Sorting Out Democracy – Entirely

What reformers – particularly progressive reformers – most tend to overlook (besides the fact that money doesn’t pick the winners, but people with money like to have given to those who have won) is that a result that looks very much like the result of gerrymandering is inherent in the way Americans sort themselves out in where they live. Legislative bodies that only elect one winner per district end up wasting votes to a maximal degree, cancelling out the votes of all minority blocs within a district but also cancelling the votes of the supermajorities formed by voters in urban districts. Our pattern of reliance on single member districts produces states in which one party can have the majority of votes statewide but get only a minority of the legislative seats. This is the next civil rights battle — not only the nominal right to vote, but the right to have the same voting power for all.

Geography or Gerrymandering? The Origins of Our Polarized District Maps

Chapter five of this report explained how the rising — yet unequal — partisanship of Republican and Democratic-leaning districts has led to the troubling bias in partisan outcomes and scarcity of competition that are now built into elections for the U.S. House of Representatives. Voters, as individuals, have become more partisan: they are less willing to split their votes between parties in a given election or from one election the next, and the ideological overlap between the parties has all but disappeared in our increasingly contentious political environment.[1]

The increasing partisanship of voters has led to less competition and more bias in the U.S. House. However, it is by no means an inevitable result. Rather, the interplay of gerrymandering and geography, in the context of single-winner districts, has ensured that rising consistency in voter preferences has resulted in such negative effects.

In this chapter, we compare the relative responsibility of gerrymandering and political geography for the rising lack of competition and partisan skew that afflicts U.S. House elections, showing that gerrymandering is responsible for a small part of the decline in competition and less than half of the partisan skew. In light of this, we suggest that redistricting reform can only partially mend U.S. House elections, and other bold reforms to our single-winner system are needed in addition to redistricting reform. For the purposes of this chapter, districts are defined by their partisanship, and not by our projections for their incumbents. Incumbency is a major source of non-competition, but the figures in this chapter demonstrate that the underlying dynamics of these races are also trending towards hyper-polarization, and declining competitiveness.

Gerrymandering

Of the factors that can be blamed for increasingly uncompetitive U.S. House districts and partisan skew — gerrymandering — receives the lion’s share of public attention. Along with money in politics, gerrymandering is one of the most-discussed targets for political reform. Commentary about gerrymandering reached a crescendo after the “wrong-way” outcome of the 2012 election, in which Democrats won the most votes across the country, but Republicans came away with a large majority of House seats. In the aftermath of the election, left-leaning political news outlets were quick to identify Republican gerrymandering during the 2011-2012 round of redistricting as the principal cause of the skewed outcome.[2] Gerrymandering is also often blamed for declining electoral competition, both in the media and by respected reform organizations.[3]

While gerrymandering is not the only explanation for declining competition or the persistent GOP advantage in the House, there is no doubt that it plays a role in these phenomena. With the partisan preferences of voters now highly predictable from one election to the next, mapmakers wielding increasingly sophisticated tools have immense power to predetermine outcomes and impact competition. Republicans and Democrats alike went to great lengths during the most recent round of redistricting to draw districts that advantaged their parties, leading to the unrepresentative outcomes and indefensible district shapes seen in states like Pennsylvania and Maryland.

Geography

While gerrymandering gains the most attention, others have identified a second factor that is critical to understanding why increasing partisanship among voters leads to uncompetitive and skewed outcomes: the changing political geography of America. Bill Bishop’s 2008 book, The Big Sort, paints a comprehensive picture of the division of Americans into increasingly homogenous political communities. Decades ago, Republican and Democratic-leaning voters were more evenly distributed throughout the country, in rural, suburban, and urban areas. Today, Americans increasingly live in communities inhabited by people with similar perspectives and worldviews to their own, and have less contact with those on the other side of the ideological divide.[4]

It is worth noting here that while we tend to think of sorting as a voluntary process of seeking out like-minded individuals, sorting is as much about socio-economic constraints as it is about voluntary movement. Many of the minority communities at the center of Democratic leaning districts are still confined by the legacies of redlining, a housing policy that determined which neighborhoods mortgage companies would finance in. Even today, forces such as gentrification, and deindustrialization relocate families along racial and economic lines in ways that are almost entirely out of their control.

Still, as a result of this partisan sorting, not only has the partisanship of individual voters be- come more pronounced, so has the partisanship of different regions. Urban voters have be- come more consistently Democratic, and rural and Southern voters more consistently Republican, driving the partisan balance of districts in these areas further from 50%-50%. No matter how the district lines are drawn, this geographic polarization will inevitably result in declining competition in single-winner districts.

