Why democracy is good; why choose-one is bad; and how to fix it

Your government is in crisis. This was predictable. And there’s hope.

(paywalled version here)

Since you’re reading this in English, you probably live in a country in deep political crisis. Your governing political party probably exercises disproportionate power despite being supported by a minority, and have dropped all pretense of governing in the interests of anyone but that minority. In fact, in many respects the governing party isn’t even looking out for its voters, only for itself.

That’s certainly the case in the US and UK right now; and arguably so in Canada. (As for other English-speaking countries, that would be another article, but some of the below applies even there.)

An illustration of “crisis of democracy”: Boris Johnson, Theresa May, and Donald Trump. Sorry, I couldn’t find a photo that also included Justin Trudeau… :)

Of course, nobody could have predicted the specifics of your country’s current crisis. In the US, Trump himself couldn’t tell you what words he’ll invent in his next tweet; in the UK, even Speaker Bercow can’t get any semblance of order. But in terms of the general problem — a party with only minority support yet with the power and will to govern in the interests of an even smaller minority — this is depressingly predictable. To understand why, and how to fix it, it helps to start from the basics.

Note: I’m trying to keep this article short, so it’s inevitably dense. I could take almost any sentence here and expand it to several paragraphs. So if anything I say seems unclear or poorly-motivated, just ask, and I’d be happy to discuss it further.

What’s the point of democracy, anyway?

Democracy is supposed to be a way to get good government; that is, to ensure several characteristics.

• Institutionality. Prosperity requires some degree of predictability; yet predictability requires some degree of change, because brittle unchanging systems eventually break down catastrophically. The first job of democracy is to provide for such predictable, non-violent change, while maintaining a baseline of institutional continuity. This also includes some amount of decisiveness. If there is no democratic means for resolving some matters decisively, conflict will just simmer or be resolved through individual, unaccountable force.
• Accountability. This starts with legal accountability; nobody should be personally above the law, least of all those who have power over others. But more importantly, it means accountability for results. Politicians should have to care about their individual reputation, and parties should have to care about their collective reputation.
• Representation. Collective decisions are better when they include a diversity of voices, and a government that doesn’t represent you will tend to oppress you.
• Deliberation. Compromise and consensus-building leads to better decisions overall. It’s also key to maintaining institutional legitimacy.
• Practicality. Ordinary citizens should be able to stay reasonably informed and to participate in democracy without it being a burden.

Designing a democratic system to meet these goals, as with any design project ever, involves some tradeoffs. But you can do a hell of a lot better than we do currently. And, as I’ll explain below, the necessary reforms aren’t just pie-in-the-sky; they’re politically feasible in the US, UK, and Canada.

Our current voting method in most of the English-speaking world, choose-one voting (also called plurality voting or first-past-the-post) is pretty terrible at all of these goals except the last and least. Partisan ping-pong damages institutionality; polarization undermines individual accountability, while minority power undermines collective accountability; wasted votes and gerrymandering make a mockery of representation; and zero-sum battles spoil deliberation. But at least it’s relatively practical and easy to decide whom to vote for, so I guess there’s that.

At the heart of all these problems with choose-one voting is the issue of wasted votes. Since you’re restricted to vote for just one candidate, supporting anybody but one of the two frontrunners means you have no voice in choosing between those two. There may be broader reasons to choose to give up that voice, but from a the point of view of narrow one-shot game theory, that’s a wasted vote. And the reason it’s wasted is that the voting system unnecessarily forced you to choose between your true favorite and a lesser evil.

Even if all you care about is the one person in charge — for insance, the President — there are voting methods that are much better than choose-one. Approval voting is the simple first step to reform. It’s just like choose-one, except that you’re allowed to approve as many candidates as you want instead of just picking one. It turns a zero-sum two-way battle into a healthier situation where win-win solutions are possible. RCV (the current brand name for a “Ranked-Choice Voting” system that used to be called instant runoff) also helps and is getting adopted more and more (though it’s got some technical problems approval voting doesn’t have). And STAR voting, a new proposal from Oregon, unites the best features of both approval and RCV.

