This essay was submitted to Open Philanthropy's Cause Exploration Prizes contest.

If you're seeing this in summer 2022, we'll be posting many submissions in a short period. If you want to stop seeing them so often, apply a filter for the appropriate tag!

 

By Marcus Ogren[1]

What is the problem?

The United States[2] uses choose-one Plurality voting, in which voters may vote for only a single candidate, for most of its elections. This is a notoriously terrible way of determining leaders; it frequently leads to people who don’t reflect the preferences of the electorate getting elected, and bears much of the responsibility for the two-party system and excessive polarization. By switching to better voting methods we can avoid consequences of polarization such as legislative gridlock or a possible civil war.

What are some better voting methods we might support?

It is important to distinguish between single-winner and multi-winner voting methods. For electing someone to an executive office (e.g., governor), only a single-winner voting method can be used. For electing a larger body (such as a city council or the US House of Representatives) either a multi-winner voting method can be used or the jurisdiction can be divided into single-member districts that use a single-winner voting method. The multi-winner case can involve either at-large elections or splitting the jurisdiction into multi-member districts that each elect multiple representatives at once.

Single-winner voting methods:

  • Approval Voting: Every voter can vote for as many candidates as they like. Whoever gets the most votes wins.
  • STAR Voting: Voters score the candidates from 0 to 5; multiple candidates may receive the same score. The two candidates with the highest total score advance to an automatic runoff with the already-cast ballots. In the runoff, each ballot counts as a single vote for whichever of the finalists is scored higher on it, regardless of the difference in scores.
  • Condorcet Voting[3]: Voters rank the candidates. To determine the winner, look at each ballot to see which candidate is preferred in every possible head-to-head matchup. The candidate who beats everyone else head-to-head wins.
  • Instant Runoff Voting (IRV, the best-known form of Ranked Choice Voting): Voters rank the candidates. Each ballot counts as one vote to the candidate that is ranked the highest on it. If one candidate is the first choice on a majority of ballots, that candidate is elected. Otherwise, the candidate with the fewest voters is eliminated, and their votes are transferred to the next highest choices on their supporters’ ballots. This process of eliminations and transfers continues until a candidate has a majority of votes among ballots that display a preference between the non-eliminated candidates.

A basic summary of the relative virtues of these voting methods: Approval Voting is by far the simplest reasonable voting method and it delivers the majority of the benefits of the more complicated ones in electing better leaders[4]. STAR is somewhat better at picking the right winner than Approval and is exceptionally good at incentivizing candidates to care about the opinions of voters in opposing factions. Condorcet methods are comparable to STAR at picking the right winner and are unparalleled at reducing the importance of strategic voting and at incentivizing candidates to endorse one another as a second choice and having it be meaningful. IRV is a significant improvement over Plurality, but elects worse leaders than these other methods, is nearly as bad as Plurality when it comes to third-party visibility, and can mimic the incentives of the partisan primary system. However, IRV has gained the most traction, has by far the best name recognition (under the umbrella term “Ranked Choice Voting”), and is currently used in Maine, Alaska, NYC, and San Francisco, and numerous smaller jurisdictions.

Multi-winner voting methods worth supporting are ones that deliver proportional representation (PR). Loosely speaking[5], proportional representation means having an elected body resemble the electorate in the ways the electorate considers important. So if the voters in a jurisdiction vote consistently with their party affiliation of 20% Democrats, 20% Greens, and 60% Republicans, and they elect 5 representatives, they elect a Democrat, a Green, and 3 Republicans. The simplest voting method that yields PR and is widely used internationally is Closed Party List, where each voter votes for a single political party and each party wins seats proportional to its share of the vote.[6]

However, the conventional wisdom is that Americans would be more inclined to adopt a voting method in which voters choose between candidates with a minimal role for political parties. The only true proportional voting method used in the US is Single Transferable Vote (Proportional Ranked Choice Voting). There are also proportional voting methods based on Approval or STAR ballots, but those are not currently used in any governmental elections. All of these proportional voting methods are more complicated than the single-winner voting methods I mentioned, and I won’t get into their algorithms or the differences between them here. Nonetheless, any of these proportional voting methods is far better than any single-winner voting method at promoting the growth of third parties, combatting gerrymandering, and improving minority representation and women’s representation.

Importance

By what mechanisms can better voting methods help?

