I’ve written before (1, 2, 3) about discrepancies in the election results from New Jersey’s February 5 presidential primary. Yesterday we received yet another set of voting machine result tapes. They show a new kind of discrepancy which we haven’t seen before – and which contradicts the story told by Sequoia (the vendor) and the NJ Secretary of State about what went wrong in the election.
The new records are from three voting machines in Pennsauken, District 6. We have the result tapes printed out by all three voting machines in that district (1, 2, 3). As usual, each result tape has a “Candidate Totals” section giving the vote count for each candidate, and a separate “Option Switch Totals” section giving the voter turnout in each party. We also have the Democratic vote totals reported by the county clerk for that district (and some others), which were apparently calculated from the memory cartridges used in the three machines.
The county clerk’s totals show 279 votes in Pennsauken District 6. The per-candidate counts are Clinton 181, Obama 94, Richardson 2, Edwards 1, Kucinich 0, Biden 1, which adds up correctly to 279. The turnout sections of the three result tapes also show a total Democratic turnout of 279 (133+126+20).
But the Candidate Totals sections of the tapes tell a different story. Adding up the three tapes, the totals are Clinton 181, Obama 95, Richardson 2, Edwards 1, Kucinich 0, Biden 1, which adds up to 280. The Candidate Totals on the tapes show an extra Obama vote that doesn’t appear anywhere else.
(Everything seems to add up on the Republican side.)
The State claimed, in response to some (but not all) of the discrepancies I pointed out previously, that I had misread the tapes. This time the tapes are absolutely clear. Here are the Democratic candidate totals from the three tapes:
Here are the turnout sections of the three tapes:
(These images are all scans – the original documents Camden County sent me are even clearer.)
This is wrong. It is inconsistent with Sequoia’s explanation for the previously-noticed discrepancies. It is inconsistent with the State’s theory of what went wrong in the election.
It’s time for an independent investigation.
Diebold and Sequoia can’t even build a simple calculator. Pathetic.
A view from outside the US:
The US Government needs to address this question ASAP in a public and transparent manner. America’s flavor of “Democracy” already IS the laughing stock of the world. Yes, even your friends are laughing behind your back.
Having the core part of the democratic voting process “outsourced” to big business with apparently minimal oversight? What could possibly go wrong…
And as for voting machine manufacturers trying to prevent independent investigations into this by claiming ip infringements / trade secrets: Your product is a glorified CALCULATOR for chrissake. Did your highly paid engineers accidentally dance the magic package-test-deploy dance counter-clockwise? Sorry for asking.
US credibility and influence is dwindling … fast. If you again want to become the world leader you once were, fix the system. And this is only a tiny part of it…
If we can’t trust your electoral system, we can’t trust your “elected” leaders.
The votes listed in the Candidate Totals are
taken from “Report Source Internal Machine Memory”.
Is there a presumption that the votes printed from
“Internal Machine Memory” have all been cast?
The “Public Counter” on the report for machine 25249
is 29 whereas the number of candidates voted for
appears to be (14+7)+(5+2+2) or 30.
Could this be consistent with a scenario where the
last voter of the day tries to vote for Obama but
doesn’t bump the “Option Switch Totals” so actually
has not in fact completed their vote? And then the machine is
powered down while one uncast vote remains in memory.
Under this theory there’s a possibility of recording
one more vote in “Candidate Totals” than is recorded
in “Option Switch Totals”, i.e. a true off-by-one bug.
-mai
You will need a copy of all the source code and all the Verilog code which are the blueprints for the software and hardware to prevent fraud. Instead of the computer comparing you with a picture you should compare the computer with a picture and if the picture is correct then you know that the voting result is also correct.
Because some people are naive and think that if the source code is correct then everything is OK.
Imagine that the problem is in hardware. You press the keyboard and if you do this in a certain way so that when the timing gets right then the user is allowed to edit the results. Cool and interesting. There are many combinations with a keypad and maybe you will need a machine to do this. But anyway you get my point.
It is called a shadow government. And the politicans did not order it. Don’t blame innocent people when you cannot find the correct guilty one.
I see “Hitler” on his way back! Dissect the complete voting machine including all its chips after that the voting is done. Reduce the number of CPU instructions. Make it simple so that even a 10 year old will understand what it is doing.
