Fixing Top Two with Open General Elections: the Colorado Innovation

by Rob Richie, Drew Spencer // Published May 27, 2014



Opens the primary election?

Preserves political party association rights?

Opens the general election?*

Top Two
(California and Washington)




Oregon  Unified Primary
(with approval voting)




(“Cajun Primary”)




Colorado “Top Four”
RCV Model




* By "open general election" we mean one with a meaningful competition among diverse candidates.
** The “Cajun Primary” eliminates the primary election and adds a runoff election in December.

Opening primary elections to include all voters should not mean closing general elections to meaningful competition when most people vote. But that is exactly the impact of the top two primary systems that are used in California and Washington, as well as the two proposals being circulated as potential ballot measures in Oregon this year. Fortunately, there is a better alternative: Colorado reformer Ryan Ross’s groundbreaking proposed ballot measure to open both primary elections and general elections by advancing more than two candidates and using ranked choice voting in November.

It is problematic that many taxpayers have to pay for primary elections in which they cannot participate due to their decision to remain unaffiliated or registered with a minor party – a number of people that is growing as more and more people decide not to register as members of a major party. In states like New Jersey, nearly half the electorate is denied the right to vote in primaries because those primaries are closed to independent voters.

This disenfranchisement, along with the increasingly exaggerated influence of primary elections in our winner-take-all elections (in which fewer and fewer general elections are meaningfully competitive), has led to a movement to open the primaries.  Its high point has been the “nonpartisan blanket primary,” or, as it is usually called in California and Washington, “Top Two.” All candidates appear on the same ballot irrespective of political party, and candidates in each race qualify for the general election without regard to political party. Louisiana uses something similar in which all candidates appear on the general election ballot without a primary, and the top two candidates face off in a December runoff election if there is no first round majority winner.

Unfortunately, this opening of primaries has typically coincided with a closing of general elections. California and Washington follow a restrictive “Top Two” model, in which only the two candidates with the most votes in the primary qualify for the general election – every other candidate is eliminated and unable to appear on the general election ballot. Among other results, that has meant that only one single non-major party candidate has reached the November ballot in a congressional or state legislative race in California in which at least one candidate from each major party had fielded a primary candidate, and none have in Washington.

The Problems with Top Two

Voters often view the general election as the “real” election. The presidential election takes place in November, and November is the election that decides which candidates will actually be seated. Top two preliminaries are only “winnowing” elections that, while hugely important, are less obviously interesting to most voters and the media. Consequently, no matter how open the primaries have become in Washington and California, most registered voters simply don’t seem very interested. A May 2014 report by Eric McGhee of the Public Policy Institute of California found that “turnout in the June 2012 primary was one of the lowest on record” and that the primary electorates are generally older, whiter, and more conservative. In 2012, California hit a record-setting low for voter turnout in a presidential election year, and FairVote predicts that California’s second “top two” primary on June 4th will set an all-time low for midterm primary turnout.

McGhee’s report remains agnostic as to whether voters will adjust their perception of elections and begin voting in higher numbers in the now-critical primary election. However, after three election cycles with Top Two in Washington, turnout in the primary remains at about half of the general election turnout, including only about 38% of registered voters in August 2012. While turnout in primaries was considerably higher in the past, nationwide trends suggest that a growing number of voters will only vote in one election per cycle: the November general election.

No matter how open the primary election is, general elections matter, and Top Two closes off the decisive general elections from meaningful competition. Democracy requires a competitive election in which the most significant perspectives on governance can compete in the public arena. A single-party election disrespects this principle by leaving out significant ideas. For example, suppose the 2008 general election for President featured only Barack Obama and Hillary Clinton, to the exclusion of not just minority “third party” candidates, but also all conservative and Republican viewpoints. Such an election would have lacked the contest of ideas and chance for people to join with like-minded people in support of a candidate who represents their views. But if you look at the results in the first primary that year in New Hampshire, that’s exactly what would have happened with a "Top Two” rule. Both Obama and Clinton had more votes than top Republican John McCain, and in a top two system, they would have been the only candidates to advance to the general election even though New Hampshire is a narrowly divided state.

Despite the openness of its primary election model, the Top Two system can leave voters with this very situation in the November general election. It happened in a particularly bizarre way in 2012 in the thirty-first congressional district race in California where only Republicans advanced in a majority-nonwhite district that voted over 57% for Barack Obama in November.

Of course, most Top Two elections are still one Republican versus one Democrat – indeed, that has been the case in 46 of 47 partisan races in Washington state’s congressional and statewide elections since adopting the system. At best, the potential of an “intraparty” (two Democrats or two Republicans) general election may change the behavior of incumbents who today only fear a potential primary challenge, but it comes at a real cost of excluding all candidates from outside the two major parties from the ballot in November. In trying to expand democracy, we should not accept a cost that limits it even more.

