You probably believe Delaware's U.S. Senate nominee Christine O'Donnell defeated Mike Castle in the state's August GOP primary. You may be right, but you can't prove it. The only thing that can be proved is that Castle trounced O'Donnell by about 55 percent to 45 percent in the paper-based absentee-ballot count. On the touch-screen voting systems used on Election Day, however, O'Donnell is said to have defeated Castle by almost the same percentages, reversed. But there is no way to verify that result. As we barrel toward Election Day, Direct Recording Electronic voting systems—usually touch-screen, always entirely unverifiable—are still being used by 20 percent to 30 percent of U.S. voters. The nation will rely on the faith-based results reported by these oft-failed, easily manipulated, virus-prone voting systems for the next two to six years and probably beyond. Whoever the machines report as the "winner" will become the winner, even though no human being will be able to prove that any of those candidates received more votes than the "loser." And there will be virtually nothing we can do about it. Some DRE systems pretend to offer voters the chance to review a "paper trail" printed alongside the machine, which is supposed to show which candidates and initiatives the voter is hoping to vote for. But these paper trails are not actually counted by anyone. The numbers used to calculate winners and losers are recorded inside the computer hardware; tallies don't come from the computer's touch-screen, nor from what's printed on those little rolls of paper. Even in the rare cases in which the paper trails are reviewed, they offer no verification that voters actually meant to vote the way the little printed receipts might show. Studies by CalTech and MIT have found the vast majority of voters either don't bother to check their receipt or don't notice printed errors on it. Another study from Rice University found that when voters do bother to review the electronic summary at the end of the voting process, about two-thirds don't notice when the computer has flipped their votes. And finally, after an election, in the event that paper trails are reviewed for some reason, and they almost never are, there's no way to know which voters actually bothered to look at the paper trail or which, if any, may have noticed printing errors. (In 2008, the computer voting system in Los Angeles misprinted four out of 12 of my own votes. Luckily, I happened to notice.)
Despite these long-reported shortcomings, DRE machines will determine the results of all sorts of races with predicted razor-thin margins next Tuesday. One of the tossups is likely to be in the Nevada election between Democratic Senate Majority Leader Harry Reid and Republican Sharron Angle. Nevada's DREs were initially certified illegally. Still, the entire state will cast votes on the Sequoia AVC Edge touch-screen, which has a "VeriVote" paper-trail printer add-on. Sequoia's touch-screen systems have notoriously failed in election after election. In the days before the 2008 presidential election, I gained access to the database compiled in Nevada by the DNC and the Obama campaign, based on reports of problem calls to their election protection hotline. I discovered the Sequoia machines failed in polling place after polling place during the pre-election early voting period. The Democrats never mentioned it publicly, though, and their lawyers did almost nothing about it.
Similar problems occurred across the country, and election officials, as is their wont, ignored or downplayed them. The political parties feared pointing them out to the public, making the bizarre calculation that doing so might somehow depress turnout. And when the mainstream press covered the failures, they were generally dismissed as little more than "glitches" or "hiccups."
The same problems are already occurring in early voting this year and will certainly continue on Nov. 2. Last week, votes reportedly flipped repeatedly from Republicans to Democrats on touch-screens in several North Carolina counties, and from Republican to the Green Party (as well as from Democrat to Republican, which is far more common) on touch-screens in Texas. A few weeks ago, a couple of computer scientists hacked Sequoia's AVC Edge, replacing the voting software with a Pac-Man video game—without even breaking the machine's co-called "tamper evident" seals.
Meanwhile, in Pennsylvania, DRE voting that cannot be verified will decide the U.S. Senate nail-biter between Democratic candidate Joe Sestak and Republican Pat Toomey. * The same will be true in Illinois' Senate race between Republican Mark Kirk and Democrat Alexi Giannoulias and in West Virginia's between Democrat Joe Manchin and Republican John Raese. (In that last one, Nate Silver's FiveThirtyEight prediction model recently found the race about as close as can be, 48.749 percent to 48.748 percent.) Unverifiable DREs will also be key to the Senate results in Missouri and Arkansas and in many other U.S. House, gubernatorial, and local elections.
If you'd like to see how easy it is to hack a Sequoia DRE and flip the results of an election in a way that is virtually impossible to discover (even on a DRE that prints receipts) University of California-Santa Barbara made a handy video as part of California's landmark 2007 "Top to Bottom Review" of e-voting systems. The California secretary of state decertified Sequoia's and most other DREs after the UCSB findings. But a bevy of states continue to use them anyway.
For small-time voter fraud, try out the little yellow button on the back of Sequoia's AVC Edge DRE. Press it in the right pattern, and you can cast as many votes as you like. You may not know the pattern, but plenty of election officials and Sequoia maintenance employees do. They frequently have unfettered access to the machines during early voting and on Election Day.
One race that probably won't be close on Tuesday is the one between South Carolina's Republican Sen. Jim DeMint and his Democratic opponent Alvin Greene. DeMint is expected to destroy Greene, the candidate who, until last June's primary, had no campaign, no campaign Web site, no money, no cell phone, no job, and lived with his father. Nobody had ever heard of Greene until the state's ES&S iVotronic touch-screen machines declared him the winner over longtime legislator and former Circuit Court Judge Vic Rawl, who had a campaign with hundreds of volunteers.
Thousands of errors were later discovered on the machines. But when Rawl protested after the primary that no legitimate explanation other than voting-system malfeasance or malfunction explained Greene's reported win, or the average 11-point disparity between the verifiable paper-based absentee-ballot results and the unverifiable DRE results, state election officials told him that there was no evidence Greene was not actually the choice of the voters.
That was true. Unfortunately, there was no evidence to prove that Greene was the winner, either. It's an old question, but it needs to be asked again, anyway: Is this any way to run a democracy?
Like Slate on Facebook. Follow us on Twitter.
Correction, Oct. 29, 2010: Originally this article incorrectly stated that the U.S. Senate race in Kentucky between Republican Rand Paul and Democrat Jack Conway would be determined almost entirely on DRE machines made by ES&S. Only 23 out of the 120 counties in Kentucky use ES&S DRE equipment. The original sentence has been removed. (Return to the corrected paragraph.)