Home
Bibliography
Calendar
Columns
Dorothy Brizill
Bonnie Cain
Jim Dougherty
Gary Imhoff
Phil Mendelson
Mark David Richards
Sandra Seegars
DCPSWatch
DCWatch
Archives
Council Period 12
Council Period 13
Council Period 14
Election 1998
Election 2000
Election 2002
Elections
Election
2004
Election 2006
Government and People
ANC's
Anacostia Waterfront Corporation
Auditor
Boards and Com
BusRegRefCom
Campaign Finance
Chief Financial Officer
Chief Management Officer
City Council
Congress
Control Board
Corporation Counsel
Courts
DC2000
DC Agenda
Elections and Ethics
Fire Department
FOI Officers
Inspector General
Health
Housing and Community Dev.
Human Services
Legislation
Mayor's Office
Mental Health
Motor Vehicles
Neighborhood Action
National
Capital Revitalization Corp.
Planning and Econ. Dev.
Planning, Office of
Police Department
Property Management
Public Advocate
Public Libraries
Public Schools
Public Service Commission
Public Works
Regional Mobility Panel
Sports and Entertainment Com.
Taxi Commission
Telephone Directory
University of DC
Water and Sewer Administration
Youth Rehabilitation Services
Zoning Commission
Issues in DC Politics
Budget issues
DC Flag
DC General, PBC
Gun issues
Health issues
Housing initiatives
Mayor’s mansion
Public Benefit Corporation
Regional Mobility
Reservation 13
Tax Rev Comm
Term limits repeal
Voting rights, statehood
Williams’s Fundraising Scandals
Links
Organizations
Appleseed Center
Cardozo Shaw Neigh.Assoc.
Committee of 100
Fed of Citizens Assocs
League of Women Voters
Parents United
Shaw Coalition
Photos
Search
What Is DCWatch?
themail
archives
|
I. Introduction
The purpose of this report is to document the results of the District
of Columbia Board of Elections and Ethics (the Board) Internal Review
Committee’s (IRC) investigation into the tabulation and reporting
anomaly that occurred on the night of the September 9, 2008 Congressional
and Council Primary Election (Election).
The objective of the IRC investigation was to review the events of that
evening – and the Board’s tabulation process in general - in order to
pinpoint the precise cause of the error and determine what steps need to
be taken to insure that it does not reoccur in future elections. A
comprehensive investigation was conducted, and various issues raised by
the anomaly were explored. For example, the IRC considered the question of
whether the anomaly was the result of a deviation from the Board’s
established tabulation procedures, or the procedures themselves. The IRC
also examined whether any technological irregularities occurred during the
tabulation process and contributed to the error.
The Board staff members selected to comprise the IRC were chosen based
upon their particular areas of expertise in technology, voting and
tabulation systems in general, and/or the Board’s tabulation process
specifically. Additionally, as a result of the roles these individuals
played on Election night, the Board determined that these persons were in
the best position to analyze the circumstances that generated the anomaly,
and recommend corrective measures that can be employed to insure that the
error does not reoccur.
The IRC was comprised of the following individuals:
Vialetta Graham, Chief Technology Officer. Ms. Graham is
responsible for overseeing the development and operation of all technical
projects within the Board and the D.C. Office of Campaign Finance. In this
capacity, she provides technical leadership and guidance to all technical
staff. Ms. Graham is also a Certified Elections/Registration Administrator
(CERA). She has been employed by the Board for seven years.
Darlene Lesesne-Horton, Data Systems Manager. Mrs. Lesesne-Horton
is responsible for the management and operation of the agency’s data
processing systems, including voter registration and election processing.
Mrs. Lesesne-Horton is also a Certified Elections/Registration
Administrator. She has been employed by the Board for 28 years.
Mohammed Maeruf, Information Technology Project Manager. Mr.
Maeruf is responsible for the planning, designing, installation, and
maintenance of the Board’s computer systems and voting system election
software and hardware. He began his service to the Board as a consultant
in 2003, and joined the Board’s staff in 2005.