  1. Ending the two-party system. Ending the two-party system would make parties dependent on coalition partners in order to govern (instead of one party or the other winning an outright majority), and such deal-making would reduce polarization.[7] Also, “us vs. them” thinking should be reduced when “them” is a less cohesive bloc.
    Internationally, there is a clear pattern of countries that use PR having more viable political parties; those that use single-winner districts have among the fewest parties.[8] The obvious mechanism is that PR reduces the threshold for winning a seat, so more parties are able to meet this threshold. Switching to better single-winner voting methods can also help third parties by (a) giving them better visibility in election results[9] and (b) removing the strategic incentive under Plurality to vote for a major party candidate even if you like a minor party candidate more.
  2. Obviating gerrymandering. Gerrymandering, the act of manipulating the process of redrawing district boundaries for political advantage, involves two complementary tactics, cracking and packing. Cracking means dividing the supporters of the opposing party into districts that they will lose by a relatively small margin, so that as many of their voters as possible get expended without electing anyone. Packing means drawing districts that the opposing party wins by the largest margins possible, so that many of their votes are superfluous. Gerrymandering reduces democratic legitimacy, makes it easy to justify claims that the opposing party is stealing an election, and raises the stakes of elections when the winning party can draw the next maps.
    Proportional representation can involve at-large elections, which completely eliminate gerrymandering since there are no districts to draw. PR can also involve multiple multi-winner districts, and this too foils gerrymandering. Reducing the percentage of votes required to win a seat deals with cracking, and making it valuable to to have a large majority instead of a small majority deals with packing.
  3. Incentivizing politicians to appeal to a broader range of voters. With Plurality general elections and partisan primaries (the status quo in most of the US), Democratic candidates have essentially no incentive to appeal to reliably Republican voters and vice versa. This is an especially large concern in noncompetitive districts[10] in which the winner of one party’s primary is a shoe-in for the general election. With a combination of primary reform and voting methods that empower partisan voters to reward members of the opposing party without sacrificing their own party’s chances, it can be made politically expedient for office holders to make compromises and to take principled stands that set them against their own party. Different voting methods vary drastically in the incentives they provide to candidates, and I consider this to be the most promising avenue for reducing polarization that isn’t predicated on dismantling the two-party system.
  4. Reducing negative campaigning. When voters have the ability to support multiple candidates this disincentivizes political attacks since candidates don’t want to alienate voters who have already decided on the recipient of the attacks as their first choice. This has some empirical support, but I don’t have a good estimate for how large the effect size is or how valuable it is to reduce the frequency of political attacks by some percentage. This is presumably beneficial for depolarization, but it’s difficult to be quantitative.
  5. Electing leaders who better reflect their constituents’ preferences. Looking at example elections demonstrates this benefit qualitatively, but doesn’t suffice for quantitative comparisons between voting methods. Probably the best tool for such an assessment  is Voter Satisfaction Efficiency (VSE), which uses computer simulations[11] to determine how different voting methods compare in terms of electing people who best satisfy the aggregate preferences of the electorate.
  6. Electing more diverse leaders. There is strong evidence that proportional representation increases the representation of women and minorities; the evidence that better single-winner voting methods yield more diverse representation is far weaker, and the theoretical arguments are less compelling. However, I have no idea how beneficial it would be for an additional 10% of US House members (a plausible effective size if the US adopted proportional representation) to be women.

In addition, changing voting methods may alter perceptions about the legitimacy of elections and change how much influence politically-engaged voters have relative to less engaged, and less informed, voters. The former is most likely a disadvantage of enacting a new voting method (at least in the short term), and I am uncertain whether the latter effect would be socially harmful or beneficial.

How important is electing better leaders?

Having different people in charge obviously affects national policy on controversial issues. However, it is nigh impossible to quantitatively evaluate the influence of leadership quality on objective measures where everyone agrees on what kind of a change is desirable. (Economic growth rates,[12] for example.) This is a major issue in assessing the impacts of voting method reform since we should not expect a change in voting methods to help Democrats over Republicans or vice versa, nor should we expect it to affect how liberal trade policies are, the strictness of environmental regulations, the average person’s tax burden, etc. Enacting better voting methods may lead to better policies, more bills getting passed, and more competent governance in general, but it will do so in a manner that is mostly symmetric from a partisan perspective.

A Fermi estimate based on VSE is included in the BOTEC. It includes a correction for VSE oversampling interesting elections[13] and for a correction (by a factor of 50[14]) for voters mostly not valuing stuff that’s good from a nonpartisan humanitarian standpoint.[15]

This suggests that having Approval Voting for all elections in the US instead of Plurality would be worth around $45 billion dollars over 30 years[16] as a result of electing better leaders - very disappointing, considering the cost of such a campaign. However, this number doesn’t capture the benefits of having more diverse leadership and more women in government.

How important is depolarization?

Polarization has many negative consequences:

  • Harder to pass legislation, even if it’s widely popular
  • Legislation naturally gets optimized more for political gain relative to public benefit than it would in a less polarized society
  • More frequent government shutdowns
  • Greater chance of hitting the debt ceiling, or otherwise failing to pass “must-pass” legislation
  • Greater chance of democratic backsliding[17]
  • Greater chance of a civil war
  • Greater chance of the US turning into a stable global totalitarian state as an ultimate consequence of democratic backsliding

My crude estimate is that switching to proportional representation and having it end the two-party system would reduce the expected impact of each of these by 50%. (This number is based on informal reasoning, the role of the two-party system in causing polarization, and the sense that polarization is a major causal factor for all of these issues.)

I attempt to estimate the impact stemming from only one benefit of depolarization: reducing the likelihood of a civil war.[18] The other benefits of depolarization seem far more difficult to assess. My best guess, based on vague impressions from many years of reading the news and not on anything resembling a rigorous study, is that preventing democratic backsliding or having a more functional congress is the largest benefit. I use the extremely crude assumption that the benefits of preventing a civil war are a fifth[19] of the total benefits of depolarization, so that:

  • The annual cost from the risk of a civil war is about $370 billion
  • The expected annual cost of polarization is about $1.8 trillion
  • Having PR in the House[20] is worth roughly $13 trillion over 30 years.

