The Risks Digest, ACM Committee on Computers and Public
Policy
Forum on Risks to the Public in Computers and Related Systems
Volume 22: Issue 26, September 25, 2002
Press Releases: MIT vs Mercuri
"Rebecca Mercuri" <mercuri@acm.org>
Tue, 24 Sep 2002 13:44:34 -0400
I was forwarded the following press release from MIT/CalTech from a source
at IEEE Spectrum [the original is at http://pr.caltech.edu/media/Press_Releases/PR12284.html]
and am seriously concerned about the conclusions they have drawn regarding
the recent Florida primary election. The MIT press release is here
in its entirety, followed by my analysis/rebuttal. R. Mercuri.
>Date: Thu, 19 Sep 2002 10:56:22 -0700
>To: Recipient List Suppressed:;
>Subject: Caltech-MIT Team Find 35% Improvement in Florida's Voting Technology
>For Immediate Release
>September 19, 2002
>Caltech-MIT Team Finds 35% Improvement in Florida's Voting Technology
>PASADENA, Calif. - If one measures election success by equipment
>performance alone, Florida's push to get new voting equipment
>on-line for the 2002 election appears to have paid off.
>Compared with the performance of equipment in past Florida state
>primary elections, the new technologies for casting and counting
>ballots look like clear improvements according to experts at the
>California Institute of Technology and the Massachusetts Institute
>of Technology.
>Researchers from the Caltech/MIT Voting Technology Project
>calculated the rate of residual votes (ballots on which no votes or
>too many votes were recorded) for the largest counties in Florida
>for the 2002 Democratic Gubernatorial Primary and for the last three
>Gubernatorial General Elections in Florida (1990, 1994, and 1998).
>These counties are Brevard, Broward, Duval, Hillsborough,
>Miami-Dade, Palm Beach, and Pinellas.
>The residual vote rate, it appears, has been substantially reduced
>as a result of the election reform efforts of the past year. On
>average, 2.0 percent of Democratic voters recorded no vote for
>governor in these seven counties. In past elections, the average
>has been 3.1 percent. This is a 35 percent improvement in
>performance.
>The largest apparent improvements came in Brevard and Duval
>counties, which switched from punch cards to optically scanned paper
>ballots. The remaining counties purchased new touch screen or
>Direct Recording Electronic (DRE) machines. All of the counties
>show some improvement in their capacity to record and count votes.
>Residual Vote Rates for Governor in the 7 Largest Florida Counties
>County 2002 Voting Equipment Residual Vote Rate
> Demo Primary.
Gen. Gen. Gen.
>
2002 1998 Ave. 1998 1994 1990
>Brevard 1.0% Scanner Punch 4.2% 2.6%
4.5% 5.4%
>Broward 2.0 DRE
Punch 2.6% 2.7 1.9 3.3
>Dade 3.0 DRE
Punch 3.2% 4.0 2.7 3.2
>Duval 2.2 Scanner Punch
3.4% 3.1 2.5 4.5
>Hillsborough 1.6 DRE Punch 2.3% 2.7
1.9 N/A
>Palm Beach 2.3 DRE Punch 3.1%
3.7 2.3 3.3
>Pinellas 1.9 DRE
Punch 2.2% 2.3 1.9 2.3
>Total 2.0
3.1%
[General elections; PGN approximate reconstitution of a garbled
table]
>Source: Florida Division of Elections and county election offices of each
county.
>"These results are very encouraging," said Stephen Ansolabehere, a
>professor at the Massachusetts Institute of Technology and
>co-director of the project. "Florida made a major effort to upgrade
>its technology and, in the primary, the machines used showed clear
>gains over the technologies in past elections."
>Professor Charles Stewart, another MIT professor working on the
>Voting Technology Project, cautions that "the success of an election
>cannot be measured solely in terms of equipment performance.
>Current events in Florida also illustrate how better technology is
>just a first step in improving the functioning of democracy."
>Stewart said, "Most of the problems reported by journalists covering
>the 2002 Primary Elections in Florida did not concern equipment
>malfunctions, but problems encountered preparing for election day,
>such as training poll workers."
