The software utility, eVote and the eVote clerk, injects true democracy and deliberation into our real-world landscape.

Saturday, March 03, 2007

Fool Me Once: Checking Vote Count Integrity

A report in PDF mathematically analyzes the election audits proposed by the US Congress and finds that they would not protect all US House races; and proposes alternatives:

From: National Election Data Archive, Park City, UT
Contact: Kathy Dopp, President, NEDA 435-658-4657

To require voter-verified permanent paper ballots and increase voter confidence and accessibility, U.S. Representative Rush Holt of New Jersey and U.S. Senator Bill Nelson of Florida have sponsored virtually identical bills (HR811 and S559) that are intended to correct some of the significant, if unintended consequences of the
Help America Vote Act of 2002.

Election officials in most states have not subjected election results to independent outside audits – manual counts of randomly-selected voter-verified paper ballots.

Tiered election audits are audits which manually count more precincts in close races. When a race is close, a small number of miscounted votes can alter the outcome and a robust audit is needed to find miscounts which could be hidden in a few precincts. While there are several identified concerns with the Holt and Nelson bills, Kathy Dopp, President of National Election Data Archive, and NEDA have released a report "Fool Me Once: Checking Vote Count Integrity" evaluating the sufficiency of the Holt and Nelson audit provisions, as shown in the table.

HR811 and S559 election audit provisions ignore the variety in the amount of precincts in U.S. House districts. For example, a 10% audit in an 800-precinct House district requires an audit of 80 precincts, whereas a 10% audit in a 150-precinct House district triggers an audit of only 15 precincts, resulting in statistically insufficient power to detect miscounts in a close race where fewer than 1 in 15 miscounted precincts could alter the outcome.

A group of university professors wrote a letter in support of Holt's audit proposal, "Thoughts on Mandatory Audits". Dopp's says that her calculations agree with the professors' for the few cases that the professors evaluated but that their analysis was restricted only to U.S. House races with 400 total precinct vote counts for only a few selected margins; and did not consider House races with closer margins or smaller number of precincts.

"Fool Me Once: Checking Vote Count Integrity" shows how to evaluate the effectiveness of proposed election audits by plotting colorful charts of the probabilities for protecting race with particular margins between the leading candidates against the total number of precincts for Congressional districts.

According to Dopp, the Holt and Nelson audit provisions would not protect many U.S. House races where "protected" means that there is at least a 50% chance of discovering miscount under the assumption that sufficient miscount exists to alter an outcome.

"Fool Me Once: Checking Vote Count Integrity" suggests alternatives designed to protect all U.S. House races, including an alternative to fix the Holt and Nelson audit by requiring minimum numbers of precincts to be audited. To protect all U.S. House races, a larger audit rate is necessary for districts with a fewer precincts.

As with all legislation, cost is a consideration. Therefore an accurate cost estimate for conducting nation-wide manual audits of federal elections is needed. Dopp says that Congressional staffers asked her to calculate cost estimates for nationwide audits of federal elections, but that election data, including the margins and number of total precincts for all recent U.S. House and Senate races, is necessary for each state and for recent elections in order to perform the calculations.

Unfortunately states are not required to collect or make publicly available precinct-level election results (or other detailed vote counts, data, or election records that are necessary for public oversight over election integrity) so that this data required for calculating an accurate cost estimate for election auditing for Congress is very difficult to obtain and must be obtained from an organization like Election Data Services which has been collecting it laboriously from numerous jurisdictions for years, and can extract it from their databases.

Kathy Dopp, President of National Election Data Archive, recently met with Congressional staff in Washington DC to recommend changes to the proposed federal election integrity bills - statistically valid election audits and public access to election records and data necessary for citizen oversight.

Tax-deductible donations to defray the $1,500 cost to obtain the data for a cost estimate for conducting audits of all federal elections from Election Data Services are accepted at Please help us give Congress a good cost estimate for conducting nationwide audits of federal elections.

The report "Fool Me Once: Checking Vote Count Integrity" can be found at:

This release in pdf format with chart and table can be found at:

People are urged to contact their U.S. Senators and Representative, send them the report, and ask them to support sufficient nationwide audits of federal elections to protect all U.S. House and Senate races. See and


Post a Comment

<< Home