At the Board of Public Works Oct. 19th meeting, members passed without discussion a proposal by the State Board of Elections to pay Clear Ballot Group Inc. $275,000 for an "independent and automated solution to verify [the] accuracy" of the state's election results.
Seems reasonable, right? Especially now that the term "rigged" frequently precedes "election" in this year's campaign rhetoric. The only problem is it won't work.
We have some experience to back this judgment: Between us, we have helped audit about 20 contests in several states and designed auditable voting systems. Methods developed by one of us are in laws in two states.
It's great that Maryland voters get to vote on paper ballots this year; paper ballots that voters can check are the best evidence of "the will of the people." Maryland's ballots will be scanned and then counted electronically. As required by hard-won state legislation passed in 2007, the paper ballots will be stored securely as durable evidence of what voters wanted.
The next step in ensuring that the electronic count shows who really won is to manually review some of the paper ballots through an audit. But the recently proposed post-election "audit" falls short; it will not look at the marked paper ballots. Instead, Clear Ballots' "ClearAudit" software assumes the state's voting system scanned every ballot perfectly, and uses that information in its review. But no system is perfect; mistakes happen, equipment malfunctions. And some people want to make it look like the rightful winner lost.
There's no good reason not to use the actual ballots in the audit. Other states review the paper ballots to ensure that any tabulation errors didn't change the outcome of an election. And modern audits can be highly efficient; they review only a small random sample of ballots.
It is good that the board plans to review all votes, races and counties. The proposed auditing technology can detect many types of errors. But relying on the scans — which are as vulnerable as any other computer data — limits the kinds of problems the reviews can detect. The scans aren't like photographs; they can differ due to machine error, tampering or human error (for instance leaving out a batch of ballots or scanning the same batch twice).
A robust statistical audit of the electronic results against the paper ballots can produce strong evidence that election outcomes are correct; it can also correct incorrect outcomes. In this contentious election, it is extremely important to Maryland and the nation to audit election results against the actual paper ballots. It is not too late to plan and conduct a real audit. We would be happy to help.
Philip B. Stark (firstname.lastname@example.org) is a professor of statistics and associate dean of Mathematical and Physical Sciences at the University of California, Berkeley. Poorvi L. Vora (email@example.com) is a professor of computer science at The George Washington University.