47 votes were "corrected" and awarded to Christine Gregoire
12 votes were "corrected" and taken from Dino Rossi.
Net result Gregoire: +59!
This is analogous to flipping a coin 59 times and getting tails every single time. In this case, the coin was biased towards tails to begin with. Nevertheless, the probability of this happening, I calculate, is 3.1x10-14.
This is not attributable to a fair process for correcting random errors. Either the recount corrected an earlier counting procedure that was biased against Gregoire or the recount introduced a new bias in favor for Gregoire. The canvassing board owes us an explanation as to which it is.
UPDATE: I should have read Jim's earlier post more carefully. Obviously, this result is attributable to divine intervention!
Posted by Stefan Sharkansky at December 22, 2004 04:56 PM | Email ThisRemember also -- there was no explanation as to where the extra 336 ballots came from in the machine recount either -- an increase from 898,238 ballots to 898,574 ballots.
Also, Gregoire gained 47 votes and Rossi lost 12 votes. If a manual recount is fair, how can you possibly lose votes on a manual recount? If the machine can read the oval for Rossi, why shouldn't the canvassing board be able to do so as well?
There was also an increase in overvotes -- from 46 in the machine recount to 86 in the manual recount. It would seem most probable that the Democrat canvassing board counted just about any Rossi ballot with a stray mark as an overvote, while the same thing was not done with Gregoire ballots. That is the most likely and logical explanation.
The one thing that made sense was that undervotes fell by 45 votes -- from 21,307 to 21,262. Presumably, the manual recount was able to find votes from these, which were missed by the machines and missed by the enhancement process in the machine recount. But I would doubt that this process was fair to Rossi, especially on those ballots that were referred to the canvassing board.
I would like to see how the precinct-by-precinct ballots counted totals from the manual recount compare with those from the machine recount and the original count. It would also be nice to see exactly how many names of people are recorded as casting valid votes, compared with the number of ballots actually being reported as counted in each precinct.
Posted by: Richard Pope on December 22, 2004 04:59 PMThe Dems were counting on "Anyone But Bush" to win the presidency. It didn't work. We can't count on "Gregoire stole the Governorship" to affect a change in leadership in this state.
Next time, let's widen the margin so Ukraine County (of which I am a resident) can't get away with theft.
As far as other counties that may have votes that weren't counted through no fault of the voter. TOUGH - those counties have already CERTIFIED their hand recounts.
It's my opinion that the certification period is still open until Sam Reed certifies for the 3rd recount. What are your thoughts?
Posted by: CP on December 22, 2004 05:09 PMIn fact, in no county where there is more than 30,000 votes for either candidate is there a negative result - except for King County where they apparently took 12 votes from Rossi.
FOUL!
Posted by: MC on December 22, 2004 05:12 PMI would recommend that Rossi concede quickly and give a hell of a good speech--is Peggy Noonan available?
The Democrats have done enormous damage to themselves here, and in the long term this is a chance to pick up at the very least a US Senate seat.
Posted by: David on December 22, 2004 05:17 PMHowever, the WA Supreme Court has now said that "certification date" means the (future) date to certify the manual recount by Sam Reed, and not the earlier date of November 17, 2004 when the counties originally certified results at the county level.
This would require that all absentee votes received after the November 17, 2004 prior "certification date" deadline must be counted. Also, absentee voters would be entitled to the same extra time to correct their signatures.
Posted by: Richard Pope on December 22, 2004 05:17 PMStop sniffin' the glue! A hand recount MIGHT be the most accurate way, if it's done properly. Were you present at this hand recount? Have you talked to anyone that was? It was a farce!
Divining voter intent? Is that accurate? A 2-to-1 Democratic canvass board that changed 59 ballots, ALL to favor one candidate? Is that accurate? Gregoire getting what was an overvote for 'Christine Rossi'? Is that fair?
Julie & Ferrous:
You can throw in the towel if you want, but most of us are going to keep up the fight. Talk about touch-feely 'We'll do better next time' all you want, but if Gregoire gets in - what will change? What will be different next time?
