If we had range voting, who would win this election?

July 16, JDN 2457586

The nomination of Donald Trump is truly a terrible outcome, and may be unprecedented in American history. One theory of its causation, taken by many policy elites (reviewed here by the Brookings Institution), is that this is a sign of “too much democracy”, a sentiment such elites often turn to, as The Economist did in the wake of the Great Recession. Even Salon has published such a theory. Yet as Michael Lind of the New York Times recognized, the problem is clearly not too much democracy but too little. “Too much democracy” is not an outright incoherent notion—it is something that I think in principle could exist—but I have never encountered it. Every time someone claims a system is too democratic, I have found that deeper digging shows that what they really mean is that it doesn’t privilege their interests enough.

Part of the problem, I think, is that even democracy as we know it in the real world is really not all that democratic, especially not in the United States, where it is totally dominated by a plurality vote system that forces us to choose between two parties. Most of the real decision-making happens in Senate committees, and when votes are important they are really most important in primaries. To be clear, I’m not saying that votes don’t count in the US or you shouldn’t vote; they do count, and you should vote. But anyone saying this system is “too democratic” clearly has no idea just how much more democratic it could be.

Indeed, there is one simple change that would both greatly expand democracy, weaken the two-party system, and undermine Trump in one fell swoop, and it is called range voting. I’ve sung the praises of range voting many times before, but some anvils need to be dropped; I guess it’s just this thing I have when a system is mathematically proven superior.

Today I’d like to run a little thought experiment: What would happen if we had used range voting this election? I’m going to use actual poll data, rather than making up hypotheticals like The New York Times did when they tried to make this same argument using Condorcet voting. (Condorcet voting is basically range voting lite, for people who don’t believe in cardinal utility.)

Of course, no actual range voting has been conducted, so I have to extrapolate. So here’s my simple, but I think reasonably reliable, methodology: I’m going to use aggregated favorability ratings from Real Clear Politics (except for Donald Trump, whom Real Clear Politics didn’t include for some reason; for him I’m using Washington Post poll numbers, which are comparable for Clinton). Sadly I couldn’t find good figures on favorability ratings for Jill Stein and Gary Johnson, though I’d very much like to; so sadly I had to exclude them. Had I included them, it’s quite possible one of them could have won, which would make my point even more strongly.

I score the ratings as follows: Every “unfavorable” rating counts as a 0. Every “favorable” rating counts as a 1. Other ratings will be ignored, and I’ll add 10% “unfavorable” ratings to every candidate as a “soft quorum” (here’s an explanation of why we want to do this). Technically this is really approval voting, which is a special case of range voting where you can only vote 0 or 1.

All right, here goes.

Candidate Favorable Unfavorable Overall score
Bernie Sanders 48.4% 37.9% 50.5%
Joe Biden 47.4% 36.6% 50.4%
Elizabeth Warren 36.0% 32.0% 46.2%
Ben Carson 37.8% 42.0% 42.1%
Marco Rubio 36.3% 40.3% 41.9%
Hillary Clinton 39.6% 55.3% 37.7%
Scott Walker 23.5% 29.3% 37.4%
Chris Christie 29.8% 44.5% 35.3%
Mike Huckabee 27.0% 40.7% 34.7%
Rand Paul 25.7% 41.0% 33.5%
Jeb Bush 30.8% 52.4% 33.0%
Mike O’Malley 17.5% 27.0% 32.1%
Bobby Jindal 18.7% 30.3% 31.7%
Rick Santorum 24.0% 42.0% 31.6%
Rick Perry 21.0% 39.3% 29.9%
Jim Webb 10.3% 15.0% 29.2%
Donald Trump 29.0% 70.0% 26.6%

Joe Biden and Elizabeth Warren aren’t actually running, but it would be great if they did (and of course people like them, what’s not to like?). Ben Carson does surprisingly well, which I confess is baffling; he’s a nice enough guy, I guess, but he’s also crazypants. Hopefully if he’d campaigned longer, his approval ratings would have fallen as people heard him talk, much like Sarah Palin and for the same reasons—but note that even if this didn’t happen, he still wouldn’t have won. Marco Rubio was always the least-scary Republican option, so it’s nice to see him come up next. And then of course we have Hillary Clinton, who will actually be our next President. (6th place ain’t so bad?)

