ERIE COUNTY NY: The problem wasn't the memory cards -- just the software that READS them!

[Bill Bored]

That's what the Erie County (Buffalo), NY Democratic Election Commissioner Dennis Ward said to try to convince his fellow Democrats NOT to seek a hand count of the paper ballots in the NY Senate District 60 race that had a margin of only 597 votes.

As a fellow Democrat, I feel much better now!:bounce:
How about the rest of you? Read more and...:puke:

New wrinkle in Senate recount
By Phil Fairbanks

The prospect of a hand count of every single ballot in the 60th State Senate race arose Tuesday after Democrats questioned the results of Erie County’s new voting machines.

County elections officials dismissed the challenge as a “red herring.”

-snip-

“A red herring,” said Dennis Ward, the county’s Democratic elections commissioner, when asked about challenges to the machine results.

Ward said the problem was not the memory sticks, which contain the machine results, but rather the software that reads the sticks.

“The data was never corrupted,” he said Tuesday. “There’s nothing wrong with any of the sticks.”

The problem with the software, Ward said, was fixed by the software manufacturer (ES&S), a remedy Senate Democrats were quick to criticize.

-snip-

Read the whole story and:puke:some more!:
http://www.buffalonews.com/city/communities/erie-county/article255921.ece

[Dragonfli]

1. I live in Buffalo, looks like voting was a waste of my time, soon NY will go RED

Just as planned....

[Bill Bored]

3. Voting is not a waste of time. There's a chance our votes may be counted as cast - just no proof. nt

[Dragonfli]

4. Without proof, money gets to win. They are only kept honest by proof

[Bill Bored]

8. Don't get me wrong. Proof is good! LOVE THE PROOF! But if you don't vote...

...it only proves that your vote won't be counted. That's bad.

Otherwise I agree with your comments and I think TheWraith doesn't get that we want to know who really won the election regardless of how sleazy they may be!

[TheWraith]

5. Paranoid much?

You did read the part about how there are paper ballots for recounts and auditing, right? Ward is saying that a full recount is a waste of time, since he thinks the results they've pulled are accurate. Based on the recounts they're already doing in the Niagara County portion of the district, he's probably right.

More to the point, this is a Democrat we don't want reelected. Antoine Thompson is high-grade scum.

[Dragonfli]

7. I'm a programmer, is knowledge paranoia? I know what can be done with the software.

You need to learn a little more about it.

[TheWraith]

11. Thanks, but I've been working with computers for about 20 years now.

I know what can be done with software too. But I also understand statistics, and the fact that anything wrong in the software count is going to be reflected in the 3% recanvass. Demanding a complete recount before you even know if there's a reason to do it is unreasonable.

[Bill Bored]

13. Silly assumption: Anything wrong in the software count will show up in the 3% canvass. nt

[TheWraith]

15. Silly assumption: that 2 plus 2 equals purple.

Which is roughly the equivalent of thinking that with 3% of cast votes you can't accurately judge whether there's a statistical difference between the machine count and a hand count.

[Bill Bored]

21. The effects of cluster sampling would invalidate what I think you're trying to say.

But it's worse than that:

The effectiveness of the audit depends on the NUMBER of units audited -- not the percentage. See post #20:
http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=203&topic_id=516847&mesg_id=516871

[WillYourVoteBCounted]

14. its about Public Confidence in Outcome of Election

with a new voting system, it would behoove officials to do a full recount.

Look at Carteret County NC, famous for losing 4,400 votes permanently in 2004.
In 2006, Carteret used their new optical scanners in the Nov GE. One contest was
too close to call.

Instead of doing a machine recount and or a 3% sample count to possibly expand,
they all agreed that for the sake of non partisanship, they'd do a 100% hand to eye count.

Because democracy depends on the public's trust in the election results, and the
best way to show the results are trustworthy is a hand to eye count.

Given all of the problems with NY's voting machines, and that they are new,
and this is a drastic change for NY from mechanical to computer counting,
its smart to hand count.

Columbia County NY did a full hand to eye count.

[Bill Bored]

9. Paranoid NOT! "Paper ballots for recounts and auditing" = bait and switch!

So show me the recounts!

Look, the voters and candidates of New York did NOT have to be subjected to the burden, expense and uncertainty of software-based vote counting. But the State let the Bush (and now Obama) Dept. of Justice talk 'em into it. They agreed to it in federal court, despite all the evidence that it's not safe and that auditing the system to achieve high confidence in election outcomes (never mind actually counting every vote) can be very onerous and complex -- unless all our elections happen to be landslides.