A second effect of geography stems from the fact that, while rural and urban voters have become increasingly polarized, they have not done so to the same extent; urban areas are now more Democratic than rural areas are Republican. Democratic voters are now concentrated in small geographic areas. Consequently, Democratic districts tend to be more Democratic than Republican districts are Republican.

One consequence of these trends is that fewer and fewer counties are competitive. Pew Research Center data shows that the number of counties competitive in presidential elections has declined from almost 700 in 1992, to fewer than 300 in 2016. [4a]

At the same time as many counties have become uncompetitive, the imbalance between the geographic concentration of Democratic and Republican voters has evolved in the last several decades. The imbalance was already apparent in 1996, when Bill Clinton won less than half of counties (49%), despite winning a large majority of the major party vote (Figure 6.1). By 2012, the geographic concentration of Democrats had become far more pronounced. Barack Obama won 52% of the two-party popular vote in 2012, but won in just 22% of counties—fewer than the 26% of counties that Democrat Michael Dukakis won in 1988 while losing the popular vote by 8%. Even in 2008, when Obama won the popular vote by 7%, he carried only 28% of counties.

6-1. Counties Won by Presidential Candidates 1996 & 2016 (Inqvisitor, Wikimedia Commons).

The concentration of Democratic voters in major cities helps explain why Republicans appeared to gain more from the 2011-2012 congressional redistricting round than Democrats. Political geography makes gerrymandering far easier for the GOP. For Democrats to eliminate a single Republican seat in Maryland, they had to draw what is arguably the most convoluted district map in the country. Meanwhile, Republicans in many states gained a similar or greater partisan advantage with maps that were far less egregious.

Gerrymandering may advantage one or other political party more, but modern political geography means that Republican bias can persist even in the absence of any attempts at partisan gerrymandering. Under a completely neutral set of district maps, Democrats will still waste more votes winning highly Democratic urban seats by large margins than Republicans will waste in winning rural seats by more narrow margins, throwing off the relationship be- tween votes cast and seats won between the two parties.

As a result of the effects of both geography and gerrymandering, there are now far more Republican-leaning U.S. House districts (237) than Democratic-leaning districts (198), despite a national electorate that is evenly divided. The number of districts with competitive partisan- ship has also declined consistently, from 134 in 1992, to 78 in the year 2000, to just 45 for 2018.

These serious problems demand action, but the most discussed avenue for reform, in- dependent redistricting, only addresses one issue — gerrymandering. The potential of redistricting reform for solving the twin problems of uncompetitive districts and partisan bias is therefore limited by the degree to which gerrymandering is their ultimate cause. If the impact of political geography is such that the partisan skew and scarce competition currently plaguing U.S. House elections would largely persist even in the absence of gerrymandering, additional bold reforms will be necessary to address them.

Weighing the Evidence

As we documented in our 2016 Monopoly Politics report, gerrymandering, at most, only partially explains the decline of competition in the U.S. House. As shown in Figure 6.2 below, the bulk of the recent decline in the number of competitive districts (those with partisanship within 3% of an even 50%-50% split) occurred from 1992 to 2000, between rounds of redistricting.[5] Despite the absence of any immediate impact from gerrymandering, fifty-six districts became uncompetitive over this period. In comparison, a total of thirty-seven competitive districts were lost immediately following redistricting in 2001 and 2011 (14 after 2001 and 23 after 2011).

Figure 6.2:
Competitive Districts in the U.S. House of Representatives, 1988 – 2016

Some major academic studies on the relationship between gerrymandering and competition suggest that partisan redistricting practices have no impact, or even a positive impact, on the level of competition in congressional elections. Abramowitz, Alexander, and Gunning (2006) find that elections immediately following redistricting saw little change in their levels of competition between 1980 and 2002, even as the overall level of competition in House races declined significantly over this period.6 Instead of gerrymandering, they suggest that the geographic polarization of America into increasingly liberal and conservative enclaves is the primary culprit, along with the increasing inability of challengers to compete with the financial resources of incumbents.[7]

Others have arrived at similar conclusions by comparing the level of competition in recent state legislative and U.S. House races in states with and without partisan redistricting. For example, Masket, Winburn and Wright find that:

“even when controlling for who draws the maps and their varied strategic interests, we do not find any support for the argument that gerrymandering is decreasing electoral competition.”[8]

