Minnesota Performs First Post-Election Review

[eridani]

http://www.electionline.org/Newsletters/tabid/87/ctl/Detail/mid/643/xmid/230/xmfid/3/Default.aspx

The county has more than 150,000 registered voters and 87 precincts and it took just under five hours to complete the review of four precincts. Of the 12 races reviewed, seven had exact matches; four races saw one vote added each to a candidate and one race saw two votes added to a candidate.

Corbid said the discrepancies were not the result of machine error but rather how the ballots had been marked. In some cases ballots either had check marks or an X marked outside the oval, or voters had circled the candidate's name, which could not be picked up by the optical-scan machine but was found during the manual count.

OK, why aren't their scanners programmed to spit those ballots out and give the voters another chance? This is the norm in WA state. With central tabulation, ballots rejected by the scanner are hand-duplicated. There is an unacceptable amount of voter disenfranchisment here IMO, and there doesn't need to be.

"I was surprised at how quickly the audit went. I was not surprised by the quality performance of the equipment and our election judges. If this is what is needed to provide some assurance to those who do not have as much confidence in the system then I have no problem continuing to do the audits," he stated.

This is the attitude we have to keep fighting. My chromatography system software does its calculations right just about all the time, but I have to do periodic manual calculations as an audit anyway to guard against bit rot and bugs that have not yet been noticed. If a business has passed accounting audits for the last 10 years, this is NOT a reason to skip year 11's audit. There is NO number of successful audits of a voting system that can EVER justify not auditing next time!

[napi21]

1. I'm glad to hear this info. I only wish all machines had the

capability to be audited. If that were the case, I doubt we'd be having any arguments.

[OnTheOtherHand]

2. you mean, to warn voters of undervotes?

Some folks would say it's a bad idea to enable undervote warnings on precinct-based optical scanners because down-ticket, probably most undervotes are intentional -- and if you force people to review intentional undervotes, they may start voting for candidates at random just to save the trouble. I'm not convinced by that argument, but it does raise some questions that I don't know the answers to. Can you comment on the experiences of some WA jurisdictions that use undervote warnings on PBOS? (A routine review of undervotes on central-count op-scan would be excellent.)

On the audit front, of course, audits are a best practice. It's too bad that an election official would portray them as a concession to the insecure.

[eridani]

4. Our scanners accept undervotes for blank ovals.

I watched the central scanner rejects being duplicated in 2005, and the election workers doing the replacement ballots read out "no vote" for races in which candidates were not selected. The trouble and expense of documenting this is the reason I favor keeping the poll option even in states that allow no-excuse absentee voting.

The ballots that are rejected are rejected for having partial marks of any sort in the oval, or marks any other place on the ballot not within one of the ovals. Sometimes scanners reject ballots for no obvious reason at all, which is probably due to the printing artwork being slightly off-register.

A national consortium charged with optimizing the scanning process specifically for voting could probably set standards that would greatly reduce these problems.

[OnTheOtherHand]

5. oh, I see

Huh. Go figure.

Yes, some common standards could be an excellent thing.

[OnTheOtherHand]

6. another question

You wrote, "Sometimes scanners reject ballots for no obvious reason at all, which is probably due to the printing artwork being slightly off-register." Do you have some idea what the background rate of unexplained, presumably innocent op-scan error is? (Obviously there isn't a universal answer.) That could have some ramifications for how an audit protocol should be designed.

[eridani]

7. That varies quite a bit

The poll scanners just do one ballot at a time, and the voter can self-correct. This greatly reduces errors. In King County, WA, anywhere from 5-30% of centrally scanned ballots have to be duplicated.

[OnTheOtherHand]

8. OK, but those dups are to correct explicable errors, yes?

Or let me put it this way, based on what I think you've said so far. The precinct-based scanners are programmed to kick out ballots that appear to be incorrectly marked (incomplete or stray marks) -- not, if I understand rightly, simple undervotes, although that isn't crucial to my question. So voters can correct those ballots on the spot. And the central count scanners are programmed the same way, and then it is up to election workers to make duplicates that reflect voters' apparent intentions.

But the case I'm curious about is the one in which the scanner spits out a ballot for "no obvious reason." How common is that?

[eridani]

9. That is most likely a problem of ballot artwork registry--

--or marks on the ballot that are not immediately obvious to the people who do the rough sorting. Sometimes people are sloppy and get a little too far outside the oval and the scanner is pickier than a human observer would be. I was the Dem observer of this second "last resort" stage.

