Jill Stein, the Green candidate in the recent US presidential elections, is right to reject accusations of “spoiling” the election in favour of Donald Trump but should be more careful in her enthusiasm for “ranked choice” voting (also known as “alternative vote”, “preferential voting”, and “instant run-off”).
Let’s get the accusations of “spoiling” out of the way first. Green supporters can give three very good replies to this:
First, as Stein points out, as a purely factual matter, there simply weren’t enough people who voted Green to tip the balance in Clinton’s favour.
Second, Green supporters are not stupid, and many of them do vote tactically for Democratic candidates. The true number of voters who would prefer Stein to Clinton is probably much greater than the number who actually voted for Stein (the phenomenon of “suppressed preferences”). For all we know, every Green supporter who could bring themselves to vote for Clinton did! The voters who voted for Stein may simply have been indifferent between Clinton and Trump. Now, you may complain that those people have the wrong preferences, but that’s different from complaining that they voted the wrong way, given their preferences. If you’re going to complain about voters preferring the wrong candidate, start with the voters who preferred Trump.
Third, the accusation of spoiling depends on voters being confident that their vote for Stein wouldn’t enable her to win. However, while the chance of Stein winning was very low, it was much greater than the chance of any one tactical vote making the difference between Trump or Clinton winning. Though it may be collectively rational for Green supporters to vote tactically, there is nothing irrational about individual voters defecting to vote their conscience.
Now, let’s turn to Stein’s advocacy of ranked choice voting, as practiced in Maine (and Australia, among other places).
Alternative vote (AV) as I’ll call it, is undoubtedly superior to first past the post (FPP) in many ways. Most notably, it allows supporters of minority parties to vote their conscience without the fear of “spoiling”, or splitting the vote of an allied majority party. Naturally, this seems attractive to parties like the Greens, who are regularly accused, in many countries, of splitting the left-wing or liberal vote. In AV, voters rank in the candidates in preference order. The candidate who receives the least number of first preferences drops out, and their 2nd preferences are used to distribute their votes among the remaining candidates; the candidate who receives the least votes after that then drops out, and the highest preferences of their voters are distributed, and so on… So, for example, if a minority of voters would like Stein to win the presidency, they can give her their first preference, and Clinton their second, and Trump their third. Then, when, as presumably will happen, Stein is knocked out, their votes go to Clinton to support her against Trump.
That example shows that AV is great for registering your approval of a losing candidate. Unfortunately it’s not so great for practical Green politics. The real beneficiaries of AV are centrist “compromise candidates” like the UK’s Lib Dems, and Australia’s Democrats.
Consider the following made up situation: the Serious Party, the Silly Party, and the Slightly Silly Party are contesting the election under AV. The parties’ share of first preferences looks this:
The Silly Party have a plurality of first preferences - they would win an FPP election. The Serious and Slightly Silly parties, taken together, have a majority, if they could just agree to co-operate.
Now suppose voters’ actual second preferences break down like this:
If, in an AV election, every voter votes their conscience, the Silly Party wins with 55000 votes. But the Serious Party voters are better off voting tactically. If they all reversed their preferences, putting Slightly Silly ahead of Serious, they could get the Slightly Silly Party to win, which is a better outcome by their own lights. The Silly Party can do nothing to resist this; there is no reason for them to vote tactically; nor is there for the Slightly Silly party, who stand to gain big time from convincing the Serious voters not to vote their conscience.
This shows the two important caveats about AV:
OK, let’s try a live-ish example. Suppose that you are in Utah:
(We’ll assume the Greens aren’t a force in Utah). Clinton voters who vote their conscience give McMullin their second preferences; but that’s a bad policy. They are far better off giving McMullin their first preferences, and Clinton their second. That way, if Clinton is going to lose anyway, they can ensure McMullin wins, rather than Trump; if McMullin is knocked out, their votes go back to Clinton.
Think about what this means for the big picture - right now, liberal voters would do anything to be able to give a mainstream Republican the presidency. If there is a serious hard-right insurgency every election cycle, that means a permanent Republican presidency.
Now, let’s take a more optimistic picture. The Greens (and the Libertarians) appear to be the first preference of only a small minority of voters. Stein rightly points out that this appearance is misleading, as many voters whose first preferences are actually for the Greens take complaints about spoiling to heart and vote tactically. In a preferential system such as AV, the Green vote would undoubtedly be greater than it is now. Moreover, if the Greens got more votes they would get more media attention, which would win them even more votes. The “actual” number of Green supporters in the US under a fairer system could be as great as the number of Democrat supporters. In the jargon, voters’ preferences for the Greens are “suppressed” by the voting system.
