Unskewing the System
Chapter 1: Introduction
Observers of recent U.S. elections could note the following somewhat surprising facts:
1. In 2016, 3 million more Americans voted for Hillary Clinton than for Donald Trump for President. Yet Donald Trump became President.
2. In 2012, more Americans voted for Democratic candidates for U.S. House seats than for Republican House candidates. Yet the GOP1 maintained majority control of the House (by 30 seats).
3. A similar comparison for the U.S. Senate is slightly more complicated because Senate terms are staggered and it takes 3 consecutive elections for all 100 Senate seats to be up for election. But in the first 3 elections in this decade (2010, 2012, and 2014), more Americans voted for Democratic Senate candidates than Republican Senate candidates (52%), and yet Democrats won only 46% of the Senate seats.
This seems to go against basic assumptions about democracy and majority rule. Our middle-school civics class understanding of American government has always been that the candidate or party who gets the most votes wins. If they get more votes than the other party, they control. They certainly shouldn’t take a back seat to a candidate or party that gets fewer votes. We can call this the “majority criterion.”
Unfortunately, the above scenarios are not freakish, once-in-a-lifetime occurrences. They happen periodically throughout our nation’s history. They have affected both parties, both nationally and at the state level. In New Jersey, for example, Republican candidates garnered 51% of the statewide vote for State Assembly in 2013 but Democrats controlled 60% of the seats.
How to explain these anomalous results? Although the scenarios above are superficially different, the same dynamic underlies all three. They all divide the electorate into lots of smaller subunits and hold a “winner-take-all” election within each—a situation sometimes called “the unit
rule.”2 Whenever you do that, you create the possibility that there will be a significant deviation between (i) the percentage of votes a party receives, and (ii) the percentage of seats the party receives (or, in the case of presidential elections, the number of states the party carries). In close elections, that “skew” between votes and seats can mean that the majority party becomes the loser and the minority party becomes the winner. But even when this “winner loses” situation doesn’t occur, there is almost always a gap between votes and seats, with the majority party either getting robbed of some seats it otherwise earned, or getting a windfall of extra seats it doesn’t deserve.
In the parlance of political scientists, the culprits are winner-take-all systems. They allow for a majority-supported candidate or party to “waste” its votes—to distribute them in a less than ideally efficient manner. This can be done in two ways: by “packing” and by “cracking.”
Presidential elections create a good example of packing, thanks to the Electoral College. Under our Constitution, we do not vote directly for our President. Instead, we vote to determine whether our state’s Electoral College votes go to one candidate or another, with the winner being the candidate who receives a nationwide majority of Electoral votes. Each state has a number of Electoral votes equal to its number of House and Senate members. Almost all states allocate these Electoral votes on a winner-take-all basis: the party that gets 50.1% or more of the votes in the state gets all its Electoral votes. Because of this, it is easy for a candidate to waste votes by running up an unnecessary and useless supermajority in some states, and coming up just short in others.
See Figure 1.1 as an example. It shows a hypothetical, simple 5-state presidential election, where each state gets 1 Electoral vote. In New York and California, the Democratic candidate gets over 80% of the vote. In Wisconsin, Texas, and Tennessee, she gets just barely under 50%. If you add up all the votes in the 5 states combined, she clearly has a majority overall (even if all states had the same population)—indeed, a 60% total, normally considered a landslide victory. But under the Electoral College system, she loses 3 States to 2.
She wasted her votes because her votes were “packed” into New York and California. In those states, every vote above a simple majority counts for nothing; it does her no good. The votes for the Republican candidate, by contrast, were distributed efficiently: just enough to make a majority in each state, and no more, in just enough red states to make the state-by-state count favor the GOP. On a larger scale, this is essentially what happened in 2016 with Trump and Clinton, and in 2000 with Bush and Gore.
Figure 1.1 A simple illustration of packing
Packing is also at work in the U.S. Senate. Democrats rack up huge supermajorities in populous states like New York and California, with lots of wasted votes. Republicans do likewise in some states like Texas, but Democrats waste more votes like this. At the same time, smaller, less populated states get overrepresented because they get the same number of Senators as larger states. Controlling elections in those smaller and mid-size states thus takes fewer votes for the same number of seats. The party that controls more of those states is more efficiently distributing its votes nationwide. That party currently happens to be the GOP.
