Statistical Analysis of the Post-Election Audit Data 2010 August Primary Election

The University of Connecticut Center for Voting Technology Research (VoTeR Center) received the data gathered in the post-election audit performed in the State of Connecticut following the August 2010 primary election. The audits involved the randomly selected 10% of the districts and the audit returns were conveyed by the Office of the Secretary of the State (SOTS) to the VoTeR Center from September 21^st (almost all returns) to October 26^th of 2010. The audit data received by the VoTeR Center contains 465 records, where each record represents information about a given candidate: date, district, machine seal number, office, candidate, machine counted total, hand counted total of the votes considered unquestionable by the auditors, hand counted total of the votes considered questionable by the auditors, and the hand counted total, that is, the sum of undisputed and questionable ballots.

The review of the audit reports prepared by the towns did not reveal any returns with high unexplained differences between hand and machine counts. The largest discrepancy is a single case of 6 votes (4.5%), the second largest is a single case of 4 votes (0.99%). The rest show discrepancies of one or less, with 95.7% of the returns showing no discrepancies. There are in total twenty records with discrepancies. We note that in seven cases out of twenty the auditors provided an explanation for the observed discrepancies. Majority of the explicated cases refer to the improper markings of the ballots by voters. One of the examples of such an improper marking is having ‘X’ mark in the bubble instead of filling in the bubble. The optical scan machine used in Connecticut, the AccuVote optical scanner, is able to recognize partially marked bubbles in some cases, but this depends on the bubble fill pattern.
Read the rest of this entry »

Pre-Election Audit of Memory Cards for the August 10, 2010 Primary Elections

The University of Connecticut Center for Voting Technology Research (VoTeR Center) performed a pre-election audit of the memory cards for the Accu-Vote Optical Scan tabulators that were to be used in the August 10, 2010 elections. The cards were programmed by LHS Associates of Methuen, Massachusetts, and shipped to Connecticut districts for use in the elections. The VoTeR Center received in total 226 memory cards from 218 districts for this audit. The cards were received during the period from August 6, 2010 to August 17, 2010. This document reports on the findings obtained during the audit.

Among these 226 cards, 192 (85%) were correctly programmed for elections. These cards contained valid ballot data and the executable code on these cards was the expected code, with no extraneous data or code on the cards. Concerning the remaining cards, 34 (15% of the total number of cards) were found to be unusable by the AV-OS. In particular, these cards contained ‘junk’ (i.e., apparently random) data. These cards were unreadable, according to the tabulators and could not have been used in an election. (We report on the causes of such card failures separately – See here.)
Read the rest of this entry »

Determining the Causes of AccuVote Optical Scan Voting Terminal Memory Card Failures

Optical scan (OS) voting systems play an increasing role in the United States elections, with over 40 states deploying such systems. The AccuVote optical scanners (AV-OS) manufactured by ES&S account for over 20% of all OS systems. OS systems typically use removable media (cards) to provide election-specific programming to the scanners and to convey precinct election results for central tabulation. Several reports document occurrences of AV-OS memory card failures, with up to 15% of all cards failing in some cases.

This paper reports on determining the causes of memory card failures that lead to complete loss of data from the card. An initial experimental analysis identified the battery discharge as a significant part of the problem. This finding led to the question of the dependability of the builtin function of the AccuVote OS system that issues a warning when the memory card contains a low-voltage battery. We identified the components used to implement this function in one type of AccuVote memory card. Using the specifications of the commodity batteries that are used in these cards, we determined the time interval from the instant when a battery warning is issued by the AccuVote to the point when the battery does not have enough voltage to retain data on the memory card. We show that such interval is about 2 weeks. Thus timely warnings cannot be provided to protect against battery discharge and loss of data during the election process. The factors contributing to the short warning interval are likely to apply to other battery-backed RAM cards, such as those used in the ES&S Model 100. Recommendations for mitigating the problem are made in light of the expected behavior of the warning system.

Research funded by the Secretary of the State of Connecticut and performed at the Center for Voting Technology Research at the University of Connecticut.

Full Paper: evt2010