Ending Spoilers & Escaping Duopoly

A voting methods primer for the Forward Party

clay shentrup
12 min readOct 12, 2021

Ranked choice voting isn’t the only game in town. Here’s what approval voting looks like (right) compared to the status quo (left).

3 Key Insights

Here’s what I want Andrew Yang and the Forward Party to understand:

  • Approval voting is better than ranked choice voting (RCV). Both in the general sense, and in the specific ways that matter to the Forward Party: ending the spoiler effect and escaping the two-party duopoly stranglehold.
  • Approval voting is more viable than RCV.
  • We need a competitive marketplace of ideas. A multi-pronged strategy is better than putting all our eggs in one basket, i.e. RCV. (So even if you aren’t convinced about the other two points, it still makes sense to support a variety of alternatives including approval voting.)

More broadly, we want to focus on cardinal (rated/scored) voting methods as opposed to ordinal (ranked) voting methods. It just happens that approval voting is by far the simplest and most politically vetted cardinal method, so that’s where this piece will focus. We’ll revisit score voting and STAR voting in the future.

Now let’s zoom in on these points by one level of magnification.

Approval voting is better

  • Approval voting more accurately reflects the will of the voters, as measured by average voter satisfaction using computer simulations.
  • Approval voting is radically simpler than RCV. It requires no change to ballots, voting machines, or counting procedures, other than to eliminate the prohibition against voting for multiple candidates. It lowers the rate of spoiled ballots, whereas RCV increases it. It lowers the risk of ties/near-ties and the ensuing recount nightmares, whereas RCV increases them. It is precinct summable, whereas RCV is not.
  • Approval voting addresses the specific issues that seem central to Yang and Forward Party’s mission. Approval voting actually fixes the spoiler effect, whereas RCV does not. (Specifically, it satisfies the favorite betrayal criterion and independence of irrelevant alternatives.) Approval voting transparently reveals support for third party and independent candidates, whereas RCV does not.

Approval voting is more viable

  • RCV has only been successful (passed, used, and not repealed) in 20 of 46 cases in the US. That’s not even counting the repeal in the mid-1900s, when two dozen US cities used it and all but one of them got rid of it. It failed in liberal Massachusetts in Nov 2020 by 54.78% to 45.22%.
  • Approval voting has a small but perfect track record so far, passing by 64% in Fargo in 2018 and 68% in St Louis in 2020. And it went so smoothly that it’s hard to envision a repeal effort.
  • Approval voting is the only voting method that can conceivably invert the “burden of proof”. Rather than asking, “why should we add more complexity to our current process?”, approval voting asks, “why do we need this superfluous extra rule that your ballot is discarded if you vote for multiple candidates?”
  • Time is of the essence! Given approval voting also requires no voting machine upgrades or other major costs, it is plausibly the only voting method which, upon achieving critical mass, could spread and replace the status quo in a matter of years. Given the urgency of climate change, and the potential for a permanent GOP trifecta, this makes approval voting uniquely suited to the task at hand.

We need a competitive marketplace of ideas

  • Even if you don’t buy these arguments, it’s good to have a variety of options in order to maximize the odds that one of them succeeds and replaces the status quo.
  • Just as the climate change pragmatists advocate R&D on all manner of carbon-free energy sources, eschewing the strident claims that “nuclear is the answer” or “solar is the answer”, democracy reformers need to support a diverse set of options.
  • Contrary to fears that having multiple options will confuse people, approval voting passed by 2:1 landslide majorities in two cities in recent years, with no indication that it was held back by the existence of RCV as an alternative.
  • Some people say that RCV “has the momentum”, but what it really has is “a head start”. And there are reasons to believe that approval voting (and cardinal methods more generally) will have more momentum once they achieve critical mass, due to the massive simplicity advantages listed above.
  • The “momentum” thus far arguably pales in comparison to what Andrew Yang and the Forward Party can bring about through their collective fame and access to political thought leaders. Thus it is imperative that they create momentum for the right policies, rather than pursue what’s perceived (perhaps wrongly!) to have the most momentum.

Now let’s zoom in two levels of magnification. It’s time for the scanning electron microscope. #MATH!

Approval Is Better

Better results

Approval voting achieves more accurate results than RCV (aka instant runoff voting or IRV) according to sophisticated computer simulations by two different math PhD’s.

Bayesian regret values from Princeton math PhD and voting expert Warren Smith, taken from William Poundstone’s Gaming the Vote. https://www.rangevoting.org/BayRegsFig
Voter satisfaction (VSE) figures from Jameson Quinn, Harvard stats PhD and former board member with the Center for Election Science. Brown (50/50 strategy/honesty) is arguably the most realistic. Approval voting is “anti-fragile”, getting even better as more voters attempt to game the system.

