This section reviews past literature investigating WTP for AV or AV technology, and states the research gaps to be addressed. Then, it is followed by variable selection and formulation of the model being tested.
2.1 Past studies investigating WTP for AV/AV technology
WTP is defined as the highest price an individual is willing to pay for a product or service. Pricing for a service can be operationalised in two ways: cost-based pricing and value-based pricing. Cost-based pricing is a pricing method that sets a selling price of goods or services by adding a fixed sum or a percentage of the total cost as profit to the cost of the product. Value-based pricing prices goods or services according to their perceived value to customers. When value-based pricing is used, the perceived values which influence one’s WTP for a service must be identified [15]. In the case of operating an AV service, the expected service quality attributes influence one’s WTP for the service. Furthermore, perceptions vary according to social-demographics and familiarity with the content and context of a service [29]. Considering valid estimates of WTP for the type of AV service with inclusion of the relevant perceptions are indeed vital for sensible pricing strategy of the AV service [3].
Asgari and Jin [2], Liu et al. [19], Bansal and Daziano [4], Talebian and Mishra [28], Jiang et al. [14], Daziano et al. [10], Bansal et al. [7], and Kyriakidis et al. [17] investigated WTP for AV or AV technology. Jiang et al. [14], Bansal et al. [7], and Kyriakidis et al. [17] examined WTP itself only, without investigating the factors affecting WTP. On the other hand, several studies looked into the factors influencing WTP for AV or AV technology. Daziano et al. [10] used mixed logit models to estimate WTP for private AV with different levels of automation for various socio-demographics. The results showed that the preferences for automation are diverse. Households who are not aware of driverless car technology are not willing to pay extra money for the technology. Out of the 1260 respondents with good representation of U.S. population, those who drive long distance, being aware of driverless car technology, own car, and with higher education have the most desire towards the technology and are willing to pay USD2,784 for a privately owned automated car with partial automation (level-3 and level-4 automation) and USD6,580 for a privately owned automated car with full automation.
Talebian and Mishra [28] included WTP in the simulation model of the adoption of CAV and found that WTP is affected by peer-to-peer communication about the technology. Bansal and Kockelman [6] regressed socio-demographic predictors, location-based predictors, travel-based predictors, tech-based predictors and safety-based predictors with WTP for private AV with different automation levels, and found that older people and experienced drivers have lower WTP for the technologies. Liu et al. [19] tested a psychological model to explain WTP for fully automated driving technology; social trust is identified to impact WTP directly rather than through perceived risk and perceived benefit.
Lastly, other than including socio-demographic variables like in the afore-mentioned studies, [2] included respondent attitudes towards driving, factors affecting mode choice selection, shared transportation, multi-tasking and new technology in the structural equation model to study their effects on WTP for four levels of automation (basic vehicles, adding advanced features, partial automation and full automation). They concluded people are willing to pay more when they believe that using the automated features/services will provide them with better utility in terms of time and cost savings, stress reduction, convenience and quality of life. This highlights the importance of understanding WTP from user perceptions point of view, particularly their level of expectation or satisfaction with the service quality attributes of an AV service.
The understanding of user expectation, perceptions and satisfaction about a product or service attributes/characteristics is crucial towards deploying a successful product and/or service delivery [18, 21]. Such knowledge is important to operators and stakeholders who are keen to provide market-driven AV services, and fosters more realistic estimation of the long-term demands and impacts from their implementations. As highlighted by Abenoza et al. [1] operational and service quality attributes especially service frequency, length of trip time, and service provider’s responsiveness in public transport services, play a more crucial role in influencing the travel demand and travel satisfaction than the physical design and technology of the vehicle itself. Herein, user expectation, perceptions, and acceptance towards such operational characteristics should not be disregarded when investigating factors that influence user WTP for AV services. Moreover, the results from Shin et al. [27] who evaluated WTP of South Koreans for private vehicles with different smart options (automated driving, connectivity, voice command, wireless internet and availability of software application) at different option prices showed that consumers are homogenously sensitive to price but vary in their preferences of the smart options. This shows the importance to understand the effects of the associated perceptions of the AV services at individual service quality attributes instead of the services as a whole bundle.
WTP values for various AV services in past studies except [2] explored mainly the effects of socio-demographic characteristics, commuting behaviours and other non-attitudinal variables on WTP for AV or AV technology. There is hardly any study that determines the impacts of service quality attributes of AV deployment towards one’s WTP for these services. Also, knowledge is still very sparse about WTP for SAV services, particularly factors associated with user WTP for SAV service as compared to PAV service. To fill these gaps, this study focuses on identifying significant explanatory factors of WTP for three types of AV services: 1) on-demand personalised AV (PAV) service, 2) demand responsive shared AV (SAV) service, and 3) first−/last-mile automated bus (AB) service that operates in the same way as existing scheduled public bus service, particularly the effects of perceived service quality attributes on WTP for each type of service. Factors that drive or hinder WTP for the services can be for: 1) WTP to try out a new service, with no intention to use it habitually, and 2) WTP to pay for a service, with intention to use it regularly. This study focuses on the later, with target on regular usage.
2.2 Variable selection and model formation
This study aims to investigate the core service quality attribute expectations affecting WTP for three types of AV services: PAV, SAV and AB services, and also how these perceptions vary with respect to socio-demographic characteristics, knowledge and experiences with AV, and existing travel modes. Herein, structural equation modelling with service quality attribute perceptions as latent variables is applied to achieve the objective.
Socio-demographic predictors which are significant to WTP for AVs such as age [7], gender and income level [7, 17] are included in the model. Additionally, significant factors influencing acceptance of AVs such as technology awareness [26], use of multiple travel modes [16], and ride experience [7] are also included. Salonen [24], Eden et al. [11], Scheltes and de Almeida Correia [25] and Piao et al. [23] investigated acceptance of using small AB. On-board safety [23, 24], comfort [11], travel time [7, 25], travel fare [7, 23] and presence of steward [23] are found to be influential to people’s acceptance of AB usage. Hence, service quality attributes including frequency, safety, ride comfort, travel time and travel cost are included as latent variables in the structural equation model applied in this study. Figure 1 shows the constructed structural equation model with service quality attribute perceptions as latent variables. The model is built using the framework of the structural modelling equation part of the Integrated Choice and Latent Variable (ICLV) model [8]. However, the whole ICLV model is not adopted because there is no choice selection involved in this study. This means that, for this study, WTP is not inferred from a stated preference study in which respondents state their preferences through their choices. Instead, the respondents directly state the amount of money they are willing to pay for the AV services.