[Paper Review] Thresholds in choice behaviour and the size of travel time savings

A. A. Obermeyer, Martin Treiber+1 more|arXiv (Cornell University)|Feb 14, 2014

Economic and Environmental Valuation36 references3 citations

TL;DR

This paper investigates how thresholds in individual choice behavior affect the perceived value of travel time savings, proposing a modeling framework using synthetic and stated choice data to demonstrate that ignoring such thresholds leads to biased estimates of time savings' value. The key contribution is a method to accurately capture non-linear valuation of time, especially for small savings, improving transport appraisal accuracy even when unit-value discounting is rejected.

ABSTRACT

Travel time savings are usually the most substantial economic benefit of transport infrastructure projects. However, questions surround whether small time savings are as valuable per unit as larger savings. Thresholds in individual choice behaviour are one reason cited for a discounted unit value for small time savings. We demonstrate different approaches for modelling these thresholds using synthetic and stated choice data. We show that the consideration of thresholds is important, even if the discounted unit value for small travel time savings is rejected for transport project appraisal. If an existing threshold is ignored in model estimation, the value of travel time savings will be biased. The presented procedure might also be useful to model thresholds in other contexts of choice behaviour.

Motivation & Objective

• To investigate whether small travel time savings are valued less per unit than larger ones due to psychological thresholds in decision-making.
• To address the methodological challenge of modeling non-linear valuation of time savings in transport choice behavior.
• To demonstrate that ignoring thresholds in model estimation leads to biased estimates of travel time savings' value.
• To develop and validate a modeling procedure applicable to choice behavior beyond transport, particularly where threshold effects may exist.

Proposed method

• The authors use synthetic and stated choice data to simulate and analyze individual travel decisions under varying time savings.
• They model thresholds in choice behavior using a piecewise linear utility function that captures non-linear valuation of time.
• The approach incorporates a threshold parameter that determines when time savings begin to be perceived as valuable, allowing for different marginal utilities below and above the threshold.
• The model is estimated using maximum likelihood, with statistical testing to assess the significance of the threshold effect.
• The method is validated by comparing model fit and parameter estimates with and without threshold assumptions.
• The framework is generalized to be applicable to other domains involving threshold effects in choice behavior.

Experimental results

Research questions

• RQ1 Do individuals assign a lower marginal value to small travel time savings compared to larger ones due to psychological thresholds in decision-making?
• RQ2 How can threshold effects in travel time savings be modeled accurately using stated choice data?
• RQ3 What is the impact of ignoring threshold effects on the estimation of travel time savings' value in transport appraisal?
• RQ4 Can the proposed modeling approach be generalized to other contexts involving threshold-based choice behavior?

Key findings

• The presence of a threshold in choice behavior significantly affects the valuation of travel time savings, with small savings being perceived as less valuable per unit.
• Ignoring the threshold in model estimation leads to biased estimates of the value of travel time savings, even when the unit-value discounting hypothesis is rejected.
• The proposed model with a threshold parameter provides a better fit to stated choice data than standard linear models, particularly when small time savings are involved.
• The threshold effect is statistically significant in the data, indicating that individuals do not value time savings linearly across all magnitudes.
• The method successfully captures non-linear valuation patterns, offering a more accurate representation of real-world decision-making.
• The framework is transferable to other domains involving threshold effects in individual choice behavior.