House of Representatives elections can illustrate both packing and cracking, the two tools of gerrymandering. A gerrymandered redistricting plan uses both tools to force the “victim” party to waste its votes. It can pack supermajorities of one party in a few districts, draining surrounding districts of the voters needed to get majorities in those neighboring districts. Or it can take a geographic concentration of voters of that party and slice it in half by drawing the district line right through it, with the result that neither remaining concentration on either side of the line is big enough to be a majority in any district.
We have had gerrymandering as long as we have had House districts. But the gerrymandering has been getting worse as computer map-drawing programs have been getting better. Since World War II, there has been an average of 6% deviation between the percentage of the countrywide vote a major party yields, and the percentage of House seats that party receives. For much of that time, the skew favored Democrats. In more recent elections, it has favored Republicans. Much of the skew (though not all) is due to gerrymandering, with almost all districts drawn in such a way that the general election result is a foregone conclusion. In a very real sense, in the House, the voters do not choose their representatives; the representatives choose their voters.
The dysfunction goes beyond the votes–seats skew of the major parties. The Electoral College encourages candidates to focus only on 10 “swing” states and to ignore the rest, a pattern that continues into governance once the candidate is elected.
In both presidential and congressional elections, winner-take-all systems lock us into the tyranny of the two-party system, freezing out any realistic chance for third parties to offer alternatives. Voting for a third party, or indeed any up-and-comer candidate who is not perceived as sufficiently well known, well funded, and mainstream to be electable, is “throwing away your vote.” Because it is winner-take-all, candidates are also encouraged to attack their opponents in a zero-sum environment.
Winner-take-all elections also tend to dilute the voting strength of racial and ethnic minorities. We partially alleviate this by drawing a few single-member districts with racial and ethnic minorities in mind. But they are too few to fairly represent such minorities, and enough could never be drawn because of the way minority voters are dispersed. The overwhelming majority of such minority voters will always live outside such districts, and that problem will only get worse as minority voters move out of the ghettos and barrios. As a civil rights strategy, minority districting perversely depends on racial segregation to be effective.
Single-member districts (SMDs) add their own special dysfunctions. Because gerrymandered districts lead to “safe” seats for Republicans and Democrats alike, general election outcomes are usually preordained, reducing competition, making voting seem futile, and depressing voter participation. The only true competition is in the primaries. This incentivizes candidates to take extreme positions on the left and the right, discouraging bipartisan compromise, increasing polarization, and baking in gridlock.
Every one of these dysfunctions occurring with the House simultaneously occurs at the state and local level in the U.S. (and indeed around the world). It is not dependent on the demographics or political culture of the city or state (or country) involved. It is a universal dynamic, an inherent problem with winner-take-all and single-member district systems.
Can anything be done about all this? The answer is yes, but in perhaps surprising ways. No federal constitutional amendment is needed. The chief hope lies in reforms to be undertaken on a state-by-state basis by each state’s legislature. The usual suspects of reform in these areas are partial solutions at best, or flat-out nonstarters at worst.
Take the Electoral College. In decades past, reformers called for a constitutional amendment to abolish it. But even with the best of causes, amending the U.S. Constitution is a Herculean task, requiring supermajority votes of both Houses of Congress and the state legislatures. With an issue like the Electoral College, which favors some states over others, such that some states would stand to lose out with reform, it is politically impossible.
Fortunately, though, there is a workaround: the National Popular Vote (NPV) Compact. This is an agreement among certain states that they will all allocate all their Electoral College votes to whoever wins the NPV. It takes effect once enough states have signed on to control the outcome of a presidential election. Currently, NPV is almost two-thirds of the way there, after only a decade or so of trying. It is very likely that in our lifetime, we will see the Electoral College not abolished, but rendered harmless.