But look, there, who is that up at the top? Why, it’s Bernie Sanders.

Let me be clear about this: Using our current poll numbers—I’m not assuming that people become more aware of him, or more favorable to him, I’m just using the actual figures we have from polls of the general American population right now—if we had approval voting, and probably if we had more expressive range voting, Bernie Sanders would win the election.

Moreover, where is Donald Trump? The very bottom. He is literally the most hated candidate, and couldn’t even beat Jim Webb or Rick Perry under approval voting.

Trump didn’t win the hearts and minds of the American people, he knew how to work the system. He knew how to rally the far-right base of the Republican Party in order to secure the nomination, and he knew that the Republican leadership would fall in line and continue their 25-year-long assault on Hillary Clinton’s character once he had.

This disaster was created by our plurality voting system. If we’d had a more democratic voting system, Bernie Sanders would be narrowly beating Joe Biden. But instead Hillary Clinton is narrowly beating Donald Trump.

Trump is not the product of too much democracy, but too little.


5 thoughts on “If we had range voting, who would win this election?

  1. > Condorcet voting is basically range voting lite, for people who don’t believe in cardinal utility.

    Hilarious! And true.

    Score Voting has seen some “real” use in a GOP straw poll for example.

    I mean, it counted for something and was highly gamed, although it did not directly decide who got into office.


    • The worst case scenario for range voting is when everyone strategically gives the top rating to their favorite candidate and the bottom rating to everyone else, which is literally equivalent to our current system. Let me repeat that: The worst-case scenario is our current system. Forcing people to give complete rank orders won’t solve that problem either; it will just result in spoiled ballots and people ticking the boxes randomly or arbitrarily. No voting system is perfect against strategic voting; but range voting is the one system that doesn’t require strategic voting just to ensure monotonicity.


      • The specific “later-do-no-harm” principle that FairVote came up with is actually quite strange, and essentially makes a virtue of violating other intuitive principles of voting.

        The only way that you can “harm” your favorite candidate in range voting by voting for your second-place candidate is if the overall utility of that second candidate is higher. That is, the only way you can cause “harm” in this sense is by making your country’s people better off.

        Here’s a very simple example. There are two candidates running, Red and Blue. The voters are highly partisan, so they come in only two possibilities:
        Reddish voters prefer Red to Blue, and Bluish voters prefer Blue to Red. 51% of voters are Bluish, while 49% of voters are Reddish. Furthermore, suppose that Reddish voters hate candidate Blue, while Bluish voters only mildly dislike candidate Red.

        If everyone voted honestly, all Reddish voters would give Red 10 and Blue 2, while all Bluish voters would give Blue 10 and Red 4. The result is that Red gets 694 points and Blue gets 608 points, so Red wins. Moreover, the total utility of society corresponds to the vote at 694 points.

        Now, if everyone voted strategically, all Reddish voters would give Red 10 and Blue 0, while all Bluish voters would give Blue 10 and Red 0. Red gets 490 points and Blue gets 510 points, so Blue wins. The total utility of society is now only 608 points, due to the fact that Blue won even though the harm to Bluish voters of Red winning is smaller than the harm to Reddish voters of Blue winning.

        So all this business about “contested elections” where “voters have a stake” is really saying when people care more about winning than they do about making their country better. If that is indeed the case, and voters are polarized in this way, then range voting can collapse into plurality voting. But actually I think IRV would also collapse into plurality voting, and so would basically everything else… so what are we even arguing about? We have this system which could in many circumstances be substantially better than our current system, and at worst ends up exactly the same—and this is somehow an argument for not using that system?


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s