Robin Hanson talks about how economists signal:

Economists like to point out there’s almost no chance that your vote is going to determine an election. So one of the things an economists like to do to show off that they’re clever economists is to not vote and to say to everybody, hey I’m smarter than all the rest of you! See, I understand that by voting, it’s not going to make any difference, anyway.

I'll trust his analysis that, psychologically, not voting is the economist version of a peacock's tail. But the more interesting question is: are these economists right? Is voting a bad idea?

It's hard to make a consequentialist case for voting in elections, but a related concept - voting with your dollars - doesn't fall victim to the same trap.

Suppose a large chicken producer like Tyson hires farmers to grow as many chickens this year as it sold last year. So if they sold 50k chickens last year, they'll buy 50k chickens to resell this year. Because they are a big company, they only track these numbers to the nearest thousand. So 999 times out of a thousand, your purchases don't even show up on Tyson's radar.

The vast majority of the time, your purchases won't affect anything. So there can't be any moral imperative to not eat chicken, since it will almost certainly have no impact, right?

The problem is: sure, there's only a one in a thousand chance that you'll impact Tyson's chicken production. But if you do have an impact, you'll lower their production by a thousand chickens.

Your expected impact is the likelihood you make any change, multiplied by size of that change. If you have a 1/1000 chance of making an impact, and you impact their production by 1000 chickens, then your expected impact is 1/1000 * 1000 = 1. The same goes for if your odds are one in two or one in a billion - your expected impact is always 1.

So stay home from the polls next November if that will get you into the cool kid crowd in the economics department. But make sure you stay home with a vegan dinner.


  1. Needs less sarcasm

  2. Yeah ok but the variance of your "impact" is (1/1000)*1000^2 - 1 = 999 so you can't claim with any certainty that what you're doing makes any difference.