At the very least you would expect 59 changed ballots to fall towards each candidate to the overall percentage of their votes. In King County Gregoire won 58% to 40% - that would indicate 35 ballots changed in Gregoire's favor, and 24 in Rossi's favor, which would be a net +11 for Gregoire. 59-0? Incredible. Unbelievable.
Posted by: Larry on December 22, 2004 05:22 PMIn the first place, your analogy to a random coin-flip process (if I was going to accept that overall), should be to getting heads 47 out of 59 times. You talk about the incredibly worse odds of 59 out of 59 times. You are off by around a factor of 4096, but there is much more to the error than I can estimate off the top of my head. You redo the math.
What you are trying to say is: of 59 corrections, Sharkansky expects them to break 50/50 -- like a coin-flip process -- so about 30/29 instead of 47/12.
Instead of breaking 50/50, the corrections broke 80/20. This is about 9 votes off the value you might expect for the 65/35 split from King County.
In other words, you should expect a ratio of other than 50:50 from an area (the most populous in the state) that does no vote 50:50 for the two candidates. You should expect a ratio close to the voting.
But even this is all a crazy, tilting at windmills kind of math goofing off. We could get a 'real' statistician to help us, and apply tests appropriate to the populations, and I am certain, we would find nothing of interest. We are talking about a delta of 59 votes on top of hundreds of thousands of ballots. 59 is a tiny, tiny sample.
Anyway, I cannot do the math for the probability of 47 heads out of 59 flips, as I am too lazy to go so far in extending your goof-ball analogy. The basic fact is, there are a lot of non-random things to consider: maybe Dems, being young, feckless, easy-to-disrespect foibleful people, had more troublesome ballots to reconsider.
its not like the RAts haven't stolen an election before...Cantwell was the recipient of all those "provisional " ballots before...I wonder if any of them actually lived here, ever...
but Gorton didn't do squat about it...
nothing will change unless we force change..
they can control any election anyway they choose if they are counting illegal ballots, enhancing ballots to fill their needs, or finding enough homeless people to fill in their names before they hop the first train south...
and for heaven's sake, they have taken the vote away from soldiers but they give it to homeless.....most who are drug/alcohol abusers....
I want Rossi to take this as far as he can...if nothing else, it shines the light of day on the corrupt democrat party and the ilk that do their cheating for them...
like some of the posters here who support the fraud....they will never go away ....
Posted by: lee on December 22, 2004 05:38 PMIt's the hubris that bothers me so much, that if I, a republican raise questions, not accustations, but honest questions, I am branded a cry-baby, but a democrat is considered 'justified because the election was stolen' (and there was absolutely no evidence of that whatsover, but it was being bandied about).
Fine, the heir apparent couped her crown, so be it. I just hope the voters of WA remember this for the next 20 years.
Posted by: Emily on December 22, 2004 05:39 PMThen we need to completely overhaul the elections process before the next major election - in particular, COMPLETELY eliminate paper ballots.
Posted by: Brian on December 22, 2004 05:45 PM(0.5 ^ 59) * C(59,47)
where C(n,k) = (n * (n-1) * (n-2) * ... * (n-k+1)) / (k * (k-1) * (k-2) * ... * 2 * 1)
Numerically, this works out to 0.000001942.
However, it's not as simple as a coin flip.
Posted by: Snowy Owl on December 22, 2004 05:46 PMThe bozos of Washington deserve this mess for voting for Democrats and liberal Republicans. Too many years of listening to Gary Locke dulls the brain.
I am now boycotting Red Hook, Starbucks, and
Windows.
It's unbelievable what has transpired. It's as if I went to sleep, and woke up in Russia! I can't conceive the level of corruption that is evident in this state!
Governor Rossi, don't give up on this now. We're behind you all the way to the end! Remember the Alamo!
Posted by: Scott on December 22, 2004 05:47 PMThat adds up to +59! Funny how THAT works out.
Questions:
Where did the undervotes go? To which candidates did these ballots that were undervotes get credited, and how many?