Approval voting makes life easy for election officials

Curtis Koch, Davis County Clerk/Auditor representing the Utah Clerks/Auditors Association to the Utah Legislature’s Political Subdivisions Interim Committee on Sept. 15, 2021:

The clerks of the state who run the elections have reviewed all of these methods…we can run Approval Voting without any change, without any additional software, and centralized tabulation is not needed in that method. 29 (of 29) of the clerks (in Utah) are willing to and support the implementation of Approval Voting.

Approval voting also reduces the rate of spoiled ballots, whereas RCV increases them.

Approval voting reduces the risk of tie (or near-tie) election recount nightmares. RCV increases it.

Approval voting is precinct summable (each precinct can just announce their local totals and those can be added together for the total result). RCV isn’t and can’t.

Approval voting uniquely aligns with Forward Party’s goals

Andrew constantly talks about the problem of the two-party system and the spoiler effect that causes it. But RCV doesn’t fix this. Approval voting does. As do the rest in the scored family of methods.

As I argued in this 2010 essay hosted on the personal blog of Nader’s former running mate, Matt Gonzalez, score voting offers a path out of the two-party system, while RCV doesn’t. Approval voting is the simplified 0–1 form of score voting, so it retains this fundamental property, for reasons we’ll now look at more closely.

Approval voting shows outsiders their true support

Approval voting reveals the support for third party and independent candidates, which makes it harder to exclude them from debates and media coverage, and allows them to incrementally slay the electability dragon. Take a look at this 2012 exit poll by Occupy Wall Street.

RCV (“IRV”, red) makes barely any difference in showing the true support for all candidates, compared to the status quo (blue).

RCV (aka instant runoff voting or IRV) barely improves the support for Jill Stein (Green Party) and other independents. Romney paradoxically comes in 2nd in the liberal stronghold of Manhattan, ahead of the Green Party! Who seriously believes the same people who prefer Obama to Romney by a landslide also prefer the GOP to every other party including the Greens? Many Obama supporters had Green as their 2nd choice, but that’s impossible to see because RCV has the later-no-harm flaw.

Now let’s look at approval voting and score voting.

Approval voting (orange bars) and score voting (green bars)

Approval voting (orange bars) and score voting (green bars) give a 27x improvement in the Green Party’s support relative to Obama, as compared with the RCV results above. And Romney moves from 2nd place to last place with approval voting and score voting, which makes a whole lot more sense given what we know about NYC demographics. I covered this issue of “transparency” in my 2015 presentation to the Colorado League of Women Voters.

For a real election, look at how little support went to “Green independent” candidate Lisa Savage and her compatriots in the 2020 Maine senate race.

The media headlines, and even the Wikipedia header for this election, didn’t even include Savage.

Now contrast this with the results from the 2021 St Louis mayoral primary using approval voting.

Source

While this was a non-partisan race, Republican Andrew Jones garnered support from over 14% of voters despite running in a city that voted 81.9% Democratic in the last presidential election, and despite the fact that Lewis Reed was far and away the best choice for conservatives, who hoped to stop the two progressives who ultimately advanced to the general.

Approval voting prevents spoilers

Approval voting satisfies the favorite betrayal criterion (it can’t possibly hurt you to support your favorite candidate) and independence of irrelevant alternatives (removing a non-winning candidate from a set of cast ballots can’t change the outcome), both of which RCV fails. Here’s a short animated explanation from Andy Jennings, a co-founder of the Center for Election Science who did his #MATH PhD thesis on voting methods.

This played out in text book form in the 2009 RCV mayoral race in Burlington, Vermont. The Democrat was eliminated, despite being preferred by a landslide both to the Progressive Party candidate on his left and to the Republican on his right. The Progressive ultimately won. Which means Republicans would have gotten a better result if some of them had insincerely/strategically voted Democrat as their 1st choice.

We call this the center squeeze effect, because broadly appealing centrists are “squeezed out” by more polarizing figures to their left and right on the spectrum. I also covered this phenomenon in my 2015 presentation in more general mathematical terms, only loosely resembling the specifics of that Burlington election.

Interestingly this is the same basic strategy my aunt in Des Moines used when she voted for Biden instead of Warren in order to minimize the risk of electing Trump. It’s called the compromise strategy. Its lesser discussed opposite is the push over strategy, which would be like Trump supporters voting for Warren to try to give Trump a weaker opponent. Both strategies work in traditional “top two” runoff elections as well as RCV.

Alaska

In his recent interview with Sam Harris, Andrew Yang specifically cited the upcoming Alaska senate race featuring Lisa Murkowski as a proof point for RCV. But the early polling does not bode well. According to a poll by Change Research, Murkowski is likely to suffer the same “center squeeze” fate as the Democrat did in Burlington.

A post from one popular Alaska news site put it bluntly:

In the ranked-choice election method to be used in Alaska for the first time in 2022, Murkowski and Gross split the liberal vote in this poll.

They specifically chose this language because, even after picking up a substantial share of Murkowski’s votes, Al Gross (D) loses badly to the “Trumpist” Tshibaka (R).

