🔬 physics

Beyond the Main Mode: The contribution of access and egress trips in door-to-door travel

Using Dutch travel diary data and a nested logit model, this study reveals that while train and public transit offer lower disutility for in-vehicle time on longer journeys, their overall attractiveness depends heavily on minimizing transfers and ensuring dense station networks to facilitate walking access and egress, particularly for travelers currently driving to dense urban areas.

Original authors: Nejc Geržinič, Mark van Hagen, Hussein Al-Tamimi, Niels van Oort, Dorine Duives

Published 2026-03-20

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Nejc Geržinič, Mark van Hagen, Hussein Al-Tamimi, Niels van Oort, Dorine Duives

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are planning a trip to a friend's house. You have two main ways to get there:

The "Door-to-Door" Drive: You hop in your car, drive straight to their driveway, and park. It's a single, continuous motion.
The "Train Adventure": You have to walk or bike to a train station, wait for the train, ride the train, get off, and then walk or bike again to your friend's house.

This paper is like a detective story trying to figure out why people often choose the car, even when the train might be faster or cheaper. The authors, Nejc Geržinič and his team, realized that most studies only look at the "middle" of the train journey (the part on the tracks). They ignored the "bookends"—the walk to the station and the walk from the station.

They decided to look at the whole sandwich, not just the filling.

The Big Discovery: The "Hidden Tax" of Train Travel

The researchers found something surprising about how our brains value time.

The Car/Active Mode Penalty: When you are driving, cycling, or walking, every minute feels heavy. You are the one doing the work. You have to steer, pedal, or watch your step. It feels like 100% effort.
The Train/BTM Discount: When you are on a train (or a bus/tram), every minute feels lighter. You can read a book, check your phone, or daydream. The study found that time spent on a train feels 60% less painful than time spent driving or cycling.

The Analogy: Imagine time is a heavy backpack.

When you drive, you carry a backpack that weighs 10kg.
When you ride the train, you carry a backpack that only weighs 4kg.

So, even if the train trip takes longer in total minutes, it might actually feel "lighter" to your brain because the "train minutes" are so much easier to carry.

The Real Problem: The "First and Last Mile"

If the train minutes are so easy, why don't more people take the train? The answer lies in the Access and Egress trips (the walk/bike to and from the station).

The study found that people have a very strong preference for walking to get to and from the station. They hate transferring (switching from a bus to a train, or walking between platforms).

The Transfer Penalty: A single transfer feels like a massive headache. The study showed that people are willing to walk 10 minutes further just to avoid one transfer.
The Urban Trap: People really, really dislike driving into crowded city centers because parking is a nightmare. They are willing to take a train that takes 28 minutes longer just to avoid the stress of finding a parking spot in a busy city.

The "Golden Rules" for Train Travel (Policy Recommendations)

Based on their findings, the authors give five simple rules for making train travel more popular:

Target the City Commuters: Since people hate driving in cities and love trains for work trips, focus on getting office workers to switch.
More Stops, But No Transfers: We need more train stations so people don't have to walk far to get there. However, we must be careful not to add stops that force people to switch trains. A direct train is worth more than a super-close station that requires a transfer.
The "Intercity" Makeover: Big, fast trains that skip small towns should stop more often. If they stop at a town 5km away, people can cycle there easily and avoid the transfer hassle.
Bike Parking is King: For the trip to the station, cycling is the hero. If you have a great bike garage at the station, people will cycle 5km to get there.
Walkability at the Destination: For the trip from the station, walking is the hero. If the station is in a walkable neighborhood, people will love it. If it's far away, they need a bus or a shared bike.

The Bottom Line

The paper concludes that the train isn't losing because the ride itself is bad. In fact, the ride is quite pleasant! The train is losing because the journey to get to the train is too hard, or because the journey from the train is too inconvenient.

To win more passengers, we don't just need faster trains; we need to make the "bookends" of the trip (the walk, the bike, the bus connection) seamless, direct, and transfer-free. If we can make the start and finish of the trip as smooth as the middle, the train will become the obvious choice.

1. Problem Statement

While the importance of access (getting to the station) and egress (getting from the station) trips in multimodal rail travel is acknowledged, most existing research focuses on isolated segments of the journey (e.g., only the train leg or only the access leg). There is a lack of comprehensive models that simultaneously analyze the entire door-to-door trip chain, including the trade-offs between access mode, boarding/alighting station choice, egress mode, and the main train leg, compared against private unimodal alternatives (car, bicycle, walking).

The core problem is that access and egress legs often constitute a significant portion of total travel time (up to 50% for short trips) and are perceived more negatively than the main leg. Understanding the full chain is critical for policymakers and operators to effectively shift travelers from cars to trains, particularly in dense urban areas where parking is difficult.

2. Methodology

Data Source:

Dataset: The study utilizes revealed preference (RP) data from the 2022 wave of the Dutch Mobility Panel (MPN), containing 3-day travel diaries.
Sample: After filtering for erroneous data, non-home-origin trips, and trips without a feasible train alternative (defined by having different nearest stations for origin and destination), the dataset was reduced to 5,276 feasible trips.
Mode Aggregation: 29 recorded modes were aggregated into five groups: Walking, Cycling, Bus/Tram/Metro (BTM), Train, and Car.