>R. Michael Alvarez, co-director of the Voting Technology Project and
>professor of political science at the California Institute of
>Technology, said "As counties and states across the country,
>especially here in California, plan out similar changes, we are
>learning important lessons about how to make such important changes
>in voting technologies."
>"The one distressing thing, though, are the reports from Florida
>that polling place workers had difficulties getting some of the new
>voting machines up and running on election day in Florida, and that
>as a result, some voters might have been turned away from the
>polling places. These reports reinforce our calls for more polling
>place workers and better training of polling place workers, as they
>provide a critical role in making sure that all votes are counted,"
>Alvarez said.
>MIT's Stewart adds "The fact that the congressional election reform
>bill is currently stalled in a House-Senate conference committee
>hasn't helped matters any."
>The Caltech/MIT Voting Technology Project is a non-partisan research
>project, formed to study election systems following the 2000
>presidential election and sponsored by the Carnegie Corporation.
>More information and copies of reports are available at
>www.vote.caltech.edu.
>MEDIA CONTACT: Jill Perry, Caltech Media Relations Director (626) 395-3226
>Sarah Wright or Ken Campbell, MIT News Office 617 253-2700
>Jill Perry, Media Relations Director
>California Institute of Technology (Caltech)
>Mailing Address: Mail Code 0-71, Pasadena, CA 91125
>Street Address: 315 S. Hill Ave., Pasadena, CA 91106
>Ph: (626) 395-3226
>Fax: (626) 577-5492
>jperry@caltech.edu
- - - - -
NEWS RELEASE, September 24, 2002
Rebecca Mercuri rebuts recent MIT/CalTech voting systems analysis and calls
for moratorium on new electronic balloting equipment purchases
After reviewing the press release issued September 19 by MIT and CalTech,
electronic voting system expert Rebecca Mercuri revealed that "the conclusion
that MIT/CalTech researchers has drawn, that Florida's new voting technology
shows a 35% improvement, is based on a flawed analysis and is likely erroneous."
She goes on to state that not only are the researchers comparing "apples
to oranges" in terms of the types of technologies surveyed (punch-cards
versus optically scanned and DRE machines), but they have misleadingly
compared Gubernatorial general election results to Gubernatorial primary
results (and only for the Democrats in the 2002 primary).
It is well known that voters in general elections turn out in far greater
numbers (in Florida it is estimated that the November election will show
a 400% increase or more) than in primaries, putting greater strain on the
performance of systems as well as on poll workers and voters. The
balloting style of the typical primary voter (usually a party insider,
and certainly a partisan with a larger interest in selecting candidates
for each race on the ballot) is quite different from the general election
voter, where independents and other non-declared or minority party affiliation
citizens are permitted to cast ballots. Thus, only in November will
we be able to ascertain whether the residual vote rate has actually "improved."
Hence, Dr. Mercuri asserts, "the conclusion is premature, as well as flawed."
Laudatory statements made by Stephen Ansolabehere, Charles Stuart and R.
Michael Alvarez regarding Florida's new voting systems are also sorely
misleading, and do not support their conclusion of 35% improvement.
MIT Professor Stuart's comment that "most of the problems covered by journalists...did
not concern equipment malfunctions" is not based on an analysis of the
numerous and severe voting system problems that occurred throughout the
state, but rather on the media reports that surfaced. Many equipment
malfunctions were reported by the Associated Press and other news bureaus,
but these were obfuscated by the public interest stories that alternatively
showed voters "pleased with the new equipment" or being "turned away from
the polls in droves."
A lot of the media attention focused on press comments by Governor Jeb
Bush and members of his staff who erroneously characterized the problems
as being based only in two counties (Miami-Dade and Broward) and blamed
the poll workers and election officials there for the situation.