Similarly, James B. Cottrill concludes that “the data… do not show that non-legislative redistricting either reduced the typical margins of incumbents’ victories of increased the likelihood that incumbents would lose. In fact, the data sometimes run contrary to expectations.”[9]

While it may seem counter-intuitive that gerrymandering does not reduce, and may even improve the level of competition in U.S. House races, this finding is less surprising when one considers the method through which district lines are drawn to create a partisan advantage. To create partisan advantage, a party is better served by dividing their voters into as many districts as possible that can be made “safe enough,” rather than a smaller number of districts that are maximally safe. As a result, mapmakers must often move some of their party’s voters from safe districts and into evenly divided ones, making some of their incumbents less secure, but flipping more districts into their control.[10] (In some instances, gerrymandering might achieve both partisan and incumbent-serving ends. The best example is North Carolina, where the Republican-drawn plan created ten safe Republican districts, and three very safe Democratic districts, in a state that is split 53% Republican to 47% Democratic.)

All this is not to say that redistricting reform cannot have any positive impact on competition in House elections; but the particular characteristics of different non-legislative redistricting mechanisms can significantly affect their impact on competition.[11] Bipartisan commissions, for example, may be more likely to protect incumbents than commissions that are truly independent. Arizona’s redistricting commission, on the other hand, has a mandate that explicitly calls for the creation of competitive districts — and may be expected to produce more competitive maps (when competitive districts can be drawn without sacrificing other goals).

Figure6.3:
The Evolution of the Republican Party’s Disproportionate Share of Safe U.S. House Districts, 1988—2012

The point remains, however, that gerrymandering has not been a principal cause of declining competition, so eliminating opportunities for gerrymandering will not provide a full solution.

Partisan Bias

As was the case with competition, a simple look at the numbers reveals that gerrymandering alone can only partially explain the entrenched structural advantage currently enjoyed by the GOP in U.S. House elections. Figure 6.3 shows the evolution of the Republican Party’s disproportionate share of safe seats in the House between 1988 and 2012, based on district partisanship.[12] While the number of safe Republican seats increased relative to Democratic seats after redistricting in 2000 and 2010, it also increased sharply between 1996 and 2000, and in- creased again between 2004 and 2008—periods in which no redistricting occurred. Given this consistency, it is unclear how much of the significant rise in Republicans’ advantage after 2010 is attributable to gerrymandering, and how much is merely a continuation of ongoing demographic trends.

Fortunately, there are ways to quantify the relative impacts of these different forces. The most direct of these is to simply create neutral, non-gerrymandered maps, and measure their partisan impact.

Disturbed by gerrymandered maps and their sprawling, irrational district shapes, Brian Olson of BDistricting.org created software for the generation of districts that are optimally compact.

Olson’s maps, shown below in Figure 6.4, are consistent with the principles of equal population and district contiguity required of current maps, but the similarities stop there. The software generates districts that are as compact as possible without violating these rules, eliminating the nefarious and complex shapes that typify current districts. With maximal compactness as the only goal, these districts are effectively an “anti-gerrymander,” with neutrality built into their design. Of course, “neutrality” does not mean the district maps are good; these districts flout county lines and communities of interest and many of them may very well violate the Voting Rights Act. But what flaws they have cannot be attributed to gerrymandering.

Figure 6.4:
Optimally Compact U.S. House Districts 13

The partisanship of these districts can be determined by sorting precincts according to the district they are placed into in the BDistricting.org map, and then aggregating the precinct-level election results. In the absence of gerrymandering, any partisan bias that persists in these neutral districts must be attributed to the geographic distribution of voters themselves. Bias that disappears under the neutral maps is likely bias attributable to gerrymandering.

Under the actual U.S. House plans adopted in 2012, there were 25 more safe GOP districts than safe Democratic districts, despite a country that is evenly divided. Under the maximally compact district maps, the GOP safe seat advantage falls by just 8 seats to 17.
If we include both safe districts and lean districts, the GOP advantage falls from 31 seats under the current plan, to 22 seats under the neutral maps. In each case, completely eliminating gerrymandering removed less than one third of Republicans’ structural advantage, strongly suggesting that political geography is the principal explanation for the partisan skew that is built-in to House elections nationally, and in many states, with gerrymandering playing a significant but secondary role.
[14]

This finding is largely consistent with a recent crop of academic research on partisan bias, motivated by the “wrong way” outcome of the 2012 House election. The most directly comparable work comes from Jowei Chen and Jonathan Rodden’s “Unintentional Gerrymandering: Political Geography and Electoral Bias in Legislatures.”[15]

Chen and Rodden employ automatically-generated neutral district maps to measure the level of partisan bias that persists under districts without any partisan gerrymandering. After establishing the degree to which Democratic voters are far more concentrated in highly homogenous precincts than Republicans, they show that the observed partisan bias in their automatically generated neutral maps is highly predictive of the skew resulting from actual state maps. In most cases, Republican drawn maps show a level of skew consistent with that of the neutral maps, and even extreme Democratic gerrymanders are often unable to overcome the natural Republican bias.

