Optical-scan equipment problems in California

[Paul Walmsley <paul@xxxxxxxxxxx>]

Diebold's optical-scan equipment miscounted thousands of paper ballots in
California last month.  From the Tri-Valley Herald (California):

   In the March primary, Alameda County workers eased the load on 
   Diebold's computer by scanning absentee ballots one party at a time.
   But San Diego County fed its absentee ballots in as a mix, and
   Diebold's software misreported almost 3,000 votes. In the worst case,
   it switched 2,747 Democratic presidential primary votes for U.S. Sen.  
   John Kerry to U.S.  Rep. Dick Gephardt, who had dropped out of the
   race.

   Diebold's latest explanation says its vote-tabulation software 
   apparently could not handle results from multiple optical-scanning
   machines, processing ballots with large numbers of candidates and
   precincts.

   That vote-tabulating software, technically known as GEMS version 
   1.18.18, is used by 18 California counties. 

(Diebold GEMS is the same optical-scan software that Boulder County used
in the 2003 mail ballot election.)

Fortunately, the counting flaws in Alameda County's election were obvious
and therefore detectable by vigilant election workers.  But what if a
software bug corrupted election results in a non-obvious way?  Unless we
test that the optical-scan machines are working the way they're supposed
to -- by hand-counting a statistically-significant sample of the live
paper ballots -- we could end up with election results that are as badly
flawed as those from a faulty DRE election.

And unless we test the machines appropriately, we might never know whether
our election results are legitimate.  We might not get lucky the way the
way Alameda County did by having thousands of votes miscounted for a
candidate that dropped out.  Those thousands of votes could have been
awarded to another active candidate.  Then, would the problem have been 
caught at all?

from the Tri-Valley Herald (California):

<http://www.trivalleyherald.com/Stories/0,1413,86~10671~2080327,00.html>

- Paul


Diebold reports multiple problems
Registrar wants reason for e-voting
By Ian Hoffman
STAFF WRITER

Tuesday, April 13, 2004 - Electronic devices that held the key to digital 
voting in Alameda County's Super Tuesday primary failed in at least a half 
dozen ways, hobbling the $12.7 million voting system at a quarter of 
polling places.

Poll workers saw unfamiliar Windows screens, frozen screens, strange error 
messages and login boxes -- none of which they'd been trained to expect.

A report released Monday by Diebold Election Systems shows that 186 of 763 
devices known as voter-card encoders failed on election day because of 
hardware or software problems or both, with only a minority of problems 
attributable to poll worker training.

Diebold's post-mortem of the March 2 election said it was "disappointed" 
in the encoder failures and that it values its ties to local elections 
officials. But the McKinney, Texas-based firm offered no fundamental 
explanation of how and why the company delivered faulty voting equipment 
to Alameda and San Diego counties -- its two largest West Coast customers 
-- on the eve of the 2004 presidential primary.

Alameda County Registrar of Voters Bradley Clark wants full answers to 
that question, plus Diebold's guaranteed fix for software that erroneously 
gave optically scanned votes to the wrong candidates, by April 29.

Otherwise, Clark says, he will consider firing Diebold. "I want to see 
some real frankness and answers to the optical scan problem. That to me is 
the biggest problem facing us," Clark said.

The faulty voter-card encoders can be fixed or replaced by older, more 
dependable devices, he said, but faulty vote-tabulating software is a more 
troubling matter.

After the Oct. 7 recall election, when Diebold's vote-tabulating software 
wrongly awarded 9,000 Democratic absentee votes to a Southern California 
Socialist, Diebold decided its computer was overwhelmed and replaced it.

In the March primary, Alameda County workers eased the load on Diebold's 
computer by scanning absentee ballots one party at a time. But San Diego 
County fed its absentee ballots in as a mix, and Diebold's software 
misreported almost 3,000 votes. In the worst case, it switched 2,747 
Democratic presidential primary votes for U.S. Sen. John Kerry to U.S. 
Rep. Dick Gephardt, who had dropped out of the race.

Diebold's latest explanation says its vote-tabulation software apparently 
could not handle results from multiple optical-scanning machines, 
processing ballots with large numbers of candidates and precincts.

That vote-tabulating software, technically known as GEMS version 1.18.18, 
is used by 18 California counties.

Clark said Diebold must offer a firm diagnosis of the software bug, devise 
a fix and figure out how to get the repaired software tested and approved 
by the November elections, or Alameda County will have to consider other 
ways to ensure a reliable election.

"If they can't give us those kinds of assurances, I don't see how we can 
count ballots," Clark said.

California faces a decisive moment next week on electronic voting. The 
voting systems panel of the Office of the Secretary of State is to 
consider findings of an investigation of Diebold, as well as approval of 
voting systems to be used in the November elections.

Secretary of State Kevin Shelley and his subordinates on the panel are 
considering a range of options including disallowing all touchscreen 
voting systems -- now used by 40 percent of California voters -- or 
selectively decertifying all or some Diebold voting systems.

The showdown joins two intersecting controversies: broad objections to 
electronic voting as unsecure and sensitive to technological failure and 
Diebold's fielding of untested or uncertified hardware and software in 
California counties.

The voter-card encoders are the latest example. A single lab hired by 
Diebold gave them cursory testing two weeks before the primary. State 
elections officials approved their use because Alameda, San Diego and 
other counties said they had no practical alternative.

"What happened was Diebold beta-tested this smart-card encoder on Alameda 
and these other counties, and the state and the counties let them and 
voters were disenfranchised because of it," said Kim Alexander, president 
of the Davis-based California Voter Foundation.

Alameda County officials estimate that about 156 voters were turned away 
from the polls March 2, the largest number from a single precinct in 
Pleasanton.

Contact Ian Hoffman at ihoffman@xxxxxxxxxxxxxxxxx