New Comment
45 comments, sorted by Click to highlight new comments since:

You're implicitly assuming that the way one candidate spends the money is completely valueless and the way the other spends it is maximally efficient. Also, that 75% influence over budget is way off, come on.

The article gives an upper limit of the expected value of a vote. Even if the lower limit is 2 orders of magnitude lower, it still is quite significant. And makes it worth the time to vote, or to try to convince other people to vote your way.

Scott Alexander estimates the value of a vote more conservatively to be about $300 to $5,000 (the higher number for if you live in a swing state.) He backs this up by pointing to specific policies and actions taken by presidents that cost trillions of dollars.

I don't think it's hard to believe that the president can be worth trillions of dollars. Presidents can start and prevent wars, and have very significant influence on domestic policy. In the worst case, a president has some risk of starting a nuclear war that costs [world GDP], or starting a world wide economic depression. In the best case they could stop said war or depression. Supreme court appointments just by itself is worth a huge amount.

"Even if the lower limit is 2 orders of magnitude lower,"

The lower limit is the negative of the upper limit, since your political opinions may be mistaken.

No that's not how this works. We are calculating the expected value of a vote. The expected value is more than just "best case - worst case". You factor in all possibilities weighted by their probability.

As long as the probability you are voting for the correct candidate is higher than 50%, the expected value of a vote is positive. And obviously it's impossible to get odds worse than that, for a binary decision. You can then multiply the probability you are voting for the correct candidate by the expected value of a correct vote, and it's likely to still be worth your time to vote.

Robin Hanson is a bit more conservative and believes the expected value of a vote is only positive if you are more informed than the average voter, and otherwise you should abstain. That's still not a very high bar to pass. The average voter is not terribly informed.

What this entire discussion has been about, is that even if you vote for the correct candidate, is the expected value of a vote really that high? OP demonstrated an upper limit for the expected value of a correct vote. I am simply arguing that even if his estimate is off by orders of magnitude, it's still large.

This makes it worthwhile to 1) take the time to vote, if you thought it was a waste of time. And 2) do more research on the candidates and become a more informed voter, which increases your odds of voting for the correct candidate. And 3) spend effort trying to get other people to do the same.

There is a pervasive attitude at Lesswrong and related communities that all politics is a waste of time. I think this calculation, even if a bit naive and imperfect, demonstrates it might not be so.

"As long as the probability you are voting for the correct candidate is higher than 50%,"

The probability may be lower than 50%, so my statement stands as it is.

That's impossible. You can't have less than 50% confidence on a binary decision. You can't literally believe that a coin flip is more likely to select the correct option than you.

Huh? Of course you can. People consistently make worse-than-random choices all the time. Especially in an information-abusive environment such as casinos, advertising, or politics.

In fact, this entire post and concept is based on the idea that without you, the voting populace would make the wrong binary decision.

It is possible to have a less than 50% track-record regarding binary decisions. However, I am doubtful that anyone could reasonably say about any particular binary decision that he/she has a less than 50% confidence in his/her choice. If he/she really thought that, then the thing to to would be to make the opposite choice at which point he/she would have a greater than 50% confidence in that particular choice.

What do you mean you can't have less than 50% confidence in a decision? The whole idea of expected value is that you can be less than 50% sure that something will have positive consequences, and do it anyway. In this very post the idea is that your vote is almost certainly worthless, but there is a very small chance of a very large effect, and therefore you should vote anyway. But you are much less than 50% sure it will have any positive effect at all. So likewise you can be much less than 50% sure your candidate is the right one.

If you really believe your candidate is less than 50% likely to be the "correct" candidate, you can just vote for the other one. Then you will necessarily have a >50% confidence you voted for the correct candidate. You can't possibly do worse on a binary decision.

You could vote for the other one, but you might not want to, say e.g. that almost all your friends think that the person is the correct candidate.

Also, when you think of the sentence, "my candidate is less than 50% likely to be the correct candidate," you are likely to dislike that assertion, and to start thinking of reasons for saying that they are more than 50% likely to be the correct candidate.

That's impossible. You can't have less than 50% confidence on a binary decision.