ASAT (As Simple As That)
In God we shall trust [and not in the chips inside the voting machine].
–HPS
Where do the county clerk’s vote totals come from? Do they come from the machine receipts, or from a separate tally kept by the poll workers?
I was a poll worker in Montgomery County, MD several years ago (after they began using the Diebold machines, but before they switched over to electronic poll books for check in). We kept a running tally of voters as they checked in, and we also counted the paper slips which voters received at check-in and later exchanged for a voting machine access card. The check in tally needed to match the slip count before we could submit our totals to the Board of Elections.
Does NJ have a system like that in place?
I second (or third) the statement that the fact of an error is what is significant here, not the apparent magnitude. In double-entry bookkeeping, if you see that your reconciliation is off by 22 cents, you don’t just dig in your pocket for the extra quarter, you go back and add everything up from scratch. The same goes for checksums on memory or communications, you repeat the whole process and hope it comes out right the next time. I am also forced to recall the old mariners’ imperative: always take one compass or three, never two, otherwise, how will you know which one is right?
In this system, we don’t have a way to go back and add everything up from scratch. The data do not exist. If we follow proper protocol, we go back to zero and toss it all out. When I worked with intelligence data, we had the same problem: when you see a discrepancy in the number of tanks reported for an enemy division between two sources, it tells you nothing about what the correct answer should be. Maybe those tanks were moved during the night between when one observer counted and the satellite went over. In that case off by one could mean that 200 tanks arrived overnight from somewhere else and 199 moved to the front. (!!!) But generally, you don’t want to say “crawl back there behind enemy lines, through that minefield and count them again” if you can avoid it. With elections, we can avoid it, by keeping a proper (and randomly validated) audit trail. It is not technically a difficult problem in this day and age with all we can bring to the task and it is certainly important.
Ed,
I missed the links to the full tapes, my apologies. I agree there needs to be a review here. It’d be nice if these numbers were off by one (not for nefarious reasons, but so that your argument would carry more “sensational” weight).
This next election is already a joke.
Sam
Sam,
Look at the Republican data from the three tapes. Everything adds up. There is no missing voter who could match up with the extra Obama vote.
In your previous posts you included the Republic side and it was shown there was a similar discrepency but in the other direction (seemingly one vote was missing). Where are the republican tallies? This is still just off by one and without the republic tallies you cannot show that this is not the same issue as before.
Yes, the control totals are off, but the net vote works out and indicates someone switched their party selection.
I heard Sequoia is changing their company slogan:
“2 + 2 = 5 (for extremely large values of 2)”
There’s even a T-Shirt over at ThinkGeek for them:
http://www.thinkgeek.com/tshirts/generic/60f5/
On a more serious note, could it be that the two sections (per candidate and per option) are stored on two separate pieces of electronic hardware? That would be a REALLY good idea for a manufacturer to do so that the system could perform a checksum (i.e. 20 != 21) and verify that the data was correct.
Hardware fails all the time, shouldn’t there be some sort of built in system to verify the calculations?
At my company we have some system boards with two processors on them from two different manufacturers. The commands are run through both processors and the results are compared to verify they are the same. That way you avoid any silly Pentium floating point problems or other possible hardware flaws.
You would think Sequoia would come out and say THANK YOU for pointing out our failsafe checksum feature! A failed checksum should prompt the county to go through and hand-count each ballot to verify the results recorded on the machine.
Once the Primaries are over, don’t these items become Public Record? Why not wait until anyone has the legal right to obtain copies from every Sequoia Machine and perform their own survey?
I doub’t Sequoia will but the Counties have to, with the right paperwork, right?
Why not just blame it on the humidity? Didn’t that work for the Diebold systems in Ohio?
The magnitude of the problem is displayed by the Idiot in Chief currently in office. We can’t prove he was supposed to be there and there is no audit capable of doing so.
+1: Has the legislature considered calling out the state militia?
//If an error occurs then it shows that something has gone wrong with the process. It is not just a question of looking for the 99c that have gone missing – it might mean that something much more serious has gone on!//
Exactly. Things like cross-check totals are not very good at identifying the magnitude of a problem; indeed, they won’t always identify when a problem exists. On the other hand, they provide a quick and easy “screening test” that will identify most of the places that require further investigation.
