Advanced models can’t replace sound judgment

By Eric Murphy

The investments made by transportation agencies are often guided by increasingly sophisticated models and forecasts, which strive to account for more factors and become more accurate in predicting travel patterns. These improvements, while promising for the future, also highlight the many ways that most agencies’ existing models miss the mark. Values-based decision-making can help agencies achieve the outcomes they have set out in their long-range plans more effectively than relying on models alone. 

These models, which typically rely on past trends, often claim to provide a high level of detail about traffic and travel decades into the future. But laws and economic, social, and cultural trends change. People develop new technologies, and a changing climate means the world itself is less and less predictable based on past trends. Human beings don’t always behave how we expect them to, or make perfectly rational decisions. 

Historically, the models have often been proven wrong. That was the case even before a global pandemic upended travel patterns, as we have noted for more than a decade. 

Actual vehicle-miles traveled (in black) compared to projections. Source: Frontier Group.

Given the uncertainties in models, some agencies are beginning to move away from trying to predict future conditions and, instead, build toward their goals, values, and priorities. This often means creating a more human-centered and adaptable transportation system, no matter what the future might hold. Sophisticated modeling can be a useful tool in the toolbox, but does not need to be the predominant decision-maker. 

In November, USDOT released a playbook called “Climate Strategies That Work.” One of the strategies they recommend is improved travel demand modeling: 

“Many commonly used travel demand models, such as traditional “four-step” models, face challenges in accounting for the full range of changes in travel behavior and land use that may result from additional roadway capacity. […] These models also may have limited sensitivity to projects and strategies that promote alternatives to driving, such as bike and pedestrian facilities.” 

The phenomenon of induced demand is a major focus of these concerns. Simply put, it means the transportation infrastructure we build shapes future transportation choices. Adding highway capacity makes driving more convenient, leading people to drive more often and even choose to live in more far-flung neighborhoods. The same goes for investments in transit and bicycle or pedestrian infrastructure, which encourage different behaviors and opposite development patterns. 

State agencies in California, Colorado, and Minnesota have taken note and they are exploring the use of simpler tools like California’s induced travel calculator to ensure new investments move the needle in the right direction toward minimizing additional emissions from car travel. This marks a more value-driven decision framework as compared to the usual “predict and provide” approach — a topic DOT leaders discussed at our recent annual meeting. 

New studies often come out highlighting additional factors that current models aren’t capturing. New research from Chile, for instance, suggests modelers should account for habit-forming tendencies, particularly among transit riders. “The addition of inertia and shock effects further improved the model fit, thus suggesting the importance of incorporating psychological measures into the modelling framework,” wrote the authors. 

Another recent study highlighted complexities in walking behavior showing that for different types of destinations, people’s perceptions play a greater role in determining whether people walk than proximity alone. “Planners should consider the potential mediating role of perceived accessibility and recognize the importance of psychological and social factors, including familiarity and attachment to place, when designing interventions to encourage walking,” say the authors. “Addressing these factors — beyond improving physical infrastructure — may result in more impactful strategies to reduce car dependency and enhance walkability.” They add: 

“Frequent walking may reshape individuals’ perceptions of accessibility as they become more familiar with their environment. This feedback loop reinforces the likelihood of walking, suggesting that repeated walking behavior makes destinations feel closer over time. Understanding this dynamic is essential for advancing future research on the interaction between accessibility and travel behavior.”

Models, no matter how sophisticated and advanced, will likely miss some important facet of reality and the human experience. Uncertainty increases dramatically over time, and existing models that rely mostly on extrapolating past trends sometimes produce unreliable results. Agencies should use these models as one tool in a broader toolbox while building toward the outcomes they want to achieve, whatever the future holds.

Photo credit: RDNE via Pexels, unmodified. License.