Well, now that that foolish decision is being implemented, the State damn well better audit and recount enough of the vote to prove who's winning and losing our elections -- don't ya think? Otherwise, we're not much better off than Florida where they don't hand count their paper ballots either.

County election officials like Ward are on the record opposing election audits greater than 3%. Do you really think that's enough to know who won all our elections? If not, which ones do you think are worth having confidence in and which not? How should our confidence in our electoral process be rationed -- and who decides this?:think:

[TheWraith]

10. Okay, how much knowledge do you have about statistics and the Board of Elections?

Because I feel I should mention that I was a poll watcher on election night, as well as being neck deep in the Wyoming County Board of Elections, watching and participating in the entire thing, and that I've also spent many hours talking with state BOE people and running the numbers on various elections. I know what I'm talking about.

Statistically speaking, recounting 3% of the vote should be plenty to give you an accurate picture of whether or not the total count is accurate, at least in larger races, say 30k votes and up. If you have a 3% recount that doesn't vary in any statistically significant way from the documented count, recounting every single ballot isn't likely to give you one either. That's the entire point of a 3% recount: to verify the machine count without having to take months to go through every single ballot.

Look, in modern elections, there is going to be NO 100% guaranteed solution that is also realistic. As much as I love and miss our old mechanical voting machines, those produced absolutely NO paper trail, so you had only to trust in the machines that they were accurate. The only solution where machinery or computers don't come into the picture at all is if you hand-count every ballot, which is completely impractical in any election larger than School Board.

[Bill Bored]

12. Thanks for being a poll watcher but the 3% audit is bullshit. Read and learn:

Thanks for being a poll watcher. But you can't tell what a computer is doing by watching it. You have to hand count the ballots.

Look at these references:
http://www.electionaudits.org/node/14

Some of them are a bit too wishy-washy for my taste, but the one called "New York Audit Graphs" seems to address the 3% problem.

[TheWraith]

16. Your link is irrelevant.

The 3% audits work on the same principle as polling to mathematically determine whether the machine count is reliable. For instance: Suppose you have a state senate race with 100,000 votes cast. A 3% audit would constitute 3,000 votes. If you randomly choose 10 precincts that add up to 3000, and each of those precincts matches exactly the vote count that was given by their machine, then you have extremely strong evidence that the vote count is accurate. If each precinct has 5 votes which were misread, then that would indicate a race-wide margin of error of about 1,666 votes. If the margin of victory was 30,000 votes, then that error is irrelevant. If it was 3,000 votes, then you would have an indication that you should run a complete hand count.

Saying that the 3% audit has no validity is like claiming that election polls are bad because they didn't call EVERY SINGLE PERSON IN THE COUNTRY.

[BeFree]

17. 3% Audits are ok

10% is even better. In fact, that's what the experts all say. 10% or bust.

So what's the hold up? Why don't the election officials always do a 10% audit?

[OnTheOtherHand]

19. that isn't "what the experts all say" at all

You just might be the only person ever to say that. I'm not sure. But it takes some mad skilled chutzpah to try to pass it off as an expert consensus.

Sometimes 10% would be far larger than necessary to provide very strong evidence of who won; sometimes it isn't nearly large enough. That's pretty obvious, even to the experts.

But as to your question: election officials don't always do a 10% audit because most of them aren't looking for more work to do. No mystery there.

[BeFree]

22. well

In the real world, absent the officials who don't give a crap that the machines are bad for democracy, a 10% audit would pretty well give a 90% chance of catching errors.

And thanks for the reply so that I can say that a 10% audit of each and every machine is what I was getting at, and the reread has brought to my attention my mis-communication. Cheerio!

[Bill Bored]

24. A 10% audit of each and every machine...

...might not find problems with any particular machine, but might find problems with the overall election results. But to do that, you have to aggregate all the audit results and compare the total to the election results. And you have to make sure that each 10% sample is truly random. Not as easy as it seems, but it might be possible.

Also, 10% would not be enough for smaller elections with small victory margins like the one we're discussing in this thread. As with whole-precinct audits, effectiveness depends on the number of units audited -- not the percentage. Small elections have fewer ballots. 10% of them is a smaller audit. That doesn't provide a small enough margin of error to tell who won.

[BeFree]

25. 90% confidence

Even hand counts can be finagled. But a strict 10% audit of every machine, with any deviation above
1% requiring a complete hand count, would give us a 90% chance of an accurate election.

It's what is done when an above 1% error is found. I am familiar with an audit that was off by 4%
and the election officials just said: "So what?"

They found an error and went on their merry way.!!!

