That's the claim made, again and again, by supporters of Washington state's "governor", Christine Gregoire. (They also say, again and again, that Washington's laws provide for a manual recount — which is not in dispute.) They make this claim in spite of the fact that nearly all authorities on counting ballots — including King County's election head, Dean Logan — believe that manual counts are less accurate than machine counts. And just plain common sense should tell you the same thing; we humans are better than our machines at many things, but tabulating millions of similar objects is not one of them.
To answer that question for certain, we would have to have at least one more recount. But we can get a tentative answer just by looking at the changes between the machine recount and the manual recount, and judging whether they are what we would expect. What I am doing, as some of you may already have realized, is throwing away everything we know about the strange happenings in the manual recount — for the moment — and just looking at the final numbers for the counties.
Before going through the numbers, let me lay out the ideas behind my thinking. Machines are more accurate in vote counts, but they miss some ballots that people can read. People are likely to be biased in their counts, even if they do not intend to be. Democrats are more likely to err in marking their ballots, since Democrats draw more support from illiterate and partially literate voters.
When I put all these factors, except for the bias, together, I get this generalized prediction for the changes between the machine recount and the manual recount. In each county, I would expect small gains for each candidate, almost proportional to their votes in each county. I say almost because I expect that Democrats will make a few more errors than Republicans, errors that machines miss, but people can correct. Some small counties, of course, will show no changes at all. There will be some statistical fluctuation as well, so I don't expect exact fits to the model. Finally, because of the bias (whether conscious or not), I expect Republican counters to slightly favor their party and Democratic counters to slightly favor their party.
At this point, to follow my discussion, you will want to print the election results, or open them in a separate window or tab. Let's take the easy counties first. Twelve counties, Chelan, Columbia, Douglas, Ferry, Garfield, Kittitas, Klickittat, Lincoln, Pacific, Pend Oreille, San Juan, and Stevens, had gains of two votes or fewer for each candidate, so let's eliminate them from further discussion. Skamania almost passes this test; Rossi gained three votes there and Gregoire one, so I'll set it aside, too.
There were nine more counties, Benton, Clallam, Clark, Gray's Harbor, Jefferson, Mason, Thurston, Whatcom, and Yakima, in which the gains for the candidates were both (relatively) small and close to proportional to the votes of each candidate. (By relatively, I mean less than 1 in a 1,000, compared to the candidate's vote.) Five other counties, Lewis, Island, Spokane, Walla Walla, and Whitman, had small gains that were not proportional to the candidates' votes. Since two, Island and Whitman, showed bigger gains for Gregoire than we would expect, and the other three show larger gains for Rossi than we would expect, I believe the results in these five to be just the random variations I expected, and will set them aside, too.
There were three small counties, Asotin, Franklin, and Okanogan, that showed, relative to the total number of votes, large gains for both candidates. The gains were close to proportional, so I see this as evidence of careless vote counting in the first recount, rather than something sinister.
Five counties, Adams, Cowlitz, Grant, Skagit, and Wahkiakum, managed to lose votes between the machine count and the manual count.* That shows that something was wrong with either the machine recount or the manual recount in those counties. (I think in some cases, the counties have explanations for their errors.) Rossi lost 14 votes, net, in these counties, and Gregoire lost 24 votes.
That leaves four counties, King, Kitsap, Pierce, and Snohomish. All produced results that are, on their face, suspicious. Kitsap's gains for both candidates were too large, though they were close to proportional. Snohomish county, perhaps because it uses electronic voting machines, did not make that error, but it did produce an improbably large gain for Gregoire in a county that Rossi carried.** Pierce County produced too large gains for both candidates, and a suspiciously large gain for Gregoire, again in a county that Rossi won. If I were a paid Republican investigator, I would look hard at the manual recount in Pierce.
Finally, we come to King County, saving the best for last. Let me start with what I consider the most remarkable point of all. In December, after the court fight, King County added, and I think I have this number right, 566 ballots from those that had been rejected because they did not have the signatures in the computer file. Between them, Rossi and Gregoire gained just 537 votes between the machine recount and the manual recount. No doubt a few of those 566 ballots had no votes for governor, but the difference seems too large to me, given that about 2.5 percent of the state's voters did not make a choice for governor. And it leaves no room at all for the votes that a human counter can see, but a machine can not. It seems certain that King County lost some votes between the machine count and the manual recount.