Most of Murkowski’s votes go to the Democrat

Whereas Murkowski likely would defeat Tshibaka (and Gross) in a head-to-head race. The center squeeze effect rears its head again.

Approval voting minimizes this risk because, as we showed above, additional votes are counted rather than hidden behind onion peel layers of reallocation. Many Democrats who fear a Trumpist will also cast a compromise vote for Murkowksi, and many Trumpists who fear a Democrat will cast a compromise vote for Murkowski, and thus Murkowski will have a good shot at winning. There’s a great deal of theoretical and empirical evidence for this.

Approval voting is more viable

RCV seems to have “momentum”, but it really has “a big head start”. It was used, largely in its proportional multi-winner form in the the mid-1900s, and repealed in all of those cities save for the lone exception of Cambridge, MA.

A generation before that, another ranked voting method called Bucklin voting was used in about 40 US cities, including Portland and San Francisco. It was repealed in all of them. It’s plausible that the complexity of ranking is politically toxic in the US, despite the fact that it’s stuck in Australia and Ireland.

There has been a resurgence of support for RCV in the past few decades, but it hasn’t gone that well. Here’s a summary by one Texas activist.

To restate that, RCV has only been successful (passed, used, and not repealed) in 20 of 46 contemporary cases in the US. That’s not even counting the repeal in the mid-1900s. For instance, RCV failed in liberal Massachusetts in Nov 2020 by a quite large 54.78% to 45.22% margin. See this compendium of RCV successes and failures.

Election administrators seem to like approval voting

With the backing of election administrators, Utah’s RCV pilot program will likely be expanded to include approval voting as an option.

https://www.kuer.org/politics-government/2021-10-12/is-ranked-choice-voting-too-complicated-utah-lawmakers-may-let-cities-experiment-with-approval-voting-too

Forward Momentum

In truth, the electoral reform progress thus far has had less momentum than what Andrew Yang my be able to achieve all on his own, by focusing his entire movement around a handful of innovations like electoral reform and UBI. Would South Korea be talking about UBI right now were it not for Yang 2020?

This is why I believe Andrew should pick the best electoral reforms and blaze a trail forward, without much regard for what has traditionally been seen as achievable or plausible or “having momentum”.

The urgency of climate change

After it hit 116° here in Portland this summer, I started reading more about climate change and came across a lot of headlines like this.

RCV just isn’t moving fast enough—and cannot move fast enough—to help avert this crisis in the crucial 10–20 year window ahead of us.

Permanent GOP rule

In a recent New York Times piece, Ezra Klein explores the work of David Shor, with an emphasis on the interplay between polarization and our dysfunctional Gerrymandered system. One stark quote strikes a particular chord: “Democrats are on the precipice of an era without any hope of a governing majority.”

We have a limited window of time in which to moderate our polarized politics, or risk losing democracy in a slow-moving coup.

A Competitive Marketplace of Ideas

There’s not a whole lot to add on this point. When we’re talking about investment strategies, there are legends like Charlie Munger who actually advise against diversification. Sure, you’re more likely not to lose everything, but you’re also not likely to achieve outsized returns.

But R&D isn’t about maximizing your return from every investment in your portfolio. It’s about maximizing the probability of finding the best ideas and doubling down on those. For instance, stainless steel was discovered accidentally in a discarded pile of test alloys.

His efforts lasted months, and while the pile of scrap metal next to his work bench rusted, he noticed something strange — a barrel gleaming amongst them.

I may not have convinced you that RCV is a lemon, that doesn’t have good political prospects for replacing the status quo (and wouldn’t escape duopoly even if it did). But I think there’s an incredibly strong argument for putting a lot more different test alloys into our pile, to maximize the odds that one of them proves impervious to the the corrosion currently eating away at our great republic.

Yang’s movement is already getting it

Tom from the Nerds for Yang podcast recently hosted a voting methods panel including approval voting and STAR voting advocates, and then he started using approval voting for his online polls.

Just the First Step

I don’t expect any human being to admit a mistake and completely change course overnight. Not even Andrew Yang. But I hope this can be the start of an ongoing conversation.

Voting methods are an incredibly complex and counterintuitive subject, and the rabbit hole goes far deeper than I think Andrew and his team yet realize. I believe the mission of the Forward Party is well served by the inclusion of the “election science community”, generally comprised of cardinal voting activists.

Approval voting is the proverbial gateway drug here, but there are a variety of related innovations in the space, including score voting, STAR voting, 3–2–1 voting, and majority judgement. To say nothing of their multi-winner proportional forms as well as more exotic methods such as asset voting. (And I haven’t yet dared to wade into the even more boundary-pushing idea of sortition, although I’m pleased to see that Forward Party has made a nod to the idea in the form of civic juries.)

There is an incredibly smart and passionate network of activists within this community, and we want to work with you to make your platform even better.

Sincerely,
Clay Shentrup
Portland, OR

--

--

clay shentrup

advocate of score voting and approval voting. software engineer.