If there are very few errors in handling the data, it’s unlikely that two or more errors will cancel each other out. As the number of errors increases, however, so does the likelihood of offsetting errors. A policy of tolerating errors and simply correcting the ones that are found will not ensure reliable results. To ensure that results are reliable, errors must be essentially eliminated. Here, though, they seem to be way too common.
P.S. For those who might have missed the import of my last suggestion–the state of affairs described bears all the marks of being utterly preposterous.
You have described a situation more usually expected in the vicinage of a South American banana plantation than a situation to be expected in one of the original thirteen states of our union, in the two-hundred and thirty-first year of independence.
P.P.S. If the forces of the state cannot cope, I believe you may apply to the federal government for further assistance in suppressing the insurrection.
So, in essence, you’re telling me that Sequoia has de facto sovereign power in Camden County, New Jersey. The corporation can run elections any way it damn well pleases, with no oversight.
Has the legislature considered calling out the state militia?
Practicality,
I don’t know who, as a legal matter, has the power to start an investigation. The counties tried to start an investigation at one point, but Sequoia reportedly threatened to sue them if they went ahead with it, so the counties backed off.
Charlie Strauss makes a very important point. The reason for cross casting the totals is to indicate if an error has occurred. This is the same principle as in double-entry book-keeping. If an error occurs then it shows that something has gone wrong with the process. It is not just a question of looking for the 99c that have gone missing – it might mean that something much more serious has gone on!
My first job when I graduated from college involved programming the embedded firmware for a printer. It was a different kind than this, more primitive, but the details of getting it to print a given value were pretty complicated internally. There must be some interface between the motherboard and the printer on this voting machine, probably a serial or parallel port and cable. The difference between a 6 and a 7 is 1 bit. If there were noise on the line connecting to the printer at the right time, then theoretically that could cause a misprint. (Sometimes there is parity protection but not always, and it’s probably questionable whether the software is prepared to retry the communication on bad parity error.)
Whether this is the reason or not, it looks to me like the software was in some sense thinking 6 and printing 7. This could happen in various layers, there could be race conditions in which some memory location got changed while it was still being used, many different things could go wrong to cause this. But the evidence suggests something along these lines.
I keep coming back to the larger issue of what this all means. Does it just show that human creations are imperfect? Does it mean that this one company is incompetent, and that all we need is to hire better people? What does it say about the virtues or limitations of paper trails as an audit mechanism? Suppose the fault turns out to be in the printer or the cabling. Then it seems that the main thing we have accomplished by adding the paper trail is to create yet another failure mode.
Thanks, Ed.
You didn’t answer one question I had buried in my first post, though:
Who has the power to launch an independent investigation in Camden County, New Jersey?
Another way of phrasing that question: Who has the power to go down to the county clerk’s office, inspect the public records, and proceed to take sworn statements?
Come on, lets get serious — how important is one vote?
Some history. There was a state wide contest (primary) for Governor in Illinois a number of years back where the difference between the winner and looser was ½ of 1 vote per precinct.
And then there is a whole element of this situation that is not being discussed. When all the results from these units were consolidated, which number was tallied to the final report?
As rational and thinking CITIZENS we should demand an independent review!
Hal,
The 7 looks even cleaner in the copy the County Clerk sent me. The small glitch you noted in the 7 is a scanning artifact — it’s not on the paper copy I have. (If this becomes an issue, I can post a high-res photo of the paper.)
Practicality,
I received these tapes directly from the Camden County Clerk’s office. The other tapes I received directly from their respective counties. (They gave me copies of course; they still have the originals.)
I suppose the discrepancy could be explained if the 7 for Obama in the 3rd machine were a misprint of a 6. That would fix the sum of 20 on the 3rd tape as well as the overall total. However the 7 looks pretty clean, there is a small glitch in the middle and the bottom has a hint of a curve, but it is far from a 6.
Thank you,
I have a much simpler explanation: The tapes that you’ve posted are forgeries. You’re ran an experiment to detect whether people will read an exclamation point as the numeral one. And now you’re running another experiment with bogus tapes.
Prove me wrong.
I’m sure Ed can answer questions about provenance and custody for the other tapes as well.
