Recounts Aren't Useless. They're Scientific.

Science relies on replication. Shouldn’t the same apply to counting votes?

Everything counts.

Photographer: Andy Manis/Getty Images

Election officials might not want to hear this, but the way we vote isn’t scientific. If they were conducted using the scientific method, recounts would be expected, maybe even mandatory. People would want to re-examine the raw data -- as former Green Party presidential candidate Jill Stein has been pushing to do in Michigan, Wisconsin and Pennsylvania.

Why? There’s no reason to assume elections are any more immune to errors than scientific studies, where replication is often a requirement for acceptance. And that means not just rechecking final results but either running an experiment again or re-evaluating the raw data -- akin to the hand recounts that Stein, as well as a number of computer scientists, have advocated for. Stein succeeded in Michigan, where a hand recount is expected to begin Friday, and lodged a partial victory in Wisconsin, where both hand and machine recounts started Thursday.  

This isn’t just an exercise in sore losing. Vote counts, after all, aren’t any more sacred than any other kind of measurement. It’s a key tenet of science that measurements, from weight to temperature to cholesterol counts, are imperfect reflections of reality. Scientists acknowledge this uncertainty with error bars -- if the error bars overlap, they can’t definitively declare one value bigger than the other.

Donald Trump won Michigan by a 0.28 percent margin, or about 13,000 votes. Could that be within that margin of error? Could a hand recount flip the result? Rebecca Mercuri, a computer scientist who studies voting technology, says it’s unlikely but not impossible.

A recount would let observers zero in on anomalies. In Michigan, for example, 84,000 people voted in state and local races, but apparently left the space for president blank. Typically the opposite happens, with voters leaving more blanks further down the ballot. Those people might have skipped the presidential candidates in protest , but Mercuri wonders why this wasn’t reflected in exit polls. If something was askew in the ballot-reading machines so they were looking for marks a fraction of an inch off the right spot, it could account for such a discrepancy.

Unfortunately, that kind of technical glitch would only be picked up in a hand recount -- the kind where a human eyes every ballot -- and not a machine recount. Electronic errors or deliberate hacks can go undetected by machines, Mercuri says, noting they can miss problems with scanners and vote-tabulation software. “As the saying goes with computers -- garbage in, garbage out,” she says. That could be a problem in Wisconsin, with its mix of hand and machine recounts.  

Still, an automated recount is better than none at all. What’s more, even recounts that don’t the change election results are valuable. If it weren’t for the Bush v. Gore recount in 2000, for example, we never would have known how bad punch-card machines are at measuring voter intent.

“The data from Florida 2000 has yet to be mined fully to learn all the lessons we can learn about election administration,” said Doug Jones, a computer scientist at the University of Iowa and co-author of “Broken Ballots: Will Your Vote Count?”. Even if no hacking is uncovered this time around, he said, a recount will be revealing.

Stein is still pushing for a recount in Pennsylvania. If nothing else, computer scientists say, holding one there would expose the scandal that so many of the state’s voting machines leave no paper trail. If voting were more scientific, nobody would use electronic machines that fail to record original data, retaining only totals. These touch-screen machines look technologically swanky, but from a scientific point of view, they’re unacceptable.

When scientists insist on transparency and independent verification, science can self-correct. Just a few years ago, astrophysicists from Harvard and Caltech went public with the finding that they’d recorded waves in space-time emanating from the origin of the universe. Rival groups re-evaluated their data and then used a separate observation to show the first group had only recorded a signal from foreground dust, not the big bang. The system worked.

Stein’s critics say there’s no reason for a recount because there’s no evidence of hacking. But in science, the investigation is the way you get evidence. What you need at the outset is a plausible hypothesis, and and computer scientists have shown an election hack is plausible by repeatedly hacking into voting machines. Scientists are spending billions to search for life on Mars and beyond, though no direct evidence exists. How can they get evidence if they don’t go out and look?

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.