No, Mathematicians Have Not "Solved" Traffic Jams

What kind of problem is traffic? Is it a mathematical or programming problem, a physics problem, an engineering problem, an economics problem, or a cultural, behavioral, or political one?

The correct answer is "Yes."

But that doesn't stop an endless parade of hucksters from informing us that the day we're going to innovate our way out of the misery of traffic congestion is just around the bend. Witness this absurd headline in Fast Company that caught my eye last week:

No, the article isn't any better than the headline either. It starts out promisingly (boldface emphasis mine):

“[Engineers] do not have competencies in the field of system-related increases in traffic performance,” says Alexander Krylatov, a mathematics professor at St. Petersburg University. “If engineers manage to achieve local improvements, after a while the flows rearrange and the same traffic jams appear in other places.” Burn!

That quote is dead accurate, and evidence to support it can be found any time a city adds lanes to a major urban freeway. In fact, more often than not, the same traffic jams eventually emerge in the same places too, as on Houston's Katy Freeway, where the average afternoon commute time from downtown increased by 55% after the freeway was widened to an astonishing 23 lanes.

Where it all falls apart is Krylatov's proposed solutions, which are fourfold:

1. Put all drivers on the same navigation system, with detour instructions coming from one central hub.

2. Remove on-street parking to add lanes to congested segments of busy roads.

3. Create special lanes reserved for electric cars.

4. Use digital modeling to painstakingly re-create "twins" of entire street networks to run simulated optimization experiments.

We could devote a whole article to each of those four things, and how absurd the notion is that any of them will do much to rescue drivers from traffic purgatory. But suffice it to say that Krylatov's problem here can be expressed in two words: domain dependence.

Domain dependence, according to this good read by Safal Niveshak, is the inability to "recognize the forces at play outside of the system in which we’ve learned about them." It's summed up in folk wisdom with the aphorism, "When you have a hammer, every problem looks like a nail."

What forces are at play outside of the network of nodes, links, and traffic flows which a mathematician like Krylatov sees? Or, equally, outside of the lanes and accesses and choke points which a traffic engineer sees?

Simply: human behavior. Each of us makes choices in where we're willing to live, how we get to our jobs, where we shop, send our kids to school, attend religious services and so forth. We make these choices within certain constraints: how much money can I afford to spend on transportation? And how much of my time am I willing to spend traveling each day?

“There’s a better way to deal with traffic—and “deal with” does not mean “solve.” It is to make our places resilient to congestion, so that if and when it happens, it doesn’t destroy our quality of life.”

It turns out that throughout human history, the answer to the latter question averages somewhere around an hour—the observation behind Marchetti's Constant, which suggests that cities grow to the point where average commute times hover around 20-30 minutes one way, no matter what the prevailing transportation technology of the era is. In ancient Athens, you likely lived within a 30 minute walk of your daily destinations. In modern Atlanta, you likely live within a 30 minute freeway commute. Nearly all U.S. cities have a commute time that roughly obeys Marchetti’s Constant, belying the popular notion that traffic is getting worse and worse.

When you optimize the network for car travel, a bunch of things happen. Some happen quickly: drivers switch to using different routes or traveling at different times of day (this is the part the mathematicians are supremely confident they can predict and manage), or maybe people even switch between driving and other options like public transit. Some happen slowly: land values change, and development becomes viable or economically attractive in places where it wasn’t. Settlement patterns change: people live differently. And then we alter the roads in response to those changes, too, and in response to political pressure to “Do something about traffic!” that arises from those changes.

No system-optimization approach is going to grapple adequately with that complexity. As long as we build a growing city around roads for cars, it’s a pretty sure bet that people in their cars are going to find ways to fill up those roads. We can’t build or network-engineer our way out of congestion, but we can bankrupt ourselves trying. (And destroy our neighborhoods and our communities’ wealth for a generation or more.)

There’s a better way to deal with traffic—and “deal with” does not mean “solve.” It is to make our places resilient to congestion, so that if and when it happens, it doesn’t destroy our quality of life. This means 15-minute neighborhoods: more destinations within walking distance of home. It means a range of ways to get around so nobody is forced into just one option, and a well-connected street network so there are many paths from A to B.

Just don’t let a mathematician or engineer tell you they’ve got the magic solution that will free you, the beleaguered commuter, from having to make choices and weigh trade-offs. They don’t. They never will.

(Cover photo via US Department of Defense)

---