Over the past 70 years, transportation system designers in America have used the same model when deciding what to build. The problem is, she's often wrong.
In November 2011, the Louisville and South Indiana Bridge Design Organization over the Ohio River published a 595-page document that was supposed to end a decades-long highway dispute. The project, to put it mildly, turned out to be controversial.
While many cities across the US were trying to figure out if they needed city highways in business districts, Louisville set about building them with a vengeance. He not only wanted to leave the infamous "spaghetti denouement" to them. Kennedywhere Highways 64, 65 and 71 are knotted - he wanted to build on it. The political alliance pushing the project has tried to widen I-64 to double the number of lanes, as well as build a new bridge just downstream from there. This would double the number of lanes crossing the river, from 6 to 12, for “only” $ 2.5 billion.
Spaghetti Junction in Louisville
But to get federal approval for this expensive proposal, proponents of the project needed to provide evidence that Louisville needed the expansion. They took advantage of a legally legal practice in the industry called Travel Demand Modeling (TDM) and hired an engineering firm to predict what traffic there would look like 20 years from now - that is, by 2030. The firm has concluded that the number of trips will increase by 29%. The consequences were obvious: if nothing was done, traffic jams would only worsen. As a result, the project received funding and was launched.
But two subsequent studies funded by the same project reached very different conclusions.
Two years later, engineering firm CDM Smith examined how much road traffic actually changed while the project was trying to get approval. It found that traffic across the river dropped 0.9% from 2010 to 2013.
Another study, commissioned by potential shareholders, raised more questions than answers. The conclusion indicated that by 2030 the number of trips across the river would be 132,000, 15% less than predicted in the SDEIS [project document assessing the environmental impact of construction / approx. transl.]. To make matters worse, the 12 lanes, which in total will cross the river, will carry 4,000 fewer cars per day than on the I-65 bridge alone in 2007. This refuted the argument that Louisville needed these new bridges.
Aaron Renn, a city strategist who often criticized the Ohio River bridges, documented the fuss in detail . “As crazy as this project may seem,” he wrote in 2013, when a study for potential shareholders came out, “it always manages to find ways to demonstrate that it is even crazier than I thought.”
Today the project is finished, and the people of Louisville can see for themselves which of the predictions turned out to be more accurate. In 2018 , a post-construction survey found that river crossings dropped by 2% between 2013 and 2018. As a result, Vox and others called the project "empty fun" of epic proportions.
The Louisville highway project is far from the first time move request models have missed the mark. While they are legally mandated to use them to defend any infrastructure project that claims federal budget, one of urbanists' dirtiest secrets is that these models are at best error-prone and at worst fundamentally flawed.
I recently asked Rennes how these initial pretty predictions of double-digit growth in traffic really did make the project come true.
“I think it was very influential,” Rennes said. “I don’t think they would have received permission to implement the project without all these traffic predictions that said it would increase significantly. If there is no increase in traffic, how can you justify building two bridges? "
TDMs are different. They can completely cover large cities, extend beyond state lines, or belong to a small segment of a suburban highway. They become more and more complex over time. However, they are all based on the so-called. A “four-step process” - a rough approximation of the decision-making process of people wanting to move from point A to point B. As a result, the model produces numbers that estimate the number of trips people will take along certain routes.
To return this number, the model goes through four stages. First, she draws up an algorithmic map based on the expected patterns of use of space (for example, the number of trips to places where there are commercial organizations will be more than to places where there are residential buildings) and socio-economic factors (for example, the less unemployment, the more trips there will be). The model then estimates where most people will go from and where. The third step is to figure out how they will go there, and the fourth is to plot their routes based on the travel time. The result is the number of trips that will be made in a given area, and the time to cover a given distance. Then engineers and planners add a new highway, road, bridge, or other piece of road infrastructure to the model and see what makes a difference.Or they change the numbers in the first step to account for the expected population or future employment growth. Often these numbers are then used by policy-makers to justify specific projects - be it a highway extension or a new narrow gauge railway.
