California Recall Analysis

[Dave Farber <dave () farber net>]

> Delivered-To: dfarber+ () ux13 sp cs cmu edu
> Date: Thu, 09 Oct 2003 19:41:48 -0400
> From: "Dr. R. Mercuri" <notable () mindspring com>
> Subject: California Recall Analysis
> To: Dave Farber <dave () farber net>, Mercuri Rebecca <mercuri () acm org>
>
>
> Dave --
>
> I'm sending you the same material I sent Peter regarding my California
> recall analysis.  He's running it in Risks and I'm hoping you'll run it in
> your e-newsletter as well.  The results were rather shocking!  Los Angeles
> voters skipped the recall question at a rate of 8.91% of their ballots cast.
> Of course they were using the WORST possible system, the infamous Votomatic
> punchcard.  On the other hand, the Datavote punchcard system had only a
> 1.94% skip rate, which was among the top 3 systems used in the state --
> beating all but 1 of the optically scanned and 1 of the touchscreens!
> Clearly SOME chad has taken a bum rap!!!
>
> More details below. If you'd like a full spreadsheet with a breakdown by
> county (not for publication but for your reference), let me know. Feel free
> to publish the material below.
>
> Good regards,
> Rebecca Mercuri.
>
> ****************FOR IMMEDIATE RELEASE********************
>
> Rebecca Mercuri has analyzed California's recall ballot data and reports
> that it confirms numerous doubts about election systems.  Her results
> demonstrate that the style of voting system in use (punchcard, optically
> scanned, or touchscreen) cannot be generically considered either "good or
> bad."  She asserts that "The particular model of the system, as well as the
> procedural controls in place in each county, along with the ballot layout,
> may have considerably more impact on the reliability of the election results
> than the type of system deployed."
>
> The analysis revealed some shocking details.  Of the 8,359,168 votes cast
> statewide, some 384,427 (nearly 4.6%) were not recorded for the recall
> question.  Almost half of these missing votes (over 175,000) were in Los
> Angeles, nearly 9% for that county.  Yet the Datavote punchcards used in 14
> other counties fared somewhat better, on average, than all of the optically
> scanned and touchscreen systems, with the exception of only the ES&S Optech
> Eagle (used in San Francisco and San Mateo counties) and the Diebold
> Accu-Vote-TS (used in Alameda, though with some reports of equipment
> malfunctions).  The Sequoia Edge touchscreens, currently under litigation in
> Riverside County, performed slightly worse than the Datavote punchcards.
> The ES&S iVotronic touchscreens were ranked lowest of the three touchscreen
> types in the state, and were outperformed by all other systems with the
> exception of the Sequoia Optech optically scanned systems and the Pollstar
> and Votomatic punchcards.
>
> In earlier court battles prior to the recall election, the ACLU claimed that
> voters using punchcards would be unfairly disenfranchised, as compared to
> voters using optically scanned or touchscreen systems.  As it turns out, the
> counties using Datavote punchcards had residual vote rates that were better
> than all but one of the optically scanned systems, and also lower than two
> of the three touchscreen systems.  At the other end of the scale, the
> counties using Pollstar and Votomatic punchcards (which included
> heavily-populated Los Angeles) had worse residual vote rates than any other
> type of voting system in use in the state.  "Clearly it is not the
> punchcards themselves that are to blame, since the Datavote systems
> demonstrate that punchcards can be used successfully."
>
> The residual vote technique was previously used by MIT/Caltech in their
> studies following the 2000 Presidential Election.  For the California
> analysis, she performed her calculations by comparing the difference between
> the total number of ballots cast, as reported by California Secretary of
> State Kevin Shelley's office, with the total numbers of "yes" and "no" votes
> on the recall question.  It should be noted that the residual vote tally is
> incapable of differentiating between a voter who deliberately or
> accidentally did not make a selection on the recall question, and an
> equipment failure (such as hanging chad) that could result in a cast vote
> not being counted.
>
> "The rush to fully computerized ballot casting is misguided.  Although
> supplemental technologies are needed for disabled voters, there is no clear
> evidence that touchscreen systems are substantially or consistently better
> for use by the general population than other voting methods.  The fact that
> the touchscreens in California do not provide any way to perform an
> independent recount, should make them less desirable than the paper-based
> systems that do have such capabilities.  Counties, like San Francisco, that
> are doing well with optically scanned ballots, and the smaller ones that use
> punchcards effectively, should feel no pressure to modernize."
>
> Numbers represent residual vote rate as percentage of
> total votes cast on type or model of machine.
>
> Punchcard        6.24
>      Datavote     1.94
>      Pollstar       6.02
>      Votomatic   8.17
>
> Optically Scanned                 2.68
>     ES&S Eagle                     1.87
>     Diebold Accu-Vote-OS      2.36
>     ES&S 550 and 560            2.42
>     Mark-A-Vote                    3.04
>     Sequoia Optech                 4.35
>
> Touchscreen                         1.49
>     Diebold Accu-Vote-TS      0.72
>     Sequoia Edge                    2.01
>     ES&S iVotronic                3.49
>
> Statewide  4.59
>
> For further information, contact Rebecca Mercuri via telephone at
> 1-609/895-1375 or 1-215/327-7105, email mercuri () acm org and Internet at
> http://www.notablesoftware.com/evote.html

-------------------------------------
You are subscribed as interesting-people () lists elistx com
To manage your subscription, go to
 http://v2.listbox.com/member/?listname=ip

Archives at: http://www.interesting-people.org/archives/interesting-people/