Secretary of the State Susan Bysiewicz and Dr. Alex Shvartsman of the University of Connecticut’s Voting Technology Research Center (VoTeR Center) announced the results of two reports that confirmed the security and accuracy of Connecticut’s November 2008 elections.
“My office entered into this historic partnership with the University of Connecticut VoTeR Center so that we could receive an independent, unbiased accounting of Connecticut’s optical scan voting machines,” said Bysiewicz. “The results of these two studies confirm that numbers tallied by the optical scanners were remarkably accurate on Election Day November 4, 2008. Voters should feel confident that their votes were secure and accurately counted.”
Proposed by Secretary Bysiewicz and approved by the General Assembly in 2007, the purpose of post election audits is to ensure that the optical scan voting machine(s) used in the election or primary accounted for all votes properly. A total of 84 precincts were randomly selected for the hand count audit, representing 10% of all polling places where ballots were cast during the General Election in 2008. The audits were conducted by local election officials. As part of UConn’s report, over 1,300 records of races were reviewed by the VoTeR Center following the local audit process. Within that sample, only 49 or 6% showed a discrepancy between machine counts and hand audits of more than 5 votes, with the largest discrepancy being 9 votes.
Bysiewicz added, “Overall, I am pleased to see that the audits were conducted successfully and accurately, and thank all of the local election officials, poll workers, and others who participated in this effort. Initially there were some larger than expected discrepancies between the machine totals reported on Election Day and the subsequent hand count audits, but further investigation by my office and the UConn researchers found that many of these turned out to be human error during the audit process. Moving forward, my office will continue to improve the training we give to Registrars of Voters and local election officials to reduce any further errors in counting. We will also start new training within weeks to improve the security of memory cards used by the optical scanners to record votes on Election Day.”
Commenting on the report’s findings, Dr. Alex Schvartsman of the UConn VoTeR Center said, “In the final analysis, no machine-attributable inaccuracies were identified in the 2008 General Election. In all cases where noteworthy discrepancies were initially reported, it was determined that the causes of the discrepancies are in hand counting errors or errors in vote allocation during the audit. For other audit returns, the reported discrepancies were low. Given that the average discrepancy is substantially smaller than the reported number of the votes the auditors considered questionable, it may be reasonable to conclude that the main source of reported discrepancies are ambiguously marked ballots.”
As part of an agreement with UConn, before and after each election, computer scientists at the university test a sample of memory cards used to tabulate votes in the optical scan machines to ensure that each card was programmed correctly. The purpose of the pre-election scans is to detect evidence of tampering, programming errors or any irregularities that might affect the voting process or election results. Pre-election testing of memory cards has made Connecticut a national leader in voting security. Before the election, UConn tested 620 memory cards. Results demonstrated that none of the memory cards tested showed evidence of tampering. 91% of the cards tested were found to have been properly programmed prior to the election. 8.9% of the memory cards were found to have contained “junk” data meaning they were unreadable. As such, the tabulators would not have accepted them for use in the election.
Following the election, UConn tested a total of 462 memory cards, including some 279 that were used in the General Election on November 4, 2008. Testing, once again, showed that none of the memory cards were tampered with and that they performed correctly. Of the cards used in the election, 100% were properly programmed and none showed evidence of tampering. Overall, 91% of all memory cards analyzed in the sample (421) were properly programmed, while 9% (41 cards) contained “junk” data and were unreadable by the machine tabulators. UConn researchers are currently performing additional analysis to determine the cause of this. A separate report will document their findings.
“Overall, I’m pleased that our first pre- and post-testing procedures with UConn demonstrate the security of our office’s chain of custody practices with election officials,” said Bysiewicz. “However, the percentage of unreadable cards is still too high and we await UConn’s forthcoming investigation into possible causes and recommended solutions for guidance on this issue. In the interim we will provide additional training to local election officials to make sure regulations concerning the handling and security of memory cards used by the optical scanners are uniformly followed throughout the State of Connecticut.”
For the full text of the reports visit the UConn Voting Technology Research Center online at: http://voter.engr.uconn.edu/voter/news/ or http://voter.engr.uconn.edu/voter/reports/