While there are many reasons why the said Ohio bridge project failed, one that few have paid attention to is TDM's role in embellishing the $ 2.5 billion project. One of the possible reasons for this is that experts in this field did not expect anything else.
Of course, not all experts in this field think so. Civil engineers are more likely to defend these models by saying that they are a useful tool that is sometimes misused. Greg Ehrhardt, a professor of civil engineering at the University of Kentucky, who has worked with these models for most of two decades, said they would, at best, "serve as an obstacle to wishful thinking." However, other experts I spoke with, especially urbanists, believe these models support the long-held myth that more highways and wider roads will reduce congestion.
In any case, everyone agrees that the biggest question is not whether the models can produce better results, but why we rely on them so much at all. The focus is not on the debate over TDM or models in general, but on the process of how we decide how our cities should look.
Critics say TDM is a prime example of an antiquated planning process that optimizes traffic flow and propels highway construction. They say that it is high time to change our ideas about why we are building something.
“This is a fundamental problem with travel modeling and how it is used,” said Beth Osborne, director of the non-profit organization Transportation for America. “We think the model gives us a ready-made answer. This is irresponsible. Nothing can give us the answer. Only we ourselves can give us the answer. "
In 1953, Detroit's highway agencies launched the first TDM study to develop a long-term highway development plan. Academic history of TDMsays the idea was disarmingly simple. To launch a massive community project like the highway system, planners had to roughly imagine where people would drive in the future. They realized that there was no point in spending decades building highways just to find out that they were too big, or too small, or led in the wrong direction.
The Detroit Metropolitan Traffic Survey, as the work was called, conducted 39,000 interviews with local residents and 7,200 interviews with truck drivers and taxi drivers (which is typical of the time, public transport was not included at all). Using an IBM 407 punch-card computer to partially automate the process, the researchers extrapolated recent trends and predicted future travel patterns to create a network of fast roads that would be useful to Detroit not only in 1955, when the study was published, but also in 1980. -m.
It was an innovative approach to transportation planning, and with the technology and thinking of the day, it was incredibly cutting edge. Other cities, including Chicago, San Juan, and Washington, D.C., soon adopted it as well. It wasn't long before this approach began to be exported to other countries, and it became a common tool for planning transportation around the world.
In retrospect, this concept had several obvious flaws. To begin with, the basic approach of the model assumed that what has happened in recent times will continue to happen. If Detroit's population grew, it will continue to grow. If the cost of fuel falls, it will continue to fall. But the world doesn't work that way. A lot can change in a few decades.
Take, for example, population growth and land use patterns as input from the first step of a four-stage model. These are two of the most important variables in any TDM because the more people live in a given area, the more travel they will take, and where they live and work will determine the travel routes. And both of these factors have changed radically in Detroit. Detroit experienced sudden and unprecedented growth in the 1950s. In 1950, at its peak, the population reached 1.8 million, according to historian Thomas Segru in The Origins of the Urban Crisis: Race and Inequality in Postwar Detroit". By 1970, almost a fifth of the population had left the city, in particular, due to the flight of white people to the suburbs. Many businesses also moved their headquarters or production outside the city, which radically changed the patterns of movement. Planner from 1955 is unlikely to . could have predicted something like
critics of the typical approach to modeling the note and less obvious problem - the model does not coincide with the actual behavior of people Suppose you live in Pasadena, a friend of yours from Culver City invites you to dinner at six o'clock in the work. day. Will you go there? Or will you say that he is crazy if he thinks that you will drive through the whole of Los Angeles in traffic jams? Most likely, the last - or such an invitation would not have been initially, based on banal decency - and the trip would not have taken place.
But traffic predictions don't work that way. In the models, any trip today will be endlessly repeated in the future, no matter how much the road situation deteriorates.