Clifford Tatum, Help America Vote Act (HAVA) Consultant. Mr.
Tatum is an elections attorney who is under contract to provide election
support to the Board. Formerly, he served as the Assistant Director of
Legal Affairs and concluded his tenure with the State of Georgia as the
State Elections Director in 2007. He is responsible for the continued
implementation and maintenance of the District’s election reform efforts
under the Help America Vote Act. Mr. Tatum has been associated with the
Board for five months.
II. Executive Summary
The IRC concludes that, on Election night, an unofficial results
summary report containing erroneous information was inadvertently released
prior to review and confirmation by the Board. This report indicated that
an abnormally high number (4,759) of votes were posted during the Election
Night Report. Prior to the release of the subsequent unofficial results
summary reports, the Board, having recognized the error, conducted a
precinct-by-precinct analysis of the anomaly. This analysis indicated that
the incorrect totals in Release No. 2 were isolated to precinct 141 in
Ward 2. After this determination was made, a thorough audit of all results
from the affected precinct was immediately carried out. The results of the
audit indicated that, while 4,759 votes were initially recorded in the
results for Precinct 141, only 326 votes were actually cast in that
precinct. All reports released subsequent to the detection of the error
accurately reflected the number of votes cast during the Election.
This report makes several recommendations, found in section VI on page
10.
III. Technical Overview of the Election Preparation and Tabulation
Processes
An overview of the Board’s election preparation and tabulation
processes will provide the proper perspective on the nature of the error
at issue and how this error may have occurred.
The Board deploys a dual-voting system at its 143 polling precincts on
Election Day. The two components of the dual voting system are the Optech
Eagle (the Eagle), an optical scan voting machine that utilizes paper
ballots, and the AVC Edge (the Edge), a touch screen voting machine. The
Eagle stores its votes on a Memory Pack (Memory Pack), while the Edge
stores its votes on a Results Cartridge (Results Cartridge).
A. Step One: Programming an Election
The entire election process begins with the creation of the election.
In order to create an election, several parameters must be defined
within the software used by the voting systems. This compilation of
information includes such indicators as: whether the election is a
primary, general or special election; the applicable offices; names of
qualified candidates; the ballot layout and ballot order; and the
registration totals by Ward, political party and precinct. Once an
election’s parameters have been defined, and the election has been
reviewed for accuracy, paper ballots are printed for the Eagle and
electronic ballots are created for the Edge.
B. Step Two: Pre-Logic and Accuracy Testing
After election parameters are created, tests are conducted on each
Eagle and Edge voting machine to be used on Election Day. These tests
are referred to as Pre-Logic and Accuracy (Pre-LAT) Tests. These
Pre-LAT tests include both a hardware and software diagnostic
assessment, and are designed to establish that each voting machine is
operating properly and is ready for use in the
election for which it has been programmed. The hardware
assessment tests the physical operation of the machine.
1. Hardware Assessment-Eagle
The Eagle machine is assessed by checking the auxiliary bin, the
ballot slots, rear and center ballot locks, slide key locks, power
supply, printers display panels and Memory Pack ports. Diagnostics
on the indicator lights, sensor displays, printers, beepers and
motors are also performed to ensure the machine is working properly.
2. Hardware Assessment-Edge
The Edge machine is assessed by checking the leg assembly, the
privacy curtain, AC power indicators, poll-worker information
display and the alignment of the touch screen display. Diagnostics
for this piece of equipment include a memory test; verifying the
functionality of the results port, the auxiliary port, the battery
charging indicator as well as the quality of the printouts; and
confirming that the date and time are correct. During this testing,
the functionality of the card activator devices is also verified.
The staff then completes a memory test, a test of all ports on the
machine (the results port, auxiliary port for audit trail, smart
card port reader, audio port and the printer port). Finally, the “power
on/off” and “polls open/close” knobs are tested.
3. Software Assessment-Eagle
A Memory Pack with a serial number is assigned to each precinct.
