Voters cast their ballots at a polling station set up at Grady High School for the midterm elections on November 6, 2018, in Atlanta.Jessica McGowan / Getty

The Georgia governor’s race is balanced on a knife’s edge. Local polls have the Democratic nominee Stacey Abrams and her GOP opponent, Georgia Secretary of State Brian Kemp, virtually tied. Abrams’s team in particular will scramble to make sure every provisional ballot is completed, that every person who faced challenges to registration is able to participate, that all absentee and vote-by-mail ballots are counted, and that every allegation of intimidation or unfair practices on Election Day is investigated. In a race in which a December runoff is a distinct possibility if neither candidate can secure 50 percent support, every single vote matters.

But no matter the outcome, it’s clear that voter rights and suppression will be one of the major stories of the 2018 election in Georgia. The state has become the battleground for something deeper than the ideas of the candidates themselves; it’s now emblematic of a larger struggle over voting rights that has changed party politics markedly over the past five years. The true nature of voter suppression as an accumulation of everyday annoyances, legal barriers, and confusion has come into full view. Today, voter suppression is a labyrinth, not a wall.

That labyrinth has been under construction for years. Kemp has embarked on what his opponents and critics say is a series of naked attempts to constrict the electorate. Since 2010, his office reports that it has purged upwards of 1.4 million voters from the rolls, including more than 660,000 Georgians in 2017 and almost 90,000 this year. Many of those voters found their registration canceled because they had not voted in the previous election. Additionally, under an “exact match” law passed by the state legislature that requires handwritten voter registrations to be identical to personal documents, 53,000 people had their registrations moved to “pending” status because of typos or other errors before a district court enjoined the policy. More than 80 percent of those registrations belonged to black voters.

Most of these maneuvers have rather small effects in a vacuum, and it’s difficult to track the effects of any one policy on the outcomes of elections. For example, Georgia’s early-voting period featured a record-shattering 2 million votes cast, a number that dwarfs the thousands of people who could have faced disenfranchisement under the exact-match law. But much of the research on election law and voter turnout shows that it’s the combination of major policies and minor barriers—like polling-place changes, long lines at the polls, and small bureaucratic hurdles—that have real and measurable impacts on turnout.

According to a new working paper from the Harvard University professor Desmond Ang, protections against those accumulative assaults against democracy have all but been erased in the past five years. Ang studied the effects of a provision in the Voting Rights Act (VRA) that required the federal preclearance of election laws in places where Jim Crow had kept black people from voting, a requirement that was expanded in 1975 to some states and districts outside the Deep South. As originally envisioned, preclearance forced districts with significant proportions of minority voters and low minority turnout to submit all changes to election laws for federal approval. The federal litmus test approved new provisions only if they were found not to decrease minority turnout relative to the status quo. Thus, the VRA was not only a protective shield against scorched-earth Jim Crow policies—it was also intended to guard against more subtle restrictions, all while promoting higher minority turnout as an explicit goal.

Ang found that in the districts covered by preclearance from 1975 to 2013, federal oversight was a major factor in sustained increases in minority turnout relative to counties not covered by the VRA. And that doesn’t include counties and states that had more explicit anti-black laws, like poll taxes, that the VRA outlawed. Ang’s findings indicate that the act of continued federal monitoring alone was responsible for a good deal of minority turnout across the country. “The estimated gains in voter turnout are large—ranging from 4 to 8 percentage points—and lasting—having persisted for 40 years,” Ang writes.

“The fact that the VRA was successful means that these places couldn’t have been putting in discriminatory measures all the way from when they fell under coverage ... until 2013,” Ang told me. “I think that the whole rationale behind preclearance was this idea that there are essentially infinite margins on which you can discriminate.”

The Justice Department and the U.S. Congress in the 1960s and ’70s predicted that southern election officials would employ clever incremental policies that, over time, would bring back Jim Crow—and cloak it in an even thicker aura of legitimacy than it existed under before. Ang cites a Jim Crow–era official in Mississippi who said: “What those smart fellows [at the Justice Department] don’t realize is that we can still get to these darkies in a whole lot of subtle ways.”

