For months I’ve been asked questions about “Rethinking I-94,” the Minnesota Department of Transportation’s seemingly interminable planning process for changing the intercity freeway between downtown Minneapolis and St. Paul. Most people want to know: What do you think of the boulevard proposal? Is it even possible? Where would the traffic go?
You’re not alone if you’re skeptical about removing this central interstate. It currently carries between 100,000 and 150,000 cars a day, around 10 times that of an arterial street like St. Paul’s Lexington Parkway (near my house). Looking at the overwhelming scale of the freeway, it’s hard to imagine that reducing the road would be possible.
That’s why we’re lucky that the transportation advocacy group Our Streets hired consultants earlier this year to study this problem, a rare outside analysis. The team of experts included a national planning firm, a local data analysis firm and a traffic modeling consultant that picked apart the state-led procedure. The resulting study included inspiring renderings and policy points, but my eyes were especially opened by its critique of status quo modeling practices, a topic most people never consider.
Traffic modeling has long been a “black box” carefully guarded by expert bureaucracy, something few understand and everyone else accepts at face value. Somewhere an engineer uses a complex formula and reams of data to predict a trend, and decision-makers and the public swallow a “baseline scenario” of congestion and ever-more driving.
After years of models failing to accurately predict transportation trends – see for example, repeatedly wrong predictions of increasing “vehicle miles traveled” (VMT) – that’s starting to change. Alongside other 20th century traffic engineering concepts like “level of service” or parking demand studies, models are becoming a target of scrutiny,
In this case, led by consultants Toole Design Group and Smart Mobility, a Vermont-based traffic planning firm, the study raises questions about Twin Cities regional traffic analyses and how they are used in the “Rethinking I-94” process. The consultant’s report argues that options reducing freeway capacity might have counterintuitive traffic outcomes in central Minneapolis and St. Paul, and that answering the question of “what would happen to the traffic” isn’t straightforward. If the critique is correct, nobody should pretend they know the answers about what would happen to traffic alongside changes to I-94.
How traffic models work
Diving into the weeds, MnDOT’s traffic analysis is based on the Met Council’s “activity based” regional traffic model, which uses a “static” traffic assignment approach. Engineers break highways into segments and then input a certain amount of traffic based on past trends (almost always an increasing number). In that way, you get results about congestion and free-flow conditions based on those inputs.
The problem is that this isn’t how real-world traffic works. Not only does VMT not always go up, but the speed and volume of cars entering a freeway segment are highly dependent on the traffic conditions on either side.
“For a static model, even if you have a bottleneck upstream or downstream, it’s not going to affect traffic volume,” explained Vermont-based Smart Mobility President Norm Marshall. “Even though a real person would have to wait behind a bottleneck’d section of I-94, in the model you just ‘squeeze through’ anyway. By not treating road segments as interdependent, it’s not really modeling for urban freeway congestion at all.”
According to Marshall, the static approach amounts to a fatal flaw for predicting real-world networks. Instead, expanding or reducing capacity at any given point might make a big difference or have a negligible effect: everything depends on its larger context.
Marshall and his co-author, Lucy Gibson, argue that the MnDOT study seems designed to ignore the region’s persistent bottlenecks, limiting the “study area” to a stretch of freeway from Cedar Avenue in Minneapolis to Rice Street in St. Paul. As any Twin Cities driver already knows, the actual highway choke points are on either side: around the 35W and downtown on-ramps in Minneapolis (especially the constrained Lowry Hill tunnel) and near I-35E in St.t Paul.
“The Lowry Tunnel is a perfect example of how their modeling method is gonna fail,” agreed Gibson, an engineer at Toole Design Group. “There’s going to be a limited amount of traffic coming out of the tunnel towards this part of the study area, and that’s not going to grow. [The traffic increase] can’t really get through, even though their model shows that it can.”
The study’s lack of choke points makes the resulting analysis much less useful. The static assignment means that recurring stop-and-go jams because of the existing bottlenecks won’t shape the results for the stretch of highway in between. Instead, the engineers in the Our Streets report suggest that a dynamic model that looks at the system as a whole could better analyze the relationships between the constituent parts.
“The dynamic model treats every segment as part of the network,” Marshall added. “It says, if there is traffic stopped in front of you, then you can’t go until that traffic clears. Or, if that traffic’s going at 10 miles an hour, you’re going to be going 10 miles an hour slowly into that segment. It actually models congestion the way it really happens.”
According to agency spokesperson Ricardo Lopez, MnDOT will use a dynamic model for its “Draft I Tier EIS” analysis later in 2026. By that point, however, the options for change will be limited to a smaller pool. Designs like the “at-grade boulevard” or other freeway-reduction strategies will likely have been eliminated because of their poor initial congestion scores.
Creating traffic demand
Another key dynamic often ignored by traffic analysis is something called “induced demand.” By now it’s a well-recognized trend that freeway expansions can create as much new traffic as they mitigate with extra travel capacity. This happens even with the current form of I-94, where the free-flowing highway “creates” by attracting trips that would otherwise take place elsewhere, or not at all.
