Two weeks ago I noted that I did not find Christine Gregoire's Snohomish gains in the manual recount especially suspicious.
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.
The same argument applies in reverse, of course. Because of the late Rossi surge, we should expect Rossi to do better with touch screen ballots cast on election day than he did overall. Two Democrats, Paul Lehto and Jeffrey Hoffman, disagree and find the Rossi edge in those ballots very suspicious.
I wouldn't agree, unless there is more in the Lehto-Hoffman paper than I have found in a quick skim. The apparatus in the paper is impressive, with pages of graphs and statistical tests, but the discussion of politics in the paper is not. For example, they point out that Maria Cantwell beat Slade Gorton handily in Snohomish in 2000. That's true enough, but it is also true that Cantwell lives in Snohomish county and represented part of the county while she was the 1st district Congresswoman. One would expect her to do well in her home area.
And they don't examine alternative hypotheses. If, for example, you agree that the Snohomish electronic votes are too different from the absentee ballots to be explained by chance or a late Rossi surge, then one explanation would be that someone rigged the absentee ballots. Those who have read my site for any length of time will know that people committing vote fraud in the United States almost always use absentee ballots. But Lehto and Hoffman never examine the possibility that fraud with absentee ballots explains the difference.
I continue to believe that timing, not fraud, explains the difference between the electronic and paper votes in Snohomish county, but I will take a closer look at the paper. And I think someone should look hard at the possibility that there was more fraud with absentee ballots in Snohomish county than most of us have thought so far.
(If you want to look at the Lehto-Hoffman paper yourself, you can find it here.)
Posted by Jim Miller at January 27, 2005 12:08 PM | Email ThisTheir paper shows they have a bias against touchscreen voting and are fishing for examples. Since non-touch screen counties in WA show the same trend they are simply dead wrong.
When they arrive later, it probably means they were mailed on a day that was closer to election day -- although, admittedly, it would be a rough explanation for their late arrival, since some would obviously have been mailed from the far points of the globe.
Has anyone looked to see if the split changed between the first group of absentees tallied and the later groups?
Posted by: Micajah on January 27, 2005 12:25 PMKeep an eye out for more of these.
It's harder to see the truth through muddy waters.
Posted by: smoke on January 27, 2005 12:25 PMHere's an interesting part of that article that suggests a "solution" to the "black box" problem of electronic voting -- but would make it possible for dishonest people to sell their votes:
http://www.seattleweekly.com/features/0504/050126_news_snohomish.php
Another technology, in which an encrypted printout is issued by the machine to the voter, who can later check the vote on the Internet to see if his vote was recorded correctly downstream, is being used in other states. The Legislature is weighing such options, but if it approves a system, Diepenbrock says, "it won't mean anything if they don't fund it. The counties don't have the money." [Emphasis added.]
How would the secrecy of the ballot be maintained, if the person who paid for votes could require an online proof of how the votes were cast before paying the promised fee?
Posted by: Micajah on January 27, 2005 12:37 PMSince Snohomish is as close to a swing county as we have in Washington, one would actually expect to see the largest movement in voting trends there, compared to more Rep or Dem counties where less movement is likely to occur given the comparative lack of swing votes.
But let's put aside logical thinking for a second and just stick with facts. They have their stats all wrong on Gorton v Cantwell in 2000. They claim Cantwell won the county 132,148 to 91,265. In reality, the Snohomish County website shows the returns as 123,075 to 119,292.
These guys don't even have their facts straight which is makes their flawed work even worse.
Posted by: Eric on January 27, 2005 12:38 PMThis whole things smells to me...it's as if Paul Berendt has contracted out fact-finding to Oliver Stone and has found a willing accomplice in the Seattle Weekly.
Posted by: SnoCo Voter on January 27, 2005 12:55 PMEric - Great point on the Cantwell-Gorton vote error in the paper. The numbers didn't look right to me but it didn't occur to me that they would have missed on something that easy to check. I am going to send a note to the authors after I have looked at the paper some more and will mention that mistake.
SnoCo Voter - Here's how I understand it: There was surge for Rossi in the last few weeks. As far as I know, and I would have to look at the polls again to be sure, there was no similar surge for Bush. So one would not expect the vote patterns to be similar.
