Home
Site Map
Reports
Voting News
Info
Donate
Contact Us
About Us

VotersUnite.Org
is NOT!
associated with
votersunite.com

Improving the count: Oklahoma, others pose solutions for a better election system in Boulder County (CO)
Boulder Daily Camera. November 9, 2008. By Laura Snider

At first, election watchdogs and transparent voting activists rejoiced.

In January 2004, the Boulder County Commissioners agreed to write a check for $1.4 million to buy optical scanning equipment for the November presidential election. The new system, purchased from Hart InterCivic out of Texas, allowed the county to use paper ballots for the election, which left a verifiable paper trail for auditors and election skeptics to follow.

The new Ballot Now machines replaced a punch-card system, which had been cast into the shadows of public skepticism after the hanging-chad controversy in Florida during the 2000 presidential election.

Voters in Boulder County loudly criticized buying “direct-recording electronic” devices, commonly called DREs, because they feared that computer errors could easily create faulty election results. On the other hand, optical scanners — while still computerized — could be easily checked. And, in the worst case, they could be scrapped altogether in favor of hand-counting the ballots.

But it took only eight months before problems that still dog the county’s election process showed up. Boulder County was now slow — very, very slow.

In August 2004, Boulder County lagged hours behind other Colorado counties in posting results of the primary election. But the real crunch came in November, when poorly printed ballots delayed election results for 72 long hours.

Four years later, another frustratingly slow count — this time blamed on tiny specks of paper dust — has some voters wondering if the county owns an election-system lemon. But what system — if any — should replace the county’s Ballot Now scanners isn’t so clear.

Voting system wish list

If she could buy any system she wanted to replace Hart’s Ballot Now, Boulder County Clerk and Recorder Hillary Hall wouldn’t know where to shop.

“I don’t think it exists,” she said.

In fact, all things being equal — or all things being equally flawed — optically scanning paper ballots is the technology that many counties are hoping to switch to, not away from.

According to the Pew Center on the States’ 2008 Election Preview, optical scan systems were the most prevalent voting technology in the country during last week’s election, with 58 percent of all voters casting a ballot that was read by these machines. In 2004, that was true for only 36 percent of voters.

“If the proper maintenance and everything else is being done to (the scanners), this is the voting system we should be using,” said John Gideon, co-director of VotersUnite!, a non-partisan group that has been logging errors on all kinds of voting machines.

David Wagner, a computer scientist at the University of California at Berkeley who studies electronic voting machines, agrees.

“Right now, I think optical scan systems are probably the most mature, reliable technology on the market,” he said. “Boulder got the best technology on the market. ... None of the voting systems are perfect, and they all have their limitations.”

Of course, there is more than one company that makes the optical scanners and software, and they’re not all the same. Beyond Hart InterCivic, the other big players are Sequoia Voting Systems, Electronic Systems and Software and Premier Election Solutions. But they all have problems, according to local computer scientists and voting-audit specialist Neal McBurnett.

“I haven’t seen one be demonstrably better,” he said. “Anytime you work with computers, if you look long and hard enough, you find bugs.”

It’s OK in Oklahoma

Oklahoma broke ground in the world of optical scanning. The entire state went to a standardized optical scanning system in 1992. All the equipment is owned by the state and it all uses the same software.

Oklahoma has been lauded for its system because there are rarely any election snafus worth publicizing, and the state is now so practiced at its own system that they have nearly eliminated the vendor from their process.

“We do our own maintenance, and we rarely turn to a vendor,” said Fran Roach, Oklahoma’s assistant secretary of the election board. “We have machines break — those sort of things — but we’ve never really had a major problem.”

Even in Oklahoma, where voter confidence is now unusually high, the question remains: How do they know they aren’t making any mistakes? The answer is a little thorny. In Oklahoma, they test their machines before the election to make sure they’re accurate. But that’s the last check.

There are no audits in Oklahoma; they only recount ballots if there is a recount in a close race. They also don’t “resolve” ballots, the painstaking process of having a bipartisan team look at a ballot to determine voter intent if it can’t be read by the machine.

