Statistical Analysis of the Post Election Audit Data 2007 November Elections

The Center for Voting Technology Research (VoTeR Center) at the School of Engineering of the University of Connecticut received the data gathered in the post-election audit performed in the State of Connecticut following the November 2007 election. The audits of the randomly selected 10% of the districts were conducted in November and December of 2007, and the returns were conveyed by the Office of the Secretary of the State to the VoTeR Center on January 8, 2008. The audit data received by the Center contains 958 records, where each record represents information about a specific candidate. Specifically, each record contains the following significant information: date, district, machine seal number, office, candidate, machine counted total, undisputed hand counted total, questionable hand counted total, overall hand counted total, that is, the sum of undisputed and questionable ballots. This report contains several statistical analyses of the audit returns and recommendations. Among the 958 records received by the Center, 175 records (18.3%) were incomplete, unusable, or obviously incorrect. Another 111 records (11.6%) contained usable, but incomplete data, or minor arithmetic errors. Thus about 70% of the audit records were complete and contained no obvious errors. While some problematic records were clearly due to human error (e.g., in addition), this suggests that auditors found the audit instructions to be ambiguous or insufficiently specific.Thus one immediate recommendation is to revise and improve the instructions and the audit procedures, and to refine the definitions of the data to be reported. The statistical analysis in this report deals with the 783 records that are sufficiently complete to perform the analysis. Among these records, 520 records (66.4%) show discrepancy of 0 or 1 votes between the machine counts and audit hand counts, and 700 records (89.4%) show discrepancy of 5 votes or lower, while 31 records (4.0%) show discrepancy of 10 or more votes. Given that the ballots considered questionable might be miscounted by the machine, reducing the number of such ballots in the totals results in 716 records (91.4%) showing discrepancy of 5 votes or lower. In this report no attempt was made to correct errors in audit reporting. For example, the largest reported discrepancy of 74 votes is due to an obvious error, where the actual discrepancy is 2 votes. On the average the reported machine counts tend to be overcounts by 1 vote relative to the hand counts, and the average absolute value of reported discrepancies is 2.2 votes per race, where the average count consisted of 277 votes This analysis was performed on request of the Office of the Secretary of the State.

Full report: audit07-h-080410


This entry was posted in news.