The Senate is a tougher nut to crack. The fundamental anti-democratic feature (or bug, really) of the Senate is that low-population states get the same representation as high-population states. Wyoming has the same say in ratifying treaties, confirming Supreme Court Justices, and passing laws as California, making a Wyomingan’s vote worth over 60 times more than a Californian’s. Unfortunately, there is nothing that can be done about that. Not only is this bug hard-wired into the CPU, the program cannot be altered by the user. The Constitution itself says that this part of the Constitution cannot be amended. Though there is a partial fix which could help slightly through coordinated action by Congress and state legislatures, it would be a long walk for a short beer. This small- and mid-state bias unfairly advantages states which tend to skew Republican.3
Instead, reformers should be content to change another anti-democratic feature of the Senate: the filibuster. Like gerrymandering, it has only gotten worse in recent decades, with the current practical reality that nothing can get done legislatively without 60 votes in the Senate. Worse, a Senator can single-handedly and anonymously halt all Senate action on a nomination or legislation without ever revealing herself, unless 60 Senators from both parties can come together to break the logjam. Again, no constitutional amendment is needed to change this.
The above solutions—the NPV, filibuster reform—are fairly straightforward. The solution needed to address House gerrymandering is somewhat more involved and requires a basic reexamination of the way we do elections.
Some reformers urge that courts should rule gerrymanders illegal and force states to draw fair district lines for U.S. House (and state legislative) elections. While courts have long been aggressive in policing racial gerrymanders, they have been unwilling to do so with regard to partisan gerrymanders, out of a sense that this is a “political question” unsuitable for intervention by unelected judges. This should change, but so far it hasn’t. The Supreme Court recently passed on some great opportunities to fix this problem.
But even if courts did get more aggressive in policing gerrymandering, we would only be working at the edges of the problem. Courts will continue to be hesitant to get too involved in political fights and micromanage the elected representatives drawing the maps. Perhaps more important, the legal doctrine focuses only on intentional gerrymanders, instances where the proof is clear that the party temporarily in power was deliberately trying to disadvantage the party temporarily out of power. Because of these interrelated factors, under the best scenarios, courts will only address the most extreme of the gerrymanders, leaving a lot of other cartographical skewing intact.
More promising is a solution that is more proactive: nonpartisan redistricting commissions. Such commissions have worked well around the world. Indeed, we are the only Western democracy left which allows elected legislators to draw their own district lines. All others leave it to the technocrats: appointed nonpartisan experts working with set criteria, sometimes with the help of equal number of politicos from both major parties. In recent years, a few U.S. states have been attempting similar approaches.
The results have been mixed. Some states have created plans with a minimum of partisan skew—i.e., a pretty close relationship between votes received and seats obtained. Other states have repeated significant partisan skews anyway. A 2002 study of such redistricting plans showed a median partisan bias in redistricting commissions states of 4.7%, compared to 8.6% in states with traditional politician-drawn maps. Better, to be sure, but there’s still a median skew of almost 5%; since that’s median, there are states with partisan biases higher than that, even using nonpartisan commissions. In a closely divided polity like ours, a 5% deviation can mean the difference between majority control and minority control. Even where the skew doesn’t have such a dramatic effect, it seems intuitively problematic that there is any such skew at all, unless it is necessary for other, sound, democratic reasons.
In some cases, the less-than-stellar results of nonpartisan redistricting commissions can be attributed to bad design of the makeup of the commission or bad execution by the particular individuals involved. But in other cases, deviations occur despite the best structure and best intentions of the participants. This is because of the phenomenon of demographic clustering, known more colloquially as “The Big Sort.”
Since World War II, America has seen a slow-motion interior migration. People have been concentrating in areas where people vote like themselves—not deliberately, to be sure, but as a natural by-product of other demographic and residential trends. Most notably, Democrats tend to over-cluster in urban areas, especially those on the coasts. As a result, they are naturally “packing” themselves. When redistricters draw SMDs, which must have equal populations and which normally are expected to be somewhat compact in shape, the inevitable result is that Democrats are packed, and the resulting plans underrepresent them. Just like the Senate, with its pro-small, pro-rural, and thus pro-GOP bias, the modern House has a natural, structural, anti-Democratic bias.
Even though this demographic clustering is particularly acute in 21st-century America because of the post-World War II Big Sort, it is not limited to 21st-century America. It happens naturally in any advanced, pluralistic society, such that single-member districting inevitably imposes a representational skew. Even Australia’s technocratic Election Commission, the gold standard for nonpartisan commission map-drawing, occasionally gets it wrong due to demographic clustering. In 1990, the Liberal Party won a majority of the nationwide vote, but the Labor Party won a majority of the seats. In 1998, it was the Labor Party’s turn to get robbed.