Let me remind you of how we expect a manual count to be different from a machine count. We expect small gains for each candidate, roughly proportional to their vote totals. Pierce County produced implausibly large gains in the manual count; King County produced an equally implausible loss, after we allow for the 566 additional ballots. How did that happen? I am not sure though Stefan has certainly given us some cases to look at. King County's result is remarkable; they were trying as hard as possible — so they say — to count every vote in the manual recount, and they ended up with fewer votes than in the machine recount.
(Some Republicans will find the disproportionate gain for Gregoire in the manual recount suspicious. I am not certain it should be, since it appears to have come entirely, net, from the set aside absentee ballots, so the same argument that applied to the additional Snohomish ballots** might apply here, too.)
So, was the manual recount more accurate than the machine recount? Probably not, given the results in Pierce and King counties. That said, I think the important problems are not in the count, the machine recount, or the manual recount, but in the registration and the voting before the counts.
Cross posted at Jim Miller on Politics.
(*All of the counties that lost votes between the machine recount and the manual recount use the E&SS optical scan systems. Those systems may be a little too good at counting every vote. For more on other effects of particular voting systems, see this Timothy Goddard post.
**Since Snohomish counts election day votes electronically, these gains come only from absentee ballots, which may explain why they tilted to Gregoire. There was a late surge to Rossi, according to the polls. Since those who vote with absentee ballots often vote early, I would expect to see less of the surge there. To put it another way; if we counted only the election day ballots, Rossi would probably have won easily, but would have lost handily if we counted only absentee ballots.)
Posted by Jim Miller at January 13, 2005 04:55 PM | Email This...unless you want to be stubborn and refer to a study that was commissioned regarding presidential voting.
Posted by: bmvaughn on January 13, 2005 05:01 PM2. If hand counting were less valid than machine counts, why did Rossi and other republicans vote for allowing it as a last resort after machine recounting?
Posted by: Tom on January 13, 2005 05:22 PMBetween the first count and the manual recount 4,145 addional ballots were counted statewide.
In King County between the second count and the manual count they added 553 ballots. They recanvased 735 ballots due to the Supreme Court decision. Without those additional ballots they would have a net loss in ballots where almost every other county had a net gain.
Most of the net gains can be explained by the people reading the ballots are able to do so when a machine may not have. Ballots were supposed to have been remarked between the first and second counts to allow the machines to read them. None should have been remarked for the manual recount. There is no need to. But we still wound up with more ballots counted. Why?
Posted by: Vince Callaway on January 13, 2005 05:25 PM Its real simple to answer your question.
the machines don't care who wins.end of
discussion.
One can argue this over and over again. It is generally accepted as being more accurate. Sam Reed believes it is more accurate (yes I know the orange group is trying to recall him)
The difference is that the state law says that the "hand" recount will be the binding one on the election. Not all states have this provision.
If Rossi had any concerns about the hand recount was, he could have supported legislation to remove entirely the hand recount provision. No one questioned this provision until Rossi lost.
Thus, the subject is just another way of wishing the goal posts were moved after you have punted.
Posted by: Erik on January 13, 2005 05:32 PM
Generally accepted? By whom, people who think that computers are a Republican conspiracy?
Do you use spreadsheets and calculators at work and at home? Or do you add everything by hand?
Maybe the reason we're going over this issue is because we're considering legislation or initiatives to change this travesty of a law. Maybe it wasn't apparent when it was voted on that King County would use the hand recount to steal an election.
Posted by: Larry on January 13, 2005 05:38 PMRight in the middle of the manual recount, there was an episode of The Greatest Race (or whatever it is called) - the show where people go around the world on a scavenger hunt of sorts, doing stunts and such along the way. In the episode, the contestants had a box containing several thousand items. They were to count the items and report the result. IIRC, none of the teams could come up with the correct count. Every count was different. One or more of the teams just gave up and didn't finish as they could not come up with the same answer twice.
Machines are unbiased. I say the law be changed such that if the machine cannot read it - it is not a vote. The only enhancement of a ballot takes place when the post office mangles the ballot. Then, it is done with witnesses from all political parties present, and the original is preserved for auditing purposes.
This was not done in this election for thousands of ballots.
