InTransition Magazine
Article URL:
InTransition Magazine : Transportation Planning, Practice & Progress

Archive Edition

Archived editions: 

Modeling Unruly Travel

Computer models are relied upon as technological crystal balls for seeing future travel demands on roads, rail lines and other facilities. In many cases, their predictions are the deciding factor in justifying investments of many millions of dollars.

But despite the increasing computing horsepower and data analysis capabilities being built into them, travel models confront an inherent and bedeviling complication – people are generally sloppy, and often bad, travel planners.  The transportation models assume they will rationally and successfully pursue their self interest—that is, examining all their options and making travel decisions to minimize time, cost and distance.

Archives of Michigan

A Bendix computer G-15 at the Michigan State Highway Department,

late 1950s.

Behavioral economists are not so sure. “We don’t have a supercomputer between our ears that can calculate sophisticated statistical functions and generate the level of uncertainty attached to different travel alternatives,” said Erel Avineri, reader in travel behavior at the Centre for Transport & Society, University of the West of England.

As a result, the behavioral economists say, people rely on habit, make snap decisions and consistently misjudge probabilities – in short, they can be unpredictable travelers. In addition, people can be swayed by innumerable “affective” factors—the weather, a recent traffic jam experience, a desire to see the sights, the lure of what’s trendy and cool—none of which can be  reduced to formulas in modeling software.

A more fundamental problem is that, simply put, the future is unknowable. Over the lifespan of infrastructure built today travel demand may be drastically changed by energy prices, climate change, new technologies, cultural trends and other factors that can’t be foreseen or modeled.

Despite their weaknesses, models provide useful demand estimates for well-defined problems—like the near-term traffic impacts of new development in a town. For broader applications and longer time frames, behavioral economists say, models necessarily build in assumptions that people’s travel is more regular and predictable than it really is and take leaps of faith about future conditions. If these shortcomings are forthrightly acknowledged, model estimates can still serve as a consideration in decision-making. Yet often the caveats get conveniently lost.

“What forecasts models really do,” said Jonathan Gifford, associate dean and director of the Transportation Policy, Operations and Logistics program at George Mason University, “is give decision-makers, engineers and planners some comfort that they’re spending the public’s resources wisely. [They say] ‘We’ve used the best available models. This is a worthwhile investment.’” 

Rarely do model users investigate whether demand estimates actually came to pass, since “Nobody wants to be backward-looking,” Gifford said. Attempts to do this, he said, find that models tend to “massively overshoot or undershoot” actual demand.

Efforts are underway to make models more accurately mirror real life travel by taking a “microscopic” approach to travel. But these efforts “at the very beginning,” Gifford said.

Given the limits of modeling, Gifford said, “we need to think about acting strategically, and incrementally recognizing that we are in an unpredictable domain.” An incremental approach, he said, might include building new facilities or services in stages “with an orientation towards testing, towards validation and towards results, recognizing that we don’t really know how people respond to changes.” We can also spend money upfront to leave future options open, such as acquiring rights-of-way or creating wider structures to allow lanes or rail tracks to be added if warranted.

The lack of certainty also means that the public sector may have to do much more to encourage use of the transportation systems it creates, rather than waiting for predicted demand to appear. Systems can be designed and marketed, Gifford said, in ways that resonate with people’s emotions, aesthetics, sensitivity to social groups and other “affective” factors. Gifford suggests that these “soft” factors have often been neglected in favor of “cold, hard facts of transportation—the ride quality, waiting time, cost and other objective things that we can easily plug into our models.”

“We can use marketing and other approaches to help shape behavior but we want to leave ourselves some options in case things develop in ways we don’t anticipate,” Gifford said.

Return to Article