Experts call this “fixed demand for travel,” which is essentially an oxymoron - since demand for travel is not fixed almost by definition. We always decide whether we should go or not, before we go. And one of the main factors in making this decision is how long this trip will take. TDM works on the opposite assumptions - if people want to go somewhere, they will go. And then they will calculate how long it took them.
So some urbanists ironically refer to this approach as "Lemming Demand Theory," said Joe Courtright, an urban economist at the Impresa consultancy for the City Observatory. She assumes that people will continue to crawl onto the highway regardless of traffic jams.
“The problem is not so much wrong measurements as wrong underlying assumption,” said University of Connecticut professor Norman Garrick. “You don't think about how people behave, how they use the system. You just say - this is how it was in the past and this is how it will be in the future, although you are introducing such a big change in the system. "
The flip side of the fixed travel demand problem is just as harmful. Let's say Los Angeles somehow doubles the number of lanes on Highways 110 and 10 connecting Pasadena to Culver City. Now going to dinner with a friend doesn't seem like such a bad idea anymore. Tens of thousands of other people will just think the same way. They, too, will begin to make trips they have never done before. In the long run, they may move farther, to places where it is cheaper at home, as the speed of commuting to work will be faster, which means they will drive longer. As a result, all these new lanes will fill up and traffic will become just as bad.
This phenomenon is called forced demand, and it is not just a thought experiment. This is exactly what happened in almost every case when cities built new highways or expanded old ones.
“Recent experience with freeways in major US cities suggests that congestion will be eternal,” wrote economist Anthony Downs in his 1962 paper, The Law of Freeway Traffic Jam at Rush Hour . “As you can see, no matter how many new super-roads connecting the suburbs to the business district we build, people driving cars will still crawl slowly during morning and evening rush hours.”
Experts have known about forced demand for generations, but we still continue to add new highways, like Sisyphus, unsuccessfully trying to get out of the rush hour congestion situation. To fully understand the absurdity of trying to solve this problem, just look at the $ 2.8 billion highway project in Katy, Texas.It was supposed to reduce the time to travel to and from work.23-lane highway , the widest in the world. Predictably, traffic jams there only intensified, and travel time increased.
A 2011 paper, The Fundamental Law of Traffic Congestion , concluded that “increasing the number of roads or public transport is unlikely to ease congestion” because as new lanes are added, travel distances increase proportionally. The more highways and roads we build, the more we drive. The converse is also true: on the rare occasion that highways are temporarily closed, as was the case with Seattle's Alaska Viaduct, traffic does not degrade much. And TDM ignores this completely.
“It is well known that the four-stage paradigm, developed 50-60 years ago, is just a convenient formula for calculating, and does not take into account behavioral factors,” wrotetravel planner and consultant David T. Hartgen in 2013 - and does not reflect the traveler's decision-making process.
Proof can be found on the highway. In his landmark 2007 study of traffic predictions in 14 countries and five continents, Professor Bent Flivbjörg of the University of Oxford found that half of the predictions about future traffic were more than 20% wrong - and this is happening everywhere . A 2006 study from the National Collaborative Highway Research Program found that 15 toll road projects had actual traffic on average 35% lower than predicted. Another study found that this error averaged 42%.
“I think there is a generally recognized issue with accuracy in general,” said Fred Jones, general project manager at planner Michael Baker International. "Sometimes by an order of magnitude, from 30% to 50%."
Even worse, no one learns from these mistakes. "Inaccuracy has not changed over the 30 years studied in this study," wrote Flivbjörg. "Forecasts don't improve over time."
It is not even clear whether urban planners or firms using these models think that such inaccuracy is bad. They claim that they are being asked to do the impossible and predict the future - naturally, there will be inaccuracies. It's like building a route on Google maps. If we are talking about a 20-minute trip, the maps will be able to predict its duration with good accuracy. If this is an 8-hour trip, then the estimated duration will be just a guess, since even Google cannot predict the future, and will know if there will be an accident on I-95 at the exit from O.K. when you get there in five hours. A 20-year prediction made by government standards, says Professor Chanyung Lee of the University of South Florida, roughly the same series.
