Rice University worked with Clemson University to examine two common ways to recount votes. One of them is the "read-and-mark" method, in which four auditors check each other's work as they plow through an unsorted stack of ballots. The other is called "sort-and-stack". That's when three auditors independently count ballots pre-sorted by candidate. Michael Byrne is associate professor of psychology at Rice. He says the sort-and-stack method produced a margin-of-error of up to 2 percent. That's about twice the error rate of read-and-mark.
"If something goes wrong, it's much easier to back up the procedure a little bit to a point where everybody agrees what everything should be, in read-and-mark than it is in sort-and-stack. So I think that's probably why sort-and-stack is a little bit worse than read-and-mark."
Professor Byrne says more accurate methods of recounting votes have been developed in the last few years.
"They involve checking records of single ballots. And the number you have to check is a function of what the margin of victory was. These are called 'risk-limiting audits'. But it's mathematically complicated, and it's very difficult to explain. But, in fact, we can prove we can actually be much more confident if we use these risk-limiting audits than if we use a full hand count."
Byrne adds that errors in hand recounts typically don't affect elections on a state or national scale. That's because the margin of victory from individual precincts tends so wide that a hand recount error usually doesn't change the results. The complete study will appear in an upcoming issue of the Election Law Journal.