These numbers are much higher than those for electing better leaders, and suggest that virtually all of the benefits from better voting methods stem from depolarization. This also implies that the vast majority of the benefits of adopting better voting methods occur only at the national level; national politics tend to be dramatically more polarized than local politics and many of the harms of polarization are milder or nonexistent at the state/local level.

Voting Methods and Existential Risk

Government policy has a significant effect on most existential risks, ranging from hopefully preventative (natural pandemics) to directly causal (nuclear war, AI arms race, biowarfare). Electing good leaders who will enact sensible policies concerning existential risks is immensely valuable - the question is, how much will better voting methods actually help with this?

I am mostly pessimistic here since I believe that voter preferences are only based on politicians’ general competence to a limited extent; I believe that multiplicative factors in my BOTEC for the benefits of better leaders are roughly correct for the abilities of leaders to avert existential catastrophes as well. However, it is possible that better voting methods would have a larger effect on preventing the election of exceptionally terrible leaders who would either directly cause an existential catastrophe or fail to avert one even if doing so was easy enough that the vast majority of leaders who could have been elected instead would succeed. If a politician appears to be extremely incompetent, two likely reasons for them to get elected are (a) vote splitting and (b) there being only one other viable candidate and that one looking even worse (perhaps due to being a member of a different party). Better voting methods avoid these possibilities by allowing more than two candidates to be politically viable without the risk of vote splitting.

Voting method reform is an especially promising avenue for reducing an existential risk that is relatively little-discussed in the EA community: a permanent global totalitarian state that greatly curtails humanity’s potential. From “Reducing long-term risks from malevolent actors”:

Even within established democracies, we could try to identify measures that avoid excessive polarization and instead reward cross-party cooperation and compromise. Mitigating the often highly combative nature of politics would plausibly make it harder for malevolent humans to rise to power. (For example, effective altruists have discussed electoral reform as a possible lever that could help achieve this.)

These avenues of risk reduction require different, but entirely compatible, reforms: Avoiding the risks of terrible leadership primarily requires changing how the President is elected, but preventing the US from becoming a stable totalitarian state requires depolarization, which is best achieved by PR.

Tractability and Neglectedness

Who is currently working on this?

The largest organization dedicated to working on voting methods is FairVote, which advocates for IRV and STV[21] under the umbrella term “Ranked Choice Voting”. They have an annual budget of roughly $3 million. Another notable organization is also the RCV Resource Center, which focuses on research and education rather than advocacy. There is substantial additional spending on individual campaigns to adopt RCV; in 2020 the successful Alaska campaign spent $6.8 million and the unsuccessful Massachusetts campaign spent $10 million.

The major Approval Voting organization is the Center for Election Science (CES), with an annual budget of roughly $1.5 million. Open Philanthropy is their biggest funder.

The national organization that advocates for STAR Voting is the Equal Vote Coalition, which has an annual budget of under $50,000 a year. They also support Approval and Condorcet, but STAR is their main focus. Nonetheless, they are the only organization I know of that supports Condorcet Voting.

There are many other organizations that are involved in voting method reform without it being their primary focus, such as the League of Women Voters, Unite America, and the Action Now Initiative. There are also many state and local organizations and chapters working on voting method reform.

How can we achieve large-scale changes?

The ultimate goal is securing passage of something like the Fair Representation Act (HR 3863), which:

  • Creates multi-member congressional districts if a state has more than one Representative in the US House
    • States with between 2 and 5  Representatives in the US House elect them in a statewide at-large election using STV
    • States with 6 or more Representatives establish multi-winner districts, each of which elects somewhere from 3 to 5 Representatives (preferably 5) via STV
  • Requires the election of Representatives from single-Representative states and all US Senators to be held using IRV

Note that this is a bill in Congress; enacting these reforms does not require a constitutional amendment. It is easy to imagine variants of this, such as:

  • Using a Condorcet method instead of IRV for single-winner elections
  • Using Approval Voting for single-winner elections and Sequential Proportional Approval Voting for multi-winner elections
  • Using STAR for single-winner elections and Allocated Score for multi-winner elections

For electing the president, the easiest means of achieving reform is probably through something like the National Interstate Vote Compact. Approval Voting would work the most naturally with it since Approval and Plurality have essentially the same concept of “a vote”; there would be additional challenges in making it work with other voting methods.

However, we are very far from passing the Fair Representation Act; at the moment it only has 8 cosponsors, all of whom are Democrats. To make such a landmark bill viable, the conventional wisdom in the voting methods community is we have to start small.

  1. Get a voting method adopted and used in some municipal elections. This has usually been achieved via ballot initiatives.
  2. As the voting method gets used more, we acquire more real-world evidence of voter behavior, its effects on political culture, whether it elects more diverse leaders, etc. (FairVote has already amassed a significant amount of such research.) It also gets better name recognition nationwide, and switching to it seems less weird. This makes it easier to adopt it in additional cities, and eventually city councils will start putting it on the ballot out of their own volition. (This is already happening with IRV.)
  3. Get that voting method adopted in statewide elections. It will be easier to get a state to adopt it if many voters in it are already familiar with it from their municipal elections.
  4. Finally, pass something like the Fair Representation Act.

