I'm writing this post to both correct a factual error in this afternoon's Snowjob! post and to provide more detailed evidence that King County is pulling a snowjob.
Rewind: Last Friday when King County reported an unexplained discrepancy of 1,217 more ballots counted than identified voters, they also made this statement:
This number is consistent with historic reconciliation rates for King County. In 2000 this number was 1,230.(King County later revised the 1,217 number to 1,800).
I now have more county documents and more detailed explanations from my sources. My earlier misstatement was that the 1,230 represented an excess of voters over ballots. That was an error. It represented an excess of ballots over voters after all. HOWEVER it is still a bogus comparison and a snowjob. Here's why.
I have four sets of documents that represent certified countywide ballot counts for the November elections of 2000, 2001, 2002 and 2003. In addition, other documents in the set show voter participation counts for each election produced at some point after the election was certified. The following table summarizes what was in the documents.
[The Report Date is the date that the voter participation report was generated.]
| Year | Report Date | Ballots | Voters | Discrepancy | Error Rate |
| 2000 | 1/17/2001 | 798,942 | 797,712 | 1,230 ballots | 0.15% |
| 2001 | 12/7/2001 | 459,930 | 462,997 | 3,067 voters | -0.67% |
| 2002 | 12/10/2002 | 548,353 | 551,162 | 2,809 voters | -0.51% |
| 2003 | 11/21/2003 | 369,779 | 369,173 | 606 ballots | 0.16% |
| 2004 | -- | 899,199 | 897,049 | 2,150 ballots | 0.24% |
798,942 ballots counted and 797,712 voters counted for a net difference of 1,230 more ballots then voters. So is this year's discrepancy of 1,217 (oops! make that 1,800, or to tell the truth 2,150 more ballots than voters) really consistent with historical reconciliation rates? No.
The report that came up with the 1,230 number was produced on January 17, 2001, two and a half months after the election and reflecting both (a) the normal attrition of voters who leave the rolls and (b) the courtesy credits which I mentioned in the earlier post. In other words, it's a largely meaningless number and is in no way a valid comparison for this year's discrepancy of 2,150, which was computed from an election day voter sample. Why would the county produce a report that is a meaningless comparison? Perhaps they cherrypicked a report where the discrepancy appeared to be similar to the discrepancy they thought they had this year. The only other year which shows an excess of ballots over voters is last year, 2003. Maybe this is a coinky-dink, but November 2003 election was the first general election after Ron Sims hired Dean Logan.
These reports that I looked at here are the only historical reconciliation reports that I am aware of that have been released by the county. Unfortunately, they're little more than a smokescreen (snowscreen?). If King County is serious about showing how this year's 2,150 ballot discrepancy compares with previous years, they'll produce the historical pre-certification reconciliation reports which they have in their files.
Posted by Stefan Sharkansky at January 13, 2005
11:30 PM | Email This
What was the error rate in other counties? I thought it was higher in a number of them. Is there any data on that or have they released ther final numbers?
Posted by: Erik on January 14, 2005 12:03 AMYou'd think that the "numbers" would be one of the most important parts of the certification, and therefore, would be double (if not triple or quadruple) checked before being signed under oath. Guess those things are not priority for King County elections officials. Are there ANY election numbers in King County that are not changed?
Posted by: TADD on January 14, 2005 12:10 AMYou always seem to show up, make a contrary comment, and then you're never heard from again on the post.
I'm still waiting for my answer from you regarding the hand recount:
Do you use spreadsheets and/or calculators to aggregate and sum large amounts of data, or do you do it by hand?
Easy question, should be a quick answer. I'm waiting.
ps. Obviously you do not know how to look at data. 'Lower error rate'? Only if you're looking at the absolute values, which disregards whether there are more voters or ballots.
Any person who can analyze data would quickly see that 2004 had the highest rate of ballots over voters. The other years had greater error rates - of voters over ballots.
