is an approximate answer to the question "what is the likelihood that the distribution of King County's unexplained absentee ballots is attributable to chance?"
This question arose in comments to my earlier posts (here and here) about the distribution of unexplained surplus absentee ballots by pro-Gregoire vs. pro-Rossi precincts.
Let's simplify the model and look at the distribution of the net surplus 808 absentee ballots over absentee voters. Among the half of the 565,014 absentee ballots that were cast in the most pro-Gregoire precincts, there were 574 net uncredited ballots. The remaining half of the ballots had 234 net uncredited ballots. If we assume that the 808 uncredited ballots should be evenly distributed between these two halves, then the probability that at most 234 are in the most pro-Rossi half is calculated by the Excel function
=BINOMDIST(234,808,0.5,1) = 4.6x10-34
As a point of comparison, that is approximately the same probability of predicting correctly that 5 specific people (say Christine Gregoire, Paul Berendt, Ron Sims, Dean Logan and the lady who planted the finger in the bowl of Wendy's chili) will all be killed by lightning next year. In other words, not very likely.
This doesn't prove that it was ballot-stuffing that caused the extra ballots to show up mainly in pro-Gregoire precincts. But there has to be some explanation other than chance.
Posted by Stefan Sharkansky at May 29, 2005 12:31 PM | Email ThisDid Gregoire win only half the precincts in King County? She surely got more than half the votes in that county.
If you assume the over and under differences between ballots and voters should be evenly distributed between Rossi and Gregoire (ignoring for the sake of simplicity the small number that may have gone to Bennett or may not have had votes for governor on them), why make that assumption?
So if they didn't, can anyone explain why not?
And if they didn't, what in heaven's are they doing?
Posted by: Deadwood on May 29, 2005 01:04 PMCorrect. What I did was rank the precincts in descending order by percentage of Gregoire's vote and divided them into two halves, the most pro-Gregoire, and the least pro-Gregoire. The bottom half contains all the pro-Rossi precincts as well as some that have slimmer majorities for Gregoire.
Posted by: Stefan Sharkansky on May 29, 2005 01:11 PM:-)
RM
The next question, which may be of more legal significance (or maybe not), What are the odds of random events resulting in the current count v.s.the odds of random events resulting in an outcome with 130 more votes for Rossi? More generally, of the universe of possible outcomes, if this distrubution of errors were up to random chance, how many of those outcomes would favor Rossi, and how many would favor Gregiore?
If that number is anywhere near as astronomically large as the other number is small, then at least circumstantially, this is a slam dunk. Which may or may not matter leagally.
Posted by: Dogbert on May 29, 2005 01:51 PMHmmm...You're correct. Tough choice. Of course, the "finger-in-the-chili" lady has already experienced the equivalent of being hit by lightning through having been caught, and the actual finger identified as her husband's co-worker's. So, that leaves Mr. "Merry...Whatever".
Posted by: FlyingTigres on May 29, 2005 01:56 PMPrior to the judge's ruling that the "scatter graph" wouldn't be used for any purpose in court (May 25, at 4 hr 24 min on the TVW audio recording), the recording started (having apparently skipped some testimony and questioning) with someone questioning the petitioners' witness, Bensen, about the decision not to include this kind of analysis by experts.
His answer was that the data provided by King County about the voters credited with voting by absentees was unclear. They had released three different versions of the data in late December and early January. Bensen was unable to obtain an answer to his questions as he sought to determine which, if any, of the data sets was the "locked down" version of the voter registration rolls used to verify absentee ballots during the November canvassing of the returns.
Bensen told his experts (Katz and Gill, I suppose) not to include the data about absentee ballot discrepancies in their reports since he didn't know which data set should be used -- as I understood his testimony, anyway.
Once Huennekens was deposed (approximately April 20, if I recall correctly), the petitioners finally got the answer to the question: the data set released Dec. 29 was the "locked down" version of the registration rolls -- if I understood his testimony correctly.
But, that was apparently too late for the petitioners to include an analysis of the data in their experts' reports and testimony.
Apparently, there was no way around the problem, since the petitioners didn't try to get an analysis into evidence.
Penny wise and pound foolish? I don't know if the experts could simply have been asked to prepare reports in time to meet the deadline for disclosing such things to the other parties -- and just treated the three data sets as alternatives, awaiting clarification of which data set should be used. But, I suspect that could have been done, since their analyses of voting by felons has been done with a view to revisions in the list of felons.
Bottom line: The court won't have expert testimony available to show what the numbers may mean. The attorneys apparently plan to argue that it isn't a random distribution of discrepancies, but they won't have evidence from an expert to show the court how probable or improbable the actual distribution is.
Posted by: Micajah on May 29, 2005 02:03 PMIt seems odd that the rules would not allow flexibility when the deadlines were deliberately used by KC to hide evidence.