As a result, city planners with their models downplay the importance of accurate numbers and speculate more broadly about changing trends over time. They say that ideally, policymakers should run the model with different predictions for population, land use, and employment to get a range of expectations. And then they have to assess which of the ranges of those expectations the project is suitable for.
The problem is that when the results are published, all these nuances are lost, and the results are perceived by policymakers as fact. As Cortright said, "Models are essentially tools to sell what the road departments want to do."
But with all these problems, models are getting smarter. Especially in the past ten years, more and more states have been working with dynamic travel models that reflect human behavior much better. They better take into account alternative transportation methods - bicycles, hiking, public transportation. Unlike previous versions, they are able to simulate how the expansion of one section of the road can lead to bottlenecks on other sections.
Still, experts warn that if we don't change the entire decision-making process on which these projects are based, we will not achieve anything with an improved model. Typically, these models are not even run, and the results are not published, until the state transportation department has chosen the most suitable project.
After talking with 10 experts in this field, one thing became clear to me: the existing obstacles are not technological, but social and political. After all, the bridge project in Louisville accurately modeled the travel request for shareholders. This can be predicted. The question is not why the models are wrong, but why the right models do not affect anything.
When I asked Rennes, who closely watched the Louisville project, how best to design the construction of a large transportation project, he said he was not sure. “There is an idea that we should get rid of politics in this process, reduce political decisions to objective criteria. I think a lot of our debate stems from the rivalry of value systems in our beliefs about what will be good for people. "
Here again, the Louisville project is a good example. The engineers who compiled SDEIS estimated that the population of the metropolitan area will grow by 15% by 2030. The prediction looks solid - Rennes said the population in these counties grew by 7.85% from 2007 to 2020. However, SDEIS predicts that virtually all of this increase will be in the suburbs and outskirts of the city. Because of this assumption, and a predicted 42% increase in employment, SDEIS reported a 52% increase in travel time and a 161% increase in time wasted in congestion in the current infrastructure. These assessments became critical for the acceptance of the project for the construction of the two bridges.
However, these tendencies are not immutable laws of human existence. “This is a classic self-fulfilling prophecydisguised as technical objectivity, ”Courtright said. "Population growth projections imply endless decentralization of households and businesses."
For these reasons, critics of TDM say that the accuracy of predictions - or rather, the lack of it - is practically unimportant, since any implemented project that changes the infrastructure also changes people's behavior. The question is not whether the model predictions about human behavior are accurate, but what kind of human behavior we would like to achieve.
“I don't really care if the highway model was accurate or not,” said Kevin Degood, director of infrastructure policy at the Center for American Progress, who often criticizes such models in highway planning, “because even if it was accurate, the project could fail. "
Therefore, he believes that we need to revise our goals at the planning stage, move away from the estimated speed of transport, traffic and congestion, to other issues related to the quality of life. For example: What percentage of households are within a quarter of a mile of high quality public transport? What percentage are able to commute to work without personal transport, or live near a city park?
Transport projects address the very foundations of social values. Do we want the suburbs to cut through the intertwined highways so that people can easily get from the suburbs to the downtown area? Or do we want pedestrian-friendly areas of the city with cleaner air, quieter streets, proximity to jobs and businesses - so people don't have to buy a car if they don't want to?
Depending on the answers to these questions, states will spend money on infrastructure in very different ways. In the first case, we will have many more projects like Louisville. In the second, attention will shift from roads to public transport and changes in density laws.
Rennes's words are confirmed by the fact that in most cities in America, the opinion of residents on these issues is divided. Perhaps the biggest benefit of this model is that it hides this debate behind a veil of scientific certainty. For clear, concrete numbers. “From the point of view of a resident, these numbers come from a black box,” he said. "You have no idea how these numbers are obtained, so you cannot criticize them."
In other words, the model gags people. It may not be true, but there is no more valuable quality in the world of transport policy.