Implications of automated vehicles for accessibility and location choices: Evidence from an expert-based experiment

Highlights

• Accessibility impacts of fully automated vehicles (AVs) are explored.

• Q-method is applied among a sample of seventeen international accessibility experts.

• AVs accessibility benefits are highly uncertain.

• AVs can induce densification of city center and further urban sprawl.

• AVs accessibility benefits will not be equally distributed among different social groups.

Abstract

In this paper, possible accessibility impacts of fully automated vehicles (AVs) are explored. A conceptual framework for those impacts is developed based on the model of four accessibility components (i.e. land use, transport, temporal and individual) of Geurs and van Wee (2004). Q-method is applied among a sample of seventeen international accessibility experts to explore heterogeneity among experts with respect to the impacts of AVs on accessibility, and study different views and clusters of experts. Q-method statements are deductively categorized according to four accessibility components of the conceptual framework. Three viewpoints were extracted, indicating that experts expect AVs to influence accessibility through all four accessibility components. Viewpoint A expects that accessibility benefits stemming from AVs will be highly uncertain, mainly because of induced travel demand that will likely cancel out travel time and cost savings of AVs in the long term. Viewpoint B anticipates that accessibility changes because of AVs will have two opposing implications for urban form: densification of city center and further urban sprawl. Finally, viewpoint C expects that those who can afford an AV will mainly enjoy AVs benefits, thus AVs will have more negative than positive implications for social equity.

Introduction

Automated vehicles could have significant implications for cities and transport systems. Milakis et al. (2017b) identify three stages of sequential impacts after introduction of AVs: first order (traffic, travel cost and travel choices), second-order (vehicle ownership and sharing, location choices and land use, transport infrastructure) and third-order (energy consumption, air pollution, safety, social equity, economy and public health). This paper focuses on the implications of AVs for accessibility and the location choices.

Thus far, only few studies have explored these impacts using quantitative modeling methods. Childress et al. (2015) used an activity-based model in Seattle, WA to simulate a transport system entirely based on AVs and to explore possible accessibility changes. These researchers concluded that the introduction of AVs could enhance accessibility across the region, particularly in rural areas. A second study explored land use impacts of automated driving from an urban economics perspective (Zakharenko, 2016), concluding that automated driving could induce two divergent land use dynamics in the city. Reduced transport costs could cause cities to further expand, while reduced parking requirements could enhance density of economic activity at the center of the cities. Similarly, Gelauff et al. (2017) using simulations of a spatial general equilibrium model (LUCA) in the Dutch context concluded that automated vehicles could induce both urban dispersion and concentration effects. Dispersion of population in suburban areas resulted when more productive use of car travel time was assumed in the model. Concentration of population resulted when most public transport services (i.e. bus, trams, metro) were replaced by door-to-door shared automated mobility services. Papa and Ferreira (2018) employed Geurs and van Wee's (2004) definition of accessibility to identify critical governance decisions that could steer impacts of AVs on the four accessibility components (i.e. land use, transport, temporal and individual) toward an optimistic or a pessimistic future with respect to the possible benefits for the society. Beyond these studies, some theoretical and empirical work has been done in the related area of Intelligent Transport Systems (ITS) by Argiolu et al., 2008, Argiolu et al., 2013, showing that these systems have significant impacts on location preferences of office-keeping organisations within urbanised areas. However, literature so far has not provided empirical evidence about potential impacts of AVs on accessibility and the location choices (see e.g. van Wee, 2016; Anonymous, 2017).

Our study aims to fill this knowledge gap by exploring these impacts through an expert-based approach. AVs are a radical and potentially even a disruptive innovation, and it is very difficult to forecast the implications of such innovations, as well as the transition path and penetration rates. What is going to happen, depends – among others – on path dependence, (potential) lock-in, coincidence, and many more factors, as explained by evolutionary economics (see Rammel and Van den Bergh, 2003). It is much easier to explain from hindsight what has happened and why, than it is to accurately forecast what is going to happen, especially in case of disruptive innovations. Therefore, we argue it is better to explicitly explore heterogeneity among experts, and study different views and clusters of experts.

To this end, we apply the Q-method among a sample of international accessibility experts to explore possible impacts of AVs on accessibility and the location choices. The Q-method is considered appropriate in this case because it allows capturing heterogeneity in subjective viewpoints regarding a particular topic. Other methods to explore expert opinions generally strive for reducing heterogeneity among experts. The Delphi method, for example, is even designed to reduce heterogeneity among respondents by presenting preliminary results in a second (or even third) round of expert elicitation, aiming to explore reasons for heterogeneity and next reduce it.

In this study, we focus on the impacts of fully automated vehicles (SAE level 5; SAE International, 2016) and we take into account possible synergistic effects of vehicle automation and vehicle sharing. Fully automated vehicles can perform all dynamic tasks of driving (e.g. monitor the driving environment, steering, acceleration/deceleration), in all conditions (e.g. highways, urban streets). They can travel both occupied and unoccupied (e.g. to park or reposition themselves in the case of shared automated vehicles). This study does not distinguish between autonomous and cooperative vehicles (i.e. vehicles that can communicate with each other and/or with the infrastructure). Below, we analyze our conceptual framework on accessibility and location choice impacts of AVs (Section 2), we describe the Q-method and how we applied it in this study (Section 3), and we present the results of our expert-based experiment (Section 4). We close this paper with the conclusions (Section 5).