I seem to recall that Benson's data was entered into the record even though his scatterplot was not. Can't Katz, who was accepted as a expert, have commented on Benson's data?
Many if not all of the statistical analyses we have seen in the trial were brought to light by Stefan a long while back. One would think, since Chris Vance proclaims himself a reader of this site, that the R's experts would have done this analysis early on and included it in their case in a timely manner.
I regret to say that I am somewhat dissapointed in that the R's legal minds don't seem to be doing a great job.
Hmmmmmmm one could hope.....
Did I say that out loud?
Posted by: Will on May 29, 2005 03:17 PMThis should have been the basis for the Republican's case.
It doesn't prove fraud, but it does prove that these weren't just random mistakes. I suppose that one (such as Westnut) could find some twisted logic to explain this non-random distribution heavily in favor of Gregoire, but as they say in science, the simplist solution is usually the best.
The easiest to imagine in this case is that Gregoire ballots were being added in pro-Gregoire precincts and ballots were being randomly (e.g., not knowing how each ballot was voted) pulled in pro-Rossi precincts. The voting patterns of precincts are well known, so a monkey - which is the quality of people working in KC elections - could have preformed it.
I am a little pessimistic about what Bridges will end up doing, based on the comments he made when he denied Dems move to dismiss, but it is hard to stomach the idea that Fraudoire and her merry band of lobotomized workers can get away with this.
Posted by: BananaLand (aka Iguana) on May 29, 2005 03:25 PMWith the totalitarians in charge, and elections being robbed, I wouldn't be surprised by either a civil war of sorts in the state or, worse, the state outlawing private property and everyone's assets.
Posted by: BananaLand (aka Iguana) on May 29, 2005 03:43 PMAll the apparent errors, irregularities, etc., occurred before the time came to issue a certificate of election declaring Gregoire to be the duly elected governor -- but had absolutely no effect on the legislature's decision to issue that certificate. (Despite the passionate arguments of Reed's detractors, he had absolutely no authority to do anything other than deliver the counties' certified returns to the legislature. It was the legislature's duty to decide whether those returns were sufficiently reliable to declare Gregoire as having been duly elected.)
Those irregularities also had no effect on the decision by King County's canvassing board to certify that the abstract of votes was a true copy of itself (being ignorant of the fact that they were supposed to certify that it was a true representation of the votes cast).
The canvassing board's performance of their required duty (if they knew what it was) was foiled by false information.
The legislature's performance of duty was foiled by a combination of the Republicans' decision not to put the Democrats on the spot (to decide the contested election) and the Democrats' partisan reliance on King County's returns which were questionable at best.
Now, the question is not whether the returns show that someone was duly elected, but whether the available (and introduced) evidence shows that an error was made when Gregoire was certified as the winner by the legislature.
It's like putting a cat back in a bag once it has been let loose.
While it may be easy to show that the cat shouldn't have been let loose, it isn't easy to put it back where it doesn't want to be.
The safe decision is for Bridges to rule insufficient evidence and send it up to the Supreme Court.
Posted by: who'dathunk? on May 29, 2005 05:08 PMI guess I am just stupid, but I still don't understand why you make the assumption that the mistakes should be divided equally between the pro-Rossi and the pro-CG precincts. I keep thinking that the 60% pro-CG vote implies that the distribution should be 60/40 rather than 50/50.
Posted by: iconoclast on May 29, 2005 05:13 PMIf you increased the confidence interval to 99%, you would say, with 99% confidence that the extra votes in either half would be between 367 and 440, inclusive.
As Stefan has indicated, 234 (in the most pro-Rossi precincts) and 574 (in the most pro-Gregoire precincts) are way, way, way outside these intervals. Any statistician would reject the hypothesis of random distribution based on this data.
If you had this much luck with the dice in Vegas, they would be taking you away for using loaded dice!
Bill
Posted by: Bill H on May 29, 2005 05:20 PMSorry.
Posted by: iconoclast on May 29, 2005 06:07 PMThe KC Canvass board used an un-scientific *pattern* analysis to discern voter intent of thousands of ballots! During their meetings - they grew increasingly frustrated and sloppy - (and silly) with those ballots. Their *pattern models* (if that's what they can be called) evolved with their frustration and the bar was raised and lowered throughout their meetings - on their understanding of voter intent!
Why do the Republicans have to waste so much time and money on experts? Heh..They should just use the *pattern model* accepted by the Democrats!
Posted by: Deborah on May 29, 2005 08:18 PMhttp://www.origins.org/articles/dembski_explanfilter.html
"Back in 1985 Nicholas Caputo was brought before the New Jersey Supreme Court. The Republican party had filed suit against him, claiming Caputo had consistently rigged the ballot line in Essex County, New Jersey where he was county clerk. It is a known fact that first position on a ballot increases one's chances of winning an election. Since in every instance but one Caputo positioned the Democrats first on the ballot line, the Republicans argued that in selecting the order of ballots Caputo had intentionally favored his own Democratic party. In short, the Republicans claimed Caputo had cheated.