At the first stage, people remove ballots that are obviously messed up for duplication, like having been partially shredded by the post office, filled out in red ink (invisible to the scanner), "x" instead of filling out the oval completely, or demonstration of editorial commentary or cartooning "talent" in the margins, etc.

I understand that not all states are "voter intent", which they should be. Even the cartoonist assholes are not disenfranchised in WA, although after having observed the tedious duplication process I can certainly understand the temptation.

[OnTheOtherHand]

10. OK, how common are the second-stage ballots?

Maybe it doesn't matter -- in the sort of voter-intent regime you describe, it doesn't seem that any of these cases would be construed as machine error. But in a live-and-let-die scan-em-and-see-what-sticks mode, they would.

I'm curious about this because in circumstances like the Cuyahoga County 2004 recount, the issue comes up that election officials have an incentive to tamper with partial recounts in order to avoid trivial errors that would trigger an expensive full recount. (I'm not saying that the errors in Cuyahoga would have turned out to be trivial -- there is really no way of knowing. I don't think their partial recount could have detected caterpillar crawl, so I'm not even considering that.) I'm looking for pointers on how to tweak an audit regime so that it catches real problems and reasonable elections officials don't hate it.

[eridani]

11. Central tabulators can require 5-30% ballot duplication.

That is why it is absolutely essential to have a process in place so that the process can be observed. Also why I think poll voting should not be abolished--it should be encouraged. Poll scanning has more like 0.01% error, and that is most likely to be a tabulation bug. (Except in 2004, my ten precinct polling station had 4 extra votes in the machine compared to the strip chart tabulation--we suspect that a couple of people stuck provisional ballots in the scanner. The poll workers were supposed to fold them twice to discourage this, but the polling place was a madhouse at rush hours and they may have skipped a couple.)

Proper auditing should be able to detect tampering, and AFAIK this was NOT done in Cuyahoga County.

[OnTheOtherHand]

14. that all makes sense

The difference between central-count and precinct-based is pretty striking. It relates back to the lawsuit about notice vs. non-notice voting technologies, although the presidential undervotes with central count don't tend to be that high as I recall.

(I'm not quite sure what to think about the "tabulation bug," although at 0.01%, I'm not too alarmed. Maybe a finicky-scanner bug? You've mentioned some of the causal mechanisms behind "finicky scanners," and I think Doug Jones has a little tutorial on this point.)

In Cuyahoga, as I understand it, the punch cards weren't even stamped as to what precincts they were for. So that audit apparently was doomed even before they started tinkering with the ballots in the back. I probably shouldn't have mentioned it as an example, since so many things were done wrong there.

Thanks for all the info, and I hope you won't bear a grudge against me for hijacking the thread. ;)

[eridani]

16. Nobody knows for sure

If you get 7 out of 10 precints to have the ballot sign-in sheets total match the strip chart total from the scanner, and a couple of extra ballots in the other 3, it's just natural to assume this is a provisional ballot problem. That's just the most obvious guess, though. Also, sometimes people sign in at one precinct and it's the wrong one, and then sign in at their correct precinct, and then the poll worker forgets to line out the error. If there was a mark on provisional ballots that caused them to get rejected unless counted by a scanner programmed to accept them after the voter's bona fides were established, that problem could be solved.

We have both human and machine error to deal with here. That's why I like polling place scanners--it's all checks and balances. With people and scanners crosschecking totals, error and hankypanky are minimized.

[dlaliberte]

13. Intentional undervote could be explicit

What about requiring voters to select a "No vote" candidate explicitly rather than merely neglecting to select any other candidate? If they don't make the explicit "No vote" selection, then they would get the warning about not having made a selection.

[eridani]

15. That would be what I would recommend for a standard

Too bad we don't have a national public consortium that could make use of ideas like this.

[FogerRox]

3. Excellent topic, KnR

[Contrite]

12. Okay, so it only takes about an hour per precinct to count.

So why in the hell not just count them all by hand in the first place?

Also, the hand counting revealed that people had made a clear choice but didn't follow the directions and fill in the oval. That is STILL a valid vote and would not have been found without the hand count!

Furthermore, these are significant errors in such a small sampling. How many votes were counted? Not many. Yet, still errors in the machine counts. Extrapolate out to the whole and how many votes might have been "missed" or wrongly tallied? Enough to throw a race?

[OnTheOtherHand]

17. nope, probably longer

I can't find info for Washington County, but if you click through on the Anoka link, you'll see that all four precincts were counted simultaneously -- three races in each. So it was one race per precinct per hour, or so.