This elides the question of how to get support for a proportional voting method. FairVote’s approach is to conflate IRV and STV under the RCV label, advocate primarily for IRV, and then either switch from IRV to STV (by also switching to multi-winner districts) or leverage voter familiarity with IRV to get STV adopted. Either way, this envisions IRV as some fashion of a stepping stone to STV. STAR advocates are going for something similar, and CES talks about PR to such a limited extent that I’m unsure how or whether they intend to pursue it.

I am skeptical but not dismissive of this approach. No US jurisdiction has switched from IRV to STV; the closest thing to a success story here is STV sometimes getting adopted alongside IRV (where city councils have both single-winner seats and at-large seats). But Australia’s senate switched from Preferential Block Voting to STV in 1948, so a change in this vein is clearly possible.

What makes a voting method tractable?

I don’t know. Most of the polling I know of[22] shows that the vast majority of Americans would rather stick with Plurality. However, CES has released polling which shows that 72% of Americans would support a ballot initiative to switch to Approval Voting,[23] but this doesn’t include a comparison to any other voting methods. Here are some theories as to what might matter:

  • Having a “strong track record”, i.e. already being used in a lot of jurisdictions. I am far more confident that this matters than that any other item on this list matters.
  • Simplicity. If voters are the most inclined to support simple voting methods, then Approval is the most tractable to advocate for. (One caveat: this depends on which kinds of simplicity are important. While Approval is exceptionally simple in most respects, it has some of the lowest strategic straightforwardness; strategic considerations are much more salient when deciding how to cast your ballot under Approval Voting than under most other methods.)
  • “Ranked Choice Voting”. Maybe a lot of voters know that “Ranked Choice Voting” is something good, and want to support it, even though they don’t know what it is. In this case, supporting a Condorcet method and calling it something like “Equal Ranked Choice Voting” is especially tractable.
  • Superficially seeming good. I think Approval is the worst here. It naturally comes across as an inferior “budget” option compared to the alternatives; it seems like the smallest change, sounds the least novel, its ballot marks are the least expressive (which naively and wrongly suggests that it yields worse outcomes than IRV), and misleading “bullet voting” arguments get a lot of traction.
  • Actually being good. I am mostly pessimistic about voters being able to assess this, but insofar as it’s important it argues most strongly for PR, STAR, and Condorcet.

What can a philanthropist do?

Fund a large-scale campaign

Getting a voting method adopted nationwide requires either a megaproject or other funders jumping in. From the Center for Election Science:

To succeed in the next phase of our mission, we need to scale up our funding as we begin to target statewide efforts.  Our future plans require having the staffing necessary to coordinate a nationwide effort of multiple concurrent state-wide campaigns while simultaneously doing the necessary research to ensure a data-backed approach.

We estimate this will require an annual budget of $15M. Our current annual budget is only $1.5M. Alongside this funding, we will need to provide financial support for the campaigns through a sister 501(c)4 organization, which does not have a cap on campaign spending. We estimate that we will require at least $250 million in funding over the next 5-6 years to cover all the ballot initiative states and maintain the infrastructure to run this scale of operations.

Obviously, we should not simply assume that $250 million gets a voting method implemented nationwide. There are also states that don’t allow ballot initiatives, so I assume that $500 million would be spent on a nationwide campaign. In the BOTEC I find that, conditional on Approval Voting appearing viable for national adoption, such a nationwide campaign would have a 16,000x return on investment over 30 years.

Fund a small-scale campaign for a less established voting method

Switching to a better voting method won’t have a big impact unless it happens on a national scale; improving local elections matters little in and of itself. However, funding campaigns to get a voting method that isn’t currently used in municipal elections (probably STAR or Condorcet) adopted in just a few cities would:

  • Aid in movement building for the voting method
  • Eliminate the stigma of being “untested”
  • Give us information about how well the voting method works
  • Give us information about how receptive voters are to adopting it

The first two bullet points are basically just the start of a larger campaign (with the added bonus that it has some chance to keep going even without continued EA funding). The last two points mean that the earliest campaigns for a voting method are the most cost-effective since they help us direct future funding. In the BOTEC I use a highly simplified model[24] for assessing the cost-effectiveness of information on how tractable a voting method is and find that supporting STAR or Condorcet[25] in its early development has a return of around 100,000x. This is largely, but not entirely, due to the advantages these methods have over Approval Voting at depolarization; even if we assume they are exactly equal to Approval Voting there would still be roughly a 30,000x return stemming from the possibility that they have greater political viability (especially in competing with IRV) than Approval.

Note that it takes several campaigns to assess the relative willingness of voters to adopt a particular voting method. We should expect the first campaigns to be the most difficult as a voting method loses its “untested” stigma, and there is substantial variation in how much support the same voting method receives in different cities. For example, campaigns to switch to STV received 44% of the vote in San Francisco (1996), 48% in Santa Clara (2018), and 73% in Albany, CA (2020).

Start an organization focused on Condorcet voting

Condorcet methods provide outcomes and incentives that are roughly as good as those of STAR Voting, but are somewhat more complicated. Since most Condorcet methods use ranked ballots, the big advantage of Condorcet is the ability to build off the RCV name while being more appealing to RCV advocates than Approval or STAR. I have written about this in Equal RCV: The Political Case for Condorcet.

I believe the optimal person to start an organization would be someone with a lot of connections in the RCV community and who legitimately supports IRV while considering Condorcet to be superior and who would talk about Condorcet in the pro-RCV community.

