Critique of H.R. 550 Audit Methodology and EDA's Proposed Solution
Fundamental Shortcomings and Proposed Solutions
By Bruce O’Dell, Jonathan D. Simon, JD,
Josh Mitteldorf, PhD, and Steven Freeman, PhD
on behalf of Election Defense Alliance.
The purpose of this paper is to assess the accuracy and efficacy of the proposed election audit mechanism defined in HR 550, “Voter Confidence and Increased Accessibility Act of 2005,” sponsored by Representative Rush Holt of New Jersey. The audit requirement described in Section 5 of HR 550 calls for a mandatory hand count of 100% of the ballots in at least 2% of the precincts in each state, for each general election for federal office and, optionally, for state contests. This paper assesses the ability of the HR 550 hand count audit scheme to reliably uncover deliberate or accidental corruption of an electronic vote tally, through two complementary methods: (1) theoretical analysis of the statistical power of the HR 550 audit protocol and (2) computer simulation of the HR 550 protocol as it would be applied in auditing elections to the United States House of Representatives.
Our key finding is that in a typical U.S. Congressional race a hand-count audit of 100% of the vote in a random 2% of the precincts would fail about 40% of the time to detect vote count corruption large enough to alter the outcome. This result is derived theoretically and confirmed by computer simulation. Even in those cases where HR 550 can detect a discrepancy, there will often be only a single precinct in which corruption is detected. We question whether such a finding would be sufficient to trigger a recount, given a real world in which public perceptions have already been framed, and political
pressures to accept the initial count are substantial. HR 550 offers no guidelines or criteria for a mandatory recount. Instead, the decision is left to a commission appointed by the President and inherently subject to partisan pressure.
The problem with HR 550 is not solvable simply by sampling more precincts. In order to have a 99% confidence level of detecting outcome-altering vote count corruption affecting a small number of precincts in an average U.S. Congressional District, the HR 550 protocol would need to hand-count up to 65% of the total precincts. An alternative to HR 550 with high confidence of detection of outcomealtering vote count corruption is described, and its accuracy was also simulated; those results are presented as a potential alternative to HR 550 well worth exploring.
Since a 10% hand-count sample could be drawn in 100% of precincts on election night, the alternative has additional practical advantages: minimizing chain-of-custody concerns which are inherent to the HR 550 approach; starting the transition to universal hand-counting of elections; and placing responsibility for the integrity of the vote count in the hands of the American people, where it rightfully belongs.
1 See www.ElectionDefenseAlliance.org for further information on Election Defense Alliance and for biographies of the
authors. Josh Mitteldorf is affiliated with the Department of Ecology & Evolutionary Biology, University of Arizona;
Steven Freeman is on the faculty of the Center for Organizational Dynamics at the University of Pennsylvania.