Posted by: Scott in Carnation on January 13, 2005 05:47 PMFurthermore, before the ballots can be counted by hand, they must be separated into precincts - again by hand. Ask anyone who was present at the hand recount - this task was very difficult. Counters were continually finding ballots from other precincts in piles that were supposed to be from one certain precinct.
All the ballots have machine-readable bar codes. In my polling place there were four precincts. At the end of the night the machine tallied the totals from the four precincts spot-on. No manual separation necessary.
Posted by: Larry on January 13, 2005 05:52 PMBack on topic, provided that everything is being done honestly, the very best I can expect a hand recount to be is as accurate as the machine count.
Unfortunately, machines are subject to tampering, so a hand-recount is a necessary option. As long as it can be demanded as a last resort, the temptation to add a "little extra line" of code in the programming can and will be resisted. (Like the reason for locking your door. It helps keep the honest folk honest.)
Now, as to some improvements to make both machine and hand recounts better and more accurate. Somewhere, I read about at least one type of voting machine where the voters indicate their choices on a touch-screen, then the machine spits out a properly filled out paper ballot.
This can be checked by the voter before it goes into the reader. No "voter intent" nonsense, no circles, check-marks, "Christine Rossi" votes, hanging chads or other absurdities.
Also, a consistent, easy-to-check paper trail to reduce the temptation to play games with the programming, and a back-up in case the machine just malfunctions. (It has been known to happen, after all.)
Does anyone think this is a bad idea?
Posted by: John Barelli on January 13, 2005 06:27 PMWhat you don't like is. A. Votes counted by voter intent and B. machines in Dem precents are more apt to be underfunded.
Posted by: Ray on January 13, 2005 06:36 PMWith all the other counties gaining ballots why does King County appear to have a net loss?
They clain the 732 votes were previously rejected by staff. That means they were not tallied in either previous count. Yet they only gained 553 in the finaly tally.
This means if 100% of those 732 ballots were valid they are still 182 ballots short of having the same total between the two counts. Most other counties seemed to have gained at least a few.
Maybe I'm just not using the new math.
Posted by: Vince Callaway on January 13, 2005 06:59 PMA manual recount is provided for in Washington LAW - there was nothing at all about individuals or parties "allowing" it.
Did or did not Rossi vote for that law?
Posted by: Tom on January 13, 2005 07:07 PMRefer to my post above. Is it not possible that Rossi voted for it before he realized that King County would use it to steal an election?
Do you believe hand recounts are more accurate?
Posted by: Larry on January 13, 2005 07:11 PMWhy on Earth are we going back to 1950, when you needed the computer the size of a living room to do simple calculations?
Please refer to my posts to Erik - do you use calculators and spreadsheets to aggregate and sum large amounts of data? Or do you do it by hand?
Simple answer, right?
Posted by: Larry on January 13, 2005 07:13 PMSecondly, let's look at the accuracy of machines in another light. Remember in grocery stores when there were no bar codes? Why is it that when we shop at Safeway or Albertsons that the clerks pass the products pass a bar code scanner? Well, because once upon a time, clerks were making more errors manually entering prices into the cash register than something swiped in front of a scanner does. It doesn't get any more cut and dry than that.
I took the time to write 27th Legislative District Rep. Jeannie Darnielle recently to share this story. Her response basically was how it would get improved next time and touted Sam Reed's proposed reforms.
Only a few days later, I see a letter in the mail addressed to my wife, from the same Jeannie Darnielle. SAME EXACT LETTER!
I certainly know who I'm NOT voting for next election.
Posted by: Brent on January 13, 2005 07:15 PMTHe problem is you have to WANT reliable technology that CAN'T BE TAMPERED WITH. I sadly don't think that's the case in KC.
I'm going to disagree with that. Some individuals (both sides) probably want the ability to cheat, just to "offset the other side's cheating." Most people, even those that are strongly partisan, want to win elections honestly.
One thing I do see is that good input from the "other side" is often ignored or distrusted just because those "other" people suggested it. If a Republican wants an improvement to the system, Democrats tend to look for ways that it benefits Republicans at the expense of Democrats. The reverse also seems to be true.
Sadly, party leaders also seem to want honest elections in districts where the other party is counting the vote more than they do in their own. After all, we know that "we" are honest, it's those other fellows that are liars, cheats, etc... So, it's ok if our machines can be tampered with (because we would never do that except to offset their cheating,) but their machines need to be tamper-proof.