But those reforms, important though they would be, don’t actually fulfill the true goals of democracy. And I’ve already said why: wasted votes.

Wait — don’t the above voting methods solve that problem, allowing you to vote for your true favorite while still having a voice in the choice between the two frontrunners? Yes, they do. But get ready to have your mind blown:

Any vote that doesn’t end up helping elect a winner is wasted. It’s not just the small percentage who votes for minor parties whose votes are wasted by our choose-one election method; it’s also everyone who votes for the losing major party.

Huh? How is that possible? Isn’t the law of democracy that “majority rules (and minority drools)”? If the ideal is a world where all voters win, wouldn’t that mean that actually nobody does?

The supervillian Syndrome says: “When everyone’s super, no one will be.” But he’s wrong.

Yes, it’s true that when it comes down to actually making a decision, there are inevitably winners and losers; though a good government will try to ensure the losses are as few and as light as possible, there are inevitably some cases where they are almost as big as the gains. But the ideal democracy isn’t about making decisions directly; it’s about deciding who will decide. And on that matter, it is entirely appropriate to try to keep the group of losers as small as possible, and in practice possible to ensure that group will be a tiny fraction of the electorate.

That’s the idea of proportional representation — #ProRep. This is not a particular voting method, but rather a guiding principle for designing voting method for electing a multi-winner body such as a legislature. It’s usually explained as “a group that has X percent of the vote should get about X percent of the seats.” But I prefer to think of it in terms of avoiding wasted votes. If all votes count equally, and as many of them as possible help elect a winner, then proportional percentages are an inevitable consequence.

Let me restate that in simpler terms:

Our current choose-one voting methods waste over half of the votes. Even the best single-winner methods can waste almost half; much better, but still not good enough. Proportional multi-winner methods can waste only a tiny fraction of the votes.

How could #ProRep work?

I’m going to discuss three proportional representation methods. For each of the three, I’ll explain the basics of how it works, then see how it measures up in terms of the democratic ideals I listed above. Though I’ll be discussing the pros and cons of each method, I’d like to be very clear that all three of these would be an enormous improvement over our current choose-one voting methods. Then finally, below, I’ll briefly consider what it would take to pass these reforms in practice, focusing on the case of the US (because it has more veto points than the UK or Canada).

I’ll list the three in ascending order of my own preference, but I could imagine good arguments for favoring any of the three.

Single Transferable Vote (STV; aka RCVn, ranked choice voting with n candidates)

Here’s how this works:

In this sample ballot for STV, voters rank candidates by filling in bubbles. I like the Australian version, where they write in numbers, a bit better.

1. Voters are divided into multi-seat districts. For instance, 5 single-seat districts might be merged into one.
2. The ballot is a list of all the candidates running in your district. As a voter, you rank them in order, writing a “1” next to your first choice, a “2” next to your second, etc.
3. First choices are tallied.
4. In a 5-seat district, each seat takes at least 1/6 of the votes (leaving less than 1/6 of the votes as wasted). If any candidate already has that many, they win, and their excess ballots are transferred to the next preference of that voter.
5. If all the seats aren’t full, the candidate with the fewest votes is eliminated, their ballots are each transferred to the next preference of that voter, and you return to step 4. Otherwise, the election is done.

This is similar to the procedure for single-winner RCV (RCV1), but the threshold to win is lower (1/6 in the example above, instead of 1/2), and the rule about transferring excess votes is different. The lower threshold means that, as compared to RCV1, centrists are much less likely to be squeezed out of the race prematurely.

STV is used in Ireland, and in many elections in Australia.

How does this do by the democratic criteria listed above?