After this assignment, this Memory Pack is loaded with all the
ballot styles for the assigned precinct. Subsequently, test decks
for the election are produced and a “test election” is executed
on the machine. After this test, Memory Packs are read, results
tabulated and compared with the expected outcomes for the test
election. At this point, final Pre-LAT reports are prepared.
Machines are then sealed in preparation for deployment to the
precincts.
4. Software Assessment-Edge
Each precinct is assigned an Edge device and a card activator.
Serial numbers of both machines are logged after assignments are
made. As with the Eagle, a test Results Cartridge is prepared. The
cartridge contains the corresponding precinct’s serial number,
electronic ballot (with applicable ballot styles—in both English
and Spanish) and audio ballots for visually impaired voters are also
included in the Results Cartridge data. Again, as with the Eagle, a
test election is conducted. Results Cartridges are read and
simulated election results are compared against expected valid
results. Pre-LAT reports are prepared and the machines are then
sealed in preparation for deployment to the precincts.
C. Step Three: Public Testing
Once the voting machines have been tested, they are subjected
to a public test. The purpose of this test, which must be held
within 4 days before the election, is to demonstrate to the
public how the votes cast on Election Day will be accurately
counted and reported. This test is also, in effect, a dress
rehearsal for the actual election; the actual races to be
contested are loaded onto the machines and demonstrations of
actual ballots are executed. Once all demonstration voting has
been rendered, the “polls” are closed and results reports
are generated.
D. Step Four: Polling Place Operation
Before Election Day, the Board’s warehouse staff delivers
voting equipment and supplies to the precincts. Poll-worker
teams meet on the Monday prior to the election to set up
machines, though they are not powered up and turned on until the
morning of the election. On Election Day, the Precinct Captain
turns on the voting machines as instructed during
mandatory poll-worker training classes. The machines are used by
voters throughout the day and are shut down after the polls
close at 8:00 p.m.
E. Step Five: Election Night Tabulation Process
A ll voting precincts close at 8:00 p.m. However, the
voting machines are not shut down and no results are tallied until
all voters in line at the precinct at 8:00 p.m. are given the
opportunity to vote. Once all voters have voted in a particular
precinct, the Precinct Captain declares the polls closed and
begins the process for closing the election and preparing reports
to be submitted to election headquarters. The Precinct Captain
prints a results tape from each voting machine and submits the
same to election headquarters.
As the polls close, the Board deploys representatives
accompanied by police escorts to collect the election results from
each precinct. The Board representative collects the Memory Pack,
the Results Cartridge, and the results tapes for each machine from
each Precinct Captain. An administrative chain of custody document
is created and processed for each representative and each police
escort, demonstrating that all election results were collected
from each precinct and delivered to the election headquarters.
As the representatives deliver the election night results,
i.e., Memory Packs, Results Cartridges, and all results reports,
to election headquarters, several staff members complete a
checklist to ensure that the Board receives all Memory Packs and
Results Cartridges from each precinct so that the results can be
tabulated.
The actual vote tabulation process takes place within a secure
server room, which houses the Election Database and the Memory
Pack reader devices. This server room is secured by an electronic
lock and key code system and is equipped with security cameras to
ensure the safety of the system and the integrity of the
tabulation process. Each Memory Pack is inserted into a Memory
Pack Reader (MPR), and each Results Cartridge is inserted into a
Results Cartridge Reader (RCR). The MPR reads the votes stored on
each Memory Pack and uploads the number of votes stored therein to
the Election Database. A similar process occurs with the Results
Cartridge and the RCR. Each Memory Pack and Results Cartridge is
stored in the server room after being processed by the staff.
As the results from the Memory Packs and Results Cartridges are
processed, both reading devices indicate whether or not they were
able to load data into the Election Database by indicating “success”
on a log report. If the report fails to indicate “success,”
the operator knows to take corrective action to ensure that all
votes cast have been properly uploaded into the tabulation system.