The main way such an organization could have a high impact is by causing a Condorcet method, rather than IRV, to be adopted nationwide. The more IRV gets used for high-profile partisan elections, the more attention center squeezes will receive, and this may make RCV advocates open to fixing the problem that doesn’t require them to abandon the RCV label.

Support STV campaigns

STV is the proportional voting method that has by far the most traction in the US, and I expect it to perform no worse than proportional forms of Approval Voting. (Proportional voting methods that use a scoring ballot can probably surpass STV, but I expect STV to at least be basically adequate.) Supporting STV campaigns - especially a hypothetical statewide campaign - could make a big difference in paving the way for the Fair Representation Act and avoiding the pitfall of IRV getting adopted for multi-member bodies instead of STV. One potential issue with funding STV is fungibility; additional funding for STV campaigns specifically could make RCV supporters focus more of their spending on IRV.

Other possibilities

  • Funding research. I think this is best achieved by funding organizations such as CES and the Equal Vote Coalition since they’re well-positioned to determine what research is useful.
  • Campaigns in other countries. Canada, the UK, and India are obvious possibilities since they all use Plurality. It is also possible that optimizing the voting methods used by fragile states could be extremely effective, though I have done nothing to investigate the tractability of campaigns outside the US.
  • Starting an organization focused solely on PR. The benefits of PR make this very appealing, but it’s been tried before and other people have abandoned the “just go for PR” approach.

I recommend against funding IRV campaigns; it is by far the least neglected and does significantly less for depolarization than other non-Plurality voting methods.

Notable Uncertainties

  • Uncertainties about the importance of depolarization or better leadership:
    • How likely is a second American civil war, and how bad would it be?
    • How likely is democratic backsliding, and how valuable is democracy in general?
    • How damaging are current levels of congressional gridlock?
    • How much does good leadership matter?
    • How strongly correlated are voter’s opinions of a candidate with how much the candidate would accomplish in office, from a humanitarian perspective?
    • How beneficial is it to have more women and minorities in congress?

I have the least confidence in these numbers, but my cost-effectiveness estimates are somewhat robust to these uncertainties since the cost-effectiveness of electoral reform is proportional to the sum of these factors. If most of them are negligible, it doesn’t mean electoral reform is useless.

  • Uncertainties about the relative quality of different voting methods. These are the numbers I have the greatest confidence in, but many people would disagree vehemently.[26] The cost-effectiveness of funding voting methods work is extremely robust to these uncertainties because getting a voting method adopted in a few cities will give us information about it. The greater the uncertainties here, the stronger the case for funding multiple voting methods. Uncertainties about how likely different voting methods are to catch on are similar, but I have less confidence in my estimates.
  • Everything related to existential risk: How likely is partisan rancor to cause the US to become a stable totalitarian state? How much better would a less polarized government do at handling other existential risks? How much existential risk is due to blatantly and extraordinary incompetent leadership?
  • Leverage and fungibility: I haven’t tried to account for this, but I think there’s a large potential that other funders (likely those funding IRV campaigns) would jump in to assist with statewide initiatives once a voting method has been adopted in several cities.
  • How much could any voting methods reform reduce polarization? The number I’m using is 50%, but this is very difficult to estimate quantitatively and I consider it to be the most pivotal uncertainty.
  • How likely is it that we get a good voting method implemented in several states, but then hit a brick wall and are unable to make meaningful progress? (For example, Congress might pass a bill that requires Plurality to be used for electing members of Congress.) Such possibilities are the likeliest reasons for a costly failure in which we spend hundreds of millions on voting method reform and ultimately fail.

 

  1. ^

    I am a voting methods researcher and am highly knowledgeable about voting methods and how they compare to one another, but am only moderately knowledgeable about the politics of adopting different voting methods. I have absolutely no expertise in assessing the importance of good national leadership, the risk of a civil war or democratic backsliding, or how valuable democracy is in general. Most of what I have written about voting methods is available on my blog. I have also worked on an (as yet unpublished) updated version of Voter Satisfaction Efficiency with Jameson Quinn and Sara Wolk, and more recently on Candidate Incentive Distributions.

  2. ^
  3. ^

    Strictly speaking, Condorcet isn’t a single voting method, but rather a category of voting methods. There may not always be a candidate who beats everyone else head-to-head, and different Condorcet methods are distinguished by the tie-breaking rules they use to determine the winner in such cases.

  4. ^

    Quantifying how well voting methods do at electing leaders that reflect the will of the electorate is best done via computer simulations. Looking at well-known examples that challenge many voting methods (and not just Plurality) is useful for gaining a gears-level understanding of why some voting methods outperform others, but introduces a bias towards optimizing for well-known and easy-to-understand scenarios.

  5. ^

    Rigorously defining proportionality when the electorate isn’t cleanly partitioned into political parties is very difficult and is an area of ongoing research.

  6. ^

    There are a few complications related to rounding, and most countries require parties to get some minimum fraction of the vote in order to win any seats.