Often a good idea (example - some way of verifying a voter's identity) gets "improved" by party leaders (hey, let's require a driver's license, bank book showing at least $10,000 and two major credit cards! ;-) that would be to the detriment of the other party. Then the whole thing turns into a shouting match, and nothing gets done.
One of the best things about these blogs is that reasonable people from both sides can potentially find solutions to these issues. The next trick is getting our party leaders (both sides) to actually use them. I'm open for suggestions on that one.
The qeustion is whether a manual recount was more accurate than the machine recount...not whether you believe the election was stolen.
Anyway, it's the peak of irony to see Rossi claim a law he, himself, supported is a reason for his defeat. It is humorous to claim you've been harmed by your own actions...yet you blame others. Actually, a fool's errand.
And, even if you believe the election was "stolen" using the law..it's the law (again, the law he voted for). Change the law. But only change it for the future...quite unconstitutional to change it for the past. This is ironic again, as the Republicans want to change the rules, after the fact, after claiming they wanted to follow the rules.
New election? The mother of all rule changes.
Posted by: Tom on January 13, 2005 07:19 PMI used to work in retail at both Fred Meyer and Montgomery Wards and I remember doing yearly inventory. One year I had to count in the hardware department. The bins of loose nails and screws was the worse. No matter how many times I counted the numbers never came out the same, but I did not care if the nails or screws won. Just kidding.
Posted by: LookoutWife on January 13, 2005 07:24 PMI don't believe the election was stolen using the law. I think it was stolen by fraud, neglect, and incompetence. Understand?
Now, read my post above about the need to separate the ballots into precincts by hand before counting. Anyone who observed the recount knows this was terribly difficult.
All the ballots have bar codes that make the aggregation by precinct moot. Understand?
Yes, I know this is about the accuracy of the manual recount, which makes it interesting that you didn't answer my question.
Do you believe the manual recount, as it was implemented in King County, was more accurate? Do you use computers and calculators to aggregate and sum large amounts of data? Or do you do it by hand?
Posted by: Larry on January 13, 2005 07:26 PMOr, more specifically, that the law would be used to justify "counting" practices it expressly forbids?
A handcount could be more accurate than a machine count, if proper procedures were followed. I would suggest the following:
-1- Have a machine physically sort ballots from each precinct into five piles: candidate #1, candidate #2, no candidate, other/spoiled and "wrong precinct". Log the number of ballots in each pile. Ballots in the "wrong precinct" pile should be moved to the right precinct, with the change duly noted in both precincts' logs.
-2- Have the piles examined by election officials who would be searched to ensure that they did not have any color of pens other than purple. If an official finds a ballot whose markings they believe were not accurately recorded, they should use a purple pen to mark a unique control number on the ballot, log it, and place it in a "challenged ballots" pile.
-3- Ballots in the "disputed ballots" piles should be examined by members of all parties for disposition. They should then be examined by members of all parties for disposition.
-4- No room should contain 'open' piles of ballots from more than one machine-count classification; piles of ballots should be counted (using a rapid counting machine) when they are brought into a room and removed from it.
The net effect should be that ballots from each precinct should be sorted into separate groups for every combination of initial classification, initial objection (if any), final classification, and final objection (the potential number of piles would be quite large (over 150), but in practice most would be empty).
Conducting the hand count in this way would probably be faster than trying to examine each ballot individually to determine for whom it is cast, and would probably also be more accurate. More importantly, it would allow examination of the specific ballots which changed between the machine count and the hand count.
Posted by: supercat on January 13, 2005 07:29 PM"Rossi voted for the current election laws, therefore he should sit down and shut up."
I mean, where did this come? Marx? Stalin? Lenin?
Oh wait, I know...the Demoncrats' playbook.
If this is the best you got, do us all favor and try really, really hard to come up with something new and actually relevant to the thread.
Rant over.
Posted by: Shaun on January 13, 2005 07:39 PM"Recount"
There weren't any recounts, if "recount" means that the same collection of items was counted again.
I believe that this is what most people think of when they hear this term. The example of the contest in the Amazing Race fits this definition.
The operations performed after the original vote count do not, because the collection of items changed after the original count, and again after the machine "recount".