Where did the overvotes come from? Which candidates lost votes, and how many, by classifying them as overvotes in the third count?
How did write-ins go up by 39??
Does anyone suspect that the undervotes went to Gregoire, and the overvotes got taken from Rossi? When can we see full transparency of the decisions of the canvassing board in King County?
Inquiring minds want to know!
Posted by: Larry on December 22, 2004 05:50 PMNo, Scott, you don't want to remember the Alamo. Nobody got out of there alive. You want to remember Bush vs. Gore.
All right, let's sue their asses off!
Posted by: FedUpWithThis on December 22, 2004 05:51 PMIt was 47 broke for Gregoire... and 12 broke FROM Rossi.
A net break of +59 Gregoire.
Posted by: Al on December 22, 2004 05:52 PMI asked you before to cite your credentials as a statistician, or to ask a credentialed statistician to confirm your nonsensical data points, but you have ignored that and continue to publish your fallacious arithmetical nonsense.
Just a couple of days ago, you published a table which stated, without a doubt, that Rossi had a 94% certainty to end up ahead after the manual recount (we're not counting the 700+ plus ballots here). But guess what? Without those disputed ballots, Gregoire overcame Rossi's lead. That didn't look like any 15-1 shot to me, which is what a 94-6% probability would be. It looked to be a 50-50 chance that any recount that started with a statistically insignificant 42 vote margin out of 2.9 million votes could easily tip the other way when people are hand counting those bits of paper. Nobody with any logical brain would dispute that; particularly a credentialed statistician.
I strongly suggest that you continue with political commentary all you want, but retire your Excel program and any statistical analysis. It just makes you look foolish and ruins any credibility you might have as a political commentator.
Posted by: Nelson on December 22, 2004 05:55 PMThe Democrats probably stole the Monorail vote, too, and how many millions did that Disneyland ride cost?
Bottom line: never vote for a Democrat. Ever!
Posted by: FedUpWithThis on December 22, 2004 05:56 PMMan oh man, what farce this election has been from the standpoint that KC election office is really bad and should be held accountable.
Posted by: Emily on December 22, 2004 05:57 PMYou imply ONLY 59 votes were changed during the manual recount and that ALL of them went to Gregoires advantage. Do you know this to be true? If so, this KC hand recount cries out fraud.
However, my guess would be that several hundred votes were changed in some way with the NET RESULTS being +47 for Gregoire and -14 for Rossi. While this senerio still smells rather fishy, it's by no means as extreme as you imply.
Am I missing something obvious?
Regards,
EricR
Posted by: EricR on December 22, 2004 05:58 PMSince you're SOOOOO smart, please take a shot at answering my questions above.
Posted by: Larry on December 22, 2004 05:58 PMWe don't know if the Rossi -12 was really 14 ballots found to newly be Rossy and 26 newly found not to be. (As a made-up hypothetical set of numbers.)
So we can't know which population to think of this process occurring on. It is not the 59-vote delta, but the population of all 'reaccounted ballots'.
I've been boycotting Windows for over 20 years.....it's called a MACINTOSH! ;)
Posted by: Scott on December 22, 2004 05:59 PMYes, I would use the binomial distribution to calculate the probability that Rossi gets at most 12 and Gregoire 47 out of 59 where p(Rossi) = 0.4
in excel this is =binomdist(12,59,0.4,1)
But this problem is a little different as 12 votes were subtracted from Rossi.
So I modelled it as =binomdist(0,59,0.4,1)
If you have a better model, I'm interested.
Posted by: Stefan Sharkansky on December 22, 2004 06:00 PMSteve suggests that the number in question is 900,000; although that's obviously incorrect, it has a hint of truth.
If the pool of ballots to be "corrected" has more than 59 entries, then the model changes a bit (i.e., some ballots were examined and not corrected).
However, I don't know what the size of the correction pool is, so I can't suggest anything better at the present time.
Posted by: Snowy Owl on December 22, 2004 06:10 PM1) No one may alter a ballot in any way. (This business with poll workers filling in the rest of the circle requires them to have pens, integrity, and decent oversight. We're lucky if we have two of three.)
