Joe Biden swept the South on Super Tuesday and also won states thought to be leaning to Bernie Sanders, including Massachusetts, Minnesota, Maine, and Texas. Exit polls reflected his strength among African Americans, older voters, and in the suburbs. Amy Klobuchar’s endorsement undoubtedly helped in Minnesota.

There was also another factor. The youth vote, strongly pro-Sanders, seemingly fizzled. According to post-mortems in the press, young people failed to turn out. A report in The Guardian on March 7 states that in six states youth turnout declined, including by 20 percent in Texas. It seems the big rallies, the rock bands, and rock-star speakers counted for little. “Kids these days,” they shrugged—not for the first time.

But then, why did Sanders have a blowout win in the caucus state of Nevada? Why did youth turnout go up 36 percent in the (otherwise disastrously run) caucus state of Iowa? And how was it that the youth vote fizzle didn’t stop Sanders from winning big in California or Colorado? Caucuses take hours of time and effort. A cultural theory of how students in (say) Berkeley and Boulder differ from Austin would be absurd.

There is a simple, clear, yet difficult-to-accept explanation. It has to do with how American elections are run. Let’s call it structural fraud. Structural fraud has two components.

American elections are run by local officials. Local officials decide where to put the polling stations and how many voting machines and how many staff or volunteers to assign to each one. They also control the training, if any is given.

These matters are ripe for manipulation—but let’s leave that aside. Let’s suppose—against lots of evidence—that local election officials are entirely neutral, nonpartisan, disinterested and professional. How do they plan to allocate their resources from one election to the next?

If they plan at all—and let’s assume that they do—they must work from the evidence of past elections, especially turnout, along with new voter registrations and similar data. That is all the information they are likely to have. With it, they must project local loading from one election to the next.

The problem is glaring. Insurgencies are matters of the moment. They arise during a campaign, not before. They can’t be predicted two or four years in advance. A big insurgency like the Bernie Sanders campaign is an inherent challenge to this system.

Thus the election system we have today is inevitably biased toward established communities and voting blocs. Older voters have usually lived in the same place for some time. Suburbs are fairly stable. Even in the South—at least in Democratic primaries these days—African American neighborhoods will have (mostly) stable accommodation to their voting needs. Voting rights advocates did their work decades ago, and while protecting those gains is a constant task, they haven’t been wholly undone yet.

But new voters—young people, immigrants, college students, rapidly growing Latinx communities—may be underpredicted and therefore underserved. And this will be true even with entirely innocent official intentions. When something new happens, our system will screw it up, every single time. Our elections are designed, by inertia and habit, to cut off new movements and truncate their votes.

Thus, a paradox emerges. There are brutally long lines in youth-vote and rapidly-growing minority communities, followed, time after time, by disappointing turnout reports. On Super Tuesday I heard reports of two-hour waits in Virginia, six (even seven!) hours in Houston, and in Los Angeles there was an epic new-system breakdown, with some voters hanging on until 1 am. In Austin I personally waited for over an hour the previous Friday to vote early in an east side Fiesta Mart, where the line snaked by the grocery shelves to a pathetic corner of the big store.

Then there is a second problem, endemic and unique. When you finally get to the head of the line, you still have to vote! In Texas this involved first checking in, then waiting with your blank ballot in a second line for an open voting machine. And then you actually had to fill the damn thing out.

In Texas on Super Tuesday, we voted for nominees for the presidency, for Congress, for state legislators and county commissioners, for judges from the Supreme Court on down, and for constables and sheriffs—a vast miscellany of small-time functionaries. And then—a minor matter—there was a well-meant but time-wasting 10-point “Texas Bill of Rights.” Even with a cheat sheet, filling out page after electronic page of ballot took five to 10 minutes. Meanwhile, the people behind me waited.

These long, slow lines are the deep scandal of American elections. They occur almost nowhere else, as European-style parliamentary elections are a matter of simply marking a single choice on a paper ballot. Foreigners who have never seen a US ballot are amazed at the complexity. It’s a structural feature of our federal system, something seen in very few other democratic states.

How many voters come in, look at the line, and decide they can’t find the time or the energy to join it? How many got into line but could not hang on? We’ll never know, because there is no count of such people. There is no regular monitoring of the length of lines. There are no bathrooms. And nobody ever arrives with more voting machines to relieve the backlog.

Those machines raise other issues. After Biden’s blowout, e-mail chains filled with suspicious chat about electronic election fraud. Did turnout among elderly Democrats in South Carolina really more than double compared to 2016? Really? If so, how was it that those voters managed to vote—while younger voters in Houston and many other places got stuck waiting in long lines?

These questions are worth asking, but the answers are always elusive. Those who wish to dismiss them will call them “conspiracy theories.” Fair enough: Suspicions are not evidence. But they are corrosive; confidence erodes, and confidence cannot be restored without a deep reform of the election process.

Is there a better way? Of course there is. It actually exists—in much of California, in the state of Washington, in Colorado, in Oregon and Hawaii. It’s called vote-by-mail. Voters fill out the ballot at home, sign the back of the envelope and drop it in the mailbox, skipping the lines and the hassle. Voting is fast. Large majorities choose this route already; in Colorado, Oregon, Washington, and Hawaii it’s the only way to vote.

True, with vote-by-mail, counting is slow. Results may not be known for days, disappointing the instant-outcome press. But the system works. The count is centralized, reliable, and it can be checked. Bottlenecks against new voters and therefore against insurgents do not exist—and this may be why, in part, Bernie Sanders could win in California and Colorado but not come close in the South.

Sanders may yet recover from Super Tuesday. But as the catastrophe of a Biden-Trump election nears, progressives must finally place voting rights and fair elections at the top of their agenda. Vote-by-mail is not just the best solution. It’s the only practical way to restore true democracy to American elections.