Is 4-Step Transport Modeling Outdated? An Introduction to Agent-Based Transport Models

The Philosophy of the 4-Step Model

Four years ago, Greater Auckland published an article provocatively titled “Is Transport Modeling Junk Science?” Although the title is generalised, the piece specifically critiques the 4-step transport model’s inherent limitations, such as its assumptions on fixed demand and its neglect of behavioural responses like induced demand. Induced demand refers to the phenomenon where new infrastructure can encourage additional travel rather than alleviate congestion, undermining infrastructure goals and skewing predictive accuracy. This simplified approach can lead to flawed outcomes, reinforcing outdated planning norms and resulting in suboptimal decision-making (Figure 1).

While Greater Auckland’s article aptly diagnoses many issues, its suggestion to abandon transport modelling and cost-benefit analysis entirely may be misguided. Instead, the solution lies in refining transport models to incorporate travel behaviour. Although initiatives like NZTA's research "Environmental and accessibility transport appraisal methodology – an alternative transport appraisal methodology", in which I participated, have attempted to incorporate accessibility and environmental factors, these improvements still fall short because they remain outside the model's core assumptions.

The Role of Agent-Based Models (ABMs) in Enhancing Transport Modeling

Agent-Based Models (ABMs) represent the latest effort to address these limitations by embedding travel behaviour directly into the model. Unlike the 4-step model’s reliance on static demand assumptions, ABMs apply a utility-based approach to each individual agent. Here, each person in the model follows a utility function that calculates the benefits (utility) of activities—such as time spent at home or work—and the costs (disutility) associated with transportation to access these activities. For example, while someone may gain utility from resting at home, they experience disutility from time and monetary costs when commuting. This detailed, behaviorally grounded approach allows ABMs to mirror real-world decision-making more closely than static 4-step models (see Figures 2 and 3).

Technical Limitations of the 4-Step Model

The 4-step model faces several key technical limitations, especially around zoning and data granularity:

• Zoning Limitations: The 4-step model divides a study area into zones, aggregating population and land-use characteristics within each. This aggregation can significantly distort results, as zone size and centroid location can unpredictably impact accessibility scores. For instance, if a zone’s centroid is near a frequent bus stop, it may receive an artificially high public transport score, while a similar zone with a different centroid location may be rated poorly, despite comparable real-world access.
• OD Matrix Complexity: The 4-step model uses Origin-Destination (OD) matrices to represent travel between zones. Even with simplified assumptions—such as commuting to work, school, shopping, or leisure across a limited number of time periods (e.g., AM peak, PM peak), transport modes, income groups, and genders—the model would need 360 matrices to capture these interactions comprehensively (Figure 4). Yet, this level of complexity still misses key factors like working from home, short local trips, and other nuanced behaviours.

3 destinations X 5 time periods X 3 modes X 4 income X 2 genders = 360

• Loss of Individual and Trip Detail: Because the 4-step model focuses on aggregate zone-level data, it lacks the ability to represent the diverse, individual-level variables that influence travel behaviour. By contrast, ABMs allow each agent to have unique attributes, destinations, and travel preferences, providing flexibility in adding variables such as age, income, vehicle access, job type, and more. ABMs can also capture complex, multimodal journeys and accommodate factors like remote work and freight, areas where 4-step models struggle (Figure 5).

Why Agent-Based Models Offer a Superior Framework for Transport Economics

The primary advantage of ABMs is their advanced functionality for analysing the economic impacts of transport investments. ABMs rely on utility functions, making them suitable for a range of economic evaluations:

1. Consumer Surplus Measurement: ABMs can calculate the consumer surplus for people using transport services and evaluate changes under different scenarios (e.g., increased fuel costs or congestion pricing). By capturing individual travel choices, ABMs can more accurately estimate the social benefits of transportation investments.
2. Productivity and GDP Impact: ABMs can model the broader economic impacts of infrastructure projects by assessing changes in productivity and GDP. For example, if a new tunnel is proposed, ABMs can predict the improvement in connectivity, quantify wider economic benefits, and estimate contributions to GDP growth.
3. Tax and Subsidy Analysis: ABMs are flexible enough to simulate targeted taxes and subsidies (e.g., vehicle taxes, public transit subsidies, emissions taxes) and explore their distributional impacts. This capability allows for the evaluation of policies that might mitigate environmental impacts or support equity goals.

The Future of Transport Modeling

ABMs provide a transformative approach to transport modelling, offering a dynamic, individualised, and behaviourally-realistic framework that overcomes many of the 4-step model’s limitations. By capturing individual travel decisions and incorporating complex variables, ABMs represent a critical step forward in developing transport models that support sustainable, equitable, and economically-sound urban planning.

I have been studying Agent-Based Modeling (ABM) in transport over the past 10 months, and I’m excited to share my insights through this series of publications. This first article focuses on the theory and foundations of ABM. In future articles, I will delve into practical applications of ABM in a New Zealand context, using publicly available data, such as census information, and open-source tools, like MATSim.

If you found this article valuable, please give it a "like" and share it on social media to help spread the word. I would love to hear your thoughts, experiences, and questions—so feel free to comment or reach out. If you're interested in implementing Agent-Based Modeling for your city, I'd be more than happy to discuss how it can make a difference.