June 27, 2005

Kerr Responds to Aspect of DNC Report on DREs

Michael Kerr, who works for ITAA and its Election Technology Council, a trade
association for the electronic voting equipment and services companies, sends along the following comments via e-mail on one aspect of the DNC's recent report on Ohio:

    I am writing to raise your attention to some problems with the DNC's analysis of electronic voting and DRE usage in the recently-released Voting Rights Institute report on the 2004 Ohio election outcome.

    To begin with, we know that there are actually very few recent DRE implementations in Ohio due to the issues they've had with paper-trail legislation and procurement - which Dan Tokaji has covered in considerable detail. So discussion of the role played by DREs in a report on the Ohio election outcome is strange and misplaced.

    The section of the report on e-voting, "Electronic Voting: Accuracy, Accessibility and Fraud," is written by Dr. Dan Wallach of Rice. Dr. Wallach has never met a DRE fraud or tampering theory that he won't entertain -- in spite of the lack of any solid evidence that the machines are insecure or inaccurate and in the face of a track record of thousands of successful elections run on electronic voting equipment.

    My primary criticism of Dr. Wallach's section is that he does not undertake any analysis of available election data, or the post-election analyses done by other parties, as a basis to support any of his arguments and recommendations. He simply throws out theories and characterizations of DREs as accepted fact and proceeds to support them with thin arguments and anecdotes. This is not a scholarly work in any sense so should not be characterized as a "study" or "independent analysis."

    This first example of this approach is contained in the first section of the E-voting piece, titled "Incident reports and machine accuracy." Wallach states "Among statisticians, we have studies of voting residual rates, turnout, and other important issues, many of which have concluded that new (emph. added) DRE voting systems are less accurate than more traditional optical scan ballots."

    Here, Dr. Wallach has mis-characterized his facts or based his assertions on old information, although its hard to tell where he is getting his data as there don't appear to be any cites in his report. Dr. Wallach seems to gloss right over the MIT-Caltech study released in early-2005 stating that in the November 2004 elections, the three states (GA, MD, and NV) with the lowest residual vote rates in the nation were the only three states with statewide use of new DRE equipment at the polls.

    The Cal-Tech MIT study showed that, while the national average residual vote rate was 1.1 %, Nevada achieved a residual vote rate of just under .3%. Georgia and Maryland posted similar results. The authors highlight the improvements in residual vote rates obtained with the use of newer-generation DREs.

    (Excerpt from the MIT-Caltech Report, starts at bottom of page 14)

    Like the analysis in Columns 2 and 3, the coefficients from Column 4 are reduced substantially when the controls are added in Column 5. There is one exception, however. The coefficient associated with moving from optical scanning to DREs is only affected slightly when the other controls are added. This finding is intriguing because in previous research (VTP 2001; Ansolabehere and Stewart 2005), we discovered that optical scanners tended to have the lowest residual vote rates and that DREs tended to have higher residual vote rates. Here we discover that there may be particular gains to be had when a jurisdiction that already uses optical scanners chooses to use the newest generation of DREs. (emph. added)


    So, Dr. Wallach's analysis is blown out of the water by readily-available reporting on the 2004 election. Most of the E-Voting section goes on to repeat the litany of familiar, and unsupported, theories on DRE accuracy and security with no data or factual evidence presented in support of those theories. This is a recurring theme in the paper. Even sections with hard data (such as the survey of voter experience) misuse or misinterpret the data. Long lines at the polls are blamed on DRE use in Franklin and other counties. Other likely factors, such as unprecedented turnout, are given no consideration.

    Dr. Wallach's arguments are then briefly and badly summarized by the DNC VRI leadership and used to support their policy proposals. But without a complete analysis of the issues, and refusal to even consider data that don't support Dr. Wallach 's theories, how can these policy proposals be held out for serious consideration? Unfortunately, this report seems to be a vehicle more suited for carrying forward the DNC 's agenda than for serious analysis of election practices or the outcome in Ohio.


Thanks for writing!
UPDATE: Dan Tokaji has these thoughts on the Wallach portion of the report.

Posted by Rick Hasen at June 27, 2005 12:32 PM