Voting Systems and Political Design

There's a result in social choice theory that sounds like it should end all discussion: Arrow's Impossibility Theorem, which proves that no ranked voting system can satisfy a set of seemingly minimal fairness criteria simultaneously. For decades, this produced a kind of despair among voting theorists — if perfection is mathematically impossible, why bother improving? But the despair was premature. Arrow's theorem applies only to systems where voters rank candidates. Systems where voters score candidates — essentially filling out a report card rather than picking a order — escape the impossibility result entirely. This turns out to be crucial.¹

The Spoiler Problem

The system most of the world uses — plurality voting, where each person votes for one candidate — is, by expert consensus, the worst of all common methods. About 11% of the time, a spoiler candidate splits the vote and the second-most-popular candidate wins. The 1912 election is the textbook case: Taft and Roosevelt split the Republican vote, and Wilson won despite a majority preferring some Republican. The 2000 election is another, with Nader pulling enough from Gore in Florida to hand the state to Bush.¹

What's disturbing isn't just that spoilers happen, but that political consultants have learned to weaponize them. Republicans paid for Nader signature drives in 2004. Democrats have paid for radio ads promoting Libertarian spoilers. It's become "thoroughly bipartisan" because funding a spoiler is often more cost-effective than buying additional ads for your own candidate. The system's flaws aren't just unfortunate — they're actively exploited by the people who benefit from them.¹

The Constitution, surprisingly, doesn't guarantee anyone the right to vote for president or Congress. The Founders were aware of the French debates about voting methods, including the spoiler problem, and essentially punted. "The democracy we have now is kind of a retrofit."¹

The Menu of Alternatives

Several alternatives have been tried or proposed, each with different tradeoffs:

Instant-runoff (ranked-choice) voting asks voters to rank candidates. If your first choice is eliminated, your vote transfers to your second choice. This works well with two strong candidates and some minor ones, but degrades when three or more candidates have real support — it can suffer the same vote-splitting effects as plurality. It's been used in San Francisco, Minneapolis, and parts of North Carolina.¹

Approval voting is radically simple: vote for as many candidates as you like, most votes wins. It was used in Renaissance Venice for 500 years and is currently used for electing the UN Secretary General. There's no known serious disadvantage, and in simulations it's the second-fairest system after range voting.¹

Range voting (or score voting) asks voters to rate each candidate on a scale — one to five stars, or one to ten. In a 2000 study, mathematician Warren Smith ran simulations comparing every common voting method and found range voting produced the most satisfactory outcomes for the greatest number of voters. It's the system used by Hot or Not, YouTube star ratings, Olympic judges, and school GPAs. "Nobody has given a convincing argument that there's anything seriously wrong with it — the one thing you sometimes hear is it's complicated, but that's about it."¹

Smith's simulations ranked the methods: range voting first, approval second, Borda third (though easily manipulated), Condorcet fourth, instant-runoff second-to-last, and plurality dead last.

Why Nothing Changes

If better systems exist and experts agree on which ones they are, why do we still use the worst one? Partly because the people in politics have skill sets geared to the current system. Partly because the major parties have discovered how to exploit its flaws. And partly because there's no overwhelming popular outcry — most voters don't know that alternatives exist.

The deeper obstacle is that voting system design is a coordination problem of exactly the kind that Scott Alexander explores in his Archipelago thought experiment. The Archipelago imagines a new frontier where groups with different values can found separate communities — conservative Christians, social justice advocates, Objectivists, people who just really want to lose weight — with a minimal United Government (UniGov) that prevents war, handles externalities, stops memetic contamination, and most importantly guarantees exit rights: anyone who wants to leave their community gets a free ticket out, and any community that tries to prevent departure gets destroyed.²

The Archipelago is interesting because it's simultaneously a libertarian utopia (UniGov does almost nothing), a conservative utopia (people can build cohesive communities around shared values), and a liberal utopia (maximum freedom of association, with oppression disrupted by exit). It extends the basic principle of liberalism — solve differences by letting everyone do their own thing — one meta-level up: instead of "I can be gay and you don't have to be," it's "I can live in a sexually liberated community and you can live in a sexually conservative one."²

But Alexander identifies the key problem with this vision: we lack the vast unsettled frontier that would make it work. And even if we had one, exit rights alone aren't enough. Few Americans actually move to Canada despite it being an equally developed, English-speaking neighbor with policies many prefer. The exit costs aren't monetary — they're social, psychological, the loss of everything familiar. This is why Legibility And State Power matters: the metis embedded in a community — the practical knowledge, the social relationships, the informal institutions — is exactly what you lose when you exit, and exactly what you can't rebuild from scratch.

The practical implication is that we're stuck with reforming existing systems rather than starting fresh. And that means voting reform — boring, incremental, technically tractable — matters more than grand redesigns. Range voting or approval voting could be implemented tomorrow with existing technology. The question isn't whether better systems exist. It's whether the people who benefit from the current broken system will allow the change.

Australia as Natural Experiment

Australia is the closest thing to a controlled experiment in electoral reform. Every major alternative the theorists propose, Australia has tried — and the results are illuminating. Preferential (instant-runoff) voting was introduced for the House of Representatives in 1918, directly in response to the Country Party splitting the conservative vote and handing seats to Labor. Compulsory voting followed in 1924 and remains in force, meaning parties don't waste resources on voter turnout and campaigns focus entirely on persuasion. Proportional representation via single transferable vote was introduced for the Senate in 1948, after the previous system produced absurd results (33 government senators to 3 opposition in 1946–49).³

The consequences have been exactly what theory predicts. Preferential voting stabilised the three-party system that provoked its introduction — the Country Party could coexist with the Liberals without spoiling each other. Compulsory voting changed the character of campaigning but made the partisan impact nearly impossible to measure. And proportional representation in the Senate gave minor parties and independents the balance of power, prompting periodic elite complaints about whether "such a state of affairs is appropriate" — which is to say, proportional representation worked as designed and the major parties don't like it.³