Dan Jacoby
Testimony before the
New York State Board of Elections
Draft Regulations for voting systems certification
December 20, 2005
I'd like to focus my
testimony today on one portion of the draft Voting Systems Standards.
Section 6209.2, Paragraph A,
Subparagraph 5 reads as follows:
"The system shall contain software and hardware required to perform a diagnostic test of system status, and the means of simulating the random selection of candidates and casting of ballots in quantities sufficient to demonstrate that the system is fully operational and that all voting positions are operable." (emphasis mine)
The main problem here is
that a simulated test is not a test.
Unless all parts of the
voting system that will be used on election day are tested, a large potential
remains for serious problems. There can be discrepancies between the actual
votes cast and recorded on the voter-verified paper audit record and the vote
tallies reported by the computer. Also, election day may prove chaotic when
voters discover that the touch screen and accessibility attachments, as well as
print and minority language displays, don't work.
With Precinct-Based, Optical
Scan (PBOS) systems, a simulated test won't detect whether ballots marked in
different ways will be accurately read. Different people fill in the circles on
the paper ballot with different degrees of totality. Additionally, many
people's marks will extend outside the circles. A simulated test cannot
determine accuracy for these real-world circumstances.
With DREs, a simulated
election has many problems.
First, the mechanism used to
select a candidate, whether touch-screen or pushbutton, will not be tested.
Faulty buttons, desensitized portions of the screen, and errors in screen
calibration, can result in failure to record votes and inaccurate vote counts
on election day. Only by having people press the buttons or touch the screens
can we test whether these parts are functioning properly.
Second, in a simulated test,
the correlation between the voter verifiable paper audit record and the votes
displayed on the screen will not be checked. On election day, if voters
discover discrepancies, there will be no method in place to correct the
situation. People will lose confidence in the election, and in future elections
as well.
Third, a simulation
interacts with the voting system software in a different manner from the way a
real-time election conducted by people does. Software glitches will go
undetected until they occur on election day.
In addition to the shortcomings
of a simulated election test, many hardware and software problems - problems
too numerous to list here - cannot be detected by a self-run diagnostic test.
Anyone with diagnostic software on their computer, Norton Disk Doctor for
instance, knows that diagnostic tests miss a lot.
In short, a diagnostic
program and simulated election cannot detect hardware and software flaws, and
will not serve the purpose of determining whether the system works.
In an election run on
insufficiently tested voting systems, machine failures, high phantom votes and
undervotes, and results that are vastly different from pre-election or exit
polls, will make New York the next Florida or Ohio.
The only way to minimize
discrepancies and chaos is to test the equipment - including hardware and
software - by conducting a test election with real people entering votes
manually on the touchscreens or pushbuttons, using all the accessibility
attachments and minority language displays, examining the voter-verified paper
audit record, and extracting the vote tallies afterward to check that they
match the votes entered.
Testing needs to use live
test voters at all stages of the test. Testing needs to use live test poll
workers, who turn the machines on and extract the tallies and other records
after the test votes are cast.
Only by conducting
real-time, "live ammunition" tests of both the hardware and the
software, with real people performing all the tasks, can we do as much as
possible to ensure that the voting systems, and especially computerized
election equipment, will work on election day.
See also
http://www.wheresthepaper.org/NoAutomatedTests.htm