All in all, after a couple of elections where the audits are substantial and correctly done, it
will force the machines to comply with actions which will lead to even better original counts.

There should be a 'Lemon Law' passed which makes the machines improve or be put out of our misery.

[Bill Bored]

26. Your math is not right for all elections because 10% is not right for all elections.

But OK, let's say it would work for LOTS of elections.

Suppose the hand count differs from the machines by 1% but the margin of victory is 10%, or 20%? Would you hand count 100% in that case too?

And suppose the margin of victory is 0.5% or less? Would you still hand-count only 10% of the ballots? What about 0.6%?

The larger problem here is that those in authority want to trust the machines and they expect the rest of us to follow suit. Once that feeling is overcome, the details of the various audit protocols can be discussed with them intelligently. But not before then.

[BeFree]

28. well Bill

Where we are now is a deep hole.

What I am getting at is a new beginning. A new law. A counter measure.

If we can get a law that makes for a 10% audit of every machine, we might stand a chance of accuracy.

What we have now is nothing. And all this jibber-jabber just makes for a confusing situation.
Just look at this discourse here amongst us. We three have been at it for sometime and I'll bet a newcomer reading here is thinking: WTF?

Experts have said that a 10% audit of each and every machine count would give us all a greater confidence. It is simple, and it is doable in every voting place in the country.

[Bill Bored]

31. It's not simple and not doable. If it were, someone would do it.

To date, there hasn't even been a pilot of the audit you're advocating, and besides, it's very susceptible to fraud by faulty aggregation of either the machine tallies or the audit tallies. In theory, I agree it's OK though. I wouldn't be against it in principle, but it's not going to happen. Neither are risk-limiting audits based on machines or precincts, at least not on a large scale.

Like I said, the PTB want to trust the machines they've spent millions of our taxpayer dollars on, and that's the story they're sticking to.

We've got crucial elections here in New York right now that would have been decided by non-computerized lever machines, and hand counts of absentee ballots. Instead, so far we've got lots of software-based results, hand counts of 3% of the computers, and in the closet races, hand counts of the absentee ballots that account for about 5% of the vote. If that's how we're determining our election results, clearly the lever and hand-count system would have been far superior.

The courts may decide that 3% of scanners is a good enough hand count because the idiots in the NY Legislature decided that it would be enough a few years ago.

There is only one thing that is likely to save us and that's the state constitution. If that fails, or no one uses it in court, we're done for.

[OnTheOtherHand]

27. OK, now I know what you're talking about

Nothing like an expert consensus on that, but at least you're not the only person to suggest it! EDA put out a paper about that.

One problem with that approach is that you can't determine whether any machine actually counted its votes accurately, although you may be able to identify machines that weren't even close. If the 10% sample can be treated as random, that approach can provide pretty good evidence about who won -- although, like all fixed-percentage audits, it sometimes does more work than needed and sometimes not enough -- but it's very hard to detect small errors, much less isolate them to individual machines.

If we get to the point where we can audit individual ballots against how the machines counted them, then a random sample of individual ballots can work very well for checking the winner. But that requires us to have a way to identify individual ballots without compromising secrecy, which is a controversial subject. And it provides even less information about whether any one machine counted accurately, although at least it is capable of detecting "small" errors.

[OnTheOtherHand]

18. actually, think about this some more

If you randomly chose 10 precincts, then you basically have a sample size of 10. (I could throw in some highfalutin statistical terminology, if that helps.) If any miscounts are more or less evenly distributed, that might suffice. Otherwise, a bunch of precincts could be quite messed up, and a sample of 10 could easily miss them all.

This isn't to say that the 3% audit "has no validity," just that it can't really provide "extremely strong evidence" supporting the apparent outcome in such a close election. The count may well have been very accurate, but a sample that small can't prove it.

[Bill Bored]

20. You are not very well informed. Try this:

3% of 100 scanners = 3 scanners. The chances of finding 1 bad scanner in such an audit using a simple random sample would be only 3%. If 1 bad scanner could change the outcome of an election, as with a very close race like the one in SD 60, there would only be a 3% chance of finding it with a 3% audit. There would very likely be no evidence on which to base expanding the audit to a full recount, even if the wrong winner was declared.

In the average NYS Senate district, there might be about 130 scanners. (The exact number depends on how the scanners are allocated amongst the Election Districts.) The chance that a 3% audit of 4 scanners would find one bad scanner out of 130 would be slightly less than 3.1%.

To get to a 90% chance of detecting fraud or error due to 1 bad scanner, you'd need to audit 117 out of 130. Might as well go for the full hand count and be 100% sure who won this eace with the leadership of the whole State Senate at stake. Otherwise, we'd be much better off with lever voting machines.