The Election Database tallies the election results from each
Memory Pack and Results Cartridge, and creates an unofficial
results summary report. Pursuant to the Board’s standard
practice, each such report is reviewed for accuracy and, once
confirmed, disseminated to the public and media and uploaded to
the Board’s website. Unofficial results summary reports are
continuously produced and verified throughout election night until
100% of the precincts’ Memory Packs and Results Cartridges have
been tabulated.
F. Step Six: Canvass Process and Post-Election Audit
On the day following Election Night, the Board staff
begins a three-pronged post-election audit that includes
processing the Memory Packs and Results Cartridges onto a
back-up server, conducting a precinct machine tape review, and
executing a ballot reconciliation process. This post-election
audit is conducted to confirm the accuracy of the unofficial
election night results. Once the results are found to be
accurate, a pre-certified unofficial results report, which also
includes the unofficial results of the absentee and special
(provisional) ballot tabulation, is generated and disseminated.
IV. September 9, 2008 Election Tabulation and Audit
A. Release of Unofficial Results Summary Reports Nos. 1 and 2
At 8:15 p.m. the Board received the first delivery of Memory
Packs and Results Cartridges from the precincts. The first set of
memory storage devices were processed at approximately 8:25 pm At
8:54 p.m., the Board released the first unofficial results summary
report (Report No. 1). Release No. 1 reflected that Board staff
and Sequoia election support personnel (the tabulation team)
processed 27 precincts (27 Memory Packs and 27 Results Cartridges)
out of 143 precincts within 29 minutes after receiving the first
set of memory devices. An analysis of the amount of time used to
process each of these memory storage devices indicates that 54
memory storage devices, all of which contained 7, 083 ballots were
processed within 32.3 seconds per memory storage device.. The vote
tally reflected in this report was unremarkable and reflected
accurate election results.
The second increment of Memory Packs and Results Cartridges began
processing after 9:00 p.m. and was processed under the same
procedures. At 9:37 p.m., the Board released the second unofficial
results summary report (Report No. 2). Report No. 2 reflects that
Board staff and Sequoia election support personnel processed 107
precincts (107 Memory Packs and 107 Results Cartridges) out of 143
precincts within 37minutes after receipt of the second set of memory
storage devices. An analysis of the amount of time used to process
each of the memory storage devices from the precincts reflected on
Report No. 2 indicates that 214 memory storage devices, all of which
contained 44,784 ballots were processed within 10.3 seconds per
memory storage device.
The tabulation team adhered to the same process used to process
the first increment of Memory Packs and Results Cartridges with one
significant exception; the speed at which the memory cards were
processed increased exponentially. The 214 memory storage devices
tabulated between the release of Reports Nos. 1 and 2 were processed
at 3 times the speed at which the first 54 memory storage devices
had been processed. Specifically, the tabulation team processed 107
precincts (214 memory storage devices) within a 37 minute time
period, or at a rate of 10.3 seconds per memory storage device. This
equates to roughly five memory storage devices per minute. In
addition, the team relied on the processed log report that indicated
that data had been successfully uploaded into the Election Database.
B. Detection of Tabulation Anomaly and Corrective Measures
After the release and posting of Report No. 2, the tabulation
team examined the figures contained therein and detected what
appeared to be a tabulation error in one of the precincts: the vote
totals in several key contests were higher than they logically
should have been. Upon noticing this error, the tabulation team
commenced a precinct-by-precinct data review. This analysis revealed
that one precinct in particular - Precinct 141 - was the source of
the tabulation anomaly. Precinct 141 is a Ward 2 precinct, located
at the Frank D. Reeves Center for Municipal Affairs, 2000 14th
Street, NW.
Once it was determined that the tabulation error was generated in
- and confined to - Precinct 141, the tabulation team reviewed the
posted results for this precinct. During this review, the tabulation
team discovered that the Memory Pack used at Precinct 141 had
inexplicably added randomly generated numbers to the totals that had
been reported. After examining the data from Precinct 141, including
the machine reports signed by the Precinct Captain, it was
determined that 326 votes were actually cast in that precinct,
instead of the 4,759 that were reflected in Report No. 2.
