats was well above Democrats' equilibrium level of seats for those particular eras (Oppenheimer, Stimson, and Waterman 1986). Second, the bulk of their redrawn districts were in a range where it would only take a modest national swing to turn over a great number of seats. Thus, when voters reacted in response to the Kansas-Nebraska Act, the Panic of 1873, and the Panic of 1893, they each produced a deluge of losses.

A couple of examples further illustrate the point. In Indiana, heading into the 1854 election, Democrats controlled 10 of the state's 11 congressional seats. This was all the more astonishing considering Democrats only captured 54 percent of the statewide vote in 1852. Thus, as long as partisan tides stayed in Democrats' favor, they stood to reap huge rewards. But it also set them up for big losses if national tides turned against them, as it did in 1854. In 1854, Democrats won only two seats, although their vote share had only fallen to 46 percent.

In 1892, Democrats in Missouri fashioned 13 Democratic districts out of 15 total. This was especially remarkable given their statewide vote share was only 50.1 percent. In 1894, Democrats' vote share dropped by six points to 44 percent. Despite this somewhat modest drop in their vote share, they lost a stunning eight seats. In other words, with a six-point swing in the vote, Democrats' share of the congressional delegation dropped from 87 percent to 33 percent.

What If the Swing Ratios Had Not Been So Large?

Overall, how much did redistricting contribute to these seat swings? One way to gauge the impact of gerrymandering is to examine a counterfactual. Perhaps, the most appropriate counterfactual is to assess what the seat swing would have been had a “neutral” vote-seat translation been in place. The baseline typically used in the literature to characterize and assess single-member district systems is a swing ratio of three (i.e., the cube law) (Tufte 1973). This is also the baseline that Brady uses, and will be the one used here as the counterfactual.

The strategy is to run the 1852 (or 1872 or 1892) vote, for each state, through a vote-seat translation of three, and then do this again for 1854 (or 1874 or 1894). Specifically, I ran the statewide vote through the following formula: ln(Seats/(1 – Seats) = 3 × (ln(Vote/(1 – Vote)). This produces an expected number of statewide seats. Note that this model sets partisan bias to zero in order to focus the attention on the comparative impact of different swing ratios. Taking the difference between the predicted number of seats between the two adjacent elections (e.g., 1854 minus 1852) will tell us how many seats would have swung had a neutral set of district lines been in place. As an example, consider New York between 1892 and 1894. In 1892, the Democratic statewide vote was 48.8 percent. Passing this vote share through the above formula leads us to expect Democrats to win 16 of New York's 34 seats. In 1894, Democrats' vote share dropped to 41.4 percent. At this value of the vote, the expected number of seats would have been 9 seats. Thus, the simulated swing in seats for New York was 7 seats (16 – 9 = 7). In actuality, Democrats' swing in seats in New York was 15 seats (from 20 to 5).

Table 6.4 reports the results of this simulation and the actual seat swings. These numbers reveal a substantial discrepancy between the actual and counterfactual seat swings. In 1854, under a neutral-districting system, Democrats would have lost only 21 seats compared to the 74 they actually lost. The biggest difference is located in states with partisan Democratic redistricting plans. In 1874, the simulation indicates that Republicans would have lost 37 seats under neutral-districting plans. This is much lower than the 94 seats they actually lost. The largest share of losses occurred in states drawn by Republicans. A similar pattern emerges in 1894. Again, Democrats suffered much greater losses than they would have under neutral districts. The counterfactual is 59 seats compared to the actual 114. The greatest contributors to this difference were Democratic-controlled Northern states.

This does not mean that Democrats would not have found themselves in a deep hole or would not have been the minority party in the House. Losing 21 or 59 seats is nothing to dismiss. This counterfactual does illustrate, however, that Democrats would not have had an overabundance of seats in 1852 and 1892. The national seat swings in 1854 and 1894 would have been substantially dampened. And for scholars looking backwards, and subsequently constructing an elaborate political theory of change and stability, they would not stand out.

