This article appears in the Fall 2018 issue of The American Prospect magazine. Subscribe here.
Americans are starting to catch on to the fact that our system of “first past the post” plurality voting in single-member districts can lead to perverse results. The citizens of Maine, for example, got stuck with an unpopular reactionary governor, Paul LePage, after he was elected in 2010 with just 38 percent of the vote in a three-way race.
Out of frustration, Mainers have since instituted an alternative system called ranked-choice voting (RCV). With RCV, voters do not just pick one candidate; they rank all the candidates in order of preference, from most favored to least favored. The candidate with the most first-choice votes wins outright only if he or she gets a majority of those votes. Otherwise, voters’ second choices come into play. (See sidebar below.) In a moderately conservative state like Maine, RCV would usually mean a more centrist or middle-of-the-road official would win, rather than the far-right LePage. In a more progressive state or district, it would also elect more representative officials—in that case, more progressive ones.
Despite repeated efforts by Maine’s Republican establishment to block RCV, citizens of the state have twice passed referenda in favor of it. They recently decided to continue to use RCVin federal congressional elections and state primary elections. Ironically, as a result of a decision by the Maine Supreme Court, the system does not apply to the general election that motivated the reform in the first place—the election of the governor.
We believe that ranked-choice voting, which a number of cities have also adopted for local elections, could help elect members of Congress who better reflect the preferences of their constituents. (RCV could also help us elect presidents more democratically, but that is a topic for another time.) For the House of Representatives, RCV would help reduce the influence of Tea Party–type extremists, thereby reducing party polarization and gridlock, and it would produce government policy better aligned with the wants and needs of all Americans.
The problems of unrepresentativeness and polarization call for further reforms as well, which should include more open ballot access for candidates and participation of all citizens at each stage of the electoral process—goals that we believe would be best achieved by abolishing primaries. In addition, Americans should consider adopting a variant of the method used by most democracies in the world for electing their national legislatures: a system of proportional representation through multi-member districts, which would help ensure that the House of Representatives actually represents every political view embraced by a substantial number of Americans in proportion to the voters who support it.
Maine Governor Paul LePage was elected in 2010 with just 38 percent of the vote in a three-way race—a result that spurred the state to approve an alternative system called ranked-choice voting.
This is a big agenda, not likely to be enacted overnight. But in the interest of overcoming widespread dissatisfaction with government, Americans may be ready to consider changes in elections that have long been off the table. Maine is not the only state to adopt a major reform; both California and Washington state have adopted an alternative election format known as “top two.” To sort out the different approaches and see how several reforms might work together in elections for Congress, it makes sense to consider how American congressional elections have been going awry.
IN A TYPICAL PRIMARY election, a small portion (usually just 15 percent to 20 percent) of each major party’s eligible voters choose two nominees for the general election. Candidates who might draw some support from both parties’ voters as well as from independents have little chance of winning either party’s primary. In states or districts that are dominated by one party, the officeholders who emerge tend to reflect the preferences of their party’s primary voters, donors, and activist organizations. As Michael Barber has shown, U.S. senators’ voting in Washington most closely reflects the preferences of their campaign donors, less closely the preferences of their same-party voters, and barely at all the preferences of their state’s voters taken as a whole. The Senate as well as the House could benefit from RCV and the abolition of primaries.
Even in states or districts with relatively even partisan divisions in the electorate, the Republican and Democratic candidates who make it through their party primaries tend not to represent the average voter in their state or district. What results instead is what Joseph Bafumi and Michael Herron call “leapfrog representation,” in which moderate districts are not represented by moderate officeholders but vacillate between Democrats and Republicans who stand to the left and the right (sometimes the far right), respectively, of the majority of voters.
In short, major party primaries, besides stifling third parties, tend to prevent the election of centrist candidates—who might appeal to the largest number of voters in their state or district—and give disproportionate power to small groups of party activists, donors, and interest groups.