2) Any ballot that fails to be read by the machine or in unclear in any way goes to the canvassing board. (No, you can't haul out a sharpie to 'fix' it.)
3) _ALL_ ballots sent to the canvassing board are scanned and presented to the public at large.
Frankly, I want to see this 'Christine Rossi' ballot and all of the other insane ballots. There's no reason to hide it from the voters - we can't figure out who did that, so their privacy is still ensured (though, frankly, anyone whose ballot is so inept it goes to the canvassing board deserves a little ridicule.)
Posted by: Al on December 22, 2004 06:10 PMAt some point, with full and accurate data, you can determine what the odds of the canvassing board decisions were. It would be better to have images of all 1600+ ballots that were decided by the canvassing board, coupled with how these ballots were decided. Did a small mark in the vicinty of Gregoire result in Rossi votes being nullified as overvotes (which increased from 46 to 86 overall), but not the other way around?
Another possible explanation -- let's suppose that 59 "new" ballots (from 59 different precincts, to make things less suspicious) were fraudulently added to the mix, and all marked clearly for Gregoire. In the absence of these 59 new ballots, both Gregoire and Rossi would have lost 12 votes -- exactly what random chance would predict.
It would have been tough to add new ballots to 59 separate precinct collections. However, it would have been a lot easier to simply slip in these 59 new ballots in the pile of ballots that were destined for the county canvassing board to consider.
Posted by: Richard Pope on December 22, 2004 06:17 PMAlso, x-tine 'won' by taking a page straight out of New Jersey--- if you're losing, just go to a judge and get the goal posts moved. She did that twice. Those provionals look even smellier now. And 2/3 of the state said that even if x-tine declares victory that they still think Dino won.
If Dino concedes now, that's just a victory for New Jersey style corrupt politics. We should not stand for that! You go, Dino! Fight on, because I think most of the state is behind you.
Posted by: Michele on December 22, 2004 06:30 PMExactly! This isn't a question any longer of merely trying to get Dino Rossi his rightful office; it's a matter of ensuring that fraud is not rewarded.
Posted by: B Knotts on December 22, 2004 06:41 PMMy hypothesis is that the increase in valid write-ins would come at the expense of "undervotes". The full intent of the voter might not have been ascertained in a machine count if the little circle next to the write in name was not blackened, but the intent could be clearly ascertained by the written name.
So +39 write-ins and minus 39 undervotes. But the total number of "undervotes" changed was only 45. So only 6 undervotes are left to convert to add to Gregoire as a result of a human seeing a mark where a non-machine readable one "existed".
In an effort to find a pdf of the sample ballot online, I found the press release from the King County election division advising that over 3,000 duplicate ballots were sent out in October. Was there any accounting for the duplicate ballots? Has anyone checked to ensure that all the persons that voted via absentee ballot had requested an absentee ballot? What follow up occured to "ensure" that duplicate ballots were not voted.
"Late last week, election officials learned some voters received duplicate ballots. Due to the ongoing surge in voter registration activity and the legal deadline to mail absentee ballots, some 3,500 King County voters may have received duplicate ballots at the same address. Those voters are advised to vote and return one ballot and discard the duplicates." - King County press release 10/17/04
Another question - as part of the "hand count", were machines used to check the stacks - e.g. the counters agree that they have a stack of 100 ballots for Gregoire. They are run through the machine and the machine reads 99 for Gregoire and 1 for Rossi.
Maybe it gets runs through again to see if it is a duplicatable error - if so, the humans do another count.
Given the stories from recount observers of "honest" mistakes of Rossi votes ending up in Gregoire piles (see free republic forums)- this can account for the otherwise inexplicable reduction in both the R and L votes ending up in the D column. (along with near 100% of the newly appearing votes)
It looks like we could learn a few things about "improving the odds of winning" elections here in NJ from the King County Dem Organization.
I'm shocked. I've been hearing for the last three years that's it's not only our right, it's our RESPONSIBILITY to question our government.
