Highway planners use a deeply flawed “lemming” model of traffic that rationalizes highway widenings
The traffic projections made as part of the Environmental Assessment for the $500 million Rose Quarter I-5 widening project make an audacious claim that the project will reduce traffic congestion and greenhouse gases, and completely unlike any other freeway expansion project, won’t induce any additional demand.
How do they know that? Because they have a model that says so.
But how realistic is that model? As it turns out, not very. As we detailed earlier, one problem is that the model (without saying so) is based on the assumption that a 12-lane, 5 mile long freeway was built (reality check: it wasn’t) between Portland and Vancouver in 2015, and is funnelling thousands and thousands more additional cars into the Rose Quarter than is actually the case.
But there’s a more fundamental problem. The traffic forecasting model used by ODOT is inherently structured to over-predict traffic congestion, and presents a distorted picture of what’s likely to happen with freeway widening.The model, a classic four-step traffic demand model (albeit with a few tacked on bells and whistles, like a microsimulation package) is decades old technique with a couple of critical and well documented weaknesses. The most important one is that it allows traffic to grow without limit in the “base” case. It predicts that as population increases, more and more people will drive more and more miles, and that they will all be oblivious in their trip making behavior to any changes in congestion. That is, they won’t be deterred by congestion from changing the timing, mode, destination, or whether they take a trip at all. That’s because four-step models, like the one used to create the Rose Quarter I-5 estimates, uses something called static trip assignment, STA that takes almost no notice of the effects of congestion and delay.
In an important sense, static trip assignment is a kind of “lemming” model of travel behavior. It assigns successively more and more trips to congested links even as they become more and more congested. Implicitly, it assumes that traveler behavior doesn’t respond at all to experienced travel conditions, especially delay. In fact, the model allows traffic to actually exceed the capacity of the roadway link–by definition of physical impossibility–and consequently over predicts traffic, and leads to forecasts of hours and hours of congestion, as more and more traffic piles obliviously onwards into the congested roadway–and does so every single day, with no learning or adaptation. Like lemmings, you might say.
This is like the classic 1958 Walt Disney nature film, White Wilderness showing hundreds of lemmings running off the edge of a cliff. The following lemmings leap to their death, even though they can see the lemmings in front of them falling.
In the film, the little rodents assemble for a mass migration, scamper across the tundra and ford a tiny stream as narrator Winston Hibbler explains that, “A kind of compulsion seizes each tiny rodent and, carried along by an unreasoning hysteria, each falls into step for a march that will take them to a strange destiny.”
That destiny is to jump into the ocean. As they approach the “sea,” (actually a river -more tight cropping) Hibbler continues, “They’ve become victims of an obsession — a one-track thought: Move on! Move on!”
The “pack of lemmings” reaches the final precipice. “This is the last chance to turn back,” Hibbler states. “Yet over they go, casting themselves out bodily into space.”
Lemmings are an apt analogy, because in real life (as opposed to 1950s nature films) lemmings don’t actually blindly leap to their death. As the Alaska Department of Fish and Wildlife explains, the classic footage which appears in White Wilderness was actually staged by the producers, who from positions off camera, chased the terrified lemmings off the cliff.
In a sense, that’s what modelers are doing with Static Traffic Assignment–no matter how bad congestion gets in the STA models, the lemming motorists, just keep pouring into the congested roadway link. But in the real world, as opposed to traffic models or 1950s Disney films, drivers (and lemmings) don’t just mindlessly drive into congestion day after day (or jump off cliffs along with hundreds of others). Instead, in the face of congestion, they change their behavior, changing the time, mode or destination of their journey, or foregoing it altogether.
What the unrealistic static traffic assignment lemming scenario does is to create a fictitious baseline forecast of traffic that looks truly horrible. And, in comparison, the “build” scenario, in this case, a road widening, looks less bad. And it also doesn’t have any “induced” demand, because the projections dramatically over-estimated the amount of traffic that would occur in the base no-build scenario. (Induced demand is almost impossible in these models because traffic demand is derived from an earlier step in the process and there’s no feedback loop of less delay to encourage greater trip-making.)
