Collaborative Public Audit
of the
November 2006 General Election
pursuant to the charge from the
Cuyahoga County Board of Elections
Final Report – Submitted on April 18, 2007 by
The Collaborative Audit Committee
Democratic Party of Cuyahoga County
Leslye Huff, J.D.
Amos Mahsua, C.P.A.
Republican Party of Cuyahoga County
Gordon Short, J.D., C.P.A.
League of Women Voters
Kathy Woodbridge (Cuyahoga Area)
Kurt F. Miller, Ph.D. (Shaker Heights)
CASE-Ohio (Citizens’ Alliance for Secure Elections)
Ron Olson
Dan Kozminski
Greater Cleveland Voter Coalition
Roslyn Talerico
Joyce Porozynski
Coordinator
The Center for Election Integrity, Cleveland State University
Candice Hoke, J.D., Project Director, Public Monitor
Abigail Horn, M.A., Audit Coordinator
Audit Methodology and Statistical Analysis
The Northern Ohio Data and Information Service (NODIS)
Ellen Cyran, M.S.
Mark Salling, Ph.D., Director, NODIS
The report can be found here:
http://urban.csuohio.edu/cei/public_monitor/cuyahoga_2006_audit_rpt.pdf Conclusions
From the limited scope of this audit, which examined the results of one race as recorded on early
absentee optical scan ballots that were part of the unofficial count, we may conclude the
following:
Conclusion 1
Election result data in the GEMS report corresponded closely to the results obtained by
the audit hand count of the optical scan ballots
Audit results either matched exactly or were discrepant in a manner and degree consistent
with the number of ballot pages misfiled for 57 of the 72 precincts included in the audit.
Fourteen of the 15 precincts that did not exactly match were discrepant by plus or minus one
vote with an aggregate of one more vote for Sykes and three more for Taylor found by the
audit. This is a low net error rate out of a total of 3628 votes. The one other discrepant
precinct was not reported in the unofficial SOVC at all and represents an apparent scanning
procedural error.
Conclusion 2
The sorting process for early absentee optical scan ballots into precinct batches prior to
scanning was neither complete nor accurate.
A total of 1,768 “decks”25 were created in which the early absentee ballots from the 1,434
precincts in Cuyahoga County were placed. Of these 1,768 decks, 201 contained ballots
from more than one precinct. The election reports also show some ballot pages of 307
precincts were separated (possibly misfiled) into more than one folder.
Prior to commencing the scanning of the early absentee ballots, the CCBOE staff hand sorted
a total of 66,228 absentee ballots into precinct decks. We were able to identify patterns of
misfiling: often the ballots were misfiled into precincts where extremely similar precincts
codes to the correct code were used. These codes often differed by only one character. This
pre-sorting was necessitated by concerns about the GEMS database’s limitations and its
stability over the period of uploading optically scanned ballots. Although the sorting was
imperfect, without it we would not have been able to conduct this audit and compression of
the GEMS database—and its consequent risks -- would have had to occur much more often.
Conclusion 3
Some discrepancies that occur at a very low frequency appear to be caused by a
scanner misreading of some of the optical scan ballots.
In several of the discrepant precincts, the correct number of ballots was identified in the
initial audit and in a subsequent follow-up visit. However, the GEMS reported results
differed in a pattern suggesting that one or more ballots that the auditors considered to have
been clearly marked in either blue or black ink had not been accurately read or recorded by
the scanners/GEMS. In these cases, one fewer vote would be reported in the GEMS data file.
Conclusion 4
Some ballots were apparently scanned at some point but were not included in the
GEMS results or on the SOVC election results report.
At least four precincts for which early absentee ballots had been returned showed no votes
recorded in the unofficial election result tabulations. In at least three cases, it appeared from
the ballot folder documentation that ballots had been scanned and must have been deleted but
not rescanned.
Conclusion 5
Some ballot batches were scanned twice with their votes double-counted when the
GEMS unofficial results and the SOVC election results report is analyzed.
The ballots in at least two precinct folders appear to have been scanned twice. The numbers
of optical scan ballots present in the folder was consistent with the number of absentees
returned while the results reported in the election result tabulation for these precincts were
approximately doubled.
Conclusion 6
Batch/deck numbers identifying specific groups of ballots may change within the
scanning process and between scanning events.
The deck/batch identification is arbitrarily assigned by the “header card” that is placed at the
front of a stack of optical scan ballots prior to their being scanned. As reflected above, a
batch can be rescanned with a different header card. Similarly, some of these ballot batches
were given different digital batch identities when they were re-scanned for the official count.
The change in ballot batch identifiers greatly impeded the accurate tracking of batches so that
they could be confirmed as having been counted, and counted only once in the election
tabulation.
LENGTHY DETAILED REPORT AT LINK
Printer-friendly format
Email this thread to a friend
Bookmark this thread
(1000+ posts)
Send PM |
Profile |
Ignore
