Florida and Michigan: use surveys

[ciphergoth]

Florida are talking about spending $10-$13 million on rather flawed June primaries to replace the discounted early primaries that the DNC have disqualified for being against the rules.

Why don't they just hold a survey? Choose 1000 registered Democrats in each state in some way that can be seen to be random in a fair way, ask them, and choose delegates that way? There's no reason not to expect the same result as a proper primary, except that they can do a much better job of it for far less money.

They won't of course, but that's because of an emotional attachment to everyone getting their ballot than because it will actually make a difference.

Comments

[olethros.livejournal.com]

A sample's only useful though when it is in some way backed up by the reality check of a census. Since full polls do show up the flaws in sampled polls from time to time, it seems unwise to rely on a sample for an actual election?

Or is it that the error margin in a vote-by-mail is probably worse than that in a sampled poll of a thousand?

[ciphergoth.livejournal.com]

But you're choosing randomly from an existing list, which is the list of registered Democrats, so you don't need the census data for demographic balancing or any such. It's exactly as fair as a real primary, except that you've randomly struck all but 1000 people off the rolls.

[djm4]

I can see it generating a bias in favour of candidates who wouldn't normally be able to get their voting base out to vote. If you only have an electorate of 1000, and they each know that they're 'representing' a much larger number, there's a fairly strong pressure for those 1000 to all vote. So I'd expect the turnout to be closer to 100% than it would be in a normal primary.

Whether or not you care about that (in fact, possibly you think it's a good thing) is another matter, but I think it could easily affect the result.

[ciphergoth.livejournal.com]

Hmm, good point! Yes, I find it hard to see it as a bad thing but it's a difference all the same.

[djm4]

...and in this specific case, it probably would be better than re-running the primary.

[drdoug.livejournal.com]

I find it hard to see it as a bad thing

Surely 'can organise and motivate supporters to actually go and vote' is a pretty important criterion in selecting a presidential candidate? With a survey you get no reliable information about that.

In this specific instance it might not matter, since both candidates have already convincingly shown extraordinary abilities to organise and motivate supporters to actually go and vote. But if we are taking this specific instance, the last thing the DNC wants now is yet more fuel for argument about the selection process, which using an entirely novel method would surely supply.

[ciphergoth.livejournal.com]

At this point any method will fuel arguments; the best you can hope for is something that both campaigns approve of and which they actually have the money to do. At the moment nothing meets all three of these criteria.

[thekumquat.livejournal.com]

There isn't yet any such list of registered Democrats - you only have to register 4 weeks before the primary (probably varies by state), and lots of people vote for the candidate of the party they don't support who they judge would do worst. In some states you don't have to register at all and can vote for Dem or Reps in the primary as you wish.

E.g. my family are in Michigan, where only the Republicans and Clinton stood, seeing as it was unofficial. My aunt registered Republican, voted Huckabee because he's a nutjob most likely to lose to any Democrat, and registered Dem again afterwards.

[henry_the_cow]

That is just bizarre.

I wonder if anyone in the UK joins political parties they disagree with in order to influence leadership elections? I guess the parties choose leaders so infrequently that it's not worth the hassle.

[thekumquat.livejournal.com]

I've heard of people joining the BNP or Green or even Monster Raving Loony parties as a kind of protest against mainstream party politics, but given you atually have to pay Real Money to join, I suspect it's not common (unlike making protest votes, which is - again especially for the BNP)

[nikolasco.livejournal.com]

The waiting period varies widely; I know MD has a three-month closing before the primaries. There's also a substantial number of states that hold open primaries (http://en.wikipedia.org/wiki/Open_primary).

See also: tactical voting (http://en.wikipedia.org/wiki/Tactical_voting)

[reddragdiva]

In US elections, turnout matters. Obama is getting people off their arses to the polls in the primaries; this is a sign he'd do the same in the election. That's important information a random sample wouldn't give you

[valkyriekaren.livejournal.com]

I'm not sure, based on doubts about previous elections, that fetishising the principles of democracy is necessarily a bad thing for America.

[ciphergoth.livejournal.com]

I don't think primaries need to he held to the same standard as general elections - the DNC has quite a bit of leeway to choose what it'll recognise, which isn't true in a general election.

[zotz]

The same logic, of course, applies to elections in general, as has been pointed out many times.

[ciphergoth.livejournal.com]

That's true, but we rightly hold general elections to a higher standard of fairness, and it would be hard work to prove that the sampling was fair. Crypto could play a role here.

[nikolasco.livejournal.com]

My favorite suggestion so far is to let people vote as usual and then use sampling to check results for evidence of fishiness. This beats the snot out of "full recounts" while preserving the "every vote counts" ideology and existing frameworks for fairness.

[lovingboth]

Because if the end result is tight - I'm thinking mostly of the final national delegate count - the loser would run to the courts over this.

[ciphergoth.livejournal.com]

There's pretty much no way forward that isn't open to such a challenge - any solution relies on the latitude of the DNC in selecting a solution. The best we can hope for is to try to find a method that both candidates will agree to in advance.

[(Anonymous)]

Because it's much easier to bribe 501 people than all the Democrats in Florida?

[ciphergoth.livejournal.com]

Doing so without getting caught could be very hard, especially if some effort was made to keep the list of 1000 secret until after voting.

(ps thanks O Bristolian cable internet user, but could you at least sign with a nym next time as per my user info?)