For technicians in the field, the problems with static traffic assignment are well documented. Norm Marshall writing in the peer-reviewed technical publication “Research in Transportation Business and Management, says:
STA has two fundamental problems that make it ill-suited at analyzing peak period congestion. First, most peak period congestion, especially on freeways, involves traffic queuing behind bottlenecks. Therefore, the roadway segments are not independent, as is assumed in STA. Second, these bottlenecks meter traffic flow to the capacity of the bottlenecks. In sharp contrast, STA allows modeled traffic volumes to exceed capacity. This misrepresents traffic not only on the over-capacity segment, but on downstream segments that the excess traffic could not really reach because it either would divert to other routes or be queued upstream.
Forecasting the impossible: The status quo of estimating traffic flows with static traffic assignment and the future of dynamic traffic assignment, published in Research in Transportation Business and Management, https://doi.org/10.1016/j.rtbm.2018.06.002
Static assignment models allow traffic forecast on specific segments of roadway to exceed the capacity of that segmenta –physical impossibility, and also inaccurately forecast actual speeds because of the non-linear relationship between capacity and speed.
There is a method to reduce this bias, called dynamic trip assignment, which adjusts travel volumes and routing based on modeled levels of congestion. Portland Metro is developing some dynamic assignment techniques, but they weren’t used in preparing the forecasts for the Rose Quarter freeway widening project.
The lemming model is pernicious for two reasons. First, it presents a distorted view of what will happen under the no-build alternative. The EA claims that there will be environmental benefits because we’ll avoid all of the congestion and pollution caused by motorist/lemmings jamming themselves into the Rose Quarter day after day.. But this claim is a fiction: in reality, traffic will never grow to the levels forecast in the model, because the model predicts more vehicles than the road can physically accommodate. By exaggerating the level of traffic in the no-build scenario, the EA has created a false baseline for estimating the environmental effects of the widened freeway, thus disguising the effect of induced demand. Because under static trip assignment, capacity makes no difference to whether or when people travel, there can’t possibly be induced demand. Induced demand is effectively assumed away.
The tendency to overestimate future traffic levels in mature travel corridors is also an endemic problem with the current methodology used to predict future transportation demand. After a careful review of the literature, the Government Accountability Office found:
. . . current travel demand models tend to predict unreasonably bad conditions in the absence of a proposed highway or transit investment. Travel forecasting, as previously discussed, does not contend well with land-use changes or effects on nearby roads or other transportation alternatives that result from transportation improvements or growing congestion. Before conditions get as bad as they are forecasted, people make other changes, such as residence or employment changes to avoid the excessive travel costs.
The weakness of transportation models in accurately predicting future traffic levels is a continuing problem. So it is not merely the Rose Quarter traffic projection model that is problematic; rather the entire class of four-step (trip generation, assignment, mode, routing models) have proved inaccurate in practice. After an exhaustive review of the state of the art, the Transportation Research Board of the National Academies wrote:
“. . . as has been true for the past four decades, these models could not provide accurate information to inform decision making on many transportation and land use policies or traffic operation projects.”
(Committee for Determination of the State of the Practice in Metropolitan Area Travel Forecasting, 2007)
While technology has allowed for faster computation, and more detailed mapping, they conclude:
“The practice of metropolitan travel forecasting has been resistant to fundamental change. Every 10 years or so there begins a cycle of research, innovation, resolve to put innovation into practice, and eventual failure to affect any appreciable change in how travel forecasting is practiced.”
Why wouldn’t highway departments implement a better method? Well, as it turns out, the static trip assignment/lemming model tells them exactly what they want to hear: If you don’t give us money to widen the highway, we’ll have endless gridlock. It also “proves” that the expansion won’t create any induced demand. From the standpoint of highway builders, this is a convenient property for a model to have.
A model, any model, is only as good as the assumptions that go into it. The fundamental assumption of static traffic assignment, that motorists behave like (fictional) lemmings, filing irrationally over a cliff (or into a gridlocked roadway, day after day), and that they don’t change their behavior in response to congestion, is dramatically wrong. It produces a gross over-estimate of how much roadways are likely to be congested, and paints capacity improvements in an unrealistic, favorable light, while structurally excluding the well-documented effects of induced demand.