California and Washington made some progress in addressing this problem when they adopted the “top two” nonpartisan nominating system. In those two states, primary elections are now open to candidates and voters of any party (or no party)—an excellent idea. The two candidates with the most votes win places on the November ballot. The hope is that in heavily one-party districts, if the top-two system produces two nominees from the dominant party, at least one of them will have broader voter appeal than the extremists and interest group–funded candidates that the old system tended to come up with. Then, in the general election, voters from the minority party will presumably join in a majority to elect the candidate who is more representative of the district as a whole.
But “top two” has not always realized this hope. For one thing, it has some of the same defects as straight plurality voting for a single winner. Each voter must pick just one candidate. So if several similar candidates split the votes of their supporters, less-popular candidates may prevail, just as LePage did in Maine.
That problem could be addressed by using ranked-choice voting in nonpartisan primary elections. If voters rank a number of candidates, both candidate A (with more first-choice votes than any other candidate) and candidate B (with the next-highest number of first choices) might well win places on the November ballot, just as in the top-two system. At least one of those two might be reasonably representative of the average voter in the district. But it is also possible that candidate C could win fewer first-choice votes than either A or B only because many voters’ first-choice votes were split between C and highly similar candidates D, E, and F. If C was the second choice of many of those voters, majorities might prefer C over both A and B. In that case, C should go to the November runoff. Top-two would eliminate C, but RCV would forward C to the general election. Consequently, the use of RCV in primaries as well as general elections would lead to more representative results than either plurality voting or even top-two.
BUT IF WE MERELY USE RCV in primaries and general elections for single-member districts, we will not have dealt with the small, heavily partisan, and unrepresentative nature of primary electorates, or with the lack of representation of minority viewpoints that is inherent in single-member districts.
In order to deal with unrepresentative primaries, it would be helpful to take a second step and eliminate primary elections altogether, nominating and electing members of Congress in one unified process through “instant runoffs” in November. Ranked-choice voting would allow citizens to evaluate candidates from any party—or no party—who qualified for the ballot by gathering petition signatures (as is currently the case in California and many other states) rather than by party endorsements. Candidates could choose to list a party affiliation on the ballot as an aid to voters, but that affiliation would not reflect any formal endorsement by the party. The winner should be the candidate who was ranked above all other candidates by majorities of voters in head-to-head comparisons.
Eliminating primary elections would advantage candidates who appeal to as broad a swath of their districts as possible. It would have other advantages as well. Americans are called upon to vote far more frequently than citizens of any other democracy, one of the sources of lower voter turnout in the United States. Eliminating primaries would reduce the burden on voters. November general elections typically see considerably higher turnout and a more representative electorate than do primaries, which badly under-represent lower-income citizens and ethnic minorities. Focusing voters’ attention on one high-stakes general election should help maximize turnout.
Further, eliminating primaries would reduce the power of small but intense cadres of extreme ideological activists. In low-visibility, low-turnout primaries, such groups can flood social media and turn out their supporters to nominate an extremist candidate. That is how Tea Party favorite Dave Brat ousted Eric Cantor—a conservative himself, but less extreme than Brat—in a Virginia Republican primary, despite Cantor’s big win in the previous general election and his closer fit with the district.
This November-only system, with open, petition-based nominations, would give voters more choices. With RCV, there would be no problem of “wasted” or counterproductive votes for third- or fourth-party candidates. A voter could rank her or his genuine first choice first, and if that candidate did poorly, the voter’s second preference would count. Third or fourth parties would no longer be blatantly discriminated against by the electoral system.
The openness of the system would put pressure on both major parties to pay more attention to all the voters in their states or districts, and less attention to their party donors and activists. Republicans would face more centrist pressures, while Democrats would have to pay more heed to progressive economic views. Large majorities of Americans hold progressive opinions on jobs, health care, the minimum wage, progressive taxation, bank regulation, and many other issues, yet those views are currently slighted by many Democratic officeholders, just as they are by nearly all Republicans.
If the major parties did not respond to the citizenry, new parties or independent candidates could challenge them. That threat would pressure the major parties toward democratic responsiveness.
Yet the two major parties would by no means be excluded from elections. We expect that they would continue to do most of the work of vetting, endorsing, and campaigning for candidates they favor, just as they do in nonpartisan primaries in California. Such candidates would generally choose to list a party affiliation on the ballot and would tend to win high rankings from most of their fellow partisans.