Suppose this rosy vision is correct. Then AV would be very bad for the Greens, because it would allow them to supplant the Democrats as the mainstream party of the left, and as we’ve already seen, AV helps centrist minority parties (as the Democrats would be in this scenario) against mainstream parties (as the Greens would be).
Consider what would happen if large number of voters who now support the Democrats were to shift to the Greens. Suppose that, nationally, first preferences are distributed like this (as they very well might be):
Suppose that Republicans and Greens give all their second preferences to the Democrats, and Democrats split evenly between the other two parties (half of them can’t stomach the Republicans, the other half are afraid the Greens will screw up the economy). What will happen?
Well, if everyone votes their conscience, the Republicans will win, despite there being a large majority against them. Just as now, however the Democrats will bitch about Greens spoiling the left-wing vote - after all, if all the Greens voted tactically by giving Democrats their first preference instead of their second, they could force a better outcome by their own lights, by enabling the Democrats to win.
So on the best scenario for the Greens under AV, things would be no different at all: the Greens would be unable to win outright, and the Republicans able to lose only if Greens vote tactically for another candidate ahead of their own.
Let’s take a closer look at tactical voting under AV. There are a few simple rules we can articulate.
First, you get more bang for your buck if you give high preferences to minority candidates. AV completely reverses the usual wisdom that you might as well throw your vote away as vote for a losing candidate. It’s in your interest as a voter to have your vote help as many of your favoured candidates as possible, and you can do that by ensuring that you vote for candidates in reverse order of their chances of winning. Suppose you like Stein and Clinton pretty much equally - you should put Stein first, so that your vote gets to help both candidates. If you put Clinton first, Stein will be knocked out by the time your second preference is counted, and she will never see it.
Second, low preferences don’t count. AV pays loads of attention to the details of who you like; it pays no attention to the details of who you don’t like. This is because your low preferences are only counted if they are between candidates who are still in the running late in the voting, when most candidates have been knocked out. Suppose you are trying to decide whether to vote “Stein > Clinton > Johnson > Trump” or “Stein > Clinton > Trump > Johnson”. It really doesn’t matter, because by the time your vote for Johnson is counted, he will be out of the running. If you really want your vote to say “I don’t like Johnson, but better him than Trump!” you need to give Johnson an artificially high preference. Unaided by tactical voting, AV would been no help in stopping Trump.
Third, the runner-up is the line of indifference. In most elections, you can have a pretty good idea concerning one or two candidates, that they will either win or be the runner-up. These candidates divide your preferences in half, because your vote will never make it past them. Suppose, if you are pretty confident that either Clinton or Trump will be runner-up, and you prefer Clinton to Trump. Then it’s simple: you are voting for candidates you put ahead of Clinton, and against candidates you put after her. The candidates you put ahead of Clinton may get your vote before they are knocked out, may even win; the candidates you put after her will definitely not get your vote, even if they win.
So you might as well do this: (a) identify which candidates you think could be the runner-up. (b) of the remaining candidates, identify the ones you think would be OK. Give your highest preferences to them, fringe candidates first. Give your highest remaining preference to the candidate you most like from set (a). Give your remaining preferences to the remaining candidates in alphabetical order.
The examples I have been using so far all have a feature that is very favorable for AV - candidates and voters have all been arranged along a single left-right axis, with voters always preferring the candidate who is closest to them: extremist voters always preferring a centrist candidate to a candidate of the opposite extreme; centrist voters splitting evenly between which extremist party they prefer.
Real life is much more complicated: political opinion has more axes than one; centrist parties are often more opposed to each other than to extremist parties; and extremist parties from different ends of politics can find themselves unlikely bedfellows.
In the recent US elections, if they had played their cards right, and been aided by a favourable voting system, the Libertarian party could easily have had a major role. The Libertarians are an anarcho-capitalist party - liberal on social issues, and free-market on economic ones. This time round, their ranks were swelled by dissident Republicans who didn’t like Donald Trump, but couldn’t bring themselves to vote Democrat. Democrats couldn’t stand Trump, to the extent that they would all rather see Gary Johnson, the Libertarian candidate, as president, rather than Trump. Trump voters, on the other hand, are a motley collection who, if Trump wasn’t on offer, would be indifferent between all other parties.
Suppose the first preference votes were like this:
And suppose that voters’ second preferences go like this:
In an FPP election, Clinton would win; but in an AV election in which everyone votes their conscience, Trump would. The Clinton voters, however, have much to gain by tactically giving their first preferences to Johnson - by doing so en bloc, they can form an anti-Trump majority.
This situation is different, however, from the ones discussed earlier. It’s not that the Johnson and Clinton voters are in any way a coalition; rather, Clinton voters stand to gain by preventing Johnson supporters’ second preferences from being expressed, and they can only do that by ensuring that Clinton loses to Johnson. Whoa, that blows my mind!
