By John Marion

It has quickly entered the folklore of Rhode Island politics: the election decided by a single vote after four different counts. What really happened in the Democratic primary in House District 58 and what lessons can we learn from it? Here’s what I saw when I attended last week’s recounts and what I think we should do differently.

On the night of Sept. 11, Rhode Islanders were told District 58 had cast 543 votes for veteran incumbent Rep. William San Bento and 540 for challenger Carlos Tobon. But the winner was not yet clear, because while that initial count included all mail (or absentee) ballots, it didn’t include 11 “provisional” ballots.

Provisional ballots are cast for a variety of reasons – for example if a voter arrives at a polling place but isn’t listed on the rolls or, starting this year, if a voter doesn’t show up with an approved form of idenitifcation. The local canvassers then must decide whether the provisional ballot should be counted. The Pawtucket Board of Canvassers counted only three of 11 provisionals that were cast in the District 58 primary – it’s not clear why the other eight were rejected.

Because the margin of victory in the San Bento-Tobon race was less than 2%, state law mandated a specific type of recount – a manual “re-feeding” of all the ballots through the optical-scan voting machine. (Races decided by margins of 2% to 5% are recounted by simply reloading the machines’ cartridges into the computer – without first re-feeding all the ballots – and getting the totals.)

At this point the potential outcomes of the District 58 primary looked pretty obvious – unless Tobon was the choice on all three provisional ballots allowed by Pawtucket’s canvassers, San Bento would win. But once the R.I. Board of Elections started its recounts on Sept. 17, the picture quickly got cloudier.

First off, there are different rules for counting different types of ballots. Ballots that are cast and counted at an actual polling place can only be recounted by re-feeding them through the optical-scan machines; mail/absentee and provisional ballots can be examined by the Board of Elections to determine “voter intent.” That means the board can scrutinize those ballots to figure out who the voter was actually trying to choose.

When the District 58 ballots were fed through the machine on Sept. 17 one fewer vote was counted than on the night of the primary – a so-called “undervote.” Two of the three provisional ballots were also for Tobon. Suddenly the race was a dead heat – tied at 543 each. There was considerable confusion at that point about what could be done to break a tie, because state law is somewhat unclear on the subject. One interpretation would have allowed the city or town committee, which includes San Bento himself, to pick the winner.

(One of those provisional ballots also showed why it matters whether they can be scrutinized by actual humans to determine voter intent. I watched over the board members’ shoulders as they examined a ballot where a voter had crossed out Tobon’s name rather than connect the arrow next to it. If the ballot had been fed through the machines instead, it would have rejected as an “undervote.”)

Immediately after the first recount, the search began for the missing ballot. It was eventually found when a high-speed scanner was used to examine all the mail ballots cast in the state; a missing District 58 one was found and put back. The board then fed through the missing ballot and all the others ones for a second time. When those results were combined with the hand-counted mail and provisional ballots, San Bento was the winner 545-543. Not only was San Bento the choice on the missing mail ballot – another ballot, previously counted for Tobon on primary night and in the first recount, was now an “undervote.”

A third recount was ordered (actually the fourth attempt to determine the election’s outcome, when the night-of results are included). Once again the totals changed – Tobon gained back a vote when the “undervote” went back to being a vote for him, making it 544-543 in San Bento’s favor.

The Board of Elections certified the 545-544 result on Sept. 19, but not before several hours of hearings on the dispute over District 58. Among the revelations at those hearings was the discovery that there was one more vote cast at the Nathanael Greene School polling place than there were ballot applications. (Those are the blue pieces of paper that voters sign before they cast their ballots after a worker affixes a sticker to it with their names.) Tobon’s attorney asked for a hand recount of every vote in the race – regular, mail and provisional – but the board rejected his request, saying machines are more accurate than humans.

* * *

What lessons can we learn from this episode?

First and foremost, both machines and humans are fallible. The system broke down despite extensive poll-worker training, the pre-election calibration of scanners and elaborate controls over completed ballots. At Nathanael Greene, one voter cast a ballot without any record of him or her checking in. At the polling place and then again at the Board of Elections a machine went back and forth on at least one ballot, counting it as a vote for Tobon or no vote at all during different counts. Another ballot was put in the wrong box and only found after an extensive search. And this week we hear from The Providence Journal’s Katherine Gregg that one mail ballot wasn’t counted because it took too long to arrive. The result of these failures was confusion for the candidates, the press, and the public.

Yet there are some silver linings in District 58′s experience.

First, Rhode Island is lucky to have verifiable paper ballots that can be recounted, unlike some other jurisdictions where voting is entirely electronic. Back in the 1990s, state leaders did right by citizens when they decided paper ballots were best and mandated their use in state law. Without paper ballots all we could do is trust that the machines work, when we know that sometimes they don’t.

A second lesson is the value of transparency in government. Transparency is a term that gets thrown around a lot, but this was transparency in action. Though I didn’t represent a party or a candidate, as a member of the public I was able to watch everything that occurred during the recounts, and even take pictures and ask questions. I have confidence that nothing devious was done because I literally watched each round of tabulations. (The same cannot necessarily be said about polling places, because the candidates select the poll-watchers – something Rhode Island should look at.)

That said, we can – and should – make improvements to how Rhode Island votes.

We need new voting machines. The Board of Elections recognizes this and is looking to buy or lease new machines in the near future. The public needs to be assured that any new machines do not take us backward by moving away from verifiable paper ballots. But with our old machines slowly failing, the time has clearly come.

We need to re-examine our practices for reconciling ballots. The issue at Nathanael Greene should have been identified on the night of the primary, instead of more than a week later – and even then only because the results were contested.

We need to implement a system of post-election audits. Rhode Island conducts extensive testing of its voting equipment prior to Election Day, but we need to put an equal emphasis on making sure the outcomes that the scanners report are accurate – by comparing them to a randomly select sample of the actual paper ballots. That’s because machines, and the humans who program and operate them, are not infallible. In fact, there are examples of the very brand of scanners Rhode Island uses – those sold by Omaha-based Electronic Systems & Software (ES&S) – malfunctioning in a way that could potentially be caught by an audit.

Common Cause, the Verified Voting Foundation and Rutgers School of Law released a report last month emphasizing these three points.

Another policy Rhode Island could embrace would be making voted ballots available for the public to audit themselves; a group of citizens in Humboldt County, Calif., uncovered 197 missing ballots when they crowd-sourced their own election audit.

There’s also a bigger picture beyond the mechanics of vote-counting. Nearly 7,000 voters were eligible to cast a ballot in the District 58 primary, but only a little more than 1,000 of them went to the polls. Why? What is it about our democracy that either discourages or prevents legitimate registered voters from exercising their constitutional right to choose their representatives? Why did the incumbent, Rep. San Bento, get to serve on the commission that redrew the state’s legislative districts last year – allowing him to essentially choose his own voters instead of the other way around?

Those are important questions that Rhode Islanders should debate in the weeks and months to come. In the meantime, there are important lessons to take from what we saw last week in District 58 – and lots of work for Common Cause and its allies to undertake as we move forward. •

John Marion is executive director of Common Cause Rhode Island.

An earlier version of this article gave an inaccurate vote total for the final recount.

(photo: John Marion)