[TheWraith]

29. You're the one who clearly doesn't understand how a 3% audit works.

You don't audit scanners, you audit the BALLOTS. 3% of, say, a state legislature election with 100,000 votes is 3,000 votes. If those votes match up exactly with how the scanners say they read, then the likelihood is extremely low that

Your "one bad scanner" theory doesn't pass the smell test. For starters, one scanner displaying radically different results than others in similar areas would trigger obvious scrutiny and a manual count anyway. If all the other machines at a given polling place go 55/45 for one candidate, and machine X goes 99 to 1 for the other, that's considered suspicious behavior.

For errors to affect the outcome of an election WITHOUT one machine being obviously out of whack, it would require many machines not counting properly. That's what the 3% audit is to test for--it indicates whether the machines are operating properly and whether a larger recount is needed.

Lastly, there's nothing about the new system that suggests your "one bad machine" theory is any more likely now than before. The old lever machines could just as easily provide a bad count. Only difference is, now there's a way to check them. While I'm not wild about the new machines, basing arguments about them on specious information and possibly deliberate misunderstandings of statistical probability is not the way to go.

[Bill Bored]

30. Dude, sorry to break this news to you but NY Election Law says to audit 3% of SCANNERS! nt

[Bill Bored]

32. Crickets. nt

[WillYourVoteBCounted]

33. definitely NY needs better audits if using computer vote counting

not saying my state's all that much better.

It will take more than a 3% audit of machines to provide reasonable confidence
in the outcome of elections.

[Bill Bored]

34. As I understand it, NC does full hand counts for the closest races.

Do you know what it takes to get this to happen, and how often it actually does?

[WillYourVoteBCounted]

35. there's an IF in here

First NC does a machine count if there is a close contest.

A candidate CAN request a "sample" hand to eye count. This is done by a
random (I think 3%). if problems are detected it can be expanded to full blown recount.

I don't think this is hard to get, there were full blown hand to eye counts in 2006 in
the NC 08 Congressional contest.

I do not think we had any full hand to eye counts this year.
Just machine.

Sorry. Its part of the law we didn't address in 2005.

[Bill Bored]

36. It seems hard to get if it relies on finding problems in a 3% sample.

That's not working here in NY for our close races, as predicted. And I'm not convinced our 3% sample is truly random either because the election law that requires a manual process for random selection has been violated.

Frankly I don't know what these so-called "election lawyers" are doing for a living these days if they're not investigating all this, or if they're using the results of these 3% audits to decide whether to count every vote in close races.

[Bill Bored]

23. More links (this time from the American Statistical Association):

http://www.amstat.org/outreach/electionauditingresources.cfm

Scroll down to the section entitled "Election Auditing-Methods."

Come back when you've had a chance to digest some of this material.

[WillYourVoteBCounted]

6. low turnout a really bad idea. Consider Bell, California

the turnout was microscopic, which allowed super crooked people to commit major
absentee ballot fraud. It is easier to rig elections that have really low turnout.

Calif. probe spotlights absentee ballot abuse
From Msnbc.com's Tom Curry
Do absentee ballots increase the potential for vote fraud and corruption?
A case that might help shed light on that question is now unfolding in California’s Bell city, a low-income Latino municipality of 37,000 where fewer than 400 people voted in a 2005 election that opened the way for city officials to arrange huge increases in their own compensation.
...
http://firstread.msnbc.msn.com/_news/2010/08/09/4851054-calif-probe-spotlights-absentee-ballot-abuse-


California's Scandal-Plagued City of Bell May Repay $22 Million in Bonds
By Christopher Palmeri - Nov 11, 2010 12:01 AM ET
...
Bell’s former city manager, Robert Rizzo, and two other officials resigned July 22, a week after the Los Angeles Times reported that they were paid a combined $1.6 million annually. One-in-four Bell residents lives below the poverty line, according the website City-Data.com. The city’s annual report listed per-capita income at $24,800 in 2008. Statewide, the figure was almost $42,700 that year, Bloomberg data show.

Rizzo, Mayor Oscar Hernandez and six other current and former city officials were arrested Sept. 21 on charges they misappropriated $5.5 million of city money, paying themselves through phantom committee meetings and illegal personal loans, according to a press release from Los Angeles County District Attorney Steve Cooley.
...
http://www.bloomberg.com/news/2010-11-11/california-city-of-bell-may-pay-22-million-to-bondholders-in-rare-move.html

[WillYourVoteBCounted]

2. oh oh!

if ONLY someone would blog about this.