About as many candidates lost votes as gained votes. There's no conclusive evidence here as to whether the hand counts were more accurate than the machine counts overall, although one would certainly hope so. However, the audit standard is that discrepancies greater than 0.5% trigger further audits. Minnesota apparently has automatic recounts if the race is within 0.5%, so that almost wipes out the possibility of throwing a race (although I think the protocol could be refined).

[Contrite]

18. I live in Minnesota and converse with our new SOS

as well as activists on this subject quite regularly. I do not agree that the audit is sufficient. If 2 million voters vote for governor and one candidate loses by 20,000 votes (which happened, and so we still have our idiot Republican governor Pawlenty instead of former AG and Democrat Mike Hatch), how is 1/2% of a discrepancy okay? In fact, I agree with Bruce O'Dell, who also lives in Minnesota, that even 10% auditing of every precinct isn't even enough to ensure accuracy--he recommends this wider audit as an "exit strategy" from counting electronically altogether. One of our representatives, Bill Hilty of Finlayson, has been active in writing the audit legislation and intends to broaden the auditing.

[OnTheOtherHand]

19. whatever

If there's a 1% margin, 1/2% is probably OK -- but I'd be fine with a more stringent audit in close races. In fact, depending on the context, I would bang the table for one myself.

What Bruce wrote about the 10% audit in his exit strategy piece was that he thinks that once the American people learn from 10% audits how much is wrong with the system, they will insist on full hand counts. I think he's wrong, but I don't see any point in arguing about it. We could just find out. However, I also don't agree with him that 10% of each precinct is an especially good way to do an audit, although it's conceivably OK if it really has the logistical advantages that the EDA report claims. The EDA audit regime actually has some real problems, which I don't think anyone has bothered to write up because no one seems prepared to introduce legislation based on it anyway. But I think there are various ways of conducting audits that could work pretty well.

[Contrite]

21. Have you read this?

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=462944&mesg_id=462944

[OnTheOtherHand]

22. sure

We've realized for some time that a 2% audit wasn't adequate to verify close House races with high confidence.

However, even at that, the folks at EDA seem to have a strange compulsion to make a 2% audit seem useless, which it certainly wouldn't be. After the election, EDA came out with a paper that suggested that the Democrats beat the Republicans by 23 points in the House total vote nationwide -- double-digit fraud nationwide (and massive exit poll error, by the way). If that were true, even a 2% audit easily would suffice to prove it.

Or suppose that the Republicans targeted fraud to steal the ten races in which the Democrats lost by the smallest margins -- all under 2%. Suppose that they could switch up to 20% of votes in any precinct at will, and assume that without fraud, the Democrats would have won each seat by the tiniest of margins (so we are positing the smallest amount of fraud that would reverse the outcome). The chances of escaping audit detection in any one of those races are better than even. But the chances of escaping audit detection in all ten would have been around 10%. Possibly higher, if fraud was targeted in large precincts -- another issue that a simple adjustment to the audit protocol could fix even without expanding the audits, although I certainly support expanding the audits.

OK, if we want to achieve 99% confidence of verifying the outcome in even the closest House races, we can do that -- and in most cases we can do it with audits substantially smaller than 100%. But even small audits of reliable paper records make it hard to get away with widespread election theft, even assuming that all other security efforts were futile.

Ah, but couldn't we close the door on all of this if we had 100% hand counts? No. Everyone knows that election theft is possible with 100% hand counts. It's not a matter of "scanners insecure, hand counts safe"; it's a matter of risk assessment across the entire system. It's hard for me to take apocalyptic rhetoric about "hand counts or the end of democracy" seriously, since we all know that most Americans haven't operated under 100% hand counts for decades.

Again, I'm not saying that 2% audits are adequate. I do say that even 2% audits -- along with the reliable paper records that they presuppose -- would shut down a lot of nightmare scenarios, and that it wouldn't be hard to shut down others. If the American people were to decide that they supported 100% hand counts for all Federal offices -- or for all offices whatsoever -- I would not stand in their way. But I see no sign that they have made either of those decisions, and I do not believe that democratic values compel them to do so. I certainly don't see how an argument against the adequacy of 2% audits in close House races constitutes an argument for 100% hand counts in all races.

