Texas’ voter ID law went on federal trial last week in Corpus Christi.* Texas Gov. Rick Perry signed the law in 2011, and it went into effect on Jan. 1, 2012. The law was blocked by a federal court later that year, but reinstated after the Supreme Court struck down Section 5 of the Voting Rights Act in 2013. This is a big test of whether Section 2 of that act, which bars racial discrimination in voting, has any teeth.

To vote in Texas, registered voters must now present one of the following: a driver’s license or state ID card, a license to carry a concealed gun, a U.S. military ID card with a photo, a U.S. citizenship certificate with a photo, a U.S. passport, or a state election certificate (a document you can obtain if you don’t have any other accepted form of ID). Student photo IDs and out-of-state driver’s licenses are not permitted. The Department of Justice, the NAACP, and various other voting and civil rights groups are challenging the law, claiming that up to 787,000 registered Texas voters lack any acceptable form of ID, and that blacks and Hispanics are more likely to be disenfranchised by the law than whites.

A federal court in Wisconsin struck down a similar voter ID measure this year, finding that it violated Section 2 of the Voting Rights Act of 1965. Until recently, most of the heavy lifting in voting rights challenges was done by Section 5—the part of the Voting Rights Act requiring designated states, counties, and municipalities with a history of discriminatory election practices to obtain federal “preclearance” before making changes to their voting laws. But that provision was struck down in June 2013 when the Supreme Court determined that the preclearance system was outdated and burdensome on the covered jurisdictions.

Section 2 is aimed at practices that make it harder for minority voters to “participate in the political process” and “elect representatives of their choice.” Although it’s gone largely untested, there is some reason to believe that Section 2 may still become a helpful tool in the voting rights wars. To defeat Texas’ ID law under Section 2, the DOJ will have to prove discrimination—that the voter ID law makes it harder for some voters to vote—which is potentially a high standard for the government to meet.* (The case is expected to take months and is unlikely to be decided before the November election.)

Until now, the national debate about voter ID has unfolded on consistent and predictable lines. Supporters of voter ID contend that these laws are necessary to prevent the supposedly rampant vote fraud that takes place at the polls. In recent years, study after study has failed to show anything resembling rampant vote fraud, and of the little fraud that does happen, most seems to happen by way of absentee ballots, which ID laws such as the one being challenged in Texas can’t stop. Despite a growing sense that the laws achieve nothing good and possibly do something very bad, these voter ID laws have nevertheless been defended on the grounds that they prevent at least some fraud and will deter a few vote fraudsters in the future.

Opponents argue that voter ID laws are thinly veiled attempts at vote suppression, modern-day poll taxes that disenfranchise the poor, the young, and minority voters—not incidentally, people who tend to skew Democratic. This is a touchy issue for Texas and the courts. Nobody likes to be called an intentional racist. Already in the trial, the plaintiffs have contended that the voter ID law disproportionately discriminates against minorities.

As the state presents its case this week, Texas has offered testimony showing that state voters support the ID requirement and that it’s needed to deter vote fraud. The state attorney general has argued that this new voter ID law has operated through two elections without glitches, irregularity, or vote suppression.

Throughout this debate, the persistent assumption has been that voter ID laws do indeed decrease vote fraud, at least by some small amount, and the question has been whether that justifies disenfranchising some voters. The Supreme Court’s ruling in Crawford v. Marion County Election Board, which upheld Indiana’s restrictive new voter ID law and paved the way for laws that came after, framed the question that way and concluded that the answer is yes. That conclusion was reached by liberal lion John Paul Stevens, who must have hated the voter ID law even as he was upholding it.

But a new study by Michael Gilbert of the University of Virginia Law School suggests that both sides are probably asking the wrong question. What if, he asks, voter ID laws in fact increase the risk of vote fraud? If they do, the lone defense of such laws—which have been passed in Virginia, North Carolina, Wisconsin, Indiana, and other places—falls apart completely. According to a report this summer by the Brennan Center, 13 states have passed highly restrictive voter ID laws since 2011. The Brennan Center’s studies and others show that approximately 11 percent of Americans don’t have a government-issued photo ID.

In a forthcoming paper to be published in the Columbia Law Review, Gilbert argues that ID laws can actually worsen the problem of voter fraud. He does so with some thought experiments. As he put it to me over email:

Imagine an election between Alice and Bill. Alice gets 13 lawful votes, and Bill gets 10 lawful votes plus a couple of fraudulent ones. Alice wins that election 13 to 12, and although fraud doesn’t determine the outcome, it’s certainly present, which is a worry. This election could use a voter ID law, right? Not necessarily.

Consider that same election with a Texas-style voter ID law in place. Now Alice and Bill each get nine lawful votes (less than before because the law has now suppressed some votes), and Bill gets one fraudulent vote (less than before because the law deterred some fraud).  So now Bill wins 10 to nine—and the election is fraudulent. Get it? In this case, the voter ID law caused the problem it was meant to solve. That example involves vote suppression:  Alice lost three lawful votes, a good chunk of her total, and Bill lost one.

But imagine a case in which vote suppression is not necessary; a voter ID law can cause fraud even if it doesn’t disenfranchise anyone. Suppose Alice got her 13 lawful votes, Bill got his 10, and they each got four fraudulent votes. Alice wins, as she should, by a vote of 17-14. Alice’s fraudulent votes came from impersonation fraud (voting in person as someone else), which is the kind ID laws target. In this hypo, Bill’s fraudulent votes came from (more common) absentee ballot fraud, which ID laws can’t stop. Now consider that election with a voter ID law in place. Alice would get her 13 lawful votes but no fraudulent ones, and Bill would get his 10 lawful votes and his four fraudulent ones. Now Bill wins 14-13, the election is a sham, and because the ID law deterred Alice’s fraudulent votes and not Bill’s, that sham was caused by the ID law. 

This fundamental problem with voter ID laws, it seems, is the result of unnecessarily binary thinking. We believe that there is only one kind of vote fraud when there are, in fact, several. And we believe that suppressing that one kind of fraud is better than suppressing none, which seems intuitively correct but creates a host of new problems. This is emblematic of a legislative approach that looks at tiny slices of problems, inflates them out of proportion, and then elects to do anything to solve them, as opposed to the right thing.

Nobody is arguing that vote fraud should be allowed to happen. The question Gilbert raises with his model is whether we are encouraging vote fraud rather than deterring it, and whether that matters in our thinking about creating fair elections. At the very least, his logic suggests that there is a mistake at the heart of the Supreme Court’s decision in Crawford. There, the court embraced a simple logic:  Voter ID laws always reduce the risk of voter fraud, so states should have the discretion to implement them. But if Gilbert is right, and if these laws may just as well worsen fraud, that logic fails, and voter ID laws should fail along with them.

Correction, Sept. 11, 2014: This article originally misspelled the city of Corpus Christi, Texas. (Return.) It also misstated that the DOJ had to prove intentional discrimination. (Return.)