(Jim Young/Reuters)

Edward B. Foley, a law professor at Ohio State University, is author of “Ballot Battles: The History of Disputed Elections in the United States.” Charles Stewart III, a political science professor at the Massachusetts Institute of Technology, is co-director of the Caltech-MIT Voting Technology Project.

After all that’s happened in this bizarre election, we need to brace ourselves for the chance that it might not end on election night, or even the next morning. The risk of that happening is higher than it used to be — and higher than most of us realize.

This is not reason to panic. No one wants to relive the 2000 recount, but the good news is that we don’t have to.

Certainly, some of the reform measures adopted in the aftermath of 2000 have had the unintended but unavoidable consequence of increasing the possibility that a presidential election remains undecided for days or weeks. Yet such uncertainty would likely be a sign of the electoral system functioning as intended, not of a massive failure. In addition, an election without a clear victor after Nov. 8 need not be as disorderly or as protracted as the 2000 mess. If the 2016 presidential election goes into overtime, the game will be played on a different — and better — field from the one back then.

The main risk factor for such a scenario is provisional voting, which Congress required as part of the Help America Vote Act of 2002, adopted to redress the defects that surfaced two years earlier. Provisional ballots are a safety net to protect eligible voters from erroneous purging of voter rolls, which happened in Florida. Rather than being turned away at the polls, voters whose eligibility is questioned are guaranteed the right to cast a provisional ballot. If upon subsequent review the voters are in fact determined to be eligible, their provisional ballots will be counted. That’s good.

But a collateral consequence of provisional voting is that an election may not be settled until all the provisional ballots are evaluated. The verification process can take anywhere from a few days to several weeks. Since the enactment of Help America Vote, we’ve seen some congressional and other down-ballot races turn on the counting of provisional ballots. It could happen in a presidential election.

It almost did in 2004, but the number of provisional ballots in Ohio (158,642) was too small in relationship to George W. Bush’s lead in election night returns (121,012) for John Kerry to have a reasonable chance of overtaking that lead. Consequently, Kerry conceded the morning after Election Day. But if Ohio in 2004 had been as close as it had been in 1976, when only 11,116 votes separated Jimmy Carter from Gerald Ford, Kerry undoubtedly would have waited for the process of reviewing the provisional ballots to play out.

Moreover, it would be reasonable for a candidate behind by several thousand votes in a state critical to winning the electoral college to think it possible to make up the difference by provisional ballots. Our research shows that in 2012, President Obama extended his margin of victory in multiple battleground states by more than 20,000 votes during the time between election night returns and final certification of the results. Obviously, these gains didn’t make a difference in the outcome, because Obama already had won. But they show that votes counted after Election Day could determine the outcome of a much closer election.

Gains of this magnitude are much larger than what routinely occurred before the enactment of Help America Vote. While other factors may play a role, such as increased reliance on mailed ballots, our statistical analysis indicates this increase is largely due to provisional voting. It also explains why the overtime vote tends to favor Democratic candidates, whose voters are disproportionately affected by the kind of circumstances (such as a change of address) that can cause the need to vote provisionally. In other words, an election that is not decided on election night is more likely than not to end up favoring Hillary Clinton.

Because no count is officially complete until it includes all valid provisional ballots, every state has a well-rehearsed operating manual for conducting this procedure. Missouri’s experience in 2008 is instructive. It took 15 days to determine that John McCain narrowly beat Obama there (although the outcome didn’t alter Obama’s overall electoral college victory). Reviewing Missouri’s provisional ballots, which kept the state “too close to call” for two weeks, proceeded without difficulty, and there is good reason to think that this model could work even when an overtime state was necessary for reaching an electoral college majority.

Thus, if this year’s presidential election ends up being closer than previously expected, we should be prepared for the possibility that provisional ballots take us into extra innings. This won’t be a sign that the system is “rigged,” but that it’s working as designed.