How are state and local governments verifying the accuracy of the vote in this critical election year? One approach designed to simplify a major challenge for election officials as the ballots arrive is the risk-limiting audit, which enables double-checking only a sample of paper ballots to make sure they match electronic results. Miles O’Brien reports.

TRANSCRIPT

Judy Woodruff:

Finally tonight, just how are states and local governments checking the accuracy of the vote? It’s obviously an enormous issue this season.

Miles O’Brien is back tonight with a report on one approach that’s designed to help simplify a big challenge awaiting officials next week.

Philip Stark:

We try to be minimal waste in the kitchen. This is the carrot tops, carrot greens.

Miles O’Brien:

At his house, amid the trees, in the hills of Berkeley, California, mathematician Philip Stark is refining a recipe.

Philip Stark:

It’s a vegetable soup. It’s a homemade stock from whatever leftover bones I have accumulated.

Miles O’Brien:

For adding ingredients of trust and verifiability to elections in the United States.

He’s the originator of a big idea to make post-election audits much easier. It’s called a risk-limiting audit, a technique that reminds him of cooking and tasting soup, or, in the lexicon of mathematics, random sampling.

Philip Stark:

So, what is a random sample and why can you learn something about a big group by looking at just a small sample from the group?

Miles O’Brien:

After a good stirring, a tablespoon is all that he needs for a taste test.

Philip Stark:

I don’t need to drink some fixed percentage of the pot to tell if it’s salty enough. I need about a tablespoon, regardless of how the big pot is, if I stir it well.

Miles O’Brien:

Colorado was the first state to mandate risk-limiting audits in 2017. Nevada, Rhode Island and Virginia have followed suit.

In all, 46 states have some sort of auditing regime in place, but most are the traditional types, which require retrieving a much larger sample of ballots.

We went to Brunswick, Georgia, in September to see how a risk-limiting audit works outside the kitchen. It was a pilot audit of a state Senate run-off election.

Monica Childers showed the ropes to local officials. She’s with a nonpartisan nonprofit called VotingWorks, focused on building secure open-source voting and auditing systems.

Monica Childers:

So, whenever we go into a state, we try and do a number of local pilots of the process, so the election administrators can get used to it, get their feet wet, and actually get experience with what we’re doing.

Miles O’Brien:

These audits typically begin with volunteer voters rolling 10-sided dice to generate a truly random number.

Philip Stark:

Why do we use dice? There are several reasons.

First of all, we want the audit to be random. So, we want unpredictability. We want genuine probability, so that we can do probability calculations. But then we also want transparency and reproducibility and, to the extent possible, public participation.

Miles O’Brien:

The random number is fed into Stark’s formula, or algorithm. So is the number of ballots for each candidate and the desired accuracy.

The formula determines how many randomly selected ballots need to be located and double-checked in order to give certainty to the outcome. In an audit like this, the total number of ballots is not what’s important. It’s the margin between the candidates.

If the margin is wide, only a very small number of ballots may need to be reviewed.

Philip Stark:

And, at some point, it will become convincing evidence that that entity, that person, that position on the ballot got more votes. You can’t say ahead of time how many ballots you’re going to have to look at and all, because it depends on what you see.

Miles O’Brien:

So, that is not actually confirming the margin, necessarily.

Philip Stark:

No.

Miles O’Brien:

It’s just confirming who won and lost, right?

Philip Stark:

It is, exactly.

Miles O’Brien:

And that’s the important point here, right?

Philip Stark:

Absolutely. You hit the nail on the head. We are not auditing the exact count. We are just auditing the political outcome, who won.

Miles O’Brien:

In Glynn County’s portion of the state Senate run-off, 8,814 ballots were cast, and the margin between the two candidates was 1,668 votes.

They stopped the risk-limiting audit after locating and hand-counting 119 ballots. It was enough to give them 99.66 percent confidence that the outcome was correct.

Monica Childers is confident the audit would have caught any meaningful discrepancy, no matter the cause.

Monica Childers:

It could be that an election was hacked. It could be the tabulator developed a bug in the programming. It could be that the eye of the scanner got some dirt on it and so it wasn’t scanning as well as it was at the beginning of the process. It doesn’t matter.

Any of those things can be the cause of a tabulation error. If that tabulation error is great enough that it would have changed the outcome, a risk-limiting audit will catch it.

Miles O’Brien:

But will it?

Georgia is among 14 states using a hybrid between paper and electronic voting machines, touch screen computers that print marked ballots. And with a computer between voter and ballot, there is no true record of what choices appeared on the screen or whether what’s in the Q.R., or bar code, read by the tabulating scanner matches the voter’s choices.

Philip Stark:

If ballot-marking devices were used to mark the bulk of the ballots, then we’re never going to know what the right outcome is. Even a full hand count of the paper isn’t going to reveal that necessarily.

That said, it’s still better to audit than not to audit, right? We ought to check whatever piece of the process we can check.

Miles O’Brien:

Unlike the soup, which looked like a sure thing to me, although, after tasting it, I thought of demanding a recount.

For the “PBS NewsHour,” I’m Miles O’Brien in Berkeley, California.

Judy Woodruff:

First time we have seen soup connected to counting the vote.

Thank you, Miles O’Brien.