We have a clarification on the Democrats' claim (see the post below) that "The error for voting machines is somewhere between 1 and 2 percent".
This is from the Assistant Elections Director for Voter Services:
Mr. Crouch,Just to reiterate Gregoire's claims:The statement that the error rate for voting machines is between 1 and 2 percent is not correct.
The Federal Election Commission standard for error rates on voting machines is one mistake in 1,000,000 ballots read by the machine. This is the standard used in Washington State for certification of voting systems.
It is the human error rate that is higher.
It is generally accepted that 1 to 2 percent of the VOTERS fail to mark their ballot so that the machine can read it. Many of these were captured during the original count or the machine recount, as the machines are set to outsort ballots that the machine cannot read. The procedure for that is that if the machine rejects the ballot, the canvassing board will determine the voter intent from the marks on it and then duplicate or enhance the ballot to reflect that determination.
If I can help with any other questions, please let me know, or contact our voting systems specialist, Paul Miller at pmiller@secstate.wa.gov or 360-664-3442. Thanks for checking with us!
Pamela Floyd
Assistant Elections Director for Voter Services
360-586-4173
Let me put this race in a little more perspective. Out of nearly 3 million votes cast, only 42 votes separate my Republican opponent and me. That's a difference of 0.0014 percent. The error for voting machines is somewhere between 1 and 2 percent, or 1,000 times as great as the vote difference.So. We are being asked to reject the accuracy rate of 1 in a million errors, for a manual recount with an expected error rate of 1 to 2 in a hundred errors. Posted by Brian Crouch at December 01, 2004 01:11 PM | Email This
This wrankles the statistician in me.
In a more perfect world the recount would have yielded only a 3 vote delta.
The fact that we had a four figure net change for both candidates means that considerable 'enhancement' has already taken place - i.e. Canvassing Boards have already eyed most anything that could be considered 'intent'.
The problem is that there are 20,000+ to 50,000+ 'unvoted' ballots still out there where no intent can be inferred.
That's precisely the danger that any counting has to Rossi.
See my post:
http://pullonsupermanscape.typepad.com/pull_on_supermans_cape/2004/11/wa_governors_ra_4.html
to see what the King County CB has already awarded to Gregoire. If they can award a pen mark on the edge of an otherwise unmarked ballot, what can't they award?
My general election thread post:
http://pullonsupermanscape.typepad.com/pull_on_supermans_cape/wa_governors_race/index.html
now contains an interesting discussion with the Spokesman-Review - I attempted to get them to run a fraud charge with me ...
Posted by: MC on December 1, 2004 01:32 PMA true fair recount would attempt to count all the outsorts by hand (as the machines have failed in that case). Ballots that have been already counted by machine (those that weren't outsorted) would not be recounted by hand, as the machines' accuracy for ballots that they do recognize is almost perfect (1/1000000 failures). No hand count could reach that level.
But the law is the law. This is all an argument for 2006. We have to face it - all the ballots are going to be handcounted.
Posted by: Frank on December 1, 2004 01:33 PMIt certainly seems plausible that machines are less likely to make the first sort of mistake; one error per 1,000,000 ballots is believable. But "evaluative" error is more troublesome. Do the machines really spit out the 1%-2% of ballots they do not count because they cannot read, and do the canvassers make the correct judgement about whether they should be counted?
I would argue, as I imagine most Republicans would, that the machine's failure to read the ballot, for whatever reason other than malfunction, should be the end of the matter. Needless to say, the Dems take a different view; they want a human being to examine each ballot before discarding it as invalid.
What is wrong with the Democratic view is not so much that human counts lend themselves to more manipulations (though they do to some extent) but rather that they do not really solve the deficiency of which they complain. Gregoire claims that the importance of public confidence in the certainty of elections is what justifies the inconvenience of yet a third count. But a hand count will tell us no more than the first two machine counts will, namely that the margin of victory is less than the anticipated margin of error (in this case "evaluative" error), and therefore cannot identify a winner with certainty.
If we continue the recounts, no doubt there is some point at which we could know the probable winner beyond a reasonable doubt, even though the exact margin of victory may always be in dispute. But suppose the margin of victory is only one vote. How many recounts would it take to reach that level of public confidence? Surely far more than could be justified by any conceivable public interest in the certainty of electoral outcomes. We are, after all, only selecting a Governor, not settling the fate of humanity.
Posted by: Tom Rekdal on December 1, 2004 02:03 PMA change is needed in this state, and that's what we voted for. The fate of the business climate depends upon it. That's what is being decided on here!