As growing dissatisfaction with Top Two becomes apparent, proponents of opening the primaries have begun to seek tweaks to the system.

The Oregon Proposals

In Oregon, two proposals to establish a top-two primary are in the petition phase.

Mark Frohnmayer, a businessman from a political family, leads a group gathering signatures for a proposal similar to Top Two, but with two key differences. The first, which we support, is that party association rights are preserved, allowing candidates to indicate their preferred party, but also allowing parties to indicate their preferred candidates. The other innovation is the introduction of approval voting in the first round of voting: voters may cast as many votes as they want to in the preliminary election that winnows the field to two. Last October, we published an article noting what these modifications would and would not change. Frohnmayer posted a comment to that article in which he made clear that he thought the use of “approval voting” would resolve all the major issues with Top Two.

We disagree. Judging by the prior uses of approval voting by private associations and universities – limited, to be sure, but the only true tests of the system to date – Frohnmayer seems to be wrong about the impact approval voting would have. Those uses typically show voters “bullet voting” for only their favorite candidate to boost that candidate’s chances more often than is intended by the system, and candidates frequently win with much less than 50% support among voters.

There are very few approval voting elections that are meaningfully contested in a single round of voting, but we found three that took place this year. At Dartmouth, the student government president was elected with less than 40% support and an average of only 1.17 votes per ballot – that’s the third straight Dartmouth student election where more than three in five voters did not approve of the winner, whereas Dartmouth previously used ranked choice voting and had winners consistently with much higher shares of support.

At the University of Colorado, fewer than 1% of student voters voted for more than one candidate for President in a three-candidate race in the university's third approval voting elections, with the winner having support from 45.6% of voters; furthermore, no winners in the multi-seat student council elections earned approval from even 40% of voters. This spring, the Bitcoin foundation held an election with approval voting for two board seats. Even with 15 candidates, two openings, and the right to approve of as many candidates as one wanted, the average number of votes cast was less than two – that is, less than the number of seats. No candidate earned 50% of the vote, triggering a runoff between the top three candidates.

Perhaps Frohnmayer is right and things will work differently under his proposal than prior experience suggests. But if approval voting were to affect outcomes, it likely would result in closing off the general elections even more than the Top Two elections in Washington and California. As an illustrative example, consider the 2008 U.S. Senate primary election in Oregon. It took place at the same time as the presidential primary in which the close contest between Barack Obama and Hilary Clinton drew far more Democrat voters than Republicans (whose nominee had already been decided).

The Democratic Senate primary featured two frontrunners, while the Republican primary featured only the incumbent, two-term Senator Gordon Smith, and one longshot challenger. About twice as many Democrats voted in the primary as Republicans, with the top two Democrats each receiving over 240,000 votes, while Sen. Smith received about 290,000 votes. Had these candidates all appeared on the same primary ballot with approval voting, even modest rates of approval voting among the Democrats (let’s say one Democrat out of five voting for both of the major Democratic candidates) would likely have resulted in them both defeating the Republican incumbent. The general election then would have excluded all Republican voices, including an incumbent officeholder who was a strong candidate and ultimately lost the general election by only 3.4%. That’s not politics as we think it should be – general elections should offer a real debate and real choices for voters.

Another feature of Frohnmayer’s plan could prove highly problematic: how it deals with deceased candidates. Such tragedies unfortunately can happen, including when Sen. Paul Wellstone (D-MN) died in a plane crash weeks before the 2002 general election and, earlier this year,  Keith Crisco died just after earning a spot in the runoff against former American Idol contestant Clay Aiken to be the Democratic nominee in North Carolina’s second congressional district.

In a traditional party system, a deceased candidate can be replaced by another candidate of the party. Washington State preserves this option in its Top Two system (accepting the candidate’s voluntary expression of preference for a party as definitive), while California simply keeps the deceased candidate’s name on the ballot and holds a special election later if that candidate wins. But in Frohnmayer’s plan, the candidate in third would advance to fill the opening. That might seem fair until one realizes that many times a strong incumbent of one party may not draw any remotely serious challengers from his or her party. This year, for example, California governor Jerry Brown is overwhelmingly favored for re-election, with two Republicans very likely to finish second and third in the primary. Under Frohnmayer’s plan, those two Republicans would likely be the only November candidates if Gov. Brown were to die or withdraw.