  7. Reasons why I’m focusing on the US:

    • The US uses extremely bad voting methods (France, for example, at least has runoffs that mitigate the problems with Plurality).
    • America’s geopolitical power means its elections have the greatest effect on the wider world
    • The US is highly influential; adopting better voting methods in the US is far more likely to inspire India to reform its elections than reform in India is to inspire the US
    • Having lived in the US for the vast majority of my life, I am far more knowledgeable about the prospects for reform there than anywhere else.
  8. ^

    For a book-length argument, see Breaking the Two-Party Doom Loop by Lee Drutman.

  9. ^

    An interesting case study is Australia, where the House uses IRV (single-winner Ranked Choice Voting) and the Senate uses STV (Proportional Ranked Choice Voting). The House has a two-party system, but the Senate does not.

  10. ^

    IRV is particularly weak in this regard since the later preferences of voters who prefer one of the top two candidates remain invisible throughout tabulation.

  11. ^

    Such districts are becoming increasingly common.

  12. ^

    There are many other simulations as well, but VSE is the most-discussed and is the only one whose source code I am intimately familiar with.

  13. ^

    One study examines changes in economic growth after a leader dies in office to assess the importance of having good leadership. It finds “evidence that the death of leaders in autocratic regimes leads to changes in growth while the death of leaders in democratic regimes does not”. The effect on autocratic regimes appears to be driven by the existence of some exceptionally bad leaders (e.g., Mao), and there are few enough such deaths that this methodology is hopelessly underpowered for discerning the importance of leadership quality in less-extreme, more democratic settings.

  14. ^

    Cases where different voting methods yield different winners happen far more often in VSE simulations than they do in real-world elections. Two countervailing factors: We aren’t just interested in real-world elections that have happened, but in the elections that would have happened under a different voting method. Notably, more candidates might have run for offices, and the structure of partisan primaries could be very different. Second, most of the data we have for Instant Runoff Voting comes from non-partisan elections, but the biggest gains from using better voting methods come from adopting them in high-profile national elections, which are always partisan. This is relevant because partisan labels help voters distinguish between different candidates and the one election where Instant Runofff Voting definitely elected a different winner than Condorcet would have was one of the partisan ones.

  15. ^

    This number is extremely uncertain and could be off by well over an order of magnitude. A possible reason for this number to be overly pessimistic: IQ is positively correlated with political accomplishment, and better voting methods would probably strengthen this correlation even if it mostly stems (for example) from smarter people being more charismatic.

  16. ^

    So, excluding charisma (irrelevant from a humanitarian perspective) and total government healthcare spending (partisan), but including the efficiency of government healthcare spending.

  17. ^

    See the “Campaigns” tab in the BOTEC.

  18. ^

    I believe that How Democracies Die by Steven Levitsky and Daniel Ziblatt gives a good argument for how polarization leads to democratic backsliding, but I have not actually read it.

  19. ^

    This was chosen because it seemed especially tractible to estimate: Metaculus has a forecast for a second American civil war, and I had estimates to draw from on the costs of civil wars.

  20. ^

    Before running the math I was planning to go with 1/10 instead of 1/5, but opted for the latter number after seeing how big a deal preventing a civil war is.

  21. ^

    I divided by two to account for the existence of the Senate

  22. ^

    They also support some other forms of ranked voting including Preferential Block Voting, a non-proportional multi-winner voting method that uses ranked ballots.

  23. ^

    In particular, see “Choosing to vote as usual” for a study that compares Plurality, IRV, Approval, and Score Voting (Score is STAR without the runoff) in which people participated in example elections for these voting methods. It finds that Plurality is by far the most popular, followed by IRV.

  24. ^

    Why is this so different from the results in “Choosing to vote as usual”? My best guess is that it’s because CES’s wording is far less explicit about removing the “vote for one” restriction and CES didn’t have respondents participate in an example election.

  25. ^

    A more fleshed-out model may have several rounds of funding, all but the last of which yield both information and the option of whether to stop the funding or keep going, analogous to a multi-round process of venture capital funding. What I have is the minimal viable product: two rounds, in which only the first yields information and the option of giving up. There is the possibility of failure on either round, and failure on the latter round (a nationwide campaign) is far more expensive.
                This simple model is not quite internally consistent; I assume that the first round of funding costs $10 million, and also that IRV has a substantial chance of failing this round and being shown not to be viable for nationwide, despite the fact that well over $10 million has been spent on IRV advocacy to date. This results from attempting to accommodate for the facts that (a) early funding costs far less than $10 million and gives a lot of info, and (b) it’s possible that IRV can’t catch on nationally whereas another voting method (probably Approval) might.

  26. ^

    Or Approval; it’s still in the early stages.

  27. ^

    A common position among RCV advocates who think about Approval Voting is that Approval Voting will lead to candidates always emphasizing “ONLY vote for me” while voters spend a lot of time agonizing over strategic voting. For additional perspectives, see comparisons between single-winner methods written by ApprovalSTAR, and RCV advocates.

36

New Comment
5 comments, sorted by Click to highlight new comments since: Today at 4:27 PM

Excellent write-up and I completely agree with the massive impact voting reform would have if it could gain momentum. As a real-life example of its impact, I like this Washington Post article How ranked-choice voting saved the Virginia GOP from itself (I apologize if you linked to this somewhere in your post, I didn't see it).

Another organization working on this is Ranked Choice Voting Resource Center, although I have no idea what their budget is.