These “naturally occurring gerrymanders” should not surprise us once we consider how many conflicting demands redistricting places on line-drawers. Under U.S. federal law, they have to draw districts that are very nearly perfectly equal in population, while also fairly reflecting racial and ethnic minority voting strength. Under state law and custom, they additionally must draw maps which keep cities and counties intact as much as possible, and are reasonably regular in shape. All these criteria can conflict with each other.
Beyond that, reformers like to have as many “competitive” or “swing” districts as possible so that general election outcomes are not always a foregone conclusion. Drawing competitive districts often conflicts with the goal of having a number of majority-Democratic and majority-Republican districts that fairly reflects the statewide partisan makeup of voters—i.e., of minimizing the skew. Put cartographic pen to paper, and something’s gotta give—usually a lot of somethings, including important things like partisan and racial fairness.
People simply do not naturally sort themselves into efficient, compact-district-shaped concentrations of voters. Their residential patterns are stubbornly complex, messy, and ever-changing. And the “ever-changing” part brings with it another headache. For, even if you were somehow to manage to balance perfectly all these many and varied criteria, at the end of the decade, you’ll have to start from scratch. (Redistricting occurs every 10 years, after the U.S. Census.)
So, while nonpartisan redistricting commissions, like court decisions policing gerrymanders, are certainly worthwhile reforms, they alone will not fully resolve the problem of unrepresentative election outcomes. This problem is inherent in the idea of carving up the polity into states or SMDs and imposing a winner-take-all construct within each. This is why the courts’ search for intentional gerrymanders addresses only part of the problem. As the failures of even the best nonpartisan redistricting commissions show, unintentional gerrymanders are commonplace. As long as you have SMDs, you will have gerrymandering.
So, what is the solution then? The solution is to avoid drawing districts in the first place. Or, at least, to draw as few as possible. Fewer districts, less gerrymandering, less fighting and litigating every 10 years. We can do that through American forms of proportional representation (PR).
PR can be used in any election where one is electing multiple legislators at the same time to fill multiple seats on a legislative assembly. Under PR, a majority of the votes yields a majority of the seats. And a minority party, or racial group, which makes up 37% of the vote will get roughly 37% of the representation. It’s not “winner-take-all”; it’s “majority take most, and minority take its fair share.”
Why should 50.1% of the vote control 100% of the power? Why should a politically cohesive minority (racial, partisan, ideological, or whatever) that consistently gets 40% of the vote repeatedly get 0% of the power? When the latter happens, that minority gets alienated. It thinks of elections as futile endeavors, and starts participating at a lower rate. Under PR, almost everyone can point to at least one winning candidate and say, “I voted for that person. She is my representative.” Elections are more competitive, and voter engagement improves.
PR is used around the world, at the national, state, and local level. Indeed, every industrialized democracy uses some form of PR for at least part of its nationwide elections—except for the U.S. and Canada, which inherited our outdated election system from the U.K. Traditionalists sometimes ask, “If PR is so good, how come our Founding Fathers didn’t adopt it?” The simple answer: it hadn’t been invented yet.
One can use a variety of electoral mechanisms to implement PR. Many countries using PR eschew primary elections; use parliamentary systems, where the chief executive is chosen by the legislature rather than being elected by the people; run elections where you vote for the party, not the candidate; and other things alien to our American experience. For a variety of reasons, those mechanisms aren’t suitable for use in the U.S. The common objections to those methods (some of them legitimate) are irrelevant to this book. This book argues for the single transferable vote method of PR elections, which in turn requires the use of Ranked Choice Voting (RCV).
Under RCV, voters rank candidates in order of preference: 1st, 2nd, 3rd, etc. Under single transferable vote (STV), these rankings are used in successive rounds of vote-counting to fill multiple legislative seats in such a way that it produces proportional results. The vote-counting rules are designed to minimize wasted votes. This includes both votes for candidates who get too few votes, and are eliminated in early rounds; and “surplus” votes, where popular candidates get way more votes than they need to qualify for a seat on the legislature. These two types of wasted votes are roughly analogous to cracking and stacking, respectively, underscoring how STV helps with gerrymandering-type issues.
In addition to permanently solving the problem of gerrymandering, STV also leads to results more accurately reflective of the popular will. It opens opportunities for third parties, as well as racial, ethnic, and gender diversity.