Take the freeway’s effect on Lexington Parkway, the main arterial near my house. According to Marshall and Gibson, one counterintuitive outcome is that expanding the freeway might make local traffic worse. The easiest way to see why is to type an address into Google Maps and look at suggested routes – say going three miles to visit my my favorite local chocolate shop. There are countless examples where the mapping software routes you onto the freeway to save a minute or two, even if city streets might be a geographically shorter route. This is exactly how people behave in the real world, often going well out of their way to use freeways.
This principle operates across all times-of-day and for thousands of people within the freeway’s catchment. As Marshall says, “no trip begins or ends on a highway.” This is why freeway expansion can “induce” highway trips, focusing even more vehicular congestion along arterial routes by interchanges, streets like Lexington, Snelling, Dale, and Cretin in St. Paul, or 26th and Cedar Avenues in Minneapolis.
The growing chorus of American highway planning critics reminds me of how different street design manuals emerged about 15 years ago. (See also the history of the National Association of City Transportation Officials.) In this case, more planners and engineers are calling into question standard 20th century practices of transportation engineering, forcing planners to look at more international and innovative approaches to urban design.
One prominent voice, an engineering professor at the University of Colorado named Wes Marshall (no relation to Norm), just released a book that debunks many long-standing highway engineering assumptions. For Wes Marshall, the limited capabilities of static traffic modeling are part and parcel of a whole list of antiquated professional practices for transportation planners.
“These models — and the conventional traffic engineering protocols that we’ve created surrounding how we use these models — make it very hard not to focus on widenings,” he told me. “On one hand, the models unrealistically overestimate how much delay we will have on a highway if we don’t do anything. On the other, they also unrealistically overestimate the benefit we will see in terms of reduced congestion.”
Before his academic role at the University of Colorado, Marshall worked in the field for years, including on the famous “Big Dig” project in Boston. That effort famously buried the “central artery” freeway through downtown, after a wildly over-budget process that set records in U.S. infrastructure. These days, similarly ambitious urban freeway widening projects are increasingly scrutinized by advocates and communities impacted by urban freeways, and projects like the I-5 expansion in Portland have stalled while other widening projects in cities like Austin or Houston go ahead.
Wes Marshall says this is all part of a system that actively promotes expansion and driving with little consideration of alternatives or consequences.
“Boston, for instance, has worse traffic today than before the Big Dig,” he said. “Would it be worse had they never done the Big Dig? Probably, but the models were still wrong on both ends. So from my perspective, we are usually asking the wrong questions when we use these models regardless of whether they are static or dynamic.”
When asked to comment on the differing modeling capabilities, MnDOT’s Lopez told me: “MnDOT is following established requirements and guidelines related to the traffic modeling. [We are] confident we are using the best practices in our evaluation of alternatives.”
Is I-94 model flawed?
The crux of the matter is the potential for change. A lot depends on whether or not, if the highway were reduced in size, traffic would dissipate into the networked grid of streets that would make travel more disbursed, efficient and predictable. That’s something that the current model seemingly cannot answer, but there’s a big difference between knowing you don’t know something and being unaware of one’s ignorance.
Last week, local journalist John Edwards, who publishes Wedge Live out of South Minneapolis, leaked part of a draft results analysis of the Rethinking I-94 project. The analysis shows how “mobility in people for motorized vehicles” fares under the 16 potential changes to the freeway. According to the rankings, the reduced and boulevard options fare poorly. Freeway reductions score yellow (“concerns”) and red (“does not meet need”) for the crucial driver mobility category.
That news comes as no surprise to members of the consultant team, who argue that the model itself is the problem. Because it fundamentally misunderstands traffic dynamics, it operates like a one-way ratchet, always predicting the need for more highway capacity.
“Well, it’s the way we’ve always done it,” said Norm Marshall. “[Engineers] get good fees for doing it this way, and it always justifies the [expansion] project.”
So what would really happen to all the traffic on I-94?
The truth seems to be that the agency hasn’t yet studied the possibilities, and the existing model, static and disconnected, cannot tell us. Instead, we can only make educated guesses by looking at other examples of changes to freeways from around the world. If you do so, it turns out that the answer is much more dynamic than we might think. It’s quite possible that reducing freeway can decrease some impacts of congestion.
“My rule of thumb in a downsizing is that the new facility will carry maybe half of the traffic, [but] it depends on how you design it,” explained Norm Marshall. “A quarter of the traffic will be diverted to roughly parallel facilities, and a quarter of the traffic, we’ve seen this all around the country, appears to disappear. That’s just because people decide, ‘If I can’t go there at 60 miles an hour, I’ll go somewhere else.’”
This is why, when people ask me “where would the traffic go,” I often demur. It’s simply hard to say. We don’t know how to precisely guess what might happen.
But if history is any guide, freeway reductions usually work out, leading to real improvements for urban neighborhoods. On top of that, the COVID pandemic upended long-standing assumptions about driving patterns, proving that our society and its travel demands are more flexible than anyone thought.
If the Rethinking I-94 effort were to live up to its name, it would be nice to figure out what might happen if we changed the freeway for the better. The jury is still out on the next step of the planning process, but I hope that changes that truly “rethink” the status quo will remain on the table.
Bill Lindeke is a lecturer in Urban Studies at the University of Minnesota’s Department of Geography, Environment and Society. He is the author of multiple books on Twin Cities culture and history, most recently St. Paul: an Urban Biography. Follow Bill on Twitter: @BillLindeke.