I'll admit I've never quite grasped the EA focus on voting systems over/vs the structural conditions under which we vote. It is a bit symptomatic of our technical solutions for social problem bias.

I don’t think voting methods is a bad EA focus, more that it is a chair with two legs. A bit too shaky while the structural conditions legs are missing to be all that confident about sitting on it.

I’m coming at this from a practical politics and campaigning expertise area as a politician who elects politicians and invests too much time tracking policy progression and maps the electoral pathway to creating the kind of parliament/senate/house etc we would need to deliver my ideal policy aspirations for fun.

If I haven't commented in this thread on it I broadly think that chunk of your argument is solid. It’s clear that your expertise in voting systems is very impressive and it’s always good to have people on side with strong expertise in neglected areas.

I think the idea that proportional systems or systems that encourage smaller parties to win will reduce policy polarisation or speed up the legislative process/prevent bottlenecks needs a bit more potential outcome consideration.

Chambers and their coalition blocks do still tend to skew Left/Right - with the us vs them mentality persisting. The largest block can be potentially larger than it would be under an FPTP system (everyone wants to be on the winning team and everyone hypothetically can be) and will broadly decide policy.

The bottlenecks simply shift to a more informal inside-the-block process vs the more open to scrutiny chamber in these circumstances. I witnessed many a late-night policy WhatsApp battle between members in blocks while working in the European Parliament that closely resemble the ones I now have in a large FPTP group.

Where the arithmetic gets a bit dicey for the potential ruling block they are incentivised to allow increasingly fringe groups into the ruling block.

Those fringe groups usually make some very cheeky demands in return for their support and can end up with an outsized impact on the administration and end up polarising the agendas further. We don’t necessarily want to be amplifying the views of fringe groups where they present risks to EA causes.

There are some safeguards against fringe group influence but most people dislike them. When I was working in the European Parliament all the groups agreed to block a handful of really right-wing/anti-EU parties from getting any meaningful committee positions or passing policy. Good for the EU, pretty rubbish if you voted for one or more of those parties.

With enough clearly EA-aligned politicians, you could consider a similar technique to lock parties that promote x-risk raising policies or are against x-risk-reducing policies out, for an EA example, but again it’s a bit dodgy to voters.

I think anti-gerrymandering work is a strong angle for us to consider investment in on the traceability and neglectedness front.

I recently ran a grassroots anti-gerrymandering campaign in my area when the boundary review came around and the vast majority of people had no clue what I was on about when I explained cared a lot but did very little. Even the easy task of writing a comment on a website wasn't taken up by a lot of people because the cause just isn't seen as important enough.

I then went to a hearing as an expert to object to a gerrymandering boundary change proposal and it was unbelievably empty. Only a handful of us spoke over the course of the day.

Controversially, I think political attacks are quite an important part of the democratic process. Particularly to prevent bad politicians, which you clearly value very highly.

The average voter does little to no background research on candidates beyond what they see over the course of the campaign. There is no reason to believe that would change under a different voting system.

‘Nice’ electoral environments and positive campaigning pledges are already pretty weaponised (used by parties with things to hide to attack other parties for bringing them up).

Voters also miss the chance to gain important information about their candidate of choice that they might not be so willing to share. MP/Senator voting records are a big one because most people do not pay enough attention to how their reps vote.

I think anyone considering investing EA cash and resources in changing voting system campaigns/referendums needs to do a cursory sweep of the UK AV referendum. A qualified disaster on all levels and a great primer on how to waste a substantial amount of money achieving nothing because of poor planning and timing.

Two points regarding the possible downsides of proportional representation and multiparty systems:

First, different ways of implementing PR lead to different numbers of parties. If you just do nationwide party list, without a threshold like requiring at least 5% support to win any seats, there will be tons of parties, and the wheeling and dealing between them will ultimately seem more impactful than the actual election. But if you have multi-winner districts with < 10 seats each, only a few parties will win representation (but probably more than two). There's an important sense in which politics is compromise. With a two-party system, voters have to compromise and support the more tolerable of two candidates. Then, whichever party wins a majority gets to govern uninhibited by the other party (aside from filibusters, the possibility of a split congress, etc.). Or with a 10+ party system, voters hardly need to compromise at all; with so many options, chances are they'll be able to find a party they absolutely love. But this doesn't remove the need for compromise, it just shifts it onto the elected officials. I think the optimal system splits the difference, putting some of the weight of compromising on voters and some of it on elected officials. Having more than two parties, but not more than six or so, should achieve this. To the extent you're worried about the effects of having significantly more than three parties, this is entirely manageable with multi-winner districts - but I think your points have a fair amount of validity even for a three-party system.

(An additional consideration is that the US has a directly elected executive branch, so you wouldn't necessarily have the dealmaking to determine who gets to form a government that you get in most places that use PR.)

Second, many of the harms of polarization are different from what you describe happening under PR systems. I don't think the risk of a civil war or democratic backsliding is driven by a handful of extremists who need to be excluded, it's driven by the widespread hatred that Democrats and Republicans have for one another and the perceived threat the other party poses to democracy. I think this has relatively little to do with the difficulties of dealmaking and controversial policy concessions to fringe parties, even though I lumped the risk of civil war and congressional deadlock together as "polarization". In my model, I assumed that electoral reform would affect all the consequences of polarization equally. This could be very wrong! Perhaps proportional representation eliminates most of the congressional gridlock caused by zero-sum partisan maneuvering, and replaces it with congressional gridlock caused by the difficulty of getting several different factions to work together. (I find Lee Drutman's argument from Breaking the Two-Party Doom Loop that having more parties will result in a more functional congress to be quite persuasive, however.)