Posted by: Jim Miller on January 27, 2005 01:39 PMLet's consider other elections, using the "since 1992" timeline like the authors:
3 of 5 county council districts are currently Republican.
In 1996 Rick White won the Snohomish County portion of the 1st CD handily. While Metcalf didn't win the Snohomish County portion of the 2nd CD that year, he did win it convincingly in 1998, the same year Inslee didn't crack 50% in defeating the then battered White.
The 1994 election isn't online but based on the data above it's a safe bet White & Metcalf did well in Snohomish that year as well. Without digging into the Leg District races myself, I bet they would also show countywide that that the county's voters are not "traditionally Democratic" though they can certainly lean that way if they percieve a Republican candidate fits a certain negative stereotype.
And that doesn't even touch on the fact the correct stats show Gorton did respectably well against Cantwell on her home turf - especially compared to other Republican candidates that year.
These guys obviously have theories about touchscreen voting favoring R's (so why are the two counties using the technology in WA run by D auditors...especially Yakima by Corky Mattingly of circle the wagons for Dean Logan fame).
Textbook lesson in why you shouldn't let your beliefs get in the way of the facts, or let your beliefs blind you from examining the facts to begin with.
Posted by: Eric on January 27, 2005 02:29 PMIn fact, this was merely the continuation of a trend which had been going on since at least 1996, and in fact was most pronounced in 2002 (when the Dems held their big Paul Wellstone Death Celebration--which Lehto and Hoffman do not mention--a few days before the election).
Posted by: ScottM on January 27, 2005 03:39 PMAs I said to the committee of the state House the other day, I regret that I published the names of political parties in the report. Perhaps I should have left everyone guessing about which party may have benefitted from malfunctions or irregularities since naming the parties merely invites (as above) snarky comments about "meds".
If the shoe were on the other foot, I wouldn't expect the Republicans to accept these facts based on a mere assertion that it "coulda" been due to some late surge, especially when that very factor is discussed at some length in the report. At least 27,500 of the ballots reported as "absentees" came in from the polling places, and thus were voted on election day. If the "late surge" for Rossi theory was true, the "true" surge was even larger and is masked by election day "absentees".
If anyone reads the report and has serious thoughts, please call me at my law office at 425-257-2297 in Everett. My law partner is a very committed Republican, yet he's still talking to me.
---Paul Lehto
Thanks for the post. My "meds" comment was a bit snarky and while I'm all for making sure touchscreen voting is foolproof, I think there are a number of notable flaws in your work.
As noted, when you don't even have the facts straight about the Cantwell/Gorton race (your report was wildly off, putting her percent of the vote 10 points off) then it doesn't put the report in good standing. Moreover, as noted, some of your underlying suppositions - such as Snohomish County being traditionally Democrat are flawed by looking at statewide & Presidential races, where many recent candidates were weak.
Also of note, I think it is regretable you're using exit poll data from the official media exit polls to support your claim debunking the Rossi surge toward election day. Those same exit polls were badly off in many respects in 2004, not to mention 2002 and 2000. Thus, another one of your key points has weak underpinnings.
In defense of the Rossi surge, you may want to track down polling done by Survey USA - which was released publicly several times before election day - that did show just such a surge. Of note, Survey USA, along with Mason-Dixon, was one of the most accurate pollsters that polled in multiple states during the 2004 cycle.
I also know for a fact from talking to consultants affiliated with the Rossi campaign that their numbers also showed them closing fast the last two weeks. And again, though the Snohomish trend may have been more stark, the same trend of Rossi improving between early absentees and election day in notable parts of the state are too large to be ignored.
Candidly, I think people forget sometimes that politics - and elections which measure human behavior (and which as we all know by know have a human element of error) - includes emotions as well as spreadsheets and charts. You can't run everything through statistics and trends, sometimes people change their minds in unexpected ways because of the emotional dynamics of the campign.
And yes, I have reviewed the report at greater length since my first point. While you have some intriguing bits of data, I still find too many of your underlying assumptions poorly supported - and thus your conclusion quite unproven.
That being said, thanks for the discourse. We're better for it.
Posted by: Eric on January 27, 2005 10:06 PMPerhaps we can agree that we both would like more questions answered by government officials, and more documents produced.