[Contrite]

23. Hand counts are much more reliable

Although as you state "Everyone knows that election theft is possible with 100% hand counts", so everyone knows that election theft is possible with ANY type of voting system. As Chuck Herrin says, whenever you have humans involved you will have cheating and/or errors. Chuck, by the way, just wrote that he is totally behind hand counted paper ballots.

But does everyone also know that hand counted paper ballots have been shown to be far more accurate, less expensive, and less easy to cheat with on any sort of grand scale?

I don't believe, as you say, that EDA is saying audits are totally useless. Their argument is that we need such a large audit to ensure any kind of accuracy we may as well hand count the races in the first place. Thus, their recommendation for a 10% audit of all precincts they view as an "exit strategy".

As for how many Americans prefer hand counted paper ballots I have seen some polls that suggest the majority do, or at least that the majority do not want their votes counted on secretly coded machines manufactured by partisan companies like Diebold and ES&S.

[OnTheOtherHand]

24. actually

I am sick of bearing the burden of knocking down vague, unsubstantiated, and often fishy claims. No, everyone does not know that HCPB has been shown to be less expensive than all the alternatives. The point about accuracy seems to be oversold as well. Sorry, I'm a bit worn out on this whole topic, but if you want to be useful, you'll find plenty of points on various recent and older threads that could bear disciplined, systematic response. (EDIT TO ADD: I'm not being facetious. There are several HCPB advocates on the board who don't seem to have any interest in being useful. I hope you will not follow their lead.)

"Their argument is that we need such a large audit to ensure any kind of accuracy we may as well hand count the races in the first place." Well, much as I like Bruce O'Dell, that's nonsense. Need I say that 10% is 90% less than 100%? Whether one supports HCPB or not, I just don't see how that is much of an argument.

"As for how many Americans prefer hand counted paper ballots" -- well, go find out. We're all familiar with how a handful of poll results can be spun. Hey, here's one:

HEADLINE: Do you think a hand recount or machine recount is a more accurate way to count votes?

BODY:

QUESTION NUMBER: 011

RESULTS:
Hand count - 35%
Machine count - 58
No opinion - 7

ORGANIZATION CONDUCTING SURVEY: GALLUP ORGANIZATION
POPULATION: National adult
NUMBER OF PARTICIPANTS: 870
INTERVIEW METHOD: Telephone
SURVEY SPONSOR: Cable News Network, U.S.A. Today
BEGINNING DATE: November 19, 2000
ENDING DATE: November 19, 2000

Do I actually think this proves anything meaningful about public support for HCPB? Nope. I don't think the Zogby factoids do either. What would be really cool would be for HCPB advocates actually to get HCPB adopted in a few metropolitan counties where it isn't currently in use. We all might learn something in the process.

[Contrite]

25. Couple of points

One, several experts in the field, such as Lynn Landes, and the folks at voters unite/ballot integrity , have said that hand counting is much more reliable and less expensive. - The "central finding" of a 2001 CalTech/MIT study was that, of all voting systems used in the United States, hand counted paper ballots have the lowest average incidence of spoiled, uncounted, and unmarked ballots. As Lynn has pointed out, even hiring temporary workers for counting would cost less than the machines, and anyway, what price do you put on making sure our democratic republic is protected by fair elections?

Second, I feel that you are being rather condescending in tone re: HCPB disciples. You even diss O'Dell and Simon's argument. What is your background that you can so glibly put down such as them?

Third, Gallup? Puh-leeze.

[OnTheOtherHand]

26. OK, three points

I'm sorry, but I don't give a damn what Lynn Landes says; I want to know what she can document.

Did you read the CalTech/MIT study yourself? have you read any study on residual votes since 2001? Try this, for instance. I think hand-counted paper is OK, but that study is hardly the last word.

I put a high price on protecting our democratic republic. I simply don't accept that as a rationale for hand counts. If you are unhappy with the tone, you might consider editing your posts to remove any possible implication that people who disagree with you love democracy less than you do, and encouraging your allies to do the same. I admit that my temper is shorter than it used to be.

Anyone can "diss" O'Dell and Simon's argument(s) if they disagree with it (them), and offer reasons. I have a Ph.D. in political science, but there are plenty of people (with or without Ph.D.s) who know more about voting systems than I do. It's about facts and arguments, not credentials.

"Gallup? Puh-leeze." Gee, and you think I'm condescending?

[Contrite]

28. Residual votes? That study assumes no fraud.

And we all know that votes were switched/stolen and "lost" in 2000 and 2004 especially on electronic machines. They vaporized.

[OnTheOtherHand]

29. we all know what?