Oklahoma is not a “voter intent” state, which means that if the voter doesn’t mark the ballot correctly, then the vote doesn’t count. That’s the law.

Only about one-third of states have a paper trail that they routinely audit, according to Berkeley’s Wagner.

If Boulder County could operate by these laws, the dust problem would be waiting for a recount to be discovered. After all, Boulder County’s pre-election test would have cleared the machines for use and no resolution teams would have been around to notice the tiny specks of dust on the digital images of the ballots.

Closer to home

This year, the city and county of Denver switched to a system similar to Boulder County’s. Denver used paper ballots almost exclusively and bought nine optical scanners from Sequoia to process the ballots.

Each of Denver’s scanners processed about 30,000 ballots. Boulder County has eight machines, each of which scanned about 20,000 ballots. Denver finished counting votes 6 1/2 hours after polls closed. Boulder County took about nine times as long, finishing its count three days after Election Night.

It appears the major difference is in the process — the way election workers interact with the software. Both counties ran tests before the election, as required by law, and both will later conduct an audit. Both tried to determine voter intent when the machines detected an “overvote,” when it appears that both “yes” and “no” are checked on a ballot question, for example, or an “undervote,” when no choice was made.

In Denver, when the machine signals an overvote, a person will look at the actual ballot, and if it really is an overvote, the person will “adjust a switch on the machine and send it back through. The machine will ignore the over or undervote,” said Alton Dillard, a spokesman for the clerk and recorder in Denver.

If the machine can’t read a ballot, but the voter’s intention was clear, a bipartisan team will actually transfer all the votes from the original ballot to a new one so the scanner can read it.

In Boulder County, an overvote triggers the software to pull a digital image of the ballot up on a computer screen. A resolution team can then look at it and decide the voter’s intent. The major difference: in Boulder County, election workers see what the machine sees. And in last week’s case, the workers saw dust specks.

The dust caused two results: an overvote, which would have flagged the ballot so a team could resolve it. (A routine procedure.) Or, it would cause a single “yes” vote when the voter meant not to vote on a particular race. This wouldn’t be caught by Denver’s system. And it wasn’t caught by Boulder’s system, either. In Boulder, election workers noticed the dust specks and dug through ballots looking for a case where an undervote turned into a “yes” vote.

In an election where many people left the boxes blank for the lesser-known races, hundreds or thousands of these phantom “yes” votes could have gone through unnoticed.

“Even if it’s an uncontested race, it’s not OK for it to be wrong,” said Hall. “I knew where I would be when I made the choice (to view every ballot). I knew it would be really uncomfortable.”

A matter of perspective

If Hall really is finding errors that might be more difficult to detect elsewhere, then the key to faster elections might be in a greater tolerance for error. All vote-counting systems have some error. Even the mechanical parts on lever systems wear out over time, sometimes causing vote counts to be off, according to Roy Saltman, who used to work for the National Institute of Standards and Technology.

The margin of error can be larger if the margin of victory is also larger without affecting the outcome of the race. A tiny margin of error could result in a recount when any errors that existed could be discovered.

But Hall said that attitude isn’t acceptable to her.

“Until we understand the impact of the errors we’re seeing, we can’t overlook them,” she said.

For some advocates of the optical-scan system, at least over other existing alternatives, the slow count in Boulder County is an example of the system working as it should: mistakes were found, and the paper ballots were there to recount.

“I think Boulder County deserves credit for using ballot-scan systems,” said Berkeley’s Wagner, who also works with ACCURATE, a project funded by the National Science Foundation to study electronic voting. “I understand the concerns about timeliness, but I think it’s much more important to get it right than to get it fast. I think it’s perfectly fine if occasionally it takes a few days.”

Contact Camera Staff Writer Laura Snider at 303-473-1327 or sniderl@dailycamera.com.

 



Previous Page
 
Favorites

Election Problem Log image
2004 to 2009



Previous
Features


Accessibility Issues
Accessibility Issues


Cost Comparisons
Cost Comparisons


Flyers & Handouts
Handouts


VotersUnite News Exclusives


Search by

Copyright © 2004-2010 VotersUnite!