The tabulation team then conducted an audit of all results in
Precinct 141. This audit included rereading and reprocessing the
Precinct 141 Memory Pack and producing a new results tape. This new
results tape was compared to the results tape generated at the
precinct and signed by the Precinct Captain. This comparison
revealed that the results rendered on the new results tape and the
precinct-generated tapes were identical.
Next, the tabulation team reprocessed the cartridge and uploaded
the correct results for Precinct 141. After the second processing,
the correct total of 326 votes was successfully integrated into the
tabulated results. At this point, the Board resumed regular
uploading of the remaining Memory Packs and Results Cartridges and,
at 11:10 p.m. issued the third unofficial results summary report
(Report No. 3).
Finally, at 11:40 p.m., the Board issued the fourth and final
unofficial results summary report (Report No. 4). This report, which
indicated that 142 of 143 precincts had been processed, accurately
reflected the Election results with respect to ballots cast on that
day.
V. Findings
Upon reviewing the circumstances surrounding the release of
Report No. 2, the IRC concludes that:
A. The Board’s staff, in its zeal to meet the expectations of
the speedy release of unofficial Election Night results, erred by
deviating from its standard practice and failing to properly
review the erroneous data contained in Report No. 2 prior to its
release.
B. The Board’s staff unwarrantedly relied upon MPR
indications that data had been uploaded successfully; the MPR
success indications establish only that data has been uploaded,
rather than establishing that correct data has been
uploaded.
C. Sequoia’s determination, as reflected in its response to
the Board’s queries, that multiple possibilities regarding the
cause of the tabulation error exist, including: the speed at which
the Memory Packs were processed leading to some type of transient
malfunction of the MPR unit; the Memory Pack not making full
contact inside the MPR socket; or some type of electrical or
static discharge taking place while inserting, reading or ejecting
the cartridges at a rapid speed, is acceptable.
D. The tabulation anomaly, in which randomly generated numbers
were inexplicably added to vote totals, was confined to Precinct
141, and that no other Memory Packs or Results Cartridges
contributed to the erroneous data included in Report No. 2.
E. The tabulation anomaly was discovered within minutes and
identified and corrected within an hour.
F. Reports Nos. 3 and 4, released at 11:10 p.m. and 11:40 p.m.
accurately reflect the unofficial Election night results.
As a result of its review of all of the procedures, facts and
circumstances surrounding the Election night reporting error on
September 9, 2008, the IRC makes the following recommendations to
the Board:
A. Conduct a hand recount of the paper ballots from
Precinct 141 prior to certification of the election results.
B. Isolate the Precinct 141 Memory Pack and discontinue its use.
C. Decrease speed at which Memory Packs and Results Cartridges
are tabulated so as to reduce and/or eradicate the risk of the
incidence of static buildup and/or electronic impulses during the
tabulation process.
D. Release results in smaller increments so as to accommodate the
review and audit procedure associated with unofficial results
summary reports.
E. Review vote totals to look for indicators of unusual results
(e.g., large write-in vote numbers, more recorded votes than
registered voters).
F. Designate staff to oversee tabulation process and confirm that
established procedures for tabulating Memory Packs and Results
Cartridges are adhered to during the process.
G. Request documentation from Sequoia as to recommended
procedures for uploading, tabulating, and reviewing results on
election nights.
H. Develop a checklist for proofing tasks before releasing
unofficial results summary reports.
VII. Conclusions
The IRC’s thorough review, culminating in this report, reflects
an accurate representation of our work in examining the facts and
circumstances surrounding the release of Report No. 2. It is our
belief that this issue has shown that the Board’s established
procedures in reporting unofficial results on Election Night must be
adhered to strictly. Further, it is vital that the Board not succumb
to any external pressures to release results faster than can be
accomplished by following established procedures to the letter. In
short, the Board must remain steadfastly committed to accuracy over
speed.
|