• STV Institutionality: great.
• STV Accountability: This is great for individual accountability of each candidate. Unfortunately, it is not good for party accountability, because each candidate simply runs as an individual.
• STV Representation. In the example above, with 5-seat districts, wasted votes are kept to under 1/6, so any united group of 17% or more of the voters is guaranteed to be represented. That’s pretty good; enormously better than choose-one voting. In practice, STV seems to be pretty good at representing all groups, including women and minorities, fairly.
• STV Deliberation. Though it’s certainly better than choose-one, this may be the biggest defect of STV. Because candidates run as individuals, they have some incentives to intransigently focus on single issues, rather than signing on to a broader multi-issue party platforms. This can impede compromise and favor hard-liners.
• STV Practicality. This is also less than ideal. In the 5-seat example above, a ballot would probably have over a dozen candidates, and a voter would probably have to rank at least half of them to avoid a heavy risk of a wasted vote. As somebody who actually uses this system for my local elections here in Cambridge, MA, I know that this can be a bit tiresome, and could end up effectively disenfranchising less-engaged voters.

Bavarian-style (open-list mixed-member proportional voting with transfers)

Here’s how this works: (Note: Feel free to skip over steps 4, 5, and 6. They sound a bit complicated, but voters don’t have to worry about them, and they make the method slightly fairer overall.)

1. Some seats will be filled locally, and some proportionally. Let’s say for this example that the ratio is 50:50. So each local district would have to be twice as big.
2. Your ballot has one column of candidates for each party (showing any independent candidates in a single column, though for the steps below each of them counts as a party of 1). At the top of each party’s column is their local candidate, if any; and below that, separated by a horizontal line across the entire ballot, are their candidates for other districts in your region (state, etc.). As a voter, you may choose one candidate above the line, and one candidate (from the same or a different party) from below the line.
3. Within each district, the local candidate with the most local votes wins, just as in choose-one voting.
4. All local ballots which choose a winning local candidate are counted as if they cast both votes for that party. (That way, you can’t get extra voting power by splitting your vote between two parties.)
5. Each party’s proportion of all votes, counting votes above and below the line equally, is calculated. Any ballots which vote only once are counted as if they’d voted for the same party twice. For instance, if I voted for a non-local candidate from party X but did not vote for any local candidates for any party, I would be counted as if I’d given both of my votes to party X.
6. If any party Y does not have enough votes for even one seat, then for any ballots that voted for both Y and some other party Z, the Y votes are transferred to party Z. (That way, you can afford to honestly vote for your favorite party on at least half your ballot, even if you think they won’t win any seats; and neither half of your vote will be wasted.)
7. Any party that has fewer seats than their proportion of votes is given extra seats, which go to whichever of their candidates have the most votes.

This system (without steps 4, 5, and 6) is used in Bavaria, a “Land” (region) within Germany.

How does this do by the democratic criteria listed above?

• “Bavarian” Institutionality: great.
• “Bavarian” Accountability: This is a good balance of individual accountability and party accountability. Independent candidates get a fair shot, but parties with a popular platform can also work together effectively. Still, individual accountability is slightly less than under STV.
• “Bavarian” Representation. Representation is excellent under this method. In an election with at least 9 seats overall, wasted votes would typically be under 10%; with at least 19 seats, under 5%.
• “Bavarian” Deliberation. Good. No voting method can ensure that politicians will compromise for the greater good, but this method at least avoids unfairly rewarding those who won’t.
• “Bavarian” Practicality. Ballots can be quite large, but as a voter you only have to choose two candidates. Also, the full rules for ballot counting are a bit complicated, but the process of voting is simple. All-in-all, this system is pretty practical; I’d say, more so than STV.

PLACE voting (Proportional Locally-Accountable Candidate Endorsement voting)

Here’s how this works:

1. Voters are divided into single-seat districts, just as today under choose-one voting.
2. Before the election, candidates publicly designate their closest allies within their own party (“faction allies”), as well as their more-distant allies in other parties (“coalition allies”). If you vote for a candidate, and they don’t win, your vote will be transferred — first to a faction ally, then to a non-ally from the same party, then to a coalition ally. Only in the unlikely event that none of these can win will your vote be wasted.
3. Your ballot lists the parties, including the local candidate for each of those parties. You have one vote, which you may use in one of three ways: choose a local candidate; choose a party without voting for its local candidate; or choose a party and write in a candidate from some other district. To help with option 3, a list of all candidates in all districts is available; and to help resolve illegible write-ins, you’re encouraged to include their district number.
4. Votes for each candidate are tallied.
5. Any candidate with under 25% of the local votes in their own district is eliminated, unless that would leave no candidates in their district. Votes for these eliminated candidates are transferred as explained above.
6. To win a seat, it takes a number of votes equal to the average precinct. Any winners are seated, and any excess votes they have are transferred as explained above.
7. Whenever a candidate wins, all other candidates in that district are eliminated, and votes are transferred as explained above.
8. If all seats are full, or if only one candidate per district remains, go to step 9. Otherwise, eliminate the candidate whose tally is the farthest from the frontrunner in their district, and transfer their votes, as explained above; then go to step 6.
9. At this point, all the seats are filled. Parties are responsible for assigning their winning candidates “extra territory” so that each district is in the territory of exactly one representative per winning party.

This process ensures:

• Each district will have one local candidate win.
• Each district will be represented by one candidate per party.
• Near-ideal representation: almost all votes end up helping to elect a representative, and very few are wasted.
• Simple ballots: Less-engaged voters can vote just as they do today, simply choosing one local candidate, and be assured of equal voting power with all other voters.
• Optimal choice: Any voter can vote for any candidate, regardless of location. If the candidate who excites me and best represents me is in some other district, I can still vote for them.
• Real voice for groups that are too small to win even one seat. If my candidate wins, great; but even if they don’t, I’m giving them power by encouraging other candidates to court them as an ally.
• Avoids excessive splinter parties (like those in Israel): the 25% local threshold means that in order to win seats, a party needs a substantial base of support in at least one district, not just tiny support everywhere. Parties which fall below that threshold will still have a fair voice during the election process, as explained in the point just above, but will not actually be seated in the legislature.
• For incumbent politicians under the current system, PLACE offers a level of predictability and stability that no other #ProRep method can. Because districts would remain the same and ballots would remain similar, popular incumbents who win fair-and-square under the current system can anticipate that they’ll still win under PLACE.

This system is not currently used anywhere.

How does this do by the democratic criteria listed above?

• PLACE Institutionality: great.
• PLACE Accountability: Excellent. Each representative is individually accountable, to both their local district and to their supporters elsewhere; each party is accountable as a whole, building a reputation over time as the 25% threshold rule discourages excessive turnover and splintering among the parties.
• PLACE Representation. Excellent. Because any voter can vote for any candidate, minority groups (whether ideological, occupational, or ethnic) can build their voters’ engagement across the entire area.
• PLACE Deliberation. Good. No voting method can ensure that politicians will compromise for the greater good, but this method at least avoids unfairly rewarding those who won’t.
• PLACE Practicality. Excellent. Ballots are reasonably short and simple, and voting is easy. Note also that the last bullet point in the group above, about stability and predictability. This makes PLACE an especially practical reform to offer to incumbent politicians.

Is this realistic? Yes!

All this discussion of improved voting methods wouldn’t be worth much if there were no chance of actually implementing these reforms. But I believe there is a chance. Voting method reform is a hot topic in Canada, where there have been referendums in several provinces; though none of these referendums has yet passed, the one in British Columbia 2005 got 58% support and only failed because of an undemocratic 60% supermajority threshold. In the UK, Wales and Scotland have already reformed locally, and there’s growing pressure for national reform.

Which brings us to the US, where national-level reform is stymied by multiple veto points and supermajority requirements. Luckily, reforming the US voting method(s) would require no constitutional amendment, and could be carried out by ordinary legislation either on a federal or a state-by-state level. State-by-state voting method changes have happened in the past, and there’s no reason to believe they can’t do so again. And federal-level proportional representation would solve the problem of gerrymandering more fully than any other measure, which makes it interesting to whichever party is currently unfairly hurt more by gerrymandering — currently, the Democrats. Finally, PLACE voting is particularly suitable as a way legislators could fix their own voting method, because it’s relatively non-disruptive and thus non-threatening to legitimately-popular incumbents.