In my days on the junior high school playground, many a sports-related crisis was averted by the offer of a do-over; after all, in the absence of referees, there really was no other way to resolve disputes. In electoral politics, the equivalent of a do-over is a revote, a concept that has received a sudden burst of attention in U.S. and international politics in recent weeks. First, in Ukraine, Viktor Yuschenko won a revote after his loss in the initial election was seen to be tainted by fraud. Then, in Washington State, Republican gubernatorial candidate Dino Rossi called for a revote after a second recount named his opponent, Democrat Christine Gregoire, the winner by a mere 129 votes. Meanwhile, revotes have played a prominent role in two local controversies: In East Chicago, a small hamlet in northern Indiana, a reform-minded candidate was installed as mayor earlier this month after winning a revote against the machine-backed incumbent. (The original race was riddled with accusations of fraud.) And in North Carolina, the State Board of Elections recently ordered a revote in the disputed race for state agriculture commissioner--though a judge has since overturned the ruling.

A do-over was the only option available in the anarchic world of playground sports. In the world of politics, however, there are referees, and those referees have a less radical option at their disposal to settle disputed elections: They can order a recount. And the availability of that option raises a question: When is a revote, as opposed to a recount, justified?

The answer lies in the distinction between fraud and error. In broad terms, revotes are the right remedy for the former, while recounts are the only (admittedly imperfect) remedy for the latter. A true democratic election should never contain fraud; but it is an uncomfortable fact that elections, like polls, do contain a margin of error, partly due to unavoidable human fallibility and partly due to serious structural flaws in the way we run them. There's a good debate to be had about whether our current margin of error is too large (I would argue it is) but simply re-rolling the dice with the same flawed system when elections fall within that margin hardly constitutes a solution--and essentially, that's what a revote does. A recount, by contrast, is not a re-rolling of the dice: During a recount, the methodology is adjusted at each stage to improve the accuracy and reduce the margin of error--hand recounts are more accurate than machine recounts because humans can use their judgment (think of the Palm Beach County election judges and the infamous dimpled chads) and catch mistakes. In fact, in Washington, it was a hand recount that revealed 732 ballots that had been erroneously discarded, and those ballots provided Gregoire's margin of victory.

Comparing Ukraine to Washington illustrates the difference between fraud and error. Rossi's supporters have sought to link the two, placing newspapers ads that say, "This isn't Ukraine, it's Washington." Unfortunately for Rossi, those poorly phrased ads are actually right. In Ukraine the original election, which was awarded to Kremlin-backed candidate Viktor Yanukovich, was tainted by systematic fraud, intimidation, and violence. International election observers universally condemned the results as invalid, and when a new election was ordered Yanukovich's slim margin of victory turned into a convincing victory for Yuschenko. In Rossi's case there has been no accusation of fraud or abuse of any kind. Rather, there's a freakishly small margin of victory and a smattering of irregularities. The argument, in the words of Rossi's lawyer Mark Braden, is essentially that "[i]t's a situation where frankly the margin's too close for our election system." In Ukraine, the machinery of democracy simply didn't work. In Washington, it worked exactly as it was supposed to--which is to say it was basically a toss-up within a certain margin of error.

A better domestic parallel to the Ukraine is the unlikely democratic revolution that recently took place with the aid of a revote in the steel town of East Chicago. For three decades, Democratic mayor Robert Pastrick presided over the city's steady decline. An unreconstructed machine boss, Pastrick allegedly ran the government through a textbook model of municipal corruption: kickbacks, patronage, and lucrative city contracts for friends and family of the mayor. Though Pastrick's grip on the city was tight for a long time, eventually he began to have problems. His son was indicted, federal and state prosecutors started investigations, and East Chicago teetered on the verge of bankruptcy. In addition to corruption, gang violence became a staple of what was once a peaceful locale--and the citizens started getting angry.

Enter George Pabey, an unassuming police commissioner-turned-city councilman. (So unassuming, in fact, that he's the only politician I've ever seen who blushes when people say nice things about him.) In 2002, Pabey ran against Pastrick in the Democratic primary, and on election day in 2003, the reform candidate's supporters were thrilled when news reports showed that he had squeaked out a narrow victory. But that was before the absentee ballots were counted; they put Pastrick ahead by 278 votes. Pabey's supporters immediately started documenting cases of phony and fraudulent absentee ballots, and his lawyers sued under an Indiana state statute that invalidates an election if a series of deliberate acts make it impossible to determine the actual outcome. It took almost two years, but eventually the State Supreme Court ruled in Pabey's favor, and thanks to an amendment to state election law (passed by a state legislature looking to put Pastrick out to pasture) a revote was ordered. This past October, a new election took place, and Pabey won by a two-to-one margin.