What you "know" about vote switching etc. in 2000 and 2004 may bear little resemblance to what I "know." A pile of votes were lost (or something) in New Mexico in 2004, but those should show up in the statistics, as should the Florida 2000 overvotes and undervotes. If you have solid evidence of widespread and statistically palpable anomalies on electronic machines in 2004, it would be a service to the country to present it.

Your response seems like a non sequitur regardless. You cited a study, I cited an apparently contradictory study, you... changed the subject? (I do realize that this is a complicated topic, so what feels like whiplash to me might be a gentle change of course to you.)

[Contrite]

30. Non-sequitur tying "fraud" to residual vote rates?

I think not, and apparently others agree (see discussion below). Re: vote switching--if votes were switched to Bush from Kerry then how do you account for the accuracy of machines by using the number of votes recorded vis a vis the number of ballots/voters? If there were "phantom" votes--i.e., not over or under votes--what is the residual rate then?

http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=337686&mesg_id=339615

[OnTheOtherHand]

31. well...

To review, you said: "The 'central finding' of a 2001 CalTech/MIT study was that, of all voting systems used in the United States, hand counted paper ballots have the lowest average incidence of spoiled, uncounted, and unmarked ballots." Ironically, the post to which you just linked describes the study more fairly. To quote from the 2001 study:

The central finding of this investigation is that paper ballots that are either hand counted or optically scanned have the lowest average incidence of residual votes in presidential elections and, down the ballot, in Senate and Governor elections.

(emphasis added)

To further quote the study, at page 10: "First, optically scanned ballots show the lowest average residual vote rate across all of the offices and years examined."

So you can probably see why I am starting to tap my foot over here. Your response to all of this, apparently, is to conjecture that maybe the decrease in DRE residual votes in some jurisdiction evinces some default-to-Republicans setting. That might even be true, although I haven't seen any evidence to support it. But it doesn't do much to salvage your original claim.

"Re: vote switching--if votes were switched to Bush from Kerry then how do you account for the accuracy of machines by using the number of votes recorded vis a vis the number of ballots/voters?" You don't. But we both knew that (right? you've read the 2001 study? at least by now?), and it has nothing to do with your original claim.

[Contrite]

32. My original claim was merely to support the accuracy of HCPB

and I actually mistrust the residual vote count with regard to e-voting accuracy since there's currently no way to determine if an e-voting machine recorded votes exactly as cast.

Residual votes can help determine if one type of voting system is better at helping voters cast their ballot correctly, but they can't determine if a machine recorded votes accurately or if a voter intended to leave a race blank.

Touch-screen voting machines, for example, are supposed to be programmed to prevent voters from choosing more candidates than allowed in a race and to ask voters to be sure they intend to leave a race blank before casting their ballot. But this works only if they're programmed correctly.

During Florida's primary, machines made by Election Systems and Software recorded that 134 voters cast blank ballots. Some officials speculated that Democratic voters cast blank ballots once they realized there were only Republican candidates on the ballot. But in a North Carolina county in 2002 that used the same voting system, officials discovered that touch-screen machines failed to record 436 ballots. The machines didn't just fail to record votes; they failed to record the ballots as being cast. The issue in that case turned out to be a programming problem. Officials were able to determine the problem only because voters filled out paper ballots before casting touch-screen ballots, a procedure sometimes used for early voting. This provided a paper trail indicating the voters' intent.

Accuracy is how well the process translates voter intent into properly counted votes.

Technologies get in the way of accuracy by adding steps. Each additional step means more potential errors, simply because no technology is perfect. Consider an optical-scan voting system. The voter fills in ovals on a piece of paper, which is fed into an optical-scan reader. The reader senses the filled-in ovals and tabulates the votes. This system has several steps: voter to ballot to ovals to optical reader to vote tabulator to centralized total.

At each step, errors can occur. If the ballot is confusing, then some voters will fill in the wrong ovals. If a voter doesn’t fill them in properly, or if the reader is malfunctioning, then the sensor won’t sense the ovals properly. Mistakes in tabulation -- either in the machine or when machine totals get aggregated into larger totals -- also cause errors. A manual system -- tallying the ballots by hand, and then doing it again to double-check -- is more accurate simply because there are fewer steps.

Bugs in software are commonplace, as any computer user knows. Computer programs regularly malfunction, sometimes in surprising and subtle ways. This is true for all software, including the software in computerized voting machines. For example:

In Fairfax County, VA, in 2003, a programming error in the electronic voting machines caused them to mysteriously subtract 100 votes from one particular candidates’ totals.