Australia is a natural experiment to test this proposition. For the last 70 years, it has simultaneously had a single-member-district, winner-take-all system for the federal House, and a geographically overlapping STV system for its Senate, used at the same time and place. The Senate has consistently had more proportional representation of the major parties; has the majority party control; has substantially more proportional representation of third parties; and has a higher percentage of women. The same is true of Germany. Closer to home, STV is used in both Minneapolis, Minnesota, and Cambridge, Massachusetts, with similar results for minorities and women.
The RCV involved in STV has further advantages—not only in House elections, but in elections for the U.S. Senate and President, and at the state and local level as well. RCV creates more opportunities for third parties and lesser-known, lesser-funded candidates. Under a winner-take-all system, voters are legitimately reluctant to vote for such candidates, even if such candidates are their personal favorite, lest they “throw away their vote” on candidates that don’t seem to have a good chance of winning. This of course triggers a vicious cycle: such voter reluctance perpetuates their perceived loser status, so they never get a chance.
Worse, under winner-take-all, a vote for a favored long-shot candidate can actually backfire and end up helping the major candidate you dislike the most: a vote for Green Party candidate Jill Stein is really a vote for Trump, and a vote for Libertarian Gary Johnson is really a vote for Clinton. RCV breaks this cycle. You can list Ralph Nader as your first choice and Al Gore as your second choice. If Nader doesn’t get enough votes to advance, your vote will transfer to Gore. You won’t be helping Bush.
And lest the above examples suggest a pro-Democratic bias, it should be made clear that PR helps all minority groups everywhere. It would help Republicans in Massachusetts and conservatives in San Francisco. Legislatures in all jurisdictions would be more diverse, by race, ethnicity, gender, party, and ideology.
STV is a natural for local elections, in cities small enough where you can run the election at-large. But how would one structure it for U.S. House elections? A bill currently pending in Congress has a decent answer. Under the Fair Representation Act, states with 5 or fewer House seats would run STV elections at-large: no districts at all. No gerrymandering, because no districting. For a state with 6 or more House seats, a nonpartisan redistricting commission would carve the state up into a few multimember districts, each electing between 3 and 5 House members. STV would be used within each multimember district. Districting would be kept to a minimum, as would gerrymandering. Mathematically and cartographically speaking, it is much harder to do real gerrymandering mischief in drawing 3 large multimember districts compared to 15 small SMDs. Even if that particular bill never passes, the general idea of moving to PR, STV, and RCV, on the federal as well as state/local level, is well worth considering.
This book examines the structural electoral issues implicated by the three anomalous results at the top of this chapter. It outlines in more detail the problems and solutions discussed in this chapter, including the legal issues and political practicalities involved. Chapter 2 discusses the Electoral College and the NPV Compact. Chapter 3 discusses the structural representational problems of the Senate as well as filibuster reform. Chapter 4 outlines the extent of House gerrymandering and why it’s now worse than ever. Chapters 5 and 6 discuss two potential solutions to gerrymandering, judicial policing and redistricting commissions, respectively, and explain why they are, at best, partial solutions. Chapter 7 introduces the concept of RCV by discussing its simplest form, Instant Runoff Voting, which is used to fill one office at a time (as in a mayoral race, or in a single-member district). Chapter 8 builds on Chapter 7 to explain how STV uses Ranked Choice Voting to yield proportional results, and why reforms like the Fair Representation Act deserve serious consideration. Chapter 9 discusses how these ideas can apply at the state and local level, and Chapter 10 offers concluding thoughts.
Are these ideas far-fetched? Perhaps some, at least for now. But these reforms are part of a growing national movement. Ten years ago, the NPV movement was purely academic; today its advocates are almost two-thirds of the way toward making it effective. Twenty years ago, the idea of using RCV in the U.S. seemed quixotic. It is now used in over a dozen cities, including major cities, and is used statewide in Maine for major elections like governor and U.S. Senate. STV has gained a foothold in Minneapolis and Cambridge, and the Fair Representation Act itself has seen a growing list of co-sponsors.
Like all fundamental democratic reforms, these ideas will take years or decades of sustained consideration. But such sustained attention is as appropriate as our current democracy’s structural flaws are fundamental.