[EM] Why the "jury" approach is right for large public elections

[Forest Simmons]

On Thu, 26 Feb 2004, Joe Weinstein wrote:

<snip>

>      In the real world - or even in a realistic fictional one in which
> voters have finite lifetimes and are encouraged to spend their limited
> available time making informed deliberative choices rather than blind
> guesses - the full ballot here is far too long.  Either candidate PROXY
> should be used, or else, using intuitive estimation of probabilities, a
> typical voter will and SHOULD truncate.

... or else, as you've suggested before, a random sample of registered
voters should be given the duty and assignment of studying the issues and
candidates in depth, and then represent the rest of us with their votes.

Personally, I would be a lot more willing to study the candidates and
issues if I thought there was a realistic chance that my vote would make a
difference.

I think that ten thousand voters should be the upper limit of a sample
size, and that most routine decisions should be handled by fewer than an
hundred voters.

Here's one rationale for the upper limit of ten thousand:  if a decision
between two alternatives is to be made and the probabilities for the
respective alternatives are 49.5% and 50.5%, then a random sample of size
ten thousand gives a maximum prior (i.e. before the random sample is
chosen) probability that a random registered voter will be pivotal in the
decision.

[Of course, after the random sample is chosen the members of the sample
will have an even higher probability of being pivotal, and if one is
pivotal, then they all will be pivotal.]

If the probabilities of the respective alternatives are closer than the
49.5% and 50.5% cited above, then the manifold sources of error in the
voting process are likely to effectively transform the outcome into a coin
toss no matter the sample size, whether one voter or the entire population
of registered voters.

[One could argue that a coin toss would be just as fair any time the
results were that close anyway.]

More to the point, if the sample of voters were more than ten thousand,
then the probability of the voters in the sample being pivotal would be
small enough that there would little incentive for them to study the
issues carefully, given the other constraints in their lives, even if the
pre-election polls indicated an exact 50%/50% split in opinions.

We cannot blame all of voter apathy on (our standard whipping boy)
Plurality and (the other old standby) lack of good candidates.  Part of
voter apathy is the realistic assessment that one vote isn't likely to
decide the result, and that there are other pressing needs to attend to,
especially for the working class here in the USA, whose members don't even
get time off from work to go to the polls for a presidential election.

We have laws that require employers to make allowances for jury duty. If
there were fewer than ten thousand voters in each presidential election,
what percentage of employers would be systematically injured by a law
requiring them to give employees time off for election duty?

Are there any compelling reasons to not adopt Joe's idea?

[I can think of some reasons, but not any compelling ones.]

Forest