Get a clue.
Posted by: Larry on January 14, 2005 12:34 AMDo you use spreadsheets and/or calculators to aggregate and sum large amounts of data, or do you do it by hand?
Easy question, should be a quick answer. I'm waiting.
Sorry Larry. I didn't see your posts for awhile as yours got lost in the pile of other posts on the matter and I got a little busy.
I am not an election expert.
However, my understanding is that both the hand re-counts and the machine counts use computers and spreadsheats to add the figures. Each machine or counter takes their tally and adds their component to the totals.
Yes, from what I have read, the hand recounts are more accurate:
"The central finding of this investigation is that manually counted paper ballots have the lowest average incidence of spoiled, uncounted, and unmarked ballots, followed closely by lever machines and optically scanned ballots," the study said. On average, punch cards and electronic voting machines had significantly higher average rates of problems.
The full study is located at:
Other than MIT, Caltech is considered the leading engineering school in the country.
Thus, I think the statute has some logic to it and the legislature acted reasonably when it passed the recount statute.
However, it is still only a theoretical question as the state statute dictates that it is the final hand recount which is binding for the election.
Posted by: Erik on January 14, 2005 01:01 AMjeeze...look at how many accounting firms have been in trouble from this sort of thing.
Horseshoes, hand gernades, atom bombs, and king county. The list of places where "close enough is good enough" is relatively short.
Posted by: scott158 on January 14, 2005 01:01 AMLarry, read the caltech study and see if you agree with me. If you don't tell me why.
If you believe the Caltech study takes the minority view on the subject in the rest of the published journals on the subject, that have scholarly peer review, let me know.
Otherwise, from what I can see, hand recounts are in fact more accurate as found by Caltech.
The legislature certainly came to the conclusion, rightly or wrongly that the hand recounts were more accurate or the statute would have been written differently.
Posted by: Erik on January 14, 2005 01:11 AMThe issue of hand count vs. machine count and the issue of number of voters vs. number of ballots counted are like apples and oranges. They don't have anything to do with each other.
If we have 2,150 more ballots counted in King County, than actual voters who cast them, then it won't make any difference whether these ballots are counted by hand or by machine. There will still be 2,150 excess ballots.
Posted by: Richard Pope on January 14, 2005 01:29 AMhttp://www.effwa.org/
Posted by: TomasM on January 14, 2005 01:34 AMWord. I agree.
Ok. Sorry. This is a blog so I guess I can't agree right? Not allowed? Ok here goes my effort to disagree when I don't really:
However, hand counts v machine counts is not really the same as apple v orange.
Hand counts v machine counts refers to a methodology and corresponding error ratios.
Apple and oranges are actual objects though different ones. Thus comparing methodogies to different objects is not synonymous.
I think a better statement would have been hand counts v machine counts is the same as sewing by machine and sewing by hand.
Posted by: Erik on January 14, 2005 01:39 AMEarlier I watched a replay of King 5 TV's 1998 investigation of voter problems in King Co. (Thanks to Mac at pull on superman's cape--via Michelle Malkin.) They hired team called "Fair Elections Group" who found 5,200 double registered voters. Seems like you've got a ways to go in finding that many double registered voters in 2004. Has King County, despite all the evidence to the contrary, "cleaned up its act" that much? Or are they just using less obvious ways to manipulate votes?
It was also noted that King Co. had no reports of steps taken to correct things that had subsequently been reported by King 5. This followed promises to correct things and a strong statement by the King County Council expressing their determination to see that changes were made. It seems that empty promises to deflect attention from discovered problems is their modus operandi.