Practicality Said:
The discrepancy shows up clearly on tape 3. That tape shows 14 Clinton votes, 7 Obama votes, no votes for the other DEM candidates, but only 20 DEM “option switch totalsâ€. My mental math sez 14 + 7 = 21. The total of 20 is certainly anomalous and appears to be an error.
Why did you present the per-candidate tallies from all three machines together, and then the option switch totals from all three machines together?
I think the issue here is this. In the previous tapes we saw the same thing. Namely more votes than ballots casts. The problem then was do we think the ballot totals are wrong, the vote totals are wrong or both? Sequoia pointed to a known error in the Ballot counting software so One suspects the error might possibly be on the ballot counting side not the vote tallies. (not a great argument but not implausible)
But now we have a case where the reported vote total by the county agree with the sum of the ballots cast. But they disagree with the sum of the printed totals from the machines. Hence it MUST be the vote totals are wrong. (both totals can’t be right at the same time)
Hence the sum of the tapes is needed to show this.
Oh, Ed–
Alongside my question about your choice about how you presented the tapes in the images here on your blog, I have another similar question (or two):
What is the provenance of these tapes? What is the chain of custody for the copies?
The discrepancy shows up clearly on tape 3. That tape shows 14 Clinton votes, 7 Obama votes, no votes for the other DEM candidates, but only 20 DEM “option switch totals”. My mental math sez 14 + 7 = 21. The total of 20 is certainly anomalous and appears to be an error.
Why did you present the per-candidate tallies from all three machines together, and then the option switch totals from all three machines together?
Who has the power to launch an independent investigation?
No offence intended, but it does not take a respected professor to tell us that things do not add up. A 4th grade child could point out the discrepancy and do the math – and our elected officials and this vendor cannot? How is a society to function when things such as this are allowed to stand? Is the real algorithm: A bunch of free lunches + a nice campaign contribution or two + a semi plausible lie = A few votes off is ok with us?
the Onion has a video related to the point:
it’s called “Diebold Accidentally Leaks Results Of 2008 Election Early” http://www.theonion.com/content/video/diebold_accidentally_leaks
This off-by-one stuff… are they by any chance using floating point to do their calculations? That would explain it quite thoroughly.
Is it possible that there are multiple kinds of engineering design flaws happening simultaneously. For example, suppose that someone who voted for Obama also wrote in someone’s name. The voting machine might record it a vote for obama (295) but the clerk’s office then removed the overvote by hand?
That explanation does not hold up unless one can also explain why the total ballots cast does not agree with the vote totals.
So here we need to invoke the other flaw that causes the misattribution of party that was discussed earlier.
But that still does not close the loop. Since if the republican votes total up correctly and match the cast ballot totals then it can’t be a mis-attribution.
But we saw this before in the other sets where the totals simply did not balance out no matter how one assumed misattribution.
The only distiniction here is where the error might be occuring. One can make conjectures that puch the error from one domain to another. In the end something does not add up.
Another mind trap to avoid here is the “it’s just one vote” or “it wont change the outcome”. The point is the numbers don’t add up. To conclude it’s “just one”, a reader would have to assume that one of the totals is correct. Why should any of them be correct. What if 20 votes were shifted from one candidate to another but barak just got one extra. What if all the totals are wrong and dozens of ballots were simply never entered or others were recorded in duplicate.
I’ve made this point before but perhaps it’s worth repeating. When you see an off by one error, it suggests a programming mistake not a clerical mistake.
I think the facts are that it “was” off meaning that if it is NOT a secure way to count votes – meaning accurate and a safely kept tally then it shouldn’t be used…
Maybe there was a programing code to it to do this for a reason not displayed here… are these all the facts that are available? maybe something is missing? I hope the machines aren’t off.. that could do a lot of financial damage that we cant take right now launched into investigations of possibly every machines across the US….
Another mind trap to avoid here is the “it’s just one vote” or “it wont change the outcome”. The point is the numbers don’t add up. To conclude it’s “just one”, a reader would have to assume that one of the totals is correct. Why should any of them be correct. What if 20 votes were shifted from one candidate to another but barak just got one extra. What if all the totals are wrong and dozens of ballots were simply never entered or others were recorded in duplicate. used compound bows