Departure Time Choice with Parametric Heterogeneity: Equilibrium and Instability

Hillel Bar-Gera, Stephen D. Boyles, Liron Ravner · 2026

Read Departure Time Choice with Parametric Heterogeneity: Equilibrium and Instability by Hillel Bar-Gera, Stephen D. Boyles, Liron Ravner (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-13831.txt.

Introduction

Bar-Gera, Boyles, and Ravner study departure-time choice with varied schedule-delay parameters. Vickrey's classic single-bottleneck equilibrium is unstable under many plausible daily learning dynamics, yet real traffic shows no such instability. The authors ask whether dynamics or a simple assumption explains the gap. Their variant uses a spread of schedule-delay parameters with a monotonic earliness-lateness link. Equilibrium exists and is unique here. Surprisingly, they prove that a broad family of day-to-day dynamics respecting local pressure and order keeping stays unstable here.

Catalog record

CREATORHillel Bar-Gera, Stephen D. Boyles, Liron Ravner
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE81585
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.13831

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Departure Time Choice with Parametric Heterogeneity: Equilibrium and Instability Hillel Bar-Gera, Stephen D. Boyles, Liron Ravner Abstract Vickrey’s classic single-bottleneck departure time choice equilibrium model exhibits instability under many plausible day-to-day learning dynamics. Such instability is not observed in reality — does this difference stem from the day-to-day dynamics or from one of the simplifying assumptions of the basic model? This paper explores a variant of the basic model with a continuous distribution of schedule delay parameters which we intuitively expect to have more favorable stability properties. To attain tractability we assume a monotonic relationship between earliness and lateness parameters. We first verify the existence and uniqueness of the equilibrium solution for this model. We then study a broad class of day-to-day dynamics satisfying local pressure and order preservation conditions. Our main contribution is a formal proof that, surprisingly, all such day-to-day dynamics in this context are unstable. 1 INTRODUCTION Vickrey’s work on departure time choice inspired a substantial number of subsequent studies ( 22 ) . A relatively recent bibliometric review identified 232 journal publications in 50 years that relied on Vickrey’s work ( 15 ) . These subsequent studies have proceeded along many different lines. Among these are extensions of Vickrey’s model in which travelers are heterogeneous in some way (e.g., schedule delay parameters), and investigations of the stability properties of the departure time choice equilibrium under perturbations. Stability is important both for practical and algorithmic reasons. Practically, stable equilibria are more likely to be observed in the field. Algorithmically, stability is helpful when designing iterative algorithms that converge to an equilibrium solution. While the equilibrium in Vickrey’s original model and in some extensions can be derived in closed form, numerical computations are important when embedding departure time choice in larger network models, or in other more complex extensions. As described in more detail in the literature review, most stability results to date are negative, and indicate that the original Vickrey bottleneck is unstable under many natural dynamic processes. These two streams of research (heterogeneity and stability) have proceeded more or less independently of each other, but we believe there is good reason to consider them together. In many cases, traveler heterogeneity implies a unique departure order at equilibrium, whereas in the homogeneous model the travelers are fully interchangeable. Furthermore, in the standard model, at equilibrium each traveler is indifferent among all possible departure times during the peak, whereas heterogeneous models can be formulated where each traveler’s departure time has strictly minimum cost. Both of these properties intuitively suggest that stability is easier to achieve in a heterogeneous setting. The main contribution of our paper is to investigate this intuition rigorously. Surprisingly, we find that it is not true, and that equilibrium is unstable in such settings. More specifically, we explore a variant of the basic model with a continuous distribution of schedule delay parameters, and investigate stability properties of the equilibrium solution. We model traveler heterogeneity with a monotonic relationship between the earliness and lateness parameters. As a result, the general case where the parameter space is two-dimensional is replaced with a single-dimensional parameter space. We believe such a model is plausible behaviorally, and considerably simpler to analyze than a general joint distribution of schedule delay parameters. We prove equilibrium existence and uniqueness under strict monotonicity, and show that the travelers’ order at equilibrium is known a priori . We then consider a general family of day-to-day dynamics, or iterative processes. Our analysis covers any iterative process that satisfies two assumptions: i) the a priori order is maintained throughout the iterative process; and ii) each traveler acts myopically to improve her own conditions. A formal definition of these assumptions, as well as the motivations for adopting them, will be discussed in Section 5 . Our main finding is that any such iterative process cannot guarantee convergence to equilibrium, even within arbitrarily small neighborhoods around the equilibrium. In that sense, the equilibrium solution is fundamentally unstable, despite having a unique departure order, and despite each traveler departing at a time which is uniquely minimal for her. Our main finding is thus also negative, and suggests that the cause of instability in departure time choice models is more fundamental. Implications for other variations of departure time choice models, including relaxing the monotonicity assumption, are presented in the discussion section at the end of the paper. 2 Literature review In Vickrey’s original model, all travelers are identical in their value of time, and in their scheduling preferences. Since then, researchers have generalized Vickrey’s model to consider different kinds of heterogeneity among travelers. This can include variations in the desired arrival time, as in 10 , 21 , and 5 , or in the penalty factors for queueing delay, early arrival, and late arrival, as in 17 , 2 , and 3 . Both continuous and discrete forms of heterogeneity have been studied; see Section 3.1.6 and Table 6 in 15 for a comprehensive review of traveler heterogeneity in the single bottleneck model. In many heterogeneous formulations, travelers obey an ordering principle at equilibrium. An interesting framework for demonstrating travelers’ order in heterogeneous departure time choice models was proposed by 1 , combining results of several previous studies on this topic. Their analysis relied primarily on the equivalence between the system optimum solution for departure time choice and the linear formulation of the ‘total transport’ problem, with a continuous time dimension. They discussed sufficient conditions on the schedule cost derivative for interpreting Lagrange multipliers as queue length, thus demonstrating the equivalence between the system optimum and user equilibrium solutions. They also showed that if schedule costs satisfy the Monge condition, for a certain permutation of traveler classes, then arrivals at equilibrium must follow a corresponding ‘sorting pattern’. This is true in one-dimensional heterogeneity such as: desired arrival time; earliness penalty (without late arrivals); and lateness penalty (without early arrivals). In each of these cases additional assumptions were made (e.g. convex or quasi-convex schedule cost). In the case of two-dimensional heterogeneity with respect to both earliness and lateness penalties, a sorting pattern occurs among early arrivals and among late arrivals, due to properties of two subproblems of the main system optimal problem. Although many researchers have studied heterogeneity in different forms, comparatively few have addressed the issue of solution stability. Only two of the studies listed in the review of 15 focus on the day-to-day dynamics ( 8 ; 25 ) , and both assumed bounded rationality. To our knowledge, the first to consider the stability of day-to-day learning dynamics in the context of departure time choice, and particularly in the context of Vickrey’s bottleneck model, was 7 . In his paper, de Palma proposed four dynamic evolution strategies, aiming to shift travelers towards lower-cost departure times, and evaluates their performance numerically. His experiments showed that these dynamics often lead to oscillatory behavior that fails to converge; one exception is a logit-based dynamic which was seen to converge to a stochastic user equilibrium when unobserved costs are sufficiently high relative to queueing and schedule delay (a paradoxical finding, since convergence to equilibrium seems easiest when travelers care about it the least). 11 provided numerical evaluation of the ‘proportional-switch’ dynamics. Subsequently, 12 proved

Read the complete text