You are confusing "confidence" and "the probability you are voting for the correct candidate". These are quite different things.

From your subjective view the expected value of a vote is always positive. That does not mean that it's actually positive -- see Cromwell.

Yes, from your subjective view your vote is always positive. Thus you should always vote.

You are mistaken. My choices are not to vote for candidate A or to vote for candidate B. My choices are to vote for candidate A, for candidate B, to write in somebody (e.g. Cthulhu), and to not vote at all.

No, the choice is to vote for your preferred candidate, or to not vote. Write ins count as "not voting".

From whose point of view? From my own personal perspective there might well be a noticeable difference in utility between writing in Cthulhu and just avoiding the voting station.

In any case, since there are at least three alternatives, one of them does not necessarily have to have >50% confidence.

The fact that you count it as not voting does not mean it is in fact not voting, and it especially does not mean that the person is choosing not to vote (they are not choosing that unless they think they are not voting.)

Well, suppose I think the probability that Johnson would be the best president is 40%, the probability that Clinton would be is 30%, that Stein would be is 20%, and that Trump would be is 10%...

In a first past the post election third parties are irrelevant.

More specifically, the calculations above apply to a close election. 538 gives Johnson a less than 0.01% chance of winning. Obviously the probability of you being the tie breaking vote is many many orders of magnitude smaller than is worth calculating.

In a first past the post election third parties are irrelevant.

*Looks at the UK*

Are they, now?

Third parties aren't stable. They can appear, but they inevitably split the vote. They always hurt their cause more than help it. Unless they are so popular they can outright replace one of the parties.

They always hurt their cause more than help it.

Huh? You mean for their cause it's better to just curl up and die, but refusing to do so subverts their cause..?

Basically yes. First Past The Post does not satisfy the criteria Independence of Irrelevant Alternatives.

Basically yes.

Looking at reality (as opposed to theoretical abstractions), this does not seem to be true.

There is a pervasive attitude at Lesswrong and related communities that all politics is a waste of time. I think this calculation, even if a bit naive and imperfect, demonstrates it might not be so.

I think the naivety and imperfections make it useless as a demonstration of this. Taking such ridiculous estimates of the difference between candidates and your level of knowledge about those differences (and especially the difference between you and the median voter you're hoping to override) makes me doubt the seriousness of the calculation, and makes it impossible to make decisions based on.

Heck, the range of possibilities between "$hundreds of thousands to charity (not mentioned: and many millions to non-charity cronies of the winner)" and "worth your time to vote", even if we discount the "negative value if you're wrong" option (which is real, but probably only reduces EV for this group rather than making it net negative) is enough to show that people making monetary EV arguments are on the propaganda side rather than the truth side of the calculation.

Are you referring to OP or me? I don't think my estimate of the difference between candidates is ridiculous. It's pretty clear the president can have a massive impact on the world. So large that, even when multiplied by a 1 in 10 million probability, it's still worth your time to vote.

Using dollar amounts might be a bit naive. Instead look at utility directly, perhaps some estimate like QALYs. I think something like health care reform alone has the potential to be worth tens of millions of QALYs. A major war or economic depression can easily cost similar amounts.

And again this is just the scenario where you are the tiebreaking vote. I think there is something to be said for the value of voting that goes beyond just the likelihood of being a tiebreaker. For instance, consider all the people that think similar to how you do. If you decide to not vote, they will also decide to not vote. So your decision to vote can still change the election, in a sort of acausal way.

Yes. This is quite an important point.

Yeah. If you assign even a small margin by which a Trump win increases the odds of human extinction, suddenly you end up with a really high number below the bottom line.

Yeah, 75% is pure nonsense. 50% of the budget is social security and medicare/medicaid/CHIP/marketplace subsidies, which are almost entirely locked in. Maybe at the margin policy can adjust this somewhat, especially the marketplace subsidies portion. 16% is defense spending, which are decisions made by the military bureaucracy with a little bit of pork barrel politics. Maybe they could adjust the growth rate of that up or down 5%. 10% is safety net and welfare, 6% is interest on the national debt, and 8% is pensions that have already been agreed to. 11% of the budget is for the rest of the government's services, and most of those budgets are requested by the bureaucracies and individually fulfilled plus or minus some percent. Then, the president has to share the responsibilities for setting those budgets with the senate and congress. I would be surprised if 15% of the budget went to different people under Obama than under Jon McCain, and, I think in terms of realistic impact, where very similar entities have relatively the same impact on my political views, the number is probably in the single-digit percentages.