Posted by: Susu on December 1, 2004 03:35 PMWe know the hand recount will be less accurate. We know that the Komo and King 5 polls show that a majority would prefer that the election end now. We know that several newspapers and prominent citizens are calling for concession. We know that Rossi has won twice. We know that he has won even after voter intent on many illegitimate ballots was determined by the King County canvassing board. We know that another hand recount will be costly if it goes to a full recount, which it probably will. We know that Gregoire only really won Seattle decisively and would most likely be viewed as illegitimate by much of the states citizens if a hand recount reversed the outcome. We know that voting is a simple process and that any voter who truly wishes his or her vote to count has had more than ample opportunity to vote properly. We know that partisan bias in the counting process has allowed many provisional ballots that would normally be thrown out to be counted for Gregoire, even though the authentication process for provisionals is not fraud proof.
If Gregoire manages to use the innacuracy of the hand recounting process to arrive at a circumstance that gives her a win, the win will be viewed as illegitimate by much of the WA electorate. This will severely taint Gregoire's ability to govern and her viablility as a candidate in any future political races.
Do the right thing Mrs. Gregoire, CONCEDE.
So, despite her long tenure as AG, she's still oblivious to the State-mandated machine-count standard of less than one ballot miscounted per million ballots (that's ONE TEN-THOUSANDTH of one percent, Ms. Gregoire). Or trying to bullshit us.
Her drivel about "thousands of uncounted ballots" across the state supports the BS theory. If such things existed, they'd properly be excluded from a RE-count, although her demands to "count all the ballots" after learning she was 42 votes short represent her hopes of another surprise truckload from the back rooms of King County to be added to the previous totals.
Regardless whether her mendacity exceeds her innumeracy, or vice versa, either of those flaws should be fatal to her qualifications for the Governorship of such a sensitive, progressive and well-educated state such as Washington.
Posted by: Insufficiently Sensitive on December 1, 2004 07:26 PMOh yea, and then the repugs drag us into the courts...
Posted by: sha on December 1, 2004 08:19 PMGo look at the recount fool. There was one, ya know?
The error rate was as high as .003 in some of those Repug counties.
Overall, it averaged about .001
Posted by: sha on December 1, 2004 09:05 PMPlus, the error rate of some stupid federal commision means jack crap. The actual error rate of the MACHINE RECOUNT we just had was 1 in a thousand. That's the fact, jack.
And yes, three people can do better than 1 error in a thousand, much much better.
Posted by: sha on December 1, 2004 09:21 PMIn science, reobserving the identical phenomena can reduce error... but reobserving _almost_ the identical phenomena is called 'Major Ethical Violation requiring Termination'.
Posted by: Al on December 1, 2004 09:30 PMAnd was anyone else struck by the significance of the Seattle Times article today stating that all of these decisions regarding the hand mis-count will be made by "the party"? Why did they state it this way other than to attempt to deflect criticism of Gregoire?
sha also forgets that the "3 people" she is hoping to be so proud of will include hundreds of TEMPS that King County will hire for this task. How does a County-hired temp inspire confidence that errors will not be made?
Posted by: zip on December 1, 2004 10:03 PM1)Machines are not Republicans or Democrats. People are ...
2)It is not a machine "error" when a person is too stupid to fill in a scantron bubble or punch out a chad.
The only reason there were so many "errors" were canvass boards decided what specks, marks, checks, chads, were votes and adding them to the totals.
Posted by: Marc on December 2, 2004 09:36 AMAnd Crouch...read through the letter again from Ms. Floyd - she does not say manual recounts have an error rate of 1 to 2 in 100 - she says 1 to 2 in a 100 ballots cannot be read by the machines, either though human error (or - presumably - through some unintended failure of the voting process, ballot manufacture, or technology).
And Marc...since you probably also support the NRA and may have heard their unofficial motto about guns not killing people...
"Machines are not Republicans or Democrats. People are .."
are not machines made by, programmed by, repaired by, accessed by - Republicans or Democrats?
1. Did the voter "intend" to vote for that candidate and just carelessly miss the mark?
2. Did the voter "intend" not to vote for that candidate, and his writing utensil happened to rest at that spot for a second.
2. Did the voter "intend" not to vote for that candidate, and a stray fleck of ink or lead from his writing utensil happened to land on that spot.
3. Did the voter "intend" not to vote for that candidate, and his ballot became contaminated through contact with another ballot that left a tiny mark on that spot.
4. Did the voter "intend" not to vote for that candidate, but mistakenly put his mark in that oval, before realizing his mistake?
5. Did the voter "intend" not to vote for that candidate, but nicked that oval with his writing utensil en route to marking a different oval.
How does anyone know? Certainly, if I am a partisan shill who wants my candidate to win, I am going to argue that that voter "intended" to vote for my candidate, even though there are at least four alternative possibilities.
That's why "intentions" can't be measured, and shouldn't count. Only correctly filled out ballots.
Posted by: Matt J Kurlander on December 3, 2004 06:20 AM