I want to split the difference between Marcus’ points and yours, Alisha. Let me start by limiting the scope to shifting the actions of the US government to better align with the needs and desires of the people without getting into prospects beyond representative democracy. There are many factors causing the current misalignment, but I believe the two largest in that context are vote splitting in elections and simple majority rules in legislatures.

Vote Splitting in Elections:
The simplest example of vote splitting would be a  an election with three candidates: Luke Skywalker, Leia Organa, and Darth Vader. Most voters who support Luke also support Leia and vice versa, but hardly any of them like Vader. When Election Day comes, nearly 60% of voters support both of the Skywalker twins, but since they are only allowed to support a single candidate at a time with their vote, they collectively end up splitting their votes between Luke and Leia. The final results are Luke with 35%, Leia with 25%, and Vader with 40%; Vader is elected even though a clear majority of voters dislike him.

Vote splitting can get deeper and more complex when there are more candidates, but the fundamental result of an electoral system with vote splitting is the trend toward electing polarizing candidates. This is because polarizing candidates gain stronger, more isolated support among their voters and stronger animosity from everyone else. Because voters are only able to support a single candidate at a time anyway and because that support is expressed on their ballot as either full-on or not-at-all with no nuance, polarizing candidates with huge sums of money perform the best under systems with high levels of vote splitting. The downstream effects of having our government filled with polarizing politicians is the misalignment I was referring to at the beginning; the distribution of voters in political space is, like most everything, a (multi-dimensional) bell curve. Of course, having polarizing leaders causes that distribution to shift — it may be double-peaked in the US right now — but it’s always important to consider the people who don’t vote and why they don’t vote when talking about political distribution. Many of the reasons people don’t vote in the US can be indirectly affected by eliminating vote splitting.

Fortunately, vote splitting is caused by a solvable engineering problem that I’ve already bolded twice: only allowing voters to support a single candidate at a time. This broken mechanic — potentially invented because the ancient Greeks voted with pebbles and jars instead of auditable paper trails — feeds the toxic idea that we can only advocate for one group or one set of ideas at a time. In reality, most voters would support multiple candidates if the good ones didn’t have to fear splitting the vote. It’s the math of the method, not a vice of the voters or a crime of the candidates.

So, what’s the solution? There are many, but the simplest, as Marcus highlighted, is to remove the (arbitrary) restriction on our ballots that only allows voters to support a single candidate at a time. This leads to Approval Voting. By allowing voters to support multiple candidates at the same time, polarizing candidates are systemically disempowered and unifying candidates that draw broad support across the entire electorate are boosted in comparison.

Again to Marcus’ point, Approval Voting is a great “budget” reform for jurisdictions that currently use Choose-one Voting, are resistant to change, and have limited legally viable options for voting method reform. For jurisdictions that have more options or are currently on Ranked Choice (Instant Runoff) Voting (which does not eliminate vote splitting because voters can only support one candidate at a time in each distinct runoff round), STAR Voting and Condorcet methods like Ranked Robin are even better upgrades.

Simple Majority Rules in Legislatures:
However, as you pointed out, Alisha, this is only half of the puzzle. Polarization is also amplified by the way our legislatures run. In a similar vein as candidates running under Choose-one Voting, policy is subject to the inherent polarizing effects of simple majority. When  a given proposal only needs support from half of the electorate (the electorate being legislators in this case), those proposals perform best by ignoring the needs and desires of the other half of the electorate. It is simple strategy, and the result of this strategy is the enactment of polarizing (read: fringe) policy, which does not align with the needs and desires of the people.

How do we fix this issue? Once again, it’s mechanical. Instead of having only one proposal at a time be put to an up-or-down vote, legislatures should vote on multiple different proposals simultaneously. Practically, this would look like selecting an issue that needs to be addressed and a timeline leading up to the vote. For example, the goal could be “reducing domestic gun deaths” and the timeline would include two months of drafting proposals followed by a week of “campaigning” for different proposals within the legislature by dedicating 2 hours per day on the floor through that week to the specific issue. Over the course of that week, feedback is taken, adjustments are made, and proposals with more than, say, 15% of the legislature cosponsoring make it into the “ballot”. At the end of the week, legislators use Score Voting or STAR Voting to vote on all of the different proposals at the same time. The winning proposal gets sent. As an extra measure, an additional up-or-down vote on the winning proposal could be made, but it should always require a supermajority to pass. That supermajority threshold could be anywhere from 60% to 90%, but the point is that comprehensive proposals can often easily gain an additional 10 to 20 percentage points of support once they cross a critical threshold of around 70% to 80% support — that is, the graph of effort to gain more support dependent on how much support a proposal already has is wonky and has a dip in it toward the right end.

Now, in the US, I think voting method reform is far more viable right now than changing the rules of US Congress to the concept I’ve described here, but having better representatives and educating the public on the fundamental concept that “Majority is not democracy. Democracy is about building consensus.” could potentially lead to legislatures at different levels adopt this method someday.