# steven0461 comments on A Rationalist Argument for Voting

• Naive and ex­tremely rough calcu­la­tion that doesn’t take log­i­cal cor­re­la­tions into ac­count: If you’re in the US and your un­cer­tainty about vote counts is in the tens of mil­lions and the ex­pected vote differ­ence be­tween can­di­dates is also in the tens of mil­lions, then the ex­pected num­ber of elec­tions swayed by the marginal vote might be 1 in 100 mil­lion (be­cause al­most-equal num­bers of votes have lower prob­a­bil­ity den­sity). If 0.1% of the qual­ity of our fu­ture light­cone is at stake, vot­ing wins an ex­pected 10 pi­col­ight­cones. If vot­ing takes an hour, then it’s worth it iff you’re oth­er­wise win­ning less than 10 pi­col­ight­cones per hour. If a life­time is 100,000 hours, that means less than a microlight­cone per life­time. The pop­u­lar vote doesn’t de­ter­mine the out­come, of course, so the rele­vant num­ber is much smaller in a non-swing state and larger in a swing state or if you’re trad­ing votes with some­one in a swing state.

• If vot­ing takes an hour, then it’s worth it iff you’re oth­er­wise win­ning less than 10 pi­col­ight­cones per life­time.

Do you mean “per hour” in­stead?

If a life­time is 100,000 hours, that means less than a microlight­cone per life­time.

Have you thought about how to es­ti­mate microlight­cone per life­time from our other x-risk ac­tivi­ties?

In­tu­itively it seems like ev­ery­one ex­cept maybe the most pro­duc­tive x-risk ac­tivists prob­a­bly have less than 10 pi­col­ight­cones per hour of marginal pro­duc­tivity. Does that seem right to you?

• Thanks, I did mean per hour and I’ll edit it. I think my im­pres­sion of peo­ple’s light­cones per hour is higher than yours. As a stupid model, sup­pose light­cone qual­ity has a term of 1% * ln(x) or 10% * ln(x) where x is the size/​power of the x-risk move­ment. (Var­i­ous hy­pothe­ses un­der which the x-risk move­ment has sur­pris­ingly low long-term im­pact, e.g. hu­man­ity is sur­rounded by aliens or there’s some sort of moral con­ver­gence, also im­ply elec­tions have no long-term im­pact, so maybe we should be es­ti­mat­ing some­thing like the qual­ity of hu­man­ity’s at­tempted in­puts into op­ti­miz­ing the light­cone.) Then you only need to in­crease x by 0.01% or 0.001% to win a microlight­cone per life­time. I think there are hun­dreds or thou­sands of peo­ple who can achieve this level of im­pact. (Or rather, I think hun­dreds or thou­sands of life­times’ worth of work with this level of im­pact will be done, and the num­ber of peo­ple who could add some of these hours if they chose to is greater than that.) Of course, at this point it mat­ters to es­ti­mate the pa­ram­e­ters more ac­cu­rately than to the near­est or­der of mag­ni­tude or two. (For ex­am­ple, Trump vs. Clin­ton was prob­a­bly more closely con­tested than my num­bers above, even in terms of ex­pec­ta­tions be­fore the fact.) Also, of course, putting this much anal­y­sis into de­cid­ing whether to vote is more costly than vot­ing, so the point is mostly to help us un­der­stand similar but differ­ent ques­tions.

• Den­sity for al­most-equal num­bers of votes is not lower in most high-stakes elec­tions. I’d say 1 in 5 mil­lion or so. That’s just a bit more than one or­der of mag­ni­tude and doesn’t sub­stan­tially change the over­all con­clu­sions.