This gives us our fourth principle of tactical voting under AV: reverse your preferences in favour of candidates whose second preferences you don’t like. This works in more mundane cases as well: when a Green tactically votes for the Democrats over the Greens, they are reversing their preferences to prevent those Democrats who prefer Republican over Green from expressing their second preferences.
So far we’ve seen that AV does not nothing to eliminate tactical voting or accusations of “spoiling”, especially not for the Green party. Is there any voting system other than AV that would fix this this problem?
In principle, no, there is not. Here’s why:
“Spoiling” is failing to vote tactically. If tactical voting is possible, there will be accusation of spoiling.
Tactical voting occurs when a voting system elects a candidate A, where there is another candidate B, such that a majority of voters prefer B to A. If that majority had all voted for B as strongly as possible, they would have got a better outcome than they actually did; they should have voted tactically.
A candidate who is such that they would win a pair-wise run-off against every other candidate is called a “Condorcet winner”. Tactical voting (as described above) occurs when a voting system elects someone who is not a Condorcet winner.
Some elections have no Condorcet winners. The Johnson / Trump / Clinton example about is such a case. With Johnson out of the picture, Clinton loses to Trump; with Trump out of the picture, Johnson loses to Clinton; with Clinton out of the picture, Trump loses to Johnson.
So: no voting system that always delivers a winner (or a winning tie) is immune to tactical voting or to accusations of spoiling.
So what am I saying? That nothing is wrong with the way the US conducts presidential elections? Hardly. I’m just saying that changing the voting system is no solution. The problem lies deeper than that, with the very idea of directly electing the head of government in a winner-takes-all contest. This massively exacerbates the weaknesses of any voting system, and exposes the populace to the tyranny, not merely of the majority, but of any plurality that happens to come along.
It wasn’t always so. The framers of the US constitution foresaw the problems of direct election of the president, and so formally, the presidential election is indirect - voters are voting for electors, who were originally supposed to deliberate before agreeing a president by consensus. Deliberative democracy of this kind tends to produce sober, non-partisan, outcomes - the intention seems to have been that the president would stand above partisan politics, being essentially an elected constitutional monarch.
Unfortunately this system could not withstand the rise of modern political parties. Since the electors have no stake in the government - their role is only to choose the president - their primary loyalty is to their party, and, with one or two exceptions, they can be counted on to vote along party lines, with no deliberation. Indeed, it is illegal in most states for electors to do anything else. The result is that presidents are chosen by an extremely opaque form of direct election, and it should be no surprise that this produces demagogues of exactly the kind the system was supposed to resist. Proposals to reform the electoral college, converting the presidential election into a transparent direct election, either by constitutional amendment, or, more practically, by the National Popular Vote Interstate Compact would only make matters worse.
How could it be done differently?
There’s no chance of going back to a deliberative Electoral College elected by FPP. As noted above, that simply won’t work in a modern multi-party democracy, as the electors have no stake in the outcome except the advancement of their party.
One option would be to go to a parliamentary system. Let Congress choose the president. The framers of the constitution rejected that option because they thought it would lead to a partisan presidency - but, it could hardly be worse, could it? Unlike the Electoral College, Congress is a deliberative body containing a plurality of viewpoints, and has a stake in the outcome. They may choose a partisan, but at least it will be a competent partisan.
If you think that gives parties too much power, you could require that Congress elect a president by a supermajority. That will ensure that the president is a non-partisan safe pair of hands, much like the Governor General is in the NZ, Australian, and Canadian systems.
OK, suppose you don’t like a parliamentary system. Why not reboot the Electoral College, and force it to deliberate by denying the parties a majority. Suppose each state elected its electors by single transferable vote (STV). STV is a preferential system like AV, but unlike AV, it produces multiple winners, and tends to be proportional. This would result in most parties (and possibly even independent electors) being represented in the Electoral College, where they would have to build a working coalition to elect a president.
That might be confusing for voters, who would have to choose long lists of named electors. No problem! Take a leaf from the Australian system, and allow voters the option of proxying their votes to a political party, or to another political organisation.
Designing effective democratic institutions is not easy. But I think it’s clear that the US’s institutions are not working (even Donald Trump was saying that - before he got elected). Time they were reformed.
Until September 2016 I am a Tutorial Fellow in Philosophy at Corpus Christi College and an Associate Professor in the Oxford Philosophy Faculty. From then on, I'll be a Senior Adviser at the New Zealand Ministry of Transport.
My intellectual interests are mainly in metaphysics, philosophy of language, and ethics, and of course transport policy.
