Consultant on Election Policy and Technology
5025 Broken Oak Lane, Columbia, MD 21044
Phone:  410.730.4983  /  Fax:  410.997.4355


December 5, 2001

Dr. Robert M. Berdahl
Chancellor, University of California, Berkeley
200 California Hall
Berkeley, CA 94720

Dear Dr. Berdahl:

I have studied the Survey Research Center's project proposal to improve voting systems in California and wish to convey my strong recommendation for funding.  I am familiar with the work of Alan Dechert and Dr. Henry Brady and I have been discussing this proposal with them since April of this year.  In addition to a broad review of how votes are cast and counted in California, they propose to develop a voting system based on a commercially available personal computer (PC).  I believe that this idea has significant merit and is well worth pursuing.

A major benefit of the successful completion of the project would be the significant reduction in costs that would result if voting machines had ordinary business or educational uses as PCs when not required for voting.  Another benefit of lower unit costs would be assurance of availability of a sufficiently large number of voting machines in each precinct to prevent the formation of waiting lines, an important issue with direct recording voting systems.  These results would have national importance if achieved.

I believe that my views on the proposed project are worthy of consideration because of my experience of over 25 years in election policy and technology.  My first report on these subjects, in 1975, sponsored by the US General Accounting Office, continues to be cited by the Federal Election Commission as laying the groundwork for its on-going voting systems standards program.  My second report in 1988, sponsored by the John and Mary R. Markle Foundation, was widely quoted by the media following the November, 2000, general election because of the report's recommendation that "the use of pre-scored punch card ballots should be ended."  Since retiring from the Federal Government's National Institute of Standards and Technology in 1996, I have served as a consultant on election policy and technology in Japan and in several countries in South America.  In 2001, I contributed to three US national election reform efforts, including the Caltech/MIT study and the National Commission on Federal Election Reform, and gave invited testimony to the US House of Representatives Committee on Science.

I am aware that there have been concerns raised, in discussions on the viability of the project, about the use of direct-recording electronic (DRE) voting technology.  When I first saw the results produced by the Caltech/MIT study that DREs might be harder to use than other systems, I was intuitively skeptical.  My extensive experience in reviewing the use of systems of many types suggested to me that, with voter training to improve familiarity, DREs could be highly voter-friendly.

I have had a work-experience in which this point was demonstrated.  In 1996, I was invited to Brazil by the Brazilian government and the Inter-American Development Bank to participate in the evaluation of that nation's new DRE system.  I saw the system in successful use in a major election by illiterate and otherwise poorly educated citizens.  The election was preceded by a massive training campaign by the national election authority using TV to teach citizens how to use the new system.  The training effort succeeded overwhelmingly.

Additionally, two authoritative studies have produced different results than the Caltech/MIT work.  The report by Dr. Henry Brady and colleagues, "Counting All the Votes:  The Performance of Voting Technology in the United States," examined the results of the 2000 Presidential election using data from two-thirds of all US counties.  Their report shows, with strong concern for data quality, that "optical scan and DREs appear to dominate all other systems in terms of overall performance across all counties ..."  A study by the Minority Staff of the Special Investigations Division of the US House of Representatives Committee on Government Reform, "Income and Racial Disparities in the Undercount in the 2000 Presidential Election," shows similar results.  In the latter study, comparing 20 high income, low minority percentage Congressional districts with 20 low income, high minority percentage Congressional districts, DRE and precinct-count optical scan systems had fewer uncounted votes in the Presidential contest than other system types in both sets of districts.  No specific voter training efforts that could have improved outcomes were identified by either report.

The advantages of a DRE voting system over an optical scan voting system are these:

  1. The "voter's intent" issue, which plagued Florida in November, 2000, does not exist.
  2. It is considerably easier and significantly less costly to implement each voting machine to present the voting instructions in a language other than English.  Each machine may be implemented with several languages, and each successive voter may select a different one to be used.
  3. It is possible, at a reasonable cost, to provide a visually handicapped voter with the ability to vote unassisted, in private.  No similar capability is available with optical scan voting equipment.
  4. Overvotes are automatically prevented on a DRE machine, and undervotes may be easily flagged for voter reconsideration.  The voter need not vacate the voting location and insert a ballot into a ballot-reader in order to find out that he or she has mistakenly undervoted; privacy is preserved.
  5. A reduction in election-day costs results from the elimination of hard-copy ballots.

However, DRE voting systems have no hard-copy audit trail.  The project should consider the problem of public confidence in DRE systems due to the lack of tangible ballots that may be recounted.  The use of various security procedures, software testing methods and machine logic design techniques should be evaluated to select the most cost-effective implementations that can demonstrate that the election results produced are actually those collectively chosen by the voters.  My 1988 report considered this issue extensively, and I could specifically contribute to the project in this area.  Furthermore, the use of open-source software, as proposed by Mr. Dechert and Dr. Brady, makes possible software testing by outside experts, a step that will further enhance public confidence.

Finally, I want to support Dr. Brady in his desire to undertake extensive human factors testing of existing systems and any new system developed.  In my 1975 report, I specifically recommended that "research into human engineering of voting systems" be pursued, in that "there is a lack of technical data on how individuals react to specific types of equipment, what kinds of errors they make, and how voting drop-off ... is affected by different voting systems."  There has not been an organized research program on this subject, and the extensive number of poorly conceived comparative evaluations of voting system types that were issued following the 2000 general election makes this effort essential.  More of the high-quality work carried out by Dr. Brady and colleagues would be extremely valuable.  The University of California at Berkeley would be performing an important public service and would be contributing significantly to the advancement of knowledge by undertaking this research effort.  Significant contributions could be made to the academic sub-specialty of computer-human interaction, and the results should enable California and the nation to improve the selection of its political leadership by "consent of the governed."

In closing, I see the proposed project as an important opportunity to undertake research and development on voting systems that can be thorough, because it is not being forced by private-sector considerations to achieve a return-on-investment in the immediate future.  I look forward eagerly to working with Alan Dechert, Henry Brady and the other members of the team.

Sincerely, Roy G. Saltman

 Close Window