Also, I would like to point out for Gleb's benefit that "tenure-track professor" sounds worse than simply "professor" for the same reason that "junior carpenter" sounds worse than "carpenter." Most people's availability heuristic is not that the typical professor is not on the tenure track. The idea wouldn't even be salient except that "tenure-track" was mentioned.

Budget source

Gleb, given the recent criticisms of your work on the EA forum, it would be better for your mental health, and less wasteful of our time, if you stopped posting this sort of thing here. Please do take care of yourself, but don't expect the average rationalist to be more sympathetic to you than the average EA.


Great article Gleb! You are the best writer of a rationalism and I will buy all your best selling book to be better at life! I will tell my friends of you on all the social medias! Many of thnaks to you and InIn for the saving of my life!

There's a lot of cringe-y stuff in the link. But Gleb seems like a nice guy to me. Probably overly ambitious about trying to complete whatever mission he sees himself as trying to complete (and probably a bit too anxious to grow his status), and therefore willing to stoop to do some exaggerating and weird quasi-unethical stuff online to get it done... but he seems pretty harmless for the amount of ire he always draws on here.

Writing is hard and he writes some nice entry-level rationality stuff that would be novel to some huge chunk of the population that will never give a damn about lesswrong or any other rationality blog.

Maybe I'm overestimating how mean people around here have been to Gleb. I know there have been some positive things said about him as well. If his treatment's been fair and I'm way off, cool. My bad. But I imagine a universe where he'd been embraced by LW and, for instance, asked to help work a project creating an ELI5 version of the sequences for mass consumption...and I think that is a better universe than this one.

Oh, boy, Gleb seems to have really pissed off some very polite people X-)

I don't see anything in that link that is relevant to this post.

I continue to be amused by people who do not understand how symmetric the equals sign is.

So voting is like donating hundreds of thousands to charity? Oh, great! That means I don't need to donate anything, I'll just go vote.

Isn't this a ripoff of Slate Star Codex take on voting?

I don't think it makes sense to call writing an article about a topic when another person wrote an article about the same subject a ripoff.

Have you read both? I ask because the first two paragraphs are very very similar.

Trying to calculate the expected value of voting goes back at least to public choice economists in the 1960s.

Which was in turn a ripoff of early attempts to calculate E.V. of voting.

No, I mean it has the same source, it starts off in the same way... The first two paragraphs of the two articles are isomoprhic.

There are actually many false assumptions that go into this calculation, including a generic utilitarianism, but even without arguing the particular points, the claim that you do as much good for the world by voting as by donating half a million to charity is absurd on its face.

Because of the particular points, however, I would also disagree that your vote is worth anything like $5,000. And I suspect the vast majority of people, if they could sell their vote, in a one-off situation where no one else could sell theirs, and no one would ever find out etc., would be happy to sell their vote to the party most opposed to their opinions for $5,000 (or for much less, for that matter.)

I've seen something similar on Facebook from Rob Wiblin recently.

One has to be careful of multiplying N by epsilon, especially when epsilon has a rather large range of values....

So if I have a 1 in 60 million chance of being the decisive vote, and 1,000,000 other voters who also voted for the same candidate could also be seen as the "decisive vote", wouldn't that mean that my EV was 640,000/1,000,000 = .64 cents?

Intuitively it seems like 640,000 for voting is way overvalued compared to some other actions, and this diffusion of responsibility argument seems to make some sort of sense.

I don't see why you do that division. The point of being the decisive vote, is that if you didn't show up to vote, the election would have gone the other way (lets ignore ties for the moment.) You can disregard other people entirely in this model. All that matters is the expected value of your action. Which is enormous.

Yeah, you are counting the fact that so many other people are also voting twice if you divide as described above.