Recently Aaron S. Edlin, Andrew Gelman and Noah Kaplan wrote an article in The Economists’ Voice setting out their argument that rational altruists should vote. A more careful version of the argument is here, and if you like there is also a mocking response by Andrew Leonard in Salon, and a more sensible counter-mock by Gelman on his blog.
There’s something right about the argument Edlin et al are making; it being rational for you to vote does require a degree of altruism. But I think their model (a) makes some fairly heroic assumptions, and more importantly (b) doesn’t explain why so many people in America should go vote today. Below the fold I give a slightly different reason for voting, one that applies in all 50 American states. The short version is that you should vote today because it increases your chances of getting a good outcome next time.
For purely selfish people, voting is a bizarre act. You spend a lot of time, energy and (in some cases) money for a vanishingly small chance at improving your lot in life. If you’re an altruist, as Edlin et al point out, it is a little more rational. You spend a lot of time, energy and money for a vanishingly small chance at improving the lot in life of billions of people. Provided you give enough weight to the utility of others, perhaps this could be rational.
But it isn’t clear that it is. It depends on how vanishingly small the chance of making a difference is. Some people are in states where the probability that they are the marginal vote is literally one in billions. (New York and Utah are examples of this.) Many of those people live in states where the probability that their state is the marginal state is also rather high. The probability of being the marginal vote in a marginal state is, for most people, very small. So the model suggested only gives you a reason to vote in close elections.
Indeed, Andrew Gelman in a post about last year’s local elections says that he didn’t vote because the races weren’t close. I think there’s a reason to vote even in races that aren’t that close, and where you don’t give any positive probability to being the marginal vote. (In the currently popular lingo in philosophy, your probability that you are the marginal vote may be vague over an interval including zero.)
Voting is a lot like playing an n-player Prisoners Dilemma with the other people who (loosely speaking) share the values that underlie your vote. I’m taking values to be defined loosely enough here that it includes most people who vote the same way you do. You’d prefer that all of you vote to all of you not voting. Given turnout rates in the U.S., that’s pretty much always the difference between winning and losing. But conditional on what the other people will almost certainly do, you’d prefer to not vote than to vote. And so would everyone else. We have a Prisoners Dilemma here, and the rational thing to do in a Prisoners Dilemma is to not cooperate.
Note that we can’t get out of the Prisoners Dilemma merely by assuming altruism. As Simon Blackburn pointed out a while ago (“Practical Tortoise Raising”, Mind 1995), as long as there are any divergences in preferences, even altruists will be in Prisoners Dilemma situations from time to time. And perfect convergence in preferences is a crazy assumption. So real-life Prisoners Dilemmas for moderate altruists are possible. And I think elections are such possibilities.
But wait! There is more than one election. If any election is a Prisoners Dilemma, then a series of elections over time is a repeated Prisoners Dilemma. And one of the things we know for sure is that in repeated 2 player Prisoners Dilemma, the rational thing to do is to cooperate. That’s true even though the short-term benefits from non-cooperation clearly outweigh the benefits of cooperation. So all we need to do is to show that n player Prisoners Dilemmas are the same, and we’ll have an argument for the rationality of voting.
This last step isn’t completely obvious. In the 2 person case, cooperation is required at turn t because otherwise the other player will have a bad attitude towards you, and hence not cooperate at turn t+1. But in the n-player case, few people will know that you haven’t voted, and fewer still will have their future behaviour effected by this. This isn’t quite true for parents of impressionable children, or perhaps for friends of people who are apathetic/economists, but we might worry it is true of a lot of people.
But there may be another effect of voting. It might be that the probability that others like you will vote in the next election is a more-or-less continuous function of the number of people like you who vote in this election. This doesn’t seem too implausible actually. If we imagine Democrats having received tens of millions fewer votes in 2000 and 2004, then it is hard to imagine the level of support and work for Democrats in this cycle that there actually are. If you imagine the prior Democratic vote being a little closer to actuality, but still down, it is easy to imagine support and enthusiasm levels this time being down.
If I were being more careful here, I’d try to work out the marginal efficiency of voting at producing future votes. That is, the marginal difference that an extra vote for your team in this election makes to the expected number of votes your team gets in subsequent elections. I suspect it is well over 1. That’s because dropping a party’s support by a lot, say 10,000,000 in a US context, feels like it could be so demoralisingly bad that the party struggles to recover. But this would need a lot more work. It is certainly easy to come up with not completely crazy looking mathematical models where the efficiency is over 2. If everyone like you is a little demoralised by your not voting, even if it’s just a very little bit, that could easily translate into votes lost next time.
Moreover, that effect need not be localised. It’s pretty unlikely that New York will be a swing state any time soon. It’s unlikely that it will be close, and conditional on it being close, it’s unlikely that the national race will be close. But cross-voter incentives can work across state lines. People pay attention to what the national popular vote number is, and it effects their marginal disposition to vote/campaign for candidates that you like. So a vote in New York now may translate into benefits in swing states (which probably means the Mountain West) in the future.
So I think there’s a reason for rational altruists to vote, even when they know they won’t be the marginal vote. They are signaling to like-minded voters that they will vote, and that’s crucial for sustained success. This isn’t an alternative to Edlin et al’s model. In fact I think they are right that someone with no other-related preferences shouldn’t vote, even given the dynamic effects of voting described here. But I think the dynamic effects (a) make the case for voting considerably stronger, and (b) explain why it is rational to vote even in states that won’t be close, and won’t be swing states if they are close.
In any case, I don’t think any readers of this blog have no other-related preferences. You should all go vote, as a signal to everyone else that you’re voters. Take my advice, and you’ll thank me in 2012. You really are voting for America’s future; if I’m right most people in the country are largely voting for America’s electoral future.