The East Chicago revote was a perfect example of why it's important to retain the possibility of a do-over. "The very nature of fraud is that it's hidden," says Pabey lawyer Ned Ruff. "So if you show the pattern, that is enough under the statute to bring in doubt who wins." In other words, signs of fraud suggest that there is probably more fraud. But in a mere recount of the original results, the courts could have thrown out only those absentee ballots proven to be phony by Pabey's attorneys, and Pastrick still would have won. Once a revote was ordered and an election was held with a high degree of official scrutiny, the real will of the people emerged, and Pabey's stunning margin suggested that in all likelihood there had been a whole lot more fraud back in 2002.


Meanwhile, a better parallel to the Washington State race has been unfolding in North Carolina. This past November, Britt Cobb, the incumbent agriculture commissioner, lost to his Republican challenger, Steve Troxler, by 2,287 votes out of 3.4 million cast. Soon after the election, it was revealed that 4,438 votes in Carteret County were lost due to an electronic voting machine error. The State Board of Elections at first decided to offer a revote to all 18,000 voters who voted in the county--both those whose votes had been lost and those whose votes hadn't been--but the courts struck that down. So the three Democrats on the five-person board ordered a statewide revote. And on January 11, a judge struck that down as well. The outcome is now back in the hands of the board of elections and there's talk of the state's General Assembly asserting its jurisdiction.

Tempting as a revote is in this case, it's a bad idea. Because error is part of the system, it happens every time an election takes place in some form or another. It's true that there are big errors and small errors, and the accidental loss of 4,438 votes in a close race is obviously a rather large error. But even if that error is fixed, there will be other errors when a revote is held. In the case of fraud, by contrast, a revote works because the fraud can largely be taken out of the equation in the second election. In the Ukraine and East Chicago a combination of press attention, citizen vigilance, shame, and armies of election observers made sure there was no repeat of the shenanigans that had marred the original outcomes. In the case of error, we might reasonably expect to not make the same error the second time around, but there's no reason to suppose there won't be different errors; indeed, one of the realities of our election system is that there is almost always error of some kind.

That's why in cases where error affects the outcome (like North Carolina and Washington State) recounts--not revotes--make the most sense. Of course, while the margin of error can be reduced by recounts, it can't be eliminated by them: Many errors, such as those in both Washington and North Carolina, are embedded in the way the votes are collected and not the way they are tabulated. Once those errors have happened, there's simply no way of correcting them. In light of this, it's hard to blame Britt Cobb or Dino Rossi for their frustration; after all, they quite literally got unlucky.

But to resolve their predicament by holding revotes could well introduce more unfairness into the system than it would remove. Rossi's lawyer Mark Braden has argued that the results of a revote in Washington would surely be more decisive, pointing out that "the statistical probability of an election being this close again are just too remote for words. It's just not going to happen." But would a more decisive victory in a second election necessarily make the outcome any more legitimate? It's possible people would change their votes because of a presumed taint of scandal attached to the very concept of a revote. And that would hardly be justified, since it was the random margin of error--not the taint of fraud--that led to the revote. In Ukraine and East Chicago, it's likely that Yanukovich and Pastrick did lose some votes in the second round because of the fraud during the first round; but if so, it's their own fault for having run campaigns that apparently engaged in fraud. In North Carolina and Washington, such a shift of votes in a second election would hardly be fair to the candidates who won in the first round; they are guilty only of benefiting from luck, not from fraud.

One might argue that the victims here aren't the candidates, but rather the voters who, in the case of North Carolina, had their franchise denied, or in the case of Washington possibly had their franchise diluted by the presence of unverified ballots. It's a compelling point, but if we were to scrap an election because a certain number of voters were disenfranchised, we'd have to scrap nearly all of them. In the last election alone, The Boston Globe reported at least 30,000 instances of procedural and voting machine error: voters accidentally directed to the wrong polling place (where their votes were subsequently not counted), double-counting of votes in four different states, and votes lost due to computer crashes. And that was an election that went "relatively smoothly," according to Election Assistance Commission chairman DeForest Soaries.


To be sure, I am not arguing that this rule of thumb--revotes for fraud, recounts for errors--is what the law prescribes. Each state has different election statutes that govern these questions. Braden says that while most state courts generally look for evidence of fraud and not just error, Washington doesn't require evidence of fraud for an election to be tossed out. Rick Hasen, a law professor at Loyola University in Los Angeles and author of, says he thinks the roughly 350 unverified provisional ballots that were mistakenly tabulated in King County may provide a solid legal foundation for the court to order a new election in Washington. Nor am I arguing that this distinction should be absolute: A trivial case of fraud--say, a single forged absentee ballot--shouldn't trigger a revote; and errors of a catastrophic or systematic sort--say, half of all ballots being lost or widespread computer errors in every poor precinct--should.

But if one definition of insanity is doing the same thing over and over but expecting a different result, than any attempt to remedy the deep structural flaws in our under-funded, under-staffed, byzantine electoral system by simply having more elections for the same position qualifies as lunacy. Freed from the obligations of elected office, Rossi and Cobb should join forces and lead a national, bipartisan crusade for the serious electoral reform that our country so badly needs. Then, the margin of error in any given election would be smaller. And the cry of "do over" could be reserved for where it's really needed: cases of widespread, deliberate fraud and the junior high school playground.