The question then before the New Jersey Supreme Court was, Did Caputo actually rig the order, or was it without malice and forethought on his part that the Democrats happened 40 out of 41 times to appear first on the ballot? Since Caputo denied wrongdoing, and since he conducted the drawing of ballots so that witnesses were unable to observe how he actually did draw the ballots, determining whether Caputo did in fact rig the order of ballots becomes a matter of evaluating the circumstantial evidence connected with this case..."
After it was shown that the odds of such an occurrence happening randomly, Mr. Caputo was convicted. And they all, except for Mr. Caputo, lived happily ever after. The End.
Don't you often get frustrated when having to deal with party hacks from the peanut gallery, such as Mr Goldstein, who have no formal education in mathematics yet choose to make nonsensical arguments? Did Mr Goldstein even graduate college? Did he even attend?
Contrasted with yourself, it must often feel like trying to explain the reason for not stealing a candy bar to a 4 yeard old. You have my sympathies.
Posted by: pbj on May 29, 2005 10:00 PMWTF! Reasoning like that is making this election challenge look like Vaudeville
and the republicans are playing the asses.
Man, I can't stop laughing.
Posted by: Doc on May 29, 2005 11:12 PMtc...
That pretty much sums up your entire post....
(heh..And your take on this election contest..)
'I thought I heard on Friday that all absentee ballots are counted in one place. Is this true? If so, then the analysis is moot, since it means that workers would have to somehow stuff pro-CG precincts and take away from pro-Rossi precincts at one spot where all the ballots come in at a random order. How pray-tell did workers do this? Are there reports at the central counting site of this happening? I guess I am confused by the Republicans argument here. To me the process used doesn't fit the theory they claim. Where am I missing something?"
Democrats and Republicans know well in advance of the election what areas of the county - more precisely what precincts - are their strongholds. To memorize the precinct numbers/names and manipulate the ballots (lose Rossi ballots, add to Gregoire) at the ballot processing facility or during the counts wouldn't be that hard. The evidence is circumstantial but compelling. You are never going to have a picture or a video or a confession of someone doing it, but the counts sure indicate it happened.
What other possible explanation is there for the overwhelming majority of the precincts with excess votes being ones that went for Gregoire and simultaneously, those with less ballots being Rossi precincts?
Posted by: Scott in Carnation on May 30, 2005 09:46 AMHow about counting *all* precincts in *all* counties for this? Nah, that would show Gregoire won the election fair and square so couldn't do that, eh?
Posted by: Jim Peterson on May 30, 2005 09:55 AMI have no qualms about a COMPLETE recunt of all LEGAL ballots in this state. Go ahead. Eliminate all provisional ballots run through without validation. Reconcile all mail-in ballots. Identify all felons and have their votes removed. Match a vote to a voter, verify the voter did not vote more than once, is alive, is legal to vote, and then we can see who won.
Oh, wait. We can't do that NOW. Some people managed to put these anonymous votes through the system already, and we cannot know which vote belonged to a legal voter and which did not. Hmmm. Well maybe we should just start from scratch. Run the whole count again by having everyone vote all over again. Those who voted for Gregoire should be prepared, in the spirit of the revote to cast their ballot for her again. The same for Rossi voters. If you didn't vote last time, don't vote this time. THEN, we can use the laws as they were written to VERIFY the legal votes, and THEN we can see who actually won the election.
Do you lefties have the guts to do it?
I suspect not. Your motto is "we have the Governor's mansion, so it doesn't matter how we got it as long as no one can prove we didn't steal it." If you DID won it fair and square, you should be willing to back that up rather than hide behind the "can't prove nothin, nyah, nyah, nyah." antics of an 8 YO juvenile delinquent.
Posted by: Eyago on May 30, 2005 10:54 AMThere are 2616 precincts and 605 polling places. On average, this works out to be 343 ballots per precinct and 1485 ballots per polling site.
Since all counts roll up from the precinct level, how is it that 343 ballots are too much to account for? There were a few precincts (10 or so) that had 1000 or more ballots, meaning that there were many more with less than the 343 average. Counting large numbers in smaller more controlled batches should make the end roll up to the larger number even more accurate. King County contains less than half the total ballots for the state, yet they had more errors than the rest of the state combined. What scares me about this is that they are proud of that record. Some visitors to SP see nothing wrong with that record – they are misguided.
If Judge Bridges finds that “legally” his hands are tied and rules against the Republicans, those that think King County was a model will be emboldened to muck with the election process even further – maybe in the next tax referendum, or school levy, or bond issue. Why not? The court didn’t say their actions were illegal, it just said if you don’t get caught anything goes.