Perhaps "retabulation" would be a better choice. I suppose the same point about the collection of items would apply, but at least it would give some indication that a different operation was being performed.
The MSM articles almost always refer to the "machine recount" and the "manual recount" without any mention of the extra ballots added for each.
Posted by: ewaggin on January 13, 2005 07:52 PMThe rationale for the hand count really revolves around one issue. Counting the ballots that the machine rejects and does not count.
If Washington State had a bullet proof process to handle the ballots that the machine would not count, then I would have a little more confidence in the result.
But as is the case in this state, political correctness has taken the place of common sense as witnessed in the mantra "count every vote".
We therefore have taken a simple process and degenerated it to "divining" voter intent by some ambigious, touchy feely method that at the very least is inconsistent and at worst is dishonest.
There would be no need for a hand count if the election laws were clear and unambigous when it came to making decisions on rejected ballots. these ballots would be set aside and counted by hand anfd then added to the machine counts.
I would bet you that every person that believes in hand counting also ascribes to the foolish phrase "count every vote". They are all part of the politically correct crowd that is driving this country to ruin.
Posted by: jaybo on January 13, 2005 07:57 PM
Where are you guys? All I did was ask if you aggregate and sum large amounts of data using spreadsheets and/or calculators, or if you do it by hand?
I assume they're all busy at the grocery store, waiting for the clerk to manually add the cost of their groceries.
It's called an ATM, or Automated Teller Machine. Nevermind votes, imagine how upset people would be if the ATMs were only 99.99% accurate. They aren't, they are better than that.
How hard could it possibly be to adapt ATM technology to voting machines?
I'm not talking about voting at existing ATMs, I'm talking about building voting specific machines modeled on ATM technology. Surely it can't be that hard.
Posted by: Jason on January 13, 2005 08:06 PMYou may actually be on to something (I am not suprised, since you know what you are talking about).
Don't worry about the idiots who might pooh-pooh it, they still rub sticks together to make music (or was it light a fire?)...
Posted by: smegma on January 13, 2005 08:16 PMThe argument keeps going over which is more accurate, but are they really the same thing?
I don't believe they are.
When the machine counts or "recounts", it simply tallies the data recorded on the ballots. For this type of operation, machines are far more accurate (not to mention much faster) than humans.
When a "manual recount" is performed (from what I have read...I trust that no one will hesitate to correct me if I am wrong), a tally is made, but it is made with the knowledge of what the machine count and "recount" were.
This is a critical difference, because if the manual tally doesn't match the previous one, it is immediately checked to find out why. This built-in error trapping means that "manual recounts" are potentially more accurate than "machine recounts" (though far more time-consuming).
Why only potentially more accurate? Leaving aside the obvious (intentional falsification), on any single count, the machine will perform virtually to perfection (barring a malfunction). Thus a "manual recount" could be only be as accurate, but not more so. But over a large number of counts, there will be some small amount of undetected machine malfunctions, which will result in erroneous tallies. Over this same range of counts, the "manual recount" will maintain the same level of accuracy, and thus could, in the long run, outperform the "machine recount".
The above discussion should only be applied to valid ballots. Ballots rejected by the machine as unreadable, undervotes, and overvotes are an entirely different matter.
Posted by: ewaggin on January 13, 2005 08:37 PMYou are forgetting that Ballots were "augmented" after the machine re-count, so how would the republicans know? So once again machines don't cheat people do. So have we regained our sanity?
Posted by: Adriel on January 13, 2005 08:46 PMBZZZZZT! Deek, care to try again?
Actually, there's a Rep, a Dem, and a third (supposedly neutral?) reviewing the votes. This is the great system that decided 2-to-1 that a write in for "Christine Rossi" went to Gregoire...
Yep, pretty much as you described it, Deek. Pretty much....actually, not at all, but if it helps you sleep nights.
Posted by: Ken on January 13, 2005 08:54 PMNot that I want to debate you, please recall however that prior to the hand recount begining numerous people including Ralph Monroe noted that hand recounts were not as accurate as machine counts.
Posted by: Jericho on January 13, 2005 08:55 PMHaven't you heard that profanity is the poor man's adjective?
ps. Lately I'm feeling very poor about the state of the state.