The single transferable voting system for counting ranked-choice votes has had a generally successful history of use in Australia, Ireland, Malta, and many U.S. communities and private groups. Here, voters in Australia cast ballots for the Senate.
STILL, EVEN WITH THESE reforms, certain problems will persist as long as we stick with single-member congressional districts. For one thing, electing just one candidate from each district—even electing the one who is most representative of the district’s median voter—would tend to leave Congress with many centrists but too few members with distinctly minority views or minority demographic characteristics. In order to get full representation of the whole range of diverse views and diverse groups in America and to enrich legislative deliberations in the House of Representatives, a third step would be necessary: moving to a system of proportional representation. In the United States, that can most feasibly be achieved by setting up multi-member congressional districts within each state.
Multi-member districts could also help with the nagging problem of “naturally” one-party districts that result from voters of similar political views being clustered together geographically, as they are, for example, in heavily minority urban areas. These lopsided districts waste votes for the Democrats: A district that is 90 percent Democratic can elect only one Democratic member of Congress, when considerably fewer Democrats in a less lopsided district could do that job (the rest could vote elsewhere and affect those elections). “Naturally” lopsided districts contribute to the perennial Republican edge in the number of House seats the GOP wins relative to the proportion of votes the party gets nationally. Such districts are bad for democracy regardless of their particular partisan effects.
Although independent redistricting commissions can and should help get rid of partisan gerrymanders, they cannot solve the problem of naturally lopsided districts even if they try to maximize two-party competition. In a heavily (say, 60 percent) one-party state, maximizing two-party closeness would be impossible except in a few districts (which ones?) and would leave the others extremely lopsided. Or, if the maximum amount of competitiveness were sought uniformly throughout the state, there would be a 60-40 balance in every district, every one of which would likely elect a member of Congress from the state’s majority party. The 40 percent minority party might be completely shut out of the state’s congressional delegation.
Proportional representation American-style, as we envision it, could solve this and other related problems. The system would start with all the reforms discussed above: involvement throughout the process by all citizens (no one-party primaries); open ballot access through petitions; ranked-choice voting; and instant runoffs that combine nominations and election in one November-only contest. It would add (in large states) mega-districts electing four or five representatives at once, and (in smaller states) a single statewide district electing all of the state’s representatives at once. It would use voters’ rankings more comprehensively, not just to pick a single, most-preferred winner, but to select a set of the most preferred candidates for all the available seats in Congress, through a “single transferable vote” (STV) counting system. (See sidebar below.)
In a four-seat mega-district that was evenly divided between Republican and Democratic voters, the result might be to elect to Congress one very progressive Democrat, one moderate Democrat, one moderately conservative Republican, and one libertarian, with roughly the same mix of races, ethnicities, and genders as in the mega-district as a whole. A more progressive district would elect more progressives, perhaps including minor-party progressives.
If this reform were implemented throughout the country, it should result in a House of Representatives that actually looks and thinks like America. It would be a House that represents all views and all demographic characteristics in close proportion to their shares in the whole adult population of the United States.
To be sure, reformers would need to address a variety of specific problems. It is essential to limit the number of seats to be filled and the number of candidates running for those seats in each mega-district, so that voters are not burdened with a hopelessly confusing number of choices, and so that promising second- or third-choice candidates do not get lost when the candidates with the fewest first-place preferences are eliminated by STV. This would also help each citizen keep in touch with his or her “own” representative (not based on geographical closeness, which has lost much of its relevance, but perhaps on ideological-, gender-, or ethnicity-based closeness) to deal with personal problems or requests.
The number of seats up for election can easily be specified by law. The number of candidates will be a function of the number of signatures required to qualify for the ballot—a number that must be high enough to prevent an unwieldy number of candidates but low enough to allow access to a wide range of office-seekers. California seems to have hit a plausible balance: The median number of primary candidates in each of California’s 53 U.S. congressional districts in 2018 was four.
The STV system for counting ranked-choice votes has had a generally successful history of use in Australia, Ireland, Malta, and many U.S. communities and private groups. But besides avoiding excessive numbers of offices to fill or too many candidates seeking to fill them, the precise design of ballots needs to be addressed.