In San Bernardino County, CA in 2001, a programming error caused the computer to look for votes in the wrong portion of the ballot in 33 local elections, which meant that no votes registered on those ballots for that election. A recount was done by hand.

In Volusia County, FL in 2000, an electronic voting machine gave Al Gore a final vote count of negative 16,022 votes.

The 2003 election in Boone County, IA, had the electronic vote-counting equipment showing that more than 140,000 votes had been cast in the Nov. 4 municipal elections. The county has only 50,000 residents and less than half of them were eligible to vote in this election.

There are literally hundreds of similar stories.

However, I support HCPB in top ticket races and (for now) optical scanning for down ticket races with aggressive random auditing ON ELECTION NIGHT.

[OnTheOtherHand]

33. OK, let's try to recapitulate

I feel almost snarky pointing out that actually, we have no way of knowing how accurate HCPB is in practice, unless it is audited -- and even then there is room for controversy. But I imagine that it is usually pretty close.

Residual vote is not a measure of accuracy, and I did not introduce it as such. (Actually, you introduced it -- but at any rate, we agree.)

You may be right that optical-scan ballots are less likely to be voted and/or counted correctly than hand-counted ballots. However, you haven't actually demonstrated it, much less quantified the extent of the difference. In principle, precinct-based optical scanners may be capable of attaining lower error rates than hand-counted ballots, because voters can be warned of certain errors.

Yes, I do know that software bugs are commonplace; I have even written a few of them myself.

[Contrite]

34. Right

and the "proof" of anything lies inside the machines...without a full, independent audit of machines (preferably optical scan) used in a major race somewhere, anywhere, how do we really know how anything compares? We simply don't know "what's been going on" inside the machines because we haven't been allowed to look. But there is certainly a lot of "circumstantial" evidence that leads one to reasonably conclude we simply cannot trust the machines. There is nothing "mysterious" about hand counted ballots, we have enough results of these races to fairly reliably calculate the residual rate, and there is no software to investigate. However, without knowing how over/undervotes/vaporvotes/phantom votes are occurring within the machines we can't just go blindly forward using them, IMHO.

[OnTheOtherHand]

36. you won't find much disagreement on those points here

I think you may tend to scant the fact that an awful lot of elections have been stolen on hand counted paper ballots -- so "machines bad, hand counts good" doesn't cut it as a risk analysis. But DREs are plenty scary, and op scanners have their own problems, no doubt.

[Mister Ed]

20. re:why aren't the scanners programmed to spit those ballots out and give the voters another chance?

I worked as an election judge in Washington County, Minnesota. Many ballots that the scanner deemed incorrectly marked were spit back out into the hands of the voters, who were instructed to get a fresh ballot from the judges and try again.

It sounds like the mismarked ballots described in the article were those on which a voter had marked their preferences entirely outside of the ovals next to the candidates' names. I infer from this that the scanner is blind to anything outside of those ovals. It must be that if there's no mark of any kind within the oval, then the scanner presumes the voter declined to vote in that race. Perhaps steps can be taken to improve this.

While our new law requiring audits of randomly-selected precincts cannot provide ironclad assurance of clean elections, it is surely a step in the right direction. It has the potential to discourage or expose attempted fraud at either the precinct level or the central tabulator level. I hope that the scope of the audits will be widened.

[Febble]

27. I think this is a very good point:

It has the potential to discourage or expose attempted fraud at either the precinct level or the central tabulator level.

If audits are truly random, and truly unpredictable, and the chain of custody of the ballots is ensured, and if the audits are of a type that will detect erroneous counts with a high degree of accuracy (i.e. if they are audits of entire precincts or machines), and if a failed audit triggers a full hand recount, as well as an investigation and prosecutions if necessary, then that will, as you say, not only provide a mechanism to detect fraud, but a disincentive to commit it.

But those are a lot of ifs. The nature of the audit protocols will be important to get right, as will what is mandated to follow if an election fails an audit.

[Contrite]

35. Exactly and some problems I have with our audits in MN include

the fact that they are statistically rather small, they allow for a 0.5% discrepancy (I think it should be 0% discrepancy because one wrong vote indicates something is wrong), and these so-called "random" audits are based on precincts picked a full week after the election (giving enough time to "fix" things if necessary so a full recount can't happen). If you make the audits truly random, larger (like at least 10%), require zero defects to prevent a full recount, and perform them ON election night, I would feel much more comfortable.