A discrete choice experiment on consumer's willingness-to-pay for vehicle automation in the Greater Toronto Area

The study examines how consumers value different levels of vehicle automation when choosing between private ownership and shared-use options. Motivated by the potential for both positive (safety, mobility, inclusion) and negative (increased VKT, congestion, sprawl, emissions) impacts of AVs, the research addresses the gap that prior willingness-to-pay (WTP) studies rarely model simultaneous choices of automation level and ownership type. Given the gradual deployment of automation (levels 0–5 per SAE) and today’s market presence of levels 1–2, the paper seeks to quantify WTP for intermediate and full automation levels and to understand the role of ownership (private vs shared) and behavioral inertia toward conventional vehicles in the Greater Toronto Area (GTA).

Prior work highlights wide-ranging consumer attitudes toward AVs and substantial heterogeneity in WTP. Many studies used stated preference (SP) or opinion surveys, often relying on direct/open-ended WTP elicitation, which is criticized for bias. Reported WTP values vary widely: e.g., Schoettle and Sivak (2014) found >50% with zero WTP, while Kyriakidis et al. (2015) reported mean WTPs around US$4,000–4,500 and substantial dispersion (22% zero, 5% >US$30,000). Others (Bansal & Kockelman, Daziano et al., Ellis et al., Olsen et al., Liu et al.) similarly found heterogeneity, demographic effects (younger, educated, urban more likely early adopters), and limited consideration of intermediate automation levels or shared ownership. Few studies combined ownership type with automation level in a modeling framework. The review also notes broader AV impacts on travel behavior and land use, and calls for modeling approaches capturing inertia and taste heterogeneity.

Study area: Greater Toronto Area (GTA), Canada (population >6.4M in 2016). Eligible respondents were GTA adults (≥18) planning to acquire a new vehicle within 5–10 years. Data collection used SurveyGizmo (March 2019); 238 completes, 190 usable after cleaning. The sample was reasonably aligned with 2016 Census but slightly over-represented Toronto residents, higher incomes, younger and more educated individuals. Survey design: Two parts—(1) questionnaires on household/person attributes, vehicle inventory, commuting, AV familiarity/comfort, and directly stated WTP by automation level; (2) an SP choice experiment. SP alternatives (four): - PCV: Private conventional vehicle (automation levels 0–1, i.e., driver assistance). - PAV: Private autonomous vehicle (levels 2–5). - SAV: Membership in an AV sharing program (levels 2–5; on-demand, hourly charge). - PSAV: Private vehicle plus AV sharing membership simultaneously (private vehicle level 1–3; shared vehicle level 2–5; shared level always higher than private). Attributes and levels: - Private ownership: automation level; additional cost of automation (PCV: base, +$500, +$1,000, +$2,000; PAV/PSAV private car: +$500, +$1,000, +$2,000, +$3,000, +$4,000, +$8,000, +$16,000, +$24,000, all CAN$); driving cost ($20/$30/$40/$50 per 100 km); monthly parking+insurance ($120/$230/$340/$450); monthly mileage limit (Unlimited, included for symmetry but applies to shared); general traffic condition (Free flow; Below speed limit; Stop and go; Congested). - Shared ownership: automation level; membership application fee ($0/$45/$90/$130); driving rate ($10/$15/$20/$25 per hour); monthly subscription fee ($0/$5/$7.5/$10); monthly mileage limit (1350/1600/4400/Unlimited km); general traffic condition (same 4 levels). Experimental design: D-efficient design; 8 SP scenarios per respondent; attributes varied across scenarios. A typical scenario showed all four alternatives with their attribute levels for the respondent to choose from. Stated perceptions and initial values: Average planned budget for next vehicle CAN$37,121; >38% moderately familiar with AVs; ~45% comfortable riding at least SAE level 3; directly stated mean WTPs (CAN$): L1: 1,042; L2: 1,667; L3: 2,197; L4: 2,642; L5: 3,351 (large standard deviations indicating heterogeneity). Empirical models: Heteroskedastic error-component mixed multinomial logit (ML) and mixed nested logit (NL) models were estimated to capture random taste heterogeneity and correlations among alternatives. A lagged-utility inertia formulation modeled propensity to remain with PCVs across sequential SP tasks: if PCV chosen in scenario i, utilities for non-PCV options in scenario i+1 include an inertia term parameterized by a population mean, individual deviation, and motivating factors (e.g., current private vehicle ownership, driver’s license, car-sharing membership). Estimation used GAUSS with 500 Halton draws and robust t-tests. Various model specifications (MNL base; ML with error components; ML with inertia; ML with inertia and motivators; subset by AV familiarity; mixed NL with best-fitting nest) were compared using LL, AIC/BIC, and McFadden’s rho-square.

- Strong inertia toward PCVs: A significant, positive inertia mean was identified across models with inertia, indicating reluctance to switch from PCVs absent substantial utility gains. Inertia was higher among those currently owning private vehicles and those holding driver’s licenses; current car-sharing members exhibited reduced inertia. - Preferences by automation level and ownership: - Private ownership: Consistent positive acceptance of high automation (Level 4) in private vehicles; mixed/insignificant parameters for full (Level 5) private automation, suggesting uncertainty or desire to retain manual driving capability. - Shared use: Positive acceptance of full (Level 5) automation within SAVs; Level 4 in SAVs not clearly preferred. - A mixed nested logit identified a valid nest grouping PCV and SAV (interpreted as a segment disfavoring private AV ownership), indicating PCV users may adopt AVs first via shared services. - Socio-demographic effects: Millennials (age <40) and full-time workers were more inclined to choose AVs (both PAV and SAV). In NL with cost-taste segmentation, higher WTP was associated with age <40, having a family member with disabilities, commuting >15 km one-way, and household income below the area average (<CAN$80,000). - Willingness-to-pay (WTP) for Level 4 private automation: Model-based WTP estimates (accounting for interaction with vehicle price via log(price)) ranged approximately from CAN$10,800 to CAN$29,800 depending on vehicle price and model. Example WTPs (in thousands CAN$): at vehicle price CAN$15k, WTP ~10.8–14.6; at CAN$50k, WTP ~15.5–21.1; at CAN$100k, WTP ~18.3–24.8; at CAN$250k, WTP ~21.9–29.8. Individual-level WTPs averaged around CAN$18,000, with 67.9% between CAN$10,000 and CAN$20,000 and 59.1% less than 50% of the next vehicle price. These model-based WTPs were substantially higher than directly stated WTPs (mean L4 ~CAN$2,642), aligning with literature that direct elicitation can be biased. - Stated preference behavior: Across 8 scenarios, many respondents consistently chose PCVs (around half in 7 of 8 scenarios), except when PAV L3 carried only +CAN$500 additional cost, where 40% chose PAVs, indicating potential for rapid shifts if AV pricing is competitive. - Sample descriptors: Average planned new-vehicle budget CAN$37,121; over 38% moderately familiar with AVs; ~45% comfortable riding Level 3+ AVs.

The study demonstrates that consumers’ WTP for automation is strongly contingent on both the automation level and the ownership model. Incorporating behavioral inertia materially improves model fit and reveals that many consumers exhibit status-quo bias toward PCVs, implying that adoption trajectories may be gradual unless AVs offer clear utility gains and competitive pricing