You state that you disagree with the "conclusions". Please specify. the general conclusion of the paper is to keep all possibilities open that are not disproven, as any good investigator would. Thus, we can not rule out either malfunctions favoring one party or hacking/tampering favoring one party. A computer forensics audit would help resolve this, but is denied on trade secret grounds. Essentially, the counting of the vote is a Sequoia corporate secret that we are not allowed access to. Perhaps it would show nothing, but we are left to guess....
I hope you all would agree that our elections should not be based on trust of anyone, rather they are in fact based on openness and the right of the parties and the public to OBSERVE. When vote counting is done with proprietary software, we are left having no observation of the vote, and no verification of it. No voter really knows how their vote was counted, or if it was. No party or candidate should have to trust the other party or ANYONE with the integrity of the vote. that is the central issue to which our study points.
Yep, I voted Democratic in 2004. But if the shoe were on the other foot and Gregoire had lost 98,000 to 96,000, but then Gregoire had won touch screen under the direction of a Democratic auditor in Snohomish county by 50,400 to 42,000 or so, I'm sure the parties showing the greater concern would be reversed here. We certainly don't see anyone giving up on the issue of the 3000 ballots in King County on the grounds they "Could have been" fully proper votes, even though not checked for signatures.
In other words, if we want elections with integrity we can not accept any old political excuse that seems superficially plausible, because such explanations will always be present except in cases where the error is so large as to be nonsensical (like a machine that registered negative 25 million votes in Ohio). In other words, it is the small to medium size error that can be hard to distinguish from real political forces. A true academic or research posture has to remain somewhat agnostic at this time, yet it also must not close its eyes to the possibilities, just as we would expect our national defense forces to anticipate and guard against any possible threat to our country, we should be vigilant as well about our elections.
If sound politics wants to sponsor a debate where ideas can get out about how to verify the integrity of elections and the need for greater transparency, I would be happy to participate. I'm sure we'd find points to agree and to disagree on.
Posted by: Paul Lehto on January 27, 2005 10:38 PMBy the way, internal campaign polls are often overly optimistic. If methodologically correct, they can be some evidence, but they would be weaker evidence than a poll not conducted by a party or campaign. I'm not enough of a politico to know if that's the conventional wisdom, but I'm enough of a lawyer to know that it would be easy to impeach campaign polls on the stand as self-serving.
The lesson I've learned in law is that there are literally ALWAYS counterarguments. Even the guiltiest criminal has a counterargument for each piece of evidence, but on balance we can't buy it... In addition, I've learned in law to ignore conclusory statements, which you use when you say "too many of ...underlying assumptions" are poorly supported. How so?
We do assume that the paper ballots, which were hand counted and thus removed the risk of machine error and hacking of the central tabulator, but not removing risks of ballot stuffing and felons voting, represent a population that has FEWER risks of error than touch screens, though admittedly not zero. We also see that both Republican and Democratic paper ballots form bell curves, but the touch screen votes form a smooth twin peak (a true surge would shift the bell curve for that race left or right, yet touch screens are deformed into twin peaks). This is a critical piece of data because it indicates that different forces acted on touch screens and absentees, and even more importantly, the "different forces" DID NOT ACT EVENLY through the county.
Even if you had been successful in suggesting weaknesses in "underlying assumptions", that does not DISPROVE the questions we've raised. It just indicates that there is a debate, which is a healthy thing. yet we need to design elections that are not so subject to question, and it should not be the duty of inquiring citizens to shut up instead of asking tough questions.
Thanks for your time and reading the report more carefully. I would submit that the paper is not just one long string that can be cut in any one place, but raises a series of credible questions that all have to be dealt with to remove and resolve issues.
I wrote to a local reporter in early December and told him I was going to wait for the recounts to finish since I was not anxious for my data to be put into the Gregoire/Rossi fight because my real issue is secrecy in elections. But I understand that when I'm a Democrat (and worked for a Dem state senator for a year long ago in another state) that's relevant as some will want to stop listening at that point.
I'd be extraordinarily curious to find folks who didn't care about either candidate... Another Republican friend of mine is a computer wizard concerned about felons voting, but when he read our paper he reluctantly admitted to some troubling data there.
We need more election integrity because I believe that the "sore loserman" attack is undeserved no matter who makes it. It is the SYSTEM that should prove itself of integrity, not the VOTERS who have to prove something went wrong. In the long run, the consent of the governed will cease to exist unless elections PROVE that they DESERVE confidence. Maybe we all agree on that kind of change?