The electoral data Chen and Rodden utilize comes from 2000, so they make no attempt to quantify the effects of gerrymandering or geography in recent elections, but geographic polarization of voters has only increased in the interim.

Of those publishing work on partisan bias in the House, Princeton University Professor Sam Wang gives the greatest share of blame to gerrymandered maps.[16]

Wang uses a computer simulation to compare the actual outcome from each state in 2012 with a multitude of combinations of districts randomly selected from around the country. He finds that in the nine states in which outcomes departed significantly from the projected “natural” outcome, Democrats won 9 to 10 fewer seats than predicted, and blames this discrepancy on partisan redistricting. In the remaining 41 states, Democrats won 8 fewer seats than they would have if the distribution of Democrats and Republicans had been symmetrical. Wang attributes the loss of these eight seats to political geography. Together, these departures from proportionality explain the entirety of the 33-seat margin by which Democrats fell short of a majority in 2012.

Nicholas Goedert uses similar methodology and arrives at a similar conclusion, comparing party vote and seat shares from 2012 by state to a historical average seats -to-votes relationship.[17] Like Wang, Goedert finds that gerrymandered districts cost Democrats 9 seats in 2012. However, he also estimates that geographic imbalances cost Democrats an additional 14 seats, nearly enough to make up the margin in the House.

Others have argued that these studies overestimate the impact of gerrymandering. Writing in the Washington Post, John Sides and Eric McGhee take a more straightforward approach, comparing the partisanship of districts before and after the 2011 redistricting process with the results from 2012, and find that changes in district lines cost Democrats 7 seats.[18] However, after controlling for the impact of the greater number of Republican incumbents, the impact of the new districts disappears entirely. Sides and McGhee explain the conflict between their findings and Wang’s by pointing out his failure to control for incumbency effects. Applying this methodology to past elections, they find a significant pro-GOP bias in House elections that has increased gradually since the early 1990s, and mostly between rounds of redistricting.

Taken together, the recent research on sources of partisan bias suggests that political geography, not gerrymandering, explains more than half of the Republican advantage currently built into U.S. House elections. The rest results from the indisputable fact that, across the country, the concentrated distribution of support for the Democratic Party in urban areas makes it far more difficult for them to convert votes into congressional seats.

Implications for Redistricting Reform

Polarized district maps and lower rates of ticket splitting have led to a dramatic decline in competition in House elections. The partisan imbalance of this polarization has resulted in a persistent advantage for one of the two major parties. However, the research presented here and the bulk of scholarly work on these subjects suggests that the most commonly proposed solution, ending gerrymandering through redistricting reform, fails to address the primary cause of these issues.

To be sure, partisan gerrymandering exacerbates geographic polarization, worsening its effects, and there are numerous other reasons why ending gerrymandering through redistricting reform would be a positive step — public trust in the democratic process chief among them. Redistricting reforms that establish the creation of competition or the production of representative partisan outcomes as explicit criteria could no doubt bring some improvement, but only separately. Under our current electoral structure, these goals are in direct conflict with one another; more competitive single-winner districts could allow for even more distorted partisan outcomes, and proportional results can only be guaranteed in the absence of widespread competition. Even if the sole goal of redistricting reforms is to eliminate gerrymandering for national partisan advantage, the failure of even maximally compact computer-drawn districts to significantly ameliorate the distortion in our current system underscores the need for more ambitious solutions.

The current failures of elections to the U.S. House are the result of flaws inherent in the use of winner-take-all single-winner districts. These flaws have been part of the American system from its inception, but have been exposed in dramatic fashion over the last three decades by the increasing geographic polarization of voters. Under winner-take-all, only the most balanced districts will be competitive in a polarized political climate, and the geographic distribution of party’s voters will always play an important role in determining their ability to win representation. However, reforms exist that could reshape our system into one in which far more voters participate in meaningful elections, and there is a more direct link between a party’s share of support and their strength in Congress.