To his credit, Frohnmayer has been creative in thinking outside the “vote-for-one” box. The protection of party association rights is also a notable improvement on the Washington and California Top Two model. In Washington and California, candidates completely out of step with a political party can still run with that party’s name right next to their own, just because they chose it. Under Frohnmayer’s proposal, candidates have their political party registration by their names, but the political parties also get to choose which candidates they will endorse, and the endorsement (if any) appears on the ballot as well.

The other Oregon proposal this year comes from James Kelly, a well-connected businessman, who is sponsoring petition #55, which sets up a Top Two system similar to Frohnmayer’s proposal, except that uses the “vote-for-one” method used in California and Washington. Like Frohnmayer’s plan, this proposal commendably protects association rights, but also like Frohnmayer’s plan, it unfortunately advances the third-place candidate in the event one of the top two dies or withdraws. These two changes aside, we can predict that its use would play out very similarly to the flawed Top Two election systems in Washington and California.

The Colorado Alternative: Top Four with Ranked Choice Voting

In Colorado, reformer Ryan Ross leads the Coalition for a New Colorado Election System. The Coalition is collecting signatures right now for a ballot measure to establish a nonpartisan blanket primary that corrects all the defects of other proposals. It preserves association rights, advances more than two candidates to the general election, and uses ranked choice voting in November to allow for more voter choice. We hope that Colorado voters will help put the Coalition’s plan on the ballot and support it this fall.

coloradocoalitionlogoFairVote has suggested altering Top Two to advance more than two candidates under a system we call “Top Four.” With four candidates advancing, all significant viewpoints should be represented, and the use of ranked choice voting in the general election will retain the majoritarian goal of Top Two and mitigate vote-splitting among similar candidates.

The Colorado initiative accomplishes most of these key goals. Unless one candidate earns more than 50% of the primary vote, it will advance three candidates automatically. It will also advance all candidates who poll at least 3% in the first election, which will sometimes result in more than four candidates appearing on the general election ballot in November. Then, in November, voters will be able to rank up to three candidates in a ranked choice voting election.ain the majoritarian goal of Top Two and mitigate vote-splitting among similar candidates.

Real diversity of opinion in the general election combined with ranked choice voting should create meaningfully contested general elections. Our simulations of Top Four with California and Washington election results demonstrate that nearly every election would include candidates from both major parties as well as more than one candidate from the larger of the two major parties, and independent and minor party candidates would appear on the general election ballot far more often. Also, mounting scholarly evidence suggests that ranked choice voting can lead to more civil elections, with candidates incentivized to campaign to voters rather than against other candidates. Like Frohnmayer’s Oregon plan, the Colorado proposal also preserves association rights for parties.

Though we support Ross’ proposal overall, we do have three real concerns. The first is the application of the nonpartisan blanket primary to the presidential election. The presidential election is fundamentally a national election, and major party nominations for president are inextricably tied up with how the nominees are chosen at national party conventions, which is why California, Washington, and Louisiana do not include the presidential race in their open primary systems. To be sure, the presidential nomination process should be reformed, but the change must consider the national character of the presidential office.

The second concern is that the plan limits rankings to three candidates in November, even though many more candidates could appear on the general election ballot. It is by no means a fatal flaw to the system, but  we would suggest either limiting the number of advancing candidates to four or allowing five rankings for races with six or more candidates. For example, Portland, Maine, which conducted a highly successful ranked choice voting election for mayor in 2011, has a ballot design that allows for more than three rankings.

Our third concern is that the Colorado Plan allows only one candidate to advance when that candidate secures more than 50% of the vote in the first round primary. If voter turnout were high and equitable in primaries, this provision would theoretically be defensible and might indeed boost primary turnout. For example, Louisiana holds its first election (without a primary) in November and elects any candidates with more than 50% of the vote. Since this election is held in November, turnout is high, and results are representative of what the eligible voters want. But given the likelihood that September turnout will be lower and less representative, that aspect of the Colorado plan is problematic. Further, effectively electing a congressional candidate before November is likely illegal, given the Supreme Court’s 1997 ruling against a similar plan in Louisiana. Fortunately, Ross has built in a severability clause that would mean that if this provision is struck down in court, such races would revert to the sounder approach of advancing the top three and any other candidates with more than three percent.


FairVote agrees that state-funded elections should include all voters, not just those registered with one of the major political parties. But as advocates for real and representative democracy, we must also insist that general elections include all voters by giving them real, meaningful choices, not just two candidates and certainly not just two candidates of one party.

Advocates of opening primaries should continue to experiment, and FairVote will continue to analyze and report on how those experiments are going. We look forward to working with more reformers who will take a cue from the efforts in Colorado and recognize the importance of opening both the primaries and general elections.