Posted by: Paul Lehto on January 27, 2005 11:04 PMAnd I appreciate your civil tone. I try for that myself, though I can't claim to always succeed.
As for electronic voting machines, I object to them in principle as they are currently constructed. I have argued here, and on my own site, for systems that use one set of machines to print ballots and another set to count them. There have been a number of failures of elcetronic voting machines, often because they were incorrectly used. (And perhaps badly designed.)
But -- I have not seen convincing evidence that they have been used fraudulently. In fact, I suspect in some areas they have reduced fraud, just as mechanical voting machines did when they were introduced. Nearly all the objections to electronic voting machines apply to mechanical voting machines -- and there is simply no doubt that mechanical voting machines reduced fraud.
All that said, I am afraid I have to point out that you did not respond to one of my points, that statistical tests that show a difference between absentee ballots and electronic votes are just as strong evidence that there was fraud in the absentee ballots, as in the electronic votes.
And I saw nothing in my quick skim through the paper that showed that either you or your co-author understand that when vote fraud is committed in the United States, it is almost always done with absentee ballots. There are places, for example along the Rio Grande, where this is so common that there are vote brokers who collect absentee ballots and cast them at every election.
(If you want more examples of vote fraud with absentee ballots, search my site, since I have been collecting them for several years.)
I was serious when I said your paper made me think that someone should look hard at the possibility that there were more fraudulent absentee ballots in Snohomish county than most of us thought. There certainly were a few; the question is how many.
Since I am interested in vote fraud no matter who commits it, I will take a longer look at the arguments in your paper. I hope that you, in turn, will consider the problems inherent in absentee ballots, and the possibility that fraudulent absentee ballots explain some of the difference between them and the electronic votes in Snohomish county. (You may want to look for some of the critiques of absentee ballots done by Melody Rose, a political science professor at Portland State. She is on the left, if that matters to you.)
You mentioned I may have missed a point, that being that it is equally possible that Gregoire's absentees were fraudulent. the truth is, it is POSSIBLE that some absentees were fraudulent, but the possibility of fraud in the absentees UNDER THE GUBERNATORIAL CIRCUMSTANCES OF A HAND RECOUNT is less than for electronic. This is because the hand recount removed risks of tabulator error, tabulator hacking, and rates of error, while still not removing inherent risks in voting of ballot stuffing and illegal voting. However a close read of the report will instruct you in how multiple voting is easily possible with touch screens (press a certain button twice) so we can not say electronic doesn't have multiple illegal voting risk, in addition to many other risks unique to electronic. Because the hand recount on electronic did not truly reassess anything (spits out same exact totals) we did not have any meaningful increase in accuracy on the electronic side as a result of the hand recount.
Now, before you respond I do understand that absentees are not perfect and that humans make errors too. But what's critical in this debate is that when two opposed parties that have every incentive to object if counting is going wrong are involved and motivated, the error rate drops down to very low. This is the very reason the last recount is designated as a hand recount, since the process is supposed to become more accurate as it goes along (it's not "best of three" like a short playoff series).
Also recall that the paper ballots for both parties were bell curves while electronic has twin peaks. A Rossi surge would not create a twin peak, it would shift the curve to the left. A twin peak indicates a large shift somewhere but only in a distinct part of the county. Therefore, any theory that would attempt to explain away the Rossi surge would (1) just be another competing possibility, and that's what trials or debates are for the weighing of and (2) would have to show credible reasons for a concentrate surge in certain areas capable of taking the Rossi bell curve on PAPER and turning it into Rossi TWIN PEAKS on electronic.
I would respectfully request that you consider this: although I am a business law and consumer fraud lawyer,some time ago I assisted former partners in a death penalty case. The death penalty was not imposed, yet our client convicted. Under legal theory, we are obliged to accept that our client was guilty (at least after appeals expire) even though we had a good argument on the evidence at every turn. Scott Peterson had many good arguments on the evidence at every turn. Ultimately, a careful look at the WHOLE picture needs to be made, and simply coming up with some capable or possible counterarguments (as you have) by no means resolves the issue or concludes the debate.
It just shows we have a real debate going on, and I thank you for your thoughts.
Posted by: Paul Lehto on January 28, 2005 07:04 AM