Choice Set Construction:

Approach: A heuristic labelling approach was used to construct choice sets for each observed trip, as RP data lacks explicit "consideration sets."
Alternatives: For each trip, the model generated a choice set of 112 alternatives, comprising:
- Multimodal Train Options: Combinations of 4 access modes (Walk, Bike, Car, BTM) $\times$ 3 nearest boarding stations $\times$ 3 nearest alighting stations $\times$ 3 egress modes (Walk, Bike, BTM).
- Unimodal Options: Single alternatives for Walk, Bike, Car, and BTM (using Google Directions/Distance Matrix APIs).
Data Enrichment: APIs (Google and Dutch Railways) were used to estimate travel times, distances, transfer counts, and ticket prices. Parking difficulty was proxied by urbanization levels.

Model Specification:

Technique: A Nested Logit (NL) model was estimated using Pandas Biogeme.
Rationale: NL was chosen over Mixed Logit to avoid computational costs associated with simulation while accounting for correlation between alternatives sharing the same main mode (train).
Nesting Structure: The optimal structure identified was a single nest containing all train-based alternatives, separating them from unimodal private modes. This yielded a nesting parameter ( $\mu$ ) of 2.45 (correlation of 0.83 within the nest).
Variables: The model included travel time (segmented by mode and leg), transfer penalties, urbanization interactions (for car), and socio-demographic interactions (trip purpose, car ownership).
Cost Handling: A direct cost parameter was excluded due to high collinearity (0.97) between travel time and imputed price; cost effects were effectively captured within the time parameters.

3. Key Contributions

Holistic Door-to-Door Modeling: Unlike previous studies that isolate specific legs, this study models the complete decision-making process, capturing the interplay between access/egress choices and the main train leg.
Quantification of Transfer Penalties: The study provides precise estimates of the disutility of transfers in a multimodal context, showing how much extra time travelers are willing to endure to avoid them.
Differentiated Time Valuation: It confirms that while Public Transport (PT) has a high initial disutility (constant), the in-vehicle time (IVT) within PT is perceived significantly less negatively than private mode travel time.
Policy-Relevant Trade-offs: The research translates model coefficients into actionable trade-offs (e.g., minutes of walking vs. avoiding a transfer) for specific urban contexts.

4. Key Results

A. Mode Preferences and Constants:

Main Leg: Private modes (Walking, Cycling, Car) are generally preferred over Train and BTM (indicated by negative constants for Train/BTM). However, Work trips significantly increase the attractiveness of Train and BTM compared to leisure trips.
Access/Egress: Walking is the dominant preferred mode for both access and egress. Cycling is highly preferred for access (due to parking availability at stations) but less so for egress (lack of destination bikes). BTM is the preferred egress mode for distances >2km.
Urbanization: Driving into highly urbanized destinations incurs a massive disutility (equivalent to an 11.5-minute time penalty), making train travel relatively more attractive for these trips.

B. Time Perception (Value of Time):

In-Vehicle Time (IVT): PT IVT (Train/BTM) is perceived as 60% less negative than private mode IVT (Walk/Bike/Car).
- Private Mode Coefficient: -0.082
- PT Mode Coefficient: -0.034
Implication: Travelers are willing to accept longer travel times in PT if it avoids transfers or improves accessibility.

C. Transfer Penalties:

Transfers are a major source of disutility.
Trade-off: Travelers are willing to walk/cycle/drive 10.5 minutes longer to the station to avoid one transfer.
PT Trade-off: Travelers are willing to accept 25.6 minutes of additional PT travel time to avoid a transfer.

D. Market Share Sensitivity:

Access: Adding cycling infrastructure can increase train market share by >10 percentage points for distances of 1–5 km.
Egress: Adding BTM options is more effective than cycling for egress, increasing market share by ~5 percentage points for distances >2 km.
Transfers: Introducing a transfer in the BTM leg drastically reduces its attractiveness, causing cycling to become the dominant mode for distances up to 4 km.

5. Significance and Policy Implications

The study provides a robust framework for optimizing rail networks and station accessibility:

Station Density vs. Directness: Operators must balance station density with service directness. While more stations reduce access time, they often introduce transfers. The study suggests that avoiding transfers is paramount; travelers will accept longer access times to avoid a transfer.
Targeting Urban Commuters: The highest mode-shift potential lies in work-related trips to dense urban areas, where the disutility of driving (parking/traffic) is highest.
Infrastructure Priorities:
- Origin (Access): Prioritize cycling infrastructure (lanes and parking) for distances of 1–5 km.
- Destination (Egress): Prioritize walkability (0–1.5 km) and high-quality BTM connections for longer distances. Shared bicycles are a viable alternative where BTM is unavailable.
Intercity Stops: Intercity trains should consider adding stops in areas >5 km from existing stations if it provides a direct connection, as the marginal increase in train travel time is less penalized than the access time saved.

Limitations & Future Work:
The study could not include a direct cost parameter due to data collinearity. Future research should aim to incorporate shared micromobility data and use latent class models to better account for respondent heterogeneity and consideration set formation.