But the Supreme Court’s decision in Shelby County v. Holder in 2013 reversed the federal doctrine of proactiveness when it effectively ended preclearance. Since that decision, Ang has tracked what appears to be the beginning of a troubling trend. Minority turnout in the past two federal elections has plummeted, specifically in the counties in which preclearance was once a first step for creating new election laws. The data don’t yet exist to confirm causality between Shelby County and the recent dips in turnout, but Ang writes that the numbers imply that “recently enacted election laws may have negated many of the gains made under preclearance.”

Included in those recently enacted laws are most of Kemp’s most controversial policies. In late October, U.S. District Court Judge Leigh Martin May enjoined the exact-match law, a policy that allowed election officials to reject absentee ballots because of signature mismatches. On Kemp’s watch, Georgia has lost almost a tenth of its polling places since 2012, with the majority of closings occurring in poor counties and those with significant African American populations. NYU’s Brennan Center for Justice finds that Georgia—along with Florida and North Carolina—has increased its voter-purge rates since the Shelby County decision.

And Georgia is merely representative of a nationwide trend. Kemp’s Kansan counterpart, Kris Kobach—who is also administering his own election—has been held in contempt of court for his attempts to disenfranchise Latino voters in the state; has rejected thousands of registrations and ballots; and has overseen the curious relocation of the polling place in Dodge City to a location outside of the city, a move that will force the city’s Hispanic majority to travel far in order to vote. In North Dakota, courts have upheld a voter-ID law that could disproportionately affect the state’s American Indian voters, a crucial constituency for Democratic Senator Heidi Heitkamp.

“Because it looks clinical and bureaucratic, we don’t pay attention to it,” says Carol Anderson, an African American–studies professor at Emory University and the author of One Person, No Vote, an exhaustive review of modern voter-suppression efforts.

“These states have been really good about making Jim Crow 2.0 seem reasonable,” Anderson told me. “This is massive disenfranchisement that is slow and corrosive. We don’t see it the way we see cross burnings and riots and beatings. What instead happens is you frustrate people out of their basic right.”

As Anderson told me, the thing about the suppression-by-frustration regime is that it provides dozens of potential exit points for voters burdened by bureaucracy. “You have to punch a clock to go to work,” she said. “If you go to the polls on Tuesday and your name’s not on the rolls, now what are you going to do? So you’re either going to go to work anyway because you can’t afford to lose pay, or you’re going to have to take some vacation leave.” Include in that calculus the process of obtaining identification in the first place, the potentially long lines on Election Day, and the fact that provisional ballots for people with discrepancies can require even more paperwork and time off, and the true costs of voting become clear.

On Election Day, it’s worth remembering just how Jim Crow elections worked. Yes, black people were lynched for attempting to vote—a history particularly salient in Georgia. The extralegal components of the system were real and dangerous, and they claimed countless lives.

But Jim Crow states were also administrative states, and the bureaucracies they developed came about as a result of a drive among powerful white politicians to discriminate within the bounds of federal law. That meant poll taxes and literacy tests, which were originally perfectly legal. It also meant recitations of preambles, long walks to county registrars, and frustration even among black people who somehow managed to register and vote. It meant all-white primaries and at-large districts and intense gerrymandering. It relied on obsessive tinkering along the margins to come up with a system that was de jure passable under the Constitution, but in the aggregate became a de facto impossible impediment for black people voting in any real numbers. In order to build their regime, southern officials needed to build mazes.

Those are important considerations for this election, and for elections to come. Regardless of the outcomes of individual races—and even perhaps because of them, if Republicans face major losses—the incentives for disenfranchising black and Latino voters may only be increasing as their share of the electorate increases, and as they steadily back Democratic candidates. And that’s as the main tool for protecting voters, the VRA, has been rendered partially inert. In places like Georgia, the rudiments of a labyrinth appear to already be in place.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.