I guess that only applies when you're questioning a Republican government.
Posted by: South County on December 22, 2004 06:55 PMI don't know that he should run against Cantwell. I would support him running for Governor in four years.
Posted by: South County on December 22, 2004 06:59 PMBut I do think the questions I asked in the post just below might clarify what happened. Did the Gregoire gains come as the result of the canvassing board votes? I heard several times that Rossi had actually gained in the early hand re-recount in King County. That makes me suspicious that ballots that went to the canvassing board -- which I believe were reported late -- determined the outcome.
And I think my follow up is also worth pursuing. How many 2-1 votes were there in the canvassing board decisions, and did they consistently follow party lines?
I do think Rossi's loss of 12 votes is the biggest single puzzle in the results. I would have predicted -- assuming the count was honest -- gains for both candidates, with the gains roughly proportional to their original vote shares. (Maybe a little more for Gregoire, for reasons I have explained before.)
There may be an explanation for this loss, but one doesn't occur to me, other, of course, than the tendency to err in favor of one's own party, which I discussed earlier.
Posted by: Jim Miller on December 22, 2004 07:02 PM"A court challenge is expected in a tightly contested Congressional primary which includes echoes of south Texas' rich history of political corruption and intrigue.
Congressman Ciro D. Rodriguez (D-Tx) says he will go to court to challenge a recount of the votes of last month's 28th Congressional district Democratic primary, which appears to have given the victory to challenger Henry Cuellar, an attorney from Laredo.
"There are some irregularities that occurred in the voting in Zapata and Webb Counties," Rodriguez told 1200 WOAI news.. "There was a lot of funny business down there, and we will definately go to court."
Posted by: Efrem from Sacramento on December 22, 2004 07:30 PMBut then again, given that you folks in WA will soon have the Cajun primary, I'll bet the Black Hills of SoDak (a VERY conservative area) sounds very good (and cheap) right now for you WA State GOPers:)
Posted by: Brad S on December 22, 2004 07:46 PMThe ONLY reason for hand recounting ballots is to interject human subjectivity in an effort to influence the outcome of the election. That's the ONLY reason, and why hand recounts are NOT more accurate than objective machines.
Posted by: Dave Kaplan on December 22, 2004 08:26 PMYour problem is that the sample size is not 59. THEREFORE, this is not like tossing a coin 59 times. There were almost 900,000 votes cast in King County and this recount resulted in a 59 vote shift from the previous count. That is 59 votes in the span of almost 900,000.
Now, there very well may be some sneaky things going on as the first post mentioned. In fact, I would like to know where the -12 on Rossi came from.
But, keep the blog righteous by using math and arithmetic appropriately. Do not try to skew the picture by using sloppy mathematical analysis. I would like the Republicans to come across as honest and fair even if it means losing.
Posted by: Yahudi Goloshes on December 22, 2004 08:54 PM. . .the final action of the KC canvassing board on a handful of ballots that were reviewed and counted today. Their final action was to review and count a handful of ballots that put Gregoire over the top. Awful suspicious. And smells real bad too.
Posted by: bob the builder on December 22, 2004 09:28 PMI voted for you, that was obviously a mistake.
The way in which you condescendingly and offensively questioned Harry Korrel in court today was unbecoming of a Supreme Court Justice. You could learn a lot about objectivity and conduct by watching some of the other justices on the court.
You should feel like the coward that you are.
Posted by: Jeff B. on December 22, 2004 10:43 PMYou say that the 59 ballots were 59 ballots actually changed by the canvassing board.
I don't think that is true. The reporting that I have read calls this 59 merely the error (or, difference) in the count from the previous count of total ballots. That is, the first count registered 898,574 votes and the second count registered 898,633 votes.
In this case, the sample size is 898,633 and the error from the previous count is 59. To pull that value of 59 out and treat it as some separate set of ballots that are treated differently is not correct.
If you review the recount of all the counties, and compare the errors noted (that is, difference in recount versus the original count) then you will find that the errors are not unusual. The mere size of King county is the cause of the larger number.