Central Count OpScan Shown to Lose Far More Votes than Precinct Count

[Wilms]

Not Quite a Blog

New Research Forthcoming on Racial Gap in Voting Technology

Joe Hall

05/23/06

I was recently reviewing the upcoming Brennan Center report on the extensive threat modeling of voting systems that they've done. I saw this in the text (this is reprinted in this blog with the permission of the Brennan Center):

"Central Count Optical Scanners have been shown to lose far more votes than . In precincts with over 30% African American voters, for example, the lost or "residual" vote rate for Central Count Optical Scanners has been shown to be as high as 4.1% as compared with 0.9% for PCOS. The lack of over/undervote protection on Central Count Optical Scanners may be the reason for this difference."

snip/links

http://josephhall.org/nqb2/index.php/2006/05/23/kimball2006

[Stevepol]

1. An important point. Thanks for the post Wilms. Cd explain FL et al.

I noticed in the 04 race the precincts in FL where the biggest rate of difference between the registration totals and the alleged vote count was in those precincts that used the scanners, not the precincts that used the DRE's, that is, the touchscreens.

I personally think the ES&S scanners have been involved in more fraudulent vote counting than any other device, despite the fact that the paper is there for an audit. Probably the central tabulators have always been the key problem and the type of machine providing the data for these machines is more or less irrelevant.

Until audits are REQUIRED, these kinds of problems will obviously continue.

[mmarcus]

2. Do you have a link to that study?

I went to the link but it goes to a blog that quotes from the study. Where can I get the whole study?

[Wilms]

3. Don't know where you can get it now.

According to the article, the paper is yet to be released (though Joe Hall seems to have it).

Perhaps, on the paper authors website in a few weeks.

[eridani]

4. This happens for the same reason that printers sometimes--

--pick up more than one sheet. At the precinct level, voters put through one at a time, and if the ballot doesn't get counted, it gets spit back out.

[Boredtodeath]

5. Excellent point, also

Central Count scanners are NOT in operation while the voter is present. Therefore, a ballot with errors cannot be repaired/fixed/replaced. Therefore, the option which is set in the definition file, to reject an invalid ballot is set to NO.

One of the things we should ALL consider is that many of these ballots belong to our troops overseas.

[OnTheOtherHand]

6. here's a source from Kimball's web site

http://www.umsl.edu/~kimballd/rtables.pdf

See especially the tables on pp. 4 (for 2000) and 5 (for 2004).

(EDIT TO ADD) Florida is a red herring, I believe: according to VerifiedVoting.org, all the optical scanners there were precinct-based. Whether or not there is evidence for vote miscount by those scanners, it could not be attributed to higher residual rates on central count scanners.

[Boredtodeath]

7. One other thing for consideration

Not all "precinct based" OpScans are set to reject a ballot which is unreadable.

For instance, in Georgia in 2000, none of the precinct based optical scanners had this option turned "on" in the ballot definition files. 53% of Georgia's voters were using opscan in 2000.

edit: grammar

2nd edit:
The above should be corrected to say "a majority" of the precinct based optical scanners did not have this option turned "on" in 2000.

In her The 2000 Election: A Wake-Up Call For Reform and Change Report to the Governor and Members of the General Assembly, Secretary of State Cox says:

Although optical scan systems, the newest technology used in Georgia, offer satisfactory performance in some counties, in many other locations optical scan undervote rates are extremely high…well above the averages of more antiquated systems. In fact, 21 counties that employ optical scan technology had undervote rates of 5 percent or higher. And, the mean average (the average of all the county percentages) of optical scan county undervotes is nearly identical to the now disparaged punch card system.

She neglects to tell us that 13 of the 21 counties employed her now implemented Central Count system where this occurs. Also not mentioned is the footnote in her document at page 7, which says:

Precinct count readers or tabulators have the capacity to reject defective ballots, but must be programmed to do so (O.C.G.A. 21-2-484). Although precinct readers can be programmed to reject overvotes so the voter can correct their mistake, counties frequently choose not to employ this capability.
Source: http://www.sos.state.ga.us/acrobat/elections/2000_election_report.pdf

http://www.countthevote.org/error_rates.htm

[OnTheOtherHand]

8. in Kimball and Kropf 2006

http://www.umsl.edu/~kimballd/kkspsa06.pdf

footnote 4 on page 11 indicates that their analysis there actually treats PBOS with rejection turned off as central-count -- but I don't know whether that is true in the tables I linked to above. (I'm also not sure about Florida in 2004.)

[Boredtodeath]

9. It's a good question for a Florida voter to ask

In an open records request.

But, clearly, if you want to "fabricate" bad numbers for an optical scanner, this is the easy way to do it.

[JimDandy]

11. Absolutely right. Just one of many ways to game the system.

Utah conducted a mock e lection in 2005, prior to selecting the statewide D bold T S x system. The D bold o p scanner I used there was programmed to catch only overvotes, not undervotes. If the state was using the mock elec tions to compare machine error rates as part of their criteria for choosing a system, well then, we were skewed. :think:

Three weeks later, the D bold sales rep demonstrated the o p scan system to state legislators up at the capitol. When I questioned her about their machine not catching undervotes, she said that it could be programmed to catch them, but that, if I was an election official, I wouldn't want to do that because it would slow down the voting process!!!

D bold really cares about good election results, don't they.

[Bill Bored]

15. Except that HAVA requires overvote notification.

And in fact, HAVA can be interpreted to require the voter to be told exactly which RACE(S) on the ballot were overvoted, which Diebold's software does not support. So Diebold's scanners may be in violation of HAVA.

Undervotes on the the other hand are fine under HAVA (go figure) and even Diebold has separate settings for overvote and undervote rejection, so it's really easy to make scanners perform "worse" than DREs. That said, they actually perform better on average!