For example, while party labels can be helpful to voters, making it too easy to vote for an entire one-party slate might damage proportional representation. If too many voters made rigid, party-line preference rankings there would be less chance of candidate merit overcoming party loyalty. True, ranked-choice voting would solve a big problem that bedeviled early-19th-century multi-member districts. (Back then, each citizen cast as many votes as there were seats in their state, and each party listed its candidates on a single slate. The dominant party in the state generally won every seat—a very undemocratic outcome.) Even with totally party-line voting, RCV would at least produce a congressional delegation whose partisan makeup reflected the partisan proportions among voters in the district. Still, we think it would be a mistake to let voters put a fixed list of candidates from one party in high positions in their preference rankings simply by making a single check-mark or a single click. Better to encourage voters to think about, and distinguish among, the merits of the individual candidates affiliated with their party (and, if they choose, candidates not so affiliated).
Advocates of federalism and states’ rights should not worry that a federal law mandating proportional representation would mean too much intrusion into the business of individual states. Mega-district boundaries could be left up to the states, requiring only that each small state elect all its representatives (by RCV) in a single statewide district, and that all large states carve out a mega-district to choose at least, say, three or four and no more than (perhaps) five or six members of Congress. Under proportional representation, the precise sizes and boundaries of districts would not matter much. Oddities would mostly cancel out across districts. Opportunities for gerrymandering would be minimal.
Supporters of ranked choice voting deliver petitions to the State House in Augusta, Maine.
No constitutional amendment would be required. This system of proportional representation could be adopted by a federal law (as the current system of single-member districts was in the mid-1800s).
No doubt this proposal would arouse substantial political opposition, based on incumbent protection and partisan or ideological worries, even if most of those worries are exaggerated. Achieving proportional representation American-style might require persistent pressure over a number of years from a broad, sustained social movement, just as was needed in the 20th century to win direct election of senators, voting rights for women, and enfranchisement of African Americans. But we are optimistic that major reforms will be once again possible as more people come to recognize how our electoral institutions have contributed to the problems of polarization, gridlock, and unresponsiveness that beset American democracy today.
How Ranked-Choice Voting Works
In Maine, voters assign preference rankings to as many of the candidates on the ballot as they like, as their first choice, second choice, and so on. If there are multiple candidates for a single office, the candidate with the most first-choice votes wins only if she or he has a majority of all the first-place votes. If not, the candidate with the fewest first-choice votes is eliminated. Voters who had preferred the eliminated candidate then have their votes transferred to their second-choice candidate. If no candidate still has a majority of the votes, the process is repeated as many times as necessary, with the votes cast by people who had backed each eliminated candidate transferred to the still-viable candidate that they rank highest.
How the Single Transferable Vote Works
Ranked-choice voting can be extended to fill multiple seats, as in a multi-member congressional district.
Here, too, each voter ranks as many candidates as she or he likes, assigning first choice, second choice, third choice, and so on down to the voter’s last-choice candidate. But the rankings are used to pick a setof winners, as many winners as there are seats to be filled. The best way to do this using the rankings is by the single transferable vote system.
First, a threshold is established for the number of votes needed to win a seat, which depends on the number of voters and the number of seats to be filled. (When picking two winners, the threshold is one-third of the number of voters; with three seats, it is one-quarter, and so on.) In the first round of counting, any candidate or candidates with more first-choice votes than the threshold wins a seat. If those candidates have excess votes above the threshold, their extra votes are transferred proportionally to the second-choice candidates of their supporters, and any candidates whose new totals exceed the threshold are elected. (This prevents voters from being “punished” for wasting their top preferences on a popular candidate who would have won even without their support, and helps eliminate any incentive for strategic voting.)
If all the seats are not filled in this manner, the candidate with the fewest first-choice votes is eliminated, and that candidate’s votes are transferred to their supporters’ second choices. After this transfer, any candidate who is pushed over the threshold wins a seat, those candidates’ excess votes are redistributed, and the process repeats itself until all the seats are filled.