In actuality, Miami-Dade and Broward could not have purchased the ES&S
machines had they not been pre-certified by the state for use. Sadly,
this certification failed to provide the counties or their poll workers
with sufficient notification as to the fact that the voting machines would
take 10 minutes to start up, with the ones outfitted for the visually impaired
taking an astonishing 23 minutes. Some machines also contained a
"safety feature" that did not permit them to be turned on before 6AM on
election day. Since each unit is activated sequentially, simple math
shows that a polling place containing 10 voting machines, with one outfitted
for the visually impaired, would not be fully operational until nearly
8AM (an hour after the polls opened) under the best conditions. Mercuri
states: "I certainly do not see how this can be blamed on the poll-workers,
nor how it constitutes an improvement. I'm hard pressed to think
of any computer equipment manufactured after the 1970's that takes 23 minutes
to be started, especially those deployed for use entirely in time-critical
operations. The failure by MIT/CalTech to raise serious concerns
about the engineering of these products is remiss."
MIT's Ansolabehere stated that "the machines used showed clear gains over
the technologies used in past elections." To which Dr. Mercuri replies:
"Yes perhaps, if one considers declaring a state of emergency (under threat
of lawsuit by a major candidate) and extending the election day by two
hours a "clear gain." How about in Union County, Florida, where 2,700
optically scanned ballots had to be hand counted, because the computers
were erroneously programmed to only tally votes for Republican candidates?
At least there, the ballots could be recounted because they were on paper.
What about the precinct in southern Florida that showed a 1200% voter turnout
(12 times as many voters as were registered) because the DRE activation
cards permitted voters to cast ballots on machines in the same building
that were not in their precinct? And what about some precincts in
Miami-Dade and Broward where the vote cartridges reflected over 40% residual
votes (lost or missing) and data had to be "extracted" from back-up memory
inside of the machines (one wonders how trusted the reconstructed results
can be)?"
CalTech's Alvarez states "we are learning important lessons about how to
make such important changes in voting technologies" and Mercuri asks: "Is
it fair to allow Florida and other states and communities to feel pressured
to replace their voting systems while being treated as guinea pigs?
Is the United States prepared to reimburse communities for defective and
obsolete equipment once new standards are in place (since all election
equipment is still being inspected by the National Association of State
Election Directors testing authorities to the outdated 1990 Federal Election
Commission guidelines)? Is it acceptable to certify voting equipment that
can be reprogrammed internally via a portal on the device (as some were,
only weeks before the election in Palm Beach County as well as elsewhere
in the state)? These new technologies are playing a role in electing
government officials - the confidence citizens have in the democratic process
is at stake."
Mercuri, who has testified before the U.S. House Science Committee regarding
the need for involvement of the National Institute of Standards and Technologies
in establishing criteria for the procurement and testing of election equipment,
feels that congressional election reform is sorely needed. But, she notes
that many of the laws proposed at federal and state levels, or enacted
since 2000, have been weakly worded so as to permit the production of election
equipment that does not provide an independent means whereby voters can
verify human-readable ballots that are secured and available for recounts.
"Real election reform," Mercuri says, "is only possible within a context
of adequate and enforceable standards for construction, testing, and deployment
of voting equipment."
But Mercuri worries that the trend to full automation of the voting process
could be used to conceal election fraud. She warns, "It is entirely possible
that Florida and other states may smooth out their election day problems
such that it appears that the voting systems are functioning properly,
but votes could still be shifted or lost in small percentages, enough to
affect the outcome of an election, within the self-auditing machines.
Whether this occurs maliciously or accidentally, it presents a frightening
prospect. Thankfully, new products are being developed that provide
the voter with a way to determine that their ballot has been tabulated
correctly, without revealing the contents of their vote, but deployment
of such systems is a few years down the road."
For these reasons, Dr. Mercuri has requested a moratorium on the purchase
of any new voting systems that do not provide, at minimum, a voter-verified,
hand-recountable, physical (paper) ballot while appropriate laws, standards,
and technologies are developed that will provide accurate, secure, reliable,
and auditable voting systems. She urges MIT, CalTech, and other concerned
scientists, public officials and private citizens to join her in this cause.
For further information contact:
Rebecca Mercuri, Ph.D.
P.O. Box 1166, Phila. PA 19105
609/895-1375, 215/327-7105
www.notablesoftware.com/evote.html
mercuri@acm.org