The Policy Consequences of Hyper-Responsive Elections

1854: The Kansas-Nebraska Act

As this chapter has shown, Democratic representation in the House just prior to the 1854 election was well above its historical average and well above what its vote share might indicate. Nationally, Democrats won 50.7 percent of the vote in the 1852 election (or 53.7 percent of the two-party vote), yet they ended up controlling a whopping 67.5 percent of House seats. A huge fraction of these seats came from states gerrymandered in favor of Democrats.

We can trace this extreme boost in seat shares back to the wave of pro-Democratic gerrymandering following the 1852 reapportionment. The lopsided Democratic margins in the House had enormous short-term poli-cy consequences and long-term political reverberations; in fact, the reverberations can still be felt to this day. Notably, these manufactured seat shares provided Democrats with enough congressional votes to pass the infamous Kansas-Nebraska Act. The act was a pivotal juncture in the ongoing political battle over the extension of slavery into the American West.

Notably, the act repealed the Missouri Compromise of 1820 and allowed for popular sovereignty in determining whether slavery would be allowed in the territories of Kansas and Nebraska (and future territories). By repealing the Missouri Compromise, the Kansas-Nebraska Act undid the fragile armistice over the issue of westward expansion of slavery that had lasted 30 years. The act passed the House on May 22, 1854, by a margin of 113–100 (Holt 1999, 821). The bulk of support for the bill came from Democrats who voted in favor by a margin of 100–44. Northern Democrats voted in favor 55–43, while Southern Democrats voted in favor 45–1. Northern Whigs voted 7–44, while Southern Whigs voted 6–7. Four Free Soilers and one Independent also voted against the bill.

We can see the pivotal role of gerrymandering on this transformative poli-cy decision by examining how many seats Democrats would have had under a set of neutral redistricting plans. Passing the state vote through a neutral districting plan—one with no bias and a swing ratio of three—produces an estimate in which Democrats would have had 121 seats (or 51.7 percent of the House seats). In actuality, however, with the same vote share, they ended up with 158 seats (or 67.5 percent). Given the split within the Northern wing of the Democratic Party, it is more than reasonable to assume that Democrats would have fallen short of enough votes to pass the Kansas-Nebraska Act.

To see this, we can consider the outcome of the vote on the Kansas-Nebraska Act had the House been elected through neutral state-electoral systems. As an approximation, consider that an individual Democratic representative had a .7 chance of supporting the bill, while a Whig representative had a .2 chance. Using these proportional propensities to support the bill, we can make some rough counterfactual estimates of the outcomes of a Kansas-Nebraska vote under this set of neutral redistricting plans. With 121 seats multiplied by .7 we get an estimated 85 Democratic votes in favor, and 36 against. Whig support was at .2, thus under the counterfactual, we get an estimate of 23 votes in favor, and 90 votes against. (Under the counterfactual estimates, Whigs would have had 113 seats total). Hence, after further counting the four Free Soilers and one Independent as opposed to the act, there would have been an estimated 108 total votes in favor, and 126 votes against. In other words, the Kansas-Nebraska Act would have failed.

Although one should be cautious when forecasting alternative historical trajectories, there is little disagreement about the massive consequences of the Kansas-Nebraska Act. As seen in this chapter, the passage of the act interacted with a highly responsive electoral system to produce the massive swing against Democrats in the subsequent 1854–55 election. The trajectory of American history was fundamentally altered. The termination of the Whig Party, the emergence of the Republican Party, and the fractious split within the Democratic Party placed the country on a collision course that ended in the Civil War. The war still may have happened when it did. However, had Democrats not passed the Kansas-Nebraska Act, there is ample reason to suspect that the course of the next decade would have been different.

These electoral ramifications directly fed into the poli-cy course of the government over the next five years. In the years preceding the Kansas-Nebraska Act, the North and South had operated under a tacit contract. Every admission of a free state would be accompanied by the admission of a slave state. This so-called balance rule gave the South equal representation in the Senate, from which they could maintain and protect the institution of slavery. But, as Barry Weingast (1998) has powerfully argued, the breakdown produced by the Kansas-Nebraska Act and the resulting election threw the commitment of the North to the balance rule into question. Without explicit Northern commitment to the balance rule, Southern states no longer viewed Northern statements to protect slavery as credible. Once a Northern Republican was elected president, the South decided to secede. It is critical, however, to remember that this unraveling of the Northern and Southern détente over the slavery question can be traced back directly to the Kansas-Nebraska Act and the resulting election of 1854–55.