Regardless of the outcome, we need to make sure that the laws and procedures are strengthened so that we really do achieve the goal of one vote, one voter. A voter being a) Legal U.S. Citizen AND Washington State resident b) living, and c) non-felon or felon with rights legally restored. The legislature missed the boat on this last time and we must keep after them. The public must be able to see that their governments are giving them this level of service and excellence. If the vote is really that sacred and important, it is the LEAST the government can do – as a matter of fact, it is the OBLIGATION of government at all levels to give the people this level of excellence.
Posted by: Jack on May 30, 2005 11:54 AMWhat the (*&^ do you this suit is trying to do? REVOTE! Exactly what you are proposing! Welcome to the right side of the issue.
Posted by: Dogbert on May 30, 2005 11:59 AMYou are never going to have a picture or a video or a confession....
Interesting notion. Why aren't the elections operations captured on video?
Posted by: iconoclast on May 30, 2005 01:42 PMYour telling me on election day/night that workers had to forsight to go through absentee ballots and pull out those from certain precincts and add ones to other. You sound like you believe any conspiracy theory that comes along. Why not blame this election on extraterrestrial?
To Scott,
I would agree with you, somewhat if it was after the fact, but the ballots were counted when it did not appear the election would be so close. You are giving the election workers too much credit and foreknowledge to be believable.
To all,
The question of the absentee ballots can be easily answered. My hope is the trial does not end this week with a decision. Instead, one of the course of actions out of the trial would be a manual count of all the returned ballot envelopes by precinct and compare to the precinct numbers. A manual count of the envelopes is the only way to tell how many ballots were returned. Further the count should be done by election representatives from other counties (not King County). Let's have a thorough audit of the evidence prior to jumping to conclusions.
tc...
You are attempting to confuse the issues here...
It was the *Provisional* ballots and *Poll* ballots that represented the voterless ballots in Gregoire dominated precincts and ballotless voters in Rossi dominated precincts. Provisional and poll ballots are identical and were the ones taken home by Democratic Poll Inspectors the weekend prior to the election - with no accounting during their removal, return or destruction. The Absentee ballots (though some were fed through accuvote machines )- are an entirely separate matter.
As far as the absentee ballots were concerned - Nicole Way testified that King County has numerous facilities involved with the printing, sorting, mailing, receiving, batching, verification and counting of absentee ballots. It was noted that there are at least 3 facilities that King County uses for the processing of absentee ballots!
Are we now clear on this?
In researching Stefan's information on the additional absentee ballots in Gregoire leaning precincts and missing ballots in Rossi leaning precincts - AND the questions *tc* raised as to how these ballots *found* themselves in primarily Gregoire precincts and *lost* themselves in Rossi precincts...I had to go back to the various steps involved in the printing, sorting, mailing, receiving, validating and counting absentee ballots.
BUT - then there were other factors at play that make it easier to add absentee ballots to a specific precinct and remove ballots from another. King County has the ability to print absentee ballots on demand out of their elections office and administration office. This is where the extra ballots can be produced with no accounting. Nicole Way testified that there are many voter registration cards (hard copy) that lack any signatures! This apparently does not stop a vote from being processed, validated and counted in King County. One has to wonder how those signature-less voter registration cards got into the county records, how many there are and if they were used to offset false additional absentee ballots? One way to check would be to pull all of the voter registration cards for one of the suspect precincts and see how many are lacking signatures and information!
As far as removing ballots from Rossi leaning precincts - that was easy! As we remember the numerous ballots *found* after the election that were in bundles - most likely from specific precincts. It would be easy to remove ballots - by bundle - from a designated precinct once they are sorted.
Posted by: Deborah on May 30, 2005 08:17 PMNow the second item. Here is how I see Judge Bridges decision. He can annul the certification and put the certification minefield back in the legislatures hands. He can annul KC election certification, this therefore annuls the statewide certification, back in the legislatures hands. He can annul the final recount which without a means of being accurate effectively bypasses the legislature directly into a revote. He can annul the election in total, revote. He can declare Rossi the winner. He can find fault yet abstain. He can deny the request. The chances as I see them.
30% Annul state certification
30% Annul KC certification
2% Annul last recount
16% Annul election
2% Declare Rossi winner
10% Abstain
10% Deny
When did this supposed printing of extra absentee ballots occur (before or after the election)?
I still don't see why after this week's trial the judge simply delays his order and ask that all KC absentee ballot envelopes be counted (by independent election officials--state or other counties). Until the envelopes are counted, we can not know what the true return of absentee ballots were. Until we know the true return, all this theory of extra absentee ballots is simply a theory that hasn't been proven. At least in this case there is a simple way to prove or disprove the theory. Why aren't the Republicans asking for a count of envelopes?
Posted by: tc on May 31, 2005 06:01 AM