Posted by: Jericho on January 13, 2005 09:00 PMWith respect to counting the votes, accuracy is important, but there is another matter that is just as important.
Since only legitimate ballots are supposed to be counted, accuracy doesn't mean diddly-squat if the set being counted is contaminated with illegitimate ballots.
Which, of course, brings us to the present situation. The set of ballots from King County is contaminated with illegitimate ballots, as has been admitted to by the officials.
Whether fraud was involved is irrelevant - the ballots (well over 2100, and rising) are illegitimate because one or more of the steps required to produce a legitimate ballot was omitted in each case.
What is relevant is that the number of illegitimate ballots is large with respect to the margin separating the two candidates, and the result of the election is therefore unknowable, and must be set aside.
Posted by: ewaggin on January 13, 2005 09:09 PMYou seem to have forgotten in regard to your ATM technology idea that you are dealing with the unreasonable. Implementing such a reasonable idea might mean to the tyrannist sector that they would have to give up their 40 minute water cooler discusions, their 9:17am to 4:23pm jobs, and their 89 minute lunch breaks.
But there is always hope for God says, "Come let us reason together -- though your sins be like scarlet they shall be white as snow, though they be red like crimson they shall be as wool." Isaiah 18:1
Another point is that why would the Democrats want elections to be efficient? Then they wouldn't be able to manipulate them. Reminds me of the Carlson-Shrambo interchange today on KVI. Carlson tried to pin the jello like Ken Schram to the wall on why the Democrats refused to allow the "are you a citizen?" question to be asked at motor-voter registrations. And the obvious answer John gave but Ken dodged was it would be an efficient way to weed out illegal voters (and might obviously put a damper on illegal aliens all around). Weeding out illegal voters is of course a big no - no in the dim playbook.
Hat tip Carlson and hat tip common sense.
Please apply white-out to poster intent.
Posted by: Jericho on January 13, 2005 09:23 PMI would hope that the events that have transpired during this election would be sufficient to convince anyone that electronic voting is a really, really, really bad idea.
The problems, errors, and wrong-doing would never have come to light if there hadn't been a paper trail to examine in the light of day.
And please don't try to tell me that some system that produces a hard copy is just as good, because it isn't. If the system produces the hard copy, then it isn't an independent record, and thus is vulnerable to hacking of the system.
If optical systems are better than punch cards, then by all means let's get them installed in every polling place in the state, but let's stick with systems where the voters create the physical records.
Posted by: ewaggin on January 13, 2005 09:24 PMIn the last year's Christmas hit (excuse me, winter hit) - "Cheaper by the Dozen" the oldest daughter had a boyfriend character played by Demi Moore's fornicating squeeze (He did an outstanding job in the role by the way.) Anyway, he stated at one point in complaint about his girlfriend's younger siblings, "They set me on fire."
Having your prospective inlaws set you on fire - not a pleasant experience unless of course your In-Law is The Lord - I very much like the way I'm burning. "Come on baby (Jesus) light my fire."
Mmmm, cough, cough, yes that revote thing - umm, Go Orange Crush!
Posted by: Jericho on January 13, 2005 09:31 PMBesides I'm sure the press would have a field day with the antics of first nephews Huey and Luey. I hear Duey has cleaned up his act of late.
Posted by: Jericho on January 13, 2005 09:35 PMWell, In Washington State I think we all know that we have ended up we Dewey, Cheatem, and Howell. Mo, Larry, and Curly would be better.
And just to stay on topic: I think that counting ballot by hand is NOT as accurate as machines, just as Paul Berendt stated.
Posted by: deadvoter#2 on January 13, 2005 10:17 PMIf Abe and Bob are running for chief bottlewasher, the electronic part lets you pick whatever as succinctly as possible. Separate screens for each race. When it prints the ballot, it isn't circles/marks/names etc. It would be completely unambiguous:
"Chief bottlewasher: Abe"
(with nothing else printed on that line). Or perhaps
"Chief bottlewasher: No preference"
Since the voter reviews a printed ballot before putting it into the voting box, we can be pretty sure our vote wasn't illegible. 'Stray mark' and 'voter intent' issues waxed. A serious dent put into the 'Diebold stole everything' mythos - there's a pile of paper and an independent review at any point should have an extremely low discrepancy compared to the original count.
Posted by: Al on January 13, 2005 11:09 PM