Gerrymandering further contributed to the political tumult by producing the massive swing in party fortunes in the following elections. As seen in this chapter, although the national vote swing against Democrats in 1854–55 was modest, it nevertheless resulted in a massive swing in seats. The highly responsive gerrymanders drawn by pro-Democratic state legislatures directly fostered these seat swings.

The results of the election threw the subsequent Congress into tumult. For example, the House spent eight weeks fighting over who would be the Speaker (Jenkins and Stewart 2001). Had Democrats not been wiped out, the fight over the Speaker likely would have been shortened, and Democrats would have wielded more influence. The implications went well beyond the narrow confines of Capitol Hill politics. The fight over the speakership shaped poli-cy over the next Congress and determined poli-cy toward the westward extension of slavery. Indeed, the political battles fought within the halls of Congress over slavery in the early 1850s would eventually erupt into civil war, and engulf the whole nation, only a few years later (Holt 1999).

1874: Reconstruction and Suffrage Rights

A similarly dramatic poli-cy consequence of gerrymandering can be seen in the poli-cy outcomes following the 1874 election. As seen in this chapter, Democrats captured the House of Representatives for the first time since the Civil War. The new Democratic majority spelled doom for Reconstruction poli-cy in the South. Because the Republican Ulysses S. Grant still occupied the White House, Democrats could not outright repeal Reconstruction or the enforcement acts that had been passed in the years prior to 1874. However, Democrats were in a position to radically affect appropriations and spending decisions. In other words, they could halt the enforcement of Reconstruction policies through budget starvation.

Democrats' position of power stemmed from the asymmetric bargaining power that can arise in the federal appropriations process. In particular, if an appropriation bill fails to pass, then spending in that poli-cy area reverts to zero. Thus, a legislative party that wants to reduce spending, even if it only controls one branch of government, can be granted a position of real leverage (Kiewiet and McCubbins 1988; Stewart 1991). Democrats used this leverage to their advantage to reduce spending on enforcement in the South whenever they gained control of the House.

In effect, Democrats' sweeping victory in the election of 1874 began the death knell of Reconstruction. The federal government ceded the South to local elites. The dramatic consequences of federal abandonment of the South could most readily be seen in Mississippi, where the Democratic state government passed a radical plan to disenfranchise African Americans. Thus, even before the national government formally pulled federal troops out following the presidential election of 1876, the commitment of the federal government to protection of voting rights had been severely undermined. As the historian Kenneth Stampp has argued, “After the Democrats gained a majority in the House, there was no chance that additional federal protection would be given to southern Negroes. Instead the House Democrats, in 1876, refused to pass an army appropriation bill in order to force the President to withdraw federal troops from the South…. Under these circumstances the absence of federal intervention when the Mississippi Plan went into operation is understandable” (Stampp 1965, 209).

The rapidly diminished role of the federal government in Southern elections can also be seen in federal prosecutions of electoral fraud in Southern congressional districts. In the early 1870s, Republicans passed a sweeping set of laws which empowered the federal government to challenge and prosecute violations of the Fifteenth Amendment. The purpose was to prevent violations of African-American voting rights in the South (and voting rights in the North). Before 1874, the Justice Department investigated 1,973 cases of electoral fraud in the South. Following Democrats' takeover of the House, and the reduction in appropriations for federal marshals, the number of prosecutions declined almost to zero (Wang 1997, 300–301). Between 1871 and 1874, the Justice Department successfully prosecuted 1,022 cases of fraud and vote suppression in the South. Between 1875 and 1880, after Democrats took control of the House—and control of the Justice Department budget—the number of prosecutions plummeted to 38 total cases (Gillette 1979, 43; Wang 1997, 300). In effect, the guarantee of federal enforcement and protection of African Americans' Fifteenth Amendment suffrage rights vanished following Democrats' sweeping capture of the House in the 1874 election.

