(I’ve said that recounts don’t change these elections by three figures unless there was a clear screw-up somewhere… this diary breaks that throwaway line out into a real argument. – promoted by odum)
The chance that the recount of the 5-way primary election for the Democratic nomination will change the results is virtually nil unless systematic fraud was done by people trying to prevent Racine’s election. Since nobody thinks that was the case, what the recount is going to to do is to clean up all the little random errors that creep into the process:
* hand counted ballots that are accidentally put into the wrong pile
* rejected ballots that clearly indicate a preference
* errors of transposition of numbers
* errors of transcription of numbers for one candidate into an other candidate
Errors of the last two types are highly unlikely given the closeness of the outcome. Before transmitting the official results on Wednesday, clerks around the state knew the importance of triple checking their work. Furthermore, the double blind entry of data at the state election office assured the state that it was identifying and resolving those issues right away.
Indeed, after all this careful work Shumlin’s lead increased by around a dozen votes. And the final results, when you look at them by county and town, all fit into patterns that don’t raise any red flags where one might say “golly, did Lamoile county mix up Matt Dunne’s results with Susan Bartlett’s?
Now the question becomes how many of the other two types of errors were there? Well, the state bar for demanding a recount is for the vote differential to be under 2%. This implies that the state anticipates a 1% error rate in vote counting and reporting — because a 1% error can turn a 2% victory into a tie.
This in turn means that there may be around 750 errors out there. Random errors — we’ve already decided that nobody thinks there was fraud anywhere.
How will random errors distribute themselves?
Randomly.
This means that each candidate’s error rate will be in proportion to the votes they attracted. Doug and Peter would both be expected to have around 180 errors. If they broke 50-50, they both wind up in the same place. But they won’t break 50-50 because of the other candidates in the race. Instead the votes Peter loses will go proportionally to Doug, Deb, Matt and Susan. And the votes Doug loses will go to Peter, Deb, Matt and Susan.
It’s the involvement of the other 5 candidates that make the recount extremely unlikely to change the outcome if the errors are random.
Since nobody in the Racine camp is pointing out to problems with the vote totals out of a particular county or town, I’m not sure why they think there is more than a one-in-a-thousand chance that the random errors will break in Doug’s favor. In effect, they are saying that in 180 flips of their own coin they are going to come up heads 2/3rds of the time, and that when Peter flips his coin 180 times he’s going to come up tails 2/3rds of the time and so will Deb, Matt and Susan — and that the benefit of all the ‘tails’ flips will be in Doug’s favor.
That’s a very, very, very, very, very, very steep hill of probability to climb.
I’ve done it for over 20 years.
This diary is excellent. It puts into probability terms what I know but would not have been able to explain in such detail.
Thanks.
The recount requst is a huge mistake, whose only purpose will be to bting “closure” for Doug and some of his supporters who urged him to do this.
He could still change his mind…
Nice work FX, your point about how a 5 way race is exponentially less likely to come out with significant changes transcends politics and ascends into the lofty realm of statistical geekiness.
and reassures us that the vote tally is unlikely to change. That being said, the fact that so many people are calling for it (did I hear something about 85% of voters on last nights news?) including supporters of candidates other than Doug Racine, we should adopt a positive perspective on the recount and assume that it will further validate Peter Shumlins presumed candidacy and allow that large segment of voters to invest themselves more enthusiastically in the general campaign…which to my mind has already begun.
Complaining about the recount is not going to change the fact that it is, in fact, happening now. It only represents an obstacle if we choose to dwell on it as such.
There is the “probability” that there will be a random distribution, but the mere fact that there is a random distribution, presupposes the “possibility” that all the error corrections could end up for one candidate.
The theory of Chaos demonstrates this point.
Short version – keep all “chances” on the horses; which is not a “random” game.
Since my name is being bandied about so freely I would just like to observe that the perceived closeness of the vote by members of the electorate who are neither statisticians nor students of quantum mechanics will, in the absence of a recount, leave a niggling doubt in the minds of the approximately 74% of the electorate whose candidate didn’t come out on top in the certified count. I include ALL non- winning candidate’s supporters because in a 5 way race we all run a little IRV machine in our heads.
A court supervised recount will eliminate doubts, not of fraud, but of possible human error and get all the team pulling in a straight line to victory in November. Generally, all publicity is good publicity. The press will be focused on the recount and provide all candidates the opportunity to exhibit statesmanlike behavior while Dubie remains sealed in his bunker.
And finally, even Schrödinger didn’t know if his cat was alive or dead, at least for the first few days.
Random Arrow
Last year we were all inspired by the democratic recount process in Minnesota where Al Frankin eventually won by 300+/- votes after Colman had declared himself the winner in the November 2008 election.
I say respect the process but verify that every vote is counted correctly.
Minnesota proved that there are many ways for votes to be wrongly counted, and not all of them sinister.
There is also plenty of precedent to question the validity of the optical scanner counts.
As discussed here http://hackingdemocracy.com/ and here
I hope all actual paper ballots are hand counted to assuage that worry.
What I really miss is Instant Runoff Voting.
Despite the complexity it was a real breath of democracy in action….
And applied here we would really know which candidate most democratic voters in VT preferred… and know who that was only as late as a day after the election too !
(Cross posted)
Now that the results are in, we know that:
* The towns did an exceptional job of counting originally. The error rate was .19%, or 19 votes in 10,000. That is about 20% of the expected 1% error rate. As I predicted, because it was a top-of-the-ticket race, the towns were super-careful in their initial reporting.
* The errors were random. Both candidates added and lost handfuls or no votes. It was like coin-flipping.
There was no evidence to support the recount request. Making it wasted a lot of time and effort. Future recount requests should not be supported absent a compelling case that the results will be changed.