VVPB question- how to remedy discrepancy between a DRE tally and the VVPB

[FogerRox]

Ok, its election night. Voting is done and all the Voter Verified Paper Ballots have been counted, if there is any discrepancy between a DRE tally and the VVPB tally, what should be done?

1)Expand the VVPB audit to 10% in the effected race?

2)Audit all DREs in that voting district?

3)Audit all DREs in that county?

4)For the effected race: count only the paper?

5)Expand the VVPB audit to 10% in the effected race, limited to the district the discrepancy occured?

If upon one of these procedures being completed, a larger discrepancy is found, what do we do?

Please keep your answers practical. The VVPB law in both NJ & Vermont has no mandatory audit, Activists are hoping to have meaningful input in writing these bills in NJ & Vermont.

[Wilms]

1. Since when are the VVPB's counted as a matter of course?

Other than an audit, which I'd presume would not include "all the Voter Verified Paper Ballots", and a close count between contestants, where law MAY allow recounting VVPB, you might not ever count paper.

To take a crack at the question anyway, I'd say count only the paper.

[FogerRox]

2. This is an Audit question- assume a 3% audit has been done

you want to count the paper-- where?-- count the paper in the effected race ? is that what you mean?

[Wilms]

3. Given the discrepancy,

I'd want to count ALL the paper from ALL of the races that were run on that type of machine.

Don't know how politically practical that is, but I have no way of knowing what caused such an error, either. So I've No Confidence in that elections electronic administration.

[msongs]

4. the paper should be ballot of record in ALL voting counts works

in canada so I hear and many other places around the world except at diebold HQ and the
republican national committee HQ

Msongs
www.msongs.com

[FogerRox]

5. upon further review--

If one DRE tally doesnt match the VVPB tally-
In the affected race/s

I see 2 options-
1)A further audit of 10% of DREs from the Affected race/s.
2)100% audit of DREs from the Affected race/s.

I think that a further 10% audit is politically acceptable. The problem is who makes the decision as to what 10% of DREs get the additional audit.

[Allyoop]

6. Vote verification

Sorry - I posted this in the wrong place.

www.NCvoter.net has some very good information. Look under Site Map for bill #S223. We did a lot of hard work trying to protect our state from "Diebold elections". Still fighting, but this bill is in place though frequently threatened.

[FogerRox]

7. found it--

Provide for a sample hand‑to‑eye count of the paper ballots or paper records of a statewide ballot item in every county. The presidential ballot item shall be the subject of the sampling in a presidential election. If there is no statewide ballot item, the State Board shall provide a process for selecting district or local ballot items to adequately sample the electorate. The sample chosen by the State Board shall be of full precincts, full counts of absentee ballots, and full counts of one‑stop early voting sites. The size of the sample of each category shall be chosen to produce a statistically significant result and shall be chosen after consultation with a statistician. The actual units shall be chosen at random. In the event of a material discrepancy between the electronic or mechanical count and a hand‑to‑eye count, the hand‑to‑eye count shall control, except where paper ballots or records have been lost or destroyed or where there is another reasonable basis to conclude that the hand‑to‑eye count is not the true count. If the discrepancy between the hand‑to‑eye count and the mechanical or electronic count is significant, a complete hand‑to‑eye count shall be conducted.

"material discrepancy "

is not defined.

"The size of the sample of each category shall be chosen to produce a statistically significant result".

This is weak, no % is mandated.

"If the discrepancy between the hand‑to‑eye count and the mechanical or electronic count is significant, a complete hand‑to‑eye count shall be conducted".

This is good, calling for a 100% hand count

[OnTheOtherHand]

13. hmm... the legislation is DRE-specific?

It makes a difference (perhaps) because while there are benign reasons for op-scan hand counts to differ from machine counts, a discrepancy in a DRE VVPB hand count must represent (1) counting error, (2) lost VVPB, or (3) actual fraud. (Maybe there is a (4) that I am not thinking of right now.) Some small benign counting error is probably to be expected, but I would hope that proper procedures would reduce it to a smallish fraction of a percent.

I don't know of any figures about the expected error rate in hand-counting machine-generated paper. I think there should be some tolerance range (on the order of 0.1-0.2%?) within which a discrepancy wouldn't have to trigger a full count unless the race was very close (0.5%?). So, what if the race is decided by 3% and the discrepancy is 0.4%? Well, at this point I personally would at least want to see a wider recount (10%?) if only to learn more about the error rates to be expected. We could get really subtle about the relationship between discrepancy rates and the likelihood of altering the outcome, but I think that is premature if we don't know what error rates to expect. And they should be low. I just don't know how low.

(EDIT TO ADD: I contradicted myself, because I am still 'thinking out loud.' My intuition is that if the discrepancy is more than maybe one quarter of the win margin, you at least need a much wider recount -- and you need the authority to widen it as far as necessary to be sure that indeed the discrepancy is smaller. Above some proportion of discrepancy, there should be a full recount regardless of margin, because you need to be able to learn as much as possible about what the heck is going wrong. What's the smallest race to which this rule could apply, and the largest?)

[diva77]

8. Check out the Gold Standard Audit Protocol draft (GSAP) at

http://www.califelectprotect.net/home.html

(click on "Gold Star Audit" on home page to download the pdf -- it's a bit long, but extremely thorough as well as being a work in progress)

[FogerRox]

9. In many states- when the tallies dont jive-- they have a 100% hand count

the GOld Star spec's dont-- Audit goes up by 10%--

Since Gold Star starts with a 5% Audit-- a 10% increase in Audit % brings the next Audit to 5.5%. That frankly is a waste of time-- If the first audit found problem--- Count everything, as many states VVPB law mandates.

..... shall require the State to repeat this Gold
Star Audit protocol; however, each time the sample size shall increase
by 10%.

I think that is not the intent here-- should be written to eliminate any repeat of Audits. Just hand count everything in that race. A lot of states do this, so there is no reason Cali cant.

[diva77]

10. As far as I'm concerned, your conclusion/suggestion is very

reasonable -- I hope others draw the same conclusion after reading the entire GSAP document...

[diva77]

11. Which states hand count everything?

or would it be specific counties in certain states??

[FogerRox]

12. some states--if there is a discrepancy

the effected race is 100% hand counted VVPB.

[Wilms]

14. Glad to know that. It seems logical.

A problem with tallying anywhere leaves open the possibility, if only due to (or especially because of) Ballot Definition Settings error/malfeasance, that the problem could involve every machine of that type used in the election.