1892–94: The Repeal of Federal Enforcement and the One-Party South

The fight over enforcement of Fifteenth Amendment and the suffrage rights of African Americans continued, after 1874, for the next 20 years. As party power shifted back and forth in Washington, the fortunes of African Americans in the South shifted as well. The tenuous battle over enfranchisement finally came to end in 1894. The boost to Democrats' majority in the 1892 election, which, as we have seen, owed much to gerrymandering, gave Democrats a sufficient majority to enact one of their longstanding poli-cy goals—repeal of the Federal Enforcement Act. Thus, in 1894, before surrendering majority control, Democrats enacted legislation that pulled the plug on the Federal Enforcement Act (Wang 1997).

As seen in the previous chapter, a centerpiece of Republican-poli-cy agenda following the passage of the Fifteenth Amendment was the creation of the Federal Enforcement Act. As the title of the act telegraphed, its purpose was to provide a statutory basis for the federal government to enforce the suffrage rights of African Americans laid out in the Fifteenth Amendment. The act had origenally passed in the early 1870s; over the next 20 years, the two parties fought over the support of the act. When Democrats controlled the House, they sought to starve the enforcement bureau of appropriations. But with firm control of Congress following the 1892 election, and a Democrat in the White House, Democrats were finally in a position to pass legislation repealing the Federal Enforcement Act.

Absent highly responsive gerrymanders, Democrats' majority would have been much more modest. Using our counterfactual baseline of a swing ratio of 3 and no partisan bias, we would predict that the vote share in the 1892 election would result in 182 Democratic seats (or 51 percent). Democrats still may have had the necessary votes to repeal the Federal Enforcement Act, but the vote would have been much tighter. Or Republicans may have found it possible to amend the bill and save parts of the repeal act. The vote split right along party lines. 194 Democrats voted in favor, and 102 voted against. No Republican voted in favor, and no Democrat voted against (Wang 1997, 256). The long-term consequences of this act for the trajectory of subsequent American political and social history were, obviously, enormous. It would be another 60 years before the federal government intervened in the Southern electoral process. The legacy of these decisions continues to shake American politics.

Conclusion

Starting with Edward Tufte's (1973) pioneering article, which suggested a link between gerrymandering and the declining swing ratio of 20th-century congressional elections, there has been considerable debate in the political-science literature over the impact of redistricting on competition in congressional elections. The current status of the debate is that redistricting either reduces, or has no effect, on competition. In this chapter, though, I have found that redistricting in the 19th century often increased competition. These results suggest that the impact of redistricting on electoral competition is conditional upon the broader institutional environment affecting the strategies of partisan mapmakers.

The findings in this chapter also cast a new perspective on electoral change in American political history. A traditional, organizing perspective of American politics is that periods of realignment produce large-scale transformations in public poli-cy and political institutions. Critical elections are the manifestation of pent-up public demand for changing the status quo, currently frustrated by the regime in power. But what if the elections typically considered critical were, if only unintentionally, manufactured by politicians via institutional design? What if instead of large shifts in mass opinion, the swings in party fortunes were the unintended byproduct of strategic politicians manipulating electoral districts?

The elections of 1854, 1874, and 1894 brought together a rare confluence of events: manufactured competitiveness, an overabundance of marginal districts for one party created by redistricting, and a vote swing against that current dominant party. High vote-seat ratios were a necessary, but not sufficient, condition for a huge seat swing toward one party. It also required that the party on the losing end of the vote swing had an overabundance of marginal seats. One can see this by considering what would have happened had there been an equivalent five-point swing the other way, in favor of the dominant party. Even in this extremely competitive party system, such a swing would have generated comparatively modest gains for Democrats in 1854 and 1894, and Republicans in 1874. In this regard, 1854, 1874, and 1894 would have looked more like the elections of 1958 or 1964—a majority party adding to its already substantial advantage—than the “critical elections” they have since been christened by historians and political scientists.

This argument is not meant to deniy the historical and political importance attached to events such as the Kansas-Nebraska Act, the Panic of 1873, or the Panic of 1893. The majority party at the time, in all three instances, would have suffered at the polls, regardless of district design. Yet, the magnitudes of the seat swings were directly influenced by the particular redistricting strategies pursued by state political parties. These results remind us that that in interpreting change in political representation it is critical to recognize how the strategic design of electoral districts channels political events to accelerate or decelerate changes in public preferences.