[Boredtodeath]

16. Uh, not really - read post #14 below

Sorry.

[Bill Bored]

19. You are citing a voting system standard -- that's not HAVA, the law.

Section 301 of the law only requires overvote notification

[JimDandy]

10. You'd have to turn off the undervote programing on DREs in order

to get a fair comparison of percentages of residual votes. Under such conditions no voters (DRE, CCOS, or PCOS users) would be given notice of their undervote error/choice, nor the opportunity to correct it. This type of study might not be possible now, though, due to HAVA.

[Boredtodeath]

12. Diebold machines cannot be programmed for UNDERVOTES

Only for ballots which cannot be read properly (i.e., stray marks, etc.) and overvotes.

[JimDandy]

13. The D bold rep termed it as "catching undervotes"

If the PCOSs can't be programmed to "catch undervotes" how are they complying with HAVA?

[Boredtodeath]

14. See #3 below

2.4.3.2.2 Precinct Count Paper-Based Systems

In addition to the above requirements, all paper-based precinct count systems shall:

1. Provide feedback to the voter that identifies specific contests or ballot issues for which an overvote or undervote is detected;
2. Allow the voter, at the voter’s choice, to vote a new ballot or submit the ballot ‘as is’ without correction; and
3. Allow an authorized election official to turn off the capabilities defined in ‘a’ and ‘b’ above.
-------------------------------
Also, on my post #12 above, I was referring to OpScans, not DREs. (I did not make that clear)

Diebold DREs do display an undervote in bright red on the Summary Screen before the voter hits "Cast Ballot."

[JimDandy]

20. Thanks for this!

Yes, in the first post I was referring to the programing that creates the red highlighting on the DREs and in the last to the PCOS's on/off capabilities.

WARNING: the following is simply to bounce around an idea.

# 3 should be eliminated. If it is not, this option to turn off the capability to detect undervotes on PCOS machines could give added weight, in future court cases, to the inequality arguments already raised in Bush v. Gore and Stewart v. Blackwell -- arguments which favor DREs -- UNLESS the DRE's capability to detect undervotes can also be turned off, as well.

BTD do you know of any section that states that, when used in elections, DREs must provide/retain the ability to catch undervotes (red highlighting) or is this not addressed at all?

If not, who is to say that a jurisdiction will retain that capability. Obviously catching undervotes slows down the voting process (as Utah's Dieb old sales rep stated) no matter what kind of machine is used to catch them: DRE or PCOS, and therefore election officials should also be motivated to eliminate that capability on their DREs, and not just on their PCOSs.

If all of the above is true, then, in any future litigation involving equality of the voting process, wouldn't any "it catches undervotes" argument be a wash for both DREs and PCOSs?

[Bill Bored]

18. There is NO undervote notification requirement in HAVA. nt

[JimDandy]

21. Thanks BB. n/t

[Bill Bored]

17. GEMS has checkboxes for OS reject settings: Unvervote, BlankVote, Overvote

Unless there is evidence to the contrary, this would seem to allow either over- or under- vote ballot rejects. Blank voted races is Diebold's term for an undervote. Undervote is their term for a vote-for >1 race with less than the number of selections allowed.

But none of these notifications specifies any particular race(s) to the voter.

(I know GA 2004 Primary DREs had different undervote reports and patterns than in the 2004 General and this is still an open research question.)

(Edited about a dozen times for clarity, I hope.)

[JimDandy]

22. Excellent explanation...

Can any combination of those boxes be checked, or unchecked?

Utah's 2005 Gen. Election unofficial "Election Summary Report" combined both the 'Under voted' and 'Blank voted' totals on a line titled 'Number Of Under Votes' and threw both the DRE votes (Polling votes) and Absentee votes into a category called 'Total Votes.'

Combining vote types like that makes it impossible to attribute votes to their source. So, for the official canvass, I asked the county's IT manager if he could produce a "Statement of Votes Cast" report that broke down the 'Number Of Under Votes' into separate 'Blank voted' and 'Under voted' categories, for each race, and each type of ballot cast(Polling, Absentee and Provisional.) He had to get instructions from Dieb old, on how to generate a report like that, and it took him four extra hours to do it, but he did!

(edited to remove link to post that is no longer relavent, because neither HAVA law nor any standards require voters be notified of undervotes. You're posting faster than I am!)

[Bill Bored]

23. Yes, any combination, at least for Op Scan.

And they are all turned off by default, i.e., no ballots will be rejected until one or more boxes is checked.

[Wilms]

24. Nice job on this thread, BB, BTD, JD. Thanks. n/t

[Land Shark]

25. This type of research was central to the Stewart v Blackwell opinion

that ruled central count opscans and punchcards unconstitutional while upholding (for now) DREs and precinct count opscans. Paper ballots were not litigated but the use of the 'residual votes' test might well have caused paper ballots hand counted to be declared unconstitutional -- that would depend on the numbers in that particular state.

The courts seem to be fine with ignoring human elements like allocation of machines, training, deliberate or negligently slow response times when questions come up, old machines to poorer precincts with less political clout, etc. All of these tend to increase residual votes for minorities especially, but don't really show that the technology PER SE is the problem.

In fact, it wouldn't be hard to imagine a set of facts whereby PAST INTENTIONAL DISCRIMINATION based on machine allocation for example and how settings are tweaked on machines, is then used to outlaw those machines and bring in DREs as a "solution" and a "low residual vote" technology, and a seeming godsend for minorities after decades of disparate abuse they've suffered. Wonderful. This ignores that with DREs you don't need to mess with or suppress individual voters anymore, you can just shave 1/25th of a vote off of each voter....