Posted by: RLG on January 14, 2005 01:57 AMLet’s say that each vote was a $100.00 bill. (How much is your vote worth?). This would mean that King County has an acceptable error rate of from $123,000 to $180,000 per election; give me a break. Every retail company in the world closes each day accounting for each dollar. You take Costco. One store/warehouse has more transactions and dollar activity in a Saturday than King County had in this election. If a store was short or over this amount of money on a day/week, the store manager would be fired, the police would be called in and there would be a serious investigation to find out what happened. There is no excuse for this type of sloppiness other than to allow manipulation of the numbers intentionally. There are well developed procedures and methodologies for tracking and accounting for every transaction both in business and elsewhere. There is absolutely no excuse for these procedures to not be implemented for something as serious as one persons vote. I can not think of anything more sacred and fundamental to the concept of this government.
Earl
Just to add emphasis to your post, I know for a fact that in the heavy shopping days leading up to Christmas the Costco store in Issaquah regularly did over $1 million in business each day.
Do you think they let several thousand dollars get lost each day?
Posted by: Scott in Carnation on January 14, 2005 05:31 AMwith all due respect, "error rate" is determined by an absolute value. Something can be off by 10% by being 10% over OR 10% under the correct amount.
re: the greater accuracy of machines vs humans.... that of course depends upon how do define accuracy. Machines are more accurate at counting the ballots that are executed consistent with the technical instructions for casting votes. (i.e. if you have 10,000 ballots that can be counted by machine, the odds are greater that there will be an error in the hand count than the machine count.)
But elections are not a test of how well one follows instructions, but a determination of the collective will of the voters. And every study of ballots and "voter intent" has determined that hand counts are far superior than machine counts in fulfilling the purpose of elections--determining the will of the voters.
You're incorrect, other than MIT, the University of Illinois at Urbana-Champaign is considered the best engineering school in the country. It has been ranked #2, and #1 in certain engineering departments, for decades. Sorry to nit-pick, but that sounded like typical west-coast bias, forgetting about 'fly-over' country.
That was a lot of fluff and bluster, but you didn't answer my question about which method YOU use. Notice I used the word 'aggregate' in my question, because before the ballots were counted by hand, they were aggregated by hand into precincts - which the machines OBVIOUSLY do better because of the bar-codes. Anyone who has at the hand recount knows how faulty this process was.
p.lukasiak:
With all due respect, looking at the absolute value of the error rate may suffice in certain situations, such as the clustering of temperatures around a mean.
However, in this situation, when we're worried about fraudulent ballots being cast - do you think it's useful to compare errors of more voters than ballots to errors of more ballots than voters?
Posted by: Larry on January 14, 2005 06:41 AM"The central finding of this investigation is that manually counted paper ballots have the lowest average incidence of spoiled, uncounted, and unmarked ballots, followed closely by lever machines and optically scanned ballots," the study said. On average, punch cards and electronic voting machines had significantly higher average rates of problems.
Not more accurate...more subjective. You (as well as the study) are using "accuracy" and "inclusion" interchangeably, and they are most definitely not.
The analogy between Costco and King County Elections is fundamentally flawed. A valid comparison wouldn't measure the transaction (which compares to the voting process itself) but to reconciliation. The question then becomes: if someone in each store counts a deposit three times, will they come up with the same number each time? If they don't, they would not be long in that position.
What King County hasn't been able to get straight is what happens at HQ. The excuses and rationalizations we're hearing wouldn't fly in private commerce.
Posted by: South County on January 14, 2005 06:42 AMAnd I still believe that both of you are wrong in this situation. The way the hand recount was implemented in King County was ludicrous.
MAYBE if they had used the machines to sort the ballots into precincts, I could agree with you. But they didn't, and as the hand recount progressed the counters were continually finding ballots from other precincts in the bins that were supposed to be one precinct.
MAYBE if King County did not change the canvassing rules several times during the process and kept consistent standards, the hand recount would have been better.
MAYBE if I did not see the examples of the votes that were awarded to Christine Rossi, I mean Gregoire, with scribbles in the margin going to the Democratic candidate, I could agree.
But when you have a two-to-one Democratic bias on the canvassing board, the divining of voter intent is bound to fall disproportionately to the Democratic candidate. 'Voter intent' should not be considered. If the oval is not filled in - sorry, Charlie.
Posted by: Larry on January 14, 2005 06:49 AMWhat makes anyone think or believe that your vote should "count", if you're incapable (or unwilling) to follow the instructions for voting?!?
Connect the broken arrow...
Color in the circle...
Whatever...
Sounds pretty simple to me! (but then again, what do I know? I'm just a stupid "red" voter...)
Posted by: Brian C on January 14, 2005 06:50 AM"Determining voter intent", my a$$...
Let's call this what it truly is:
"Determining the outcome of a "fixable" election".
I'm so sick of this. What is the problem with following instructions, laws, procedures, etc.??
Heck, if her highness truly, honestly, legitimately won this election, I'll kiss her feet. But, we'll never really know, will we? All I REALLY care about is that democracy (i.e. abiding by election rules and the RULE OF LAW) remains intact after all this.
Does your discrepancy of 1,800 include courtesy credits? If not, should the discrpancy be between 3,000 and 4,800 when including historical courtesy credit numbers?
Also, why are there usually more ballots than voters in Presidential election years, and more voters than ballots in other elections? Hmmm....
Posted by: LarryG on January 14, 2005 07:35 AMHe asserts an erroneous standard (to devine the people's will) against which he attempts to frame an argument. The purpose of an election is to determine, BY REASONABLE MEANS OF MEASUREMENT, the preferred choice of voters in the aggregate. An election that determines a winner using the reasonable means of measurement but does not reflect the will of the people is nonetheless a valid and legal election whereas one that does reflect the will of the people but fails to ACCURATELY EMPLOY A REASONABLE MEANS OF MEASUREMENT is not a valid election. It is not uncommon for voters, particularly in lower ticket races, to be confused about the person for whom they are voting especially where a candidate has a name familiar to them (Justice Charles Johnson's race was viewed as one such race years ago). Machines use predictable and non-varying standards, people do not. If examined, King Co. ballots will reveal multiple ballots where two ovals containing marks in them were treated differently (though no doubt lukasiak would argue that such standardless interpretations are proper). One treated as a vote another as an overvote. This is due solely to the subjectivity of the "interpreters" of the people's will. This is why studies that insist on objective measurement standards show, contrary to lukasiak's comments, that manual counts are more prone to error than machine counts.
Posted by: barchester on January 14, 2005 08:35 AMAll this will be exposed in a real election contest conducted with testimony under oath with properly deposed witnesses. If not, our hopes of restoring validity to elections in this state are considerably diminished. Meanwhile, my thanks are to Stefan and others who are raising the questions that keep the pressure on.
Posted by: RLG on January 14, 2005 10:10 AMWhat are the odds that King County had a 100% voter turnout? Only then would the error be the 1800 or so extra votes.
In reality, the error is the total number of votes cast minus the actual number of voters who turned up to vote. It would be interesting to see a reasonable calculation of actual voters, based on statewide voter turnout rates, historical county turnout rates, etc. Then one would take the total number of votes cast minus this number to get an idea of the actual error.
Posted by: Thought on January 14, 2005 12:21 PMIf they don't want to carry that burden, shouldn't they be accountable for balancing the ballots to voters?
Posted by: TADD on January 14, 2005 04:04 PMAlso, there's a big difference between having more ballots than voters, and having fewer. The latter situation could arise without malfeasance on the part of election officials if someone signed in so they could get their "I voted" sticker but didn't feel like casting a ballot (I would suggest the proper thing to do would be for one election official to cast a write-in for "BLANK", have both parties' officials initial it, and deposit it as a normal ballot). If such procedures are not in place, it would seem plausible that the number of extra votes is really greater than reported, but some of them 'replace' legitimately-missing ballots.
Posted by: supercat on January 15, 2005 12:30 AM