We have looked closely in the past at the importance of post-election audits and the essential role that they play in guaranteeing election integrity. Many state and local governments have been slow to institute adequate audit procedures, but that could soon change. Colorado's Boulder County has taken a big step forward with a comprehensive auditing strategy that leverages open source software to boost election integrity.
Boulder County adopted ElectionAudits, an open source software program that implements a "risk-limiting" audit system modeled after the concepts described in Principles and Best Practices for Post Election Audits—a guide that is endorsed by the Brennan Center for Justice, Verified Voting, and a number of various state-based election integrity advocacy groups.
In a risk-limiting audit, the sample size of randomly selected ballots in each precinct is based on the margin of victory. A higher volume of records will be analyzed in tight contests where the winner's lead was narrow. Experts say that this kind of audit is more reliable and statistically rigorous than audits that use a predetermined fixed percentage of ballots for every precinct.
ElectionAudits supports several different techniques for computing sample sizes in risk-limiting elections, including the "negative-exponential" method (NEGEXP) and the "probability proportional to error bounds, with replacement" method (PPEBWR). Boulder County used NEGEXP, which was first introduced in a paper that was published earlier this year by Northeastern University researcher Javed A. Aslam and MIT researchers Ronald L. Rivest and Raluca A. Popa. Their paper, On Auditing Elections When Precincts Have Different Sizes, compares several different methodologies and suggests that NEGEXP "appears to have near-optimal efficiency."
The ElectionAudits software, which was largely developed by Neal McBurnett, is built with the Python programming language and the Django framework. The source code is available for download from the project's version control repository and is distributed under the MIT license. Some of the source code for the election computations was adapted from a Python script written by Rivest, who is best known for his role in creating the RSA encryption algorithm and is a member of the Election Assistance Commission's technical guidelines development committee.
ElectionAudits was used in Boulder following the general election to conduct what election officials say could be the most extensive risk-limiting audit in history. The county plans to encourage broader adoption of the methodology and hopes that it will be endorsed by the Colorado Election Reform Commission.
"The overall results give us more confidence than ever in the reported results for this election, based on the close agreement between the hand counts and machine counts," a representative of Boulder County said in a statement. "The audit also set new standards for transparency and verifiability. All the detail on results to be audited was published on the web before the random selection of batches to be audited was done. Peer-reviewed procedures for ensuring verifiably random dice throws and selections were used."The development of the ElectionAudits software is a big step in the right direction for election integrity and accountable democracy. Modern voting technology poses significant challenges, but it is clear that open software and collaborative innovation can make a big difference and help ensure that the election process remains transparent.