Dan Tokaji's Blog
Professor Dan Tokaji
Election reform, the Voting Rights Act, the Help America Vote Act, and related topics -- with special attention to the voting rights of people of color, non-English proficient citizens, and people with disabilities

Dan Tokaji's Blog Links Publications & Working Papers
Equal Vote
Saturday, October 8
Electionline on Recounts
Electionline.org has released this report on recounts. Entitled "Recounts: From Punch Cards to Paper Trails," it places particular emphasis on the rules regarding voter-verifiable paper audit trails in those states that require them. Here's a summary from electionline.org's press release:
A new report finds that while 25 states will require the use of paper trails in time for the 2008 presidential election, so far only 14 states currently plan to use them as the official record in a recount of votes. How or if they would be used in recounts -- and how difficult that process might be -- are questions many states still need to answer....

"It would appear that more and more states are making the decision to require the use of voter-verifiable paper audit trails (VVPATs) with their electronic voting machines. But what they will do with them, aside from putting voters' minds at ease, is likely to be the subject of sharp debate in the immediate future," said Doug Chapin, electionline.org's director....

Among the other findings in the report:

- The most publicized instances of contested elections have been in the courts -- the Bush v. Gore decision in the 2000 election and the 2004 gubernatorial recount in Washington are perhaps the best known recent examples. Recounts, however, are far more common in local races and most states have established rules to avoid such legal battles.

- States vary in their "triggers" that initiate ballot recounts. In 39 states, losing candidates can request recounts, largely at their own expense. In 25 of those, a recount can be undertaken regardless of the margin of votes while in 14, the difference must be within a certain margin of votes.

- Sixteen states have rules that require an automatic recount of votes if contests fall within a prescribed margin.

- Four states -- California, New York, West Virginia and Kentucky -- recount a small percentage of ballots from randomly-chosen precincts to test the accuracy of the vote.

Eleven states require manual audits of VVPATs. Audits test the accuracy of electronic voting machines by comparing paper totals to digitized machine totals, but do not change the outcome of elections. The most recent audit in Nevada found a 100 percent match between paper and machine totals, but only after teams took as long as four minutes to count accurately a single ballot on a paper audit trail.
The report also features a summary of state recount laws. One of my research assistants and I prepared a table summarizing the states' laws several months ago, available here.

My take: One of the things to keep in mind is that the reasons for conducting recounts differ, depending upon the type of system used. With a paper-based system such as punch cards or optical scans, recounts may be necessary in deciphering ambiguously marked ballots in accordance with state law. This is something that a machine count may not be effective in doing. In addition, they can be useful in verifying that tabulating machines are counting accurately. This is what was done in Ohio after the last election, when a certain percentage of ballots were recounted by hand, and the totals compared to those generated by the machine count. If the two match, it's a strong indication that the tabulators are working properly.

On the other hand, the reasons for conducting a recount are different with a direct record electronic system. The concern is no longer with ambiguously marked ballots, which aren't an issue with this type of equipment. Instead, the main reason for a recount of the paper ballot copies generated by the VVPAT is to make sure that the machine is accurately recording the electronically voted ballots -- for example, that the machines aren't displaying the intended choice to the voter (say Kerry) on the review screen that comes up at the end of the voting process, while internally recording another (say Bush). With the VVPAT, the theory goes, at least some voters will actually check the paper record. On that theory, if there's a discrepancy between the paper ballot totals and the electronic totals, the paper total should be the one that counts.

There are a couple of problems with this. One is that the little research that exists suggests that many voters won't actually check the VVPAT records. See here for my prior post on research that Ted Selker of MIT has done on this.

The other problem is that, if only a small percentage of paper records (say 1 or 2% of ballots) are recounted, it's not likely that errors will be detected, at least for local and regional races. As an example, Andy Neff ran the numbers for a California congressional race. He found that if 10% of the votes were changed -- more than enough to alter the result in a reasonably close race -- then a 1% recount has only about 40% chance of catching it. While I'm no statistical expert, and thus can't confirm the accuracy of Neff's calulations, I've yet to see anything that shows them to be wrong. If he's right, or at least in the ballpark, it means that a 1% recount of VVPATs will be of limited effectiveness in catching malicious software that might turn an election. And this is aside from mechanical problems such as paper jams, and from the logistical problems of counting all those strips of paper.

Of course, there's always the possibility of a candidate or voter requesting a full manual recount, at least in those states that allow it. But those are rare and expensive. Moreover, if we hypothesize a clever hacker able to insert malicious software without detection, there might be no reason for suspicion on the losing candidate's part, and thus no reason to call for a full-scale manual recount.

My point here is that the introduction of electronic voting equipment requires that we rethink the means adopted to promote and verify the accuracy of results. A 1% or 2% recount may be effective in verifying the accuracy of tabulating equipment with paper-based systems, but probably won't do much good to ensure the accuracy of elections conducted with electronic equipment, even if there's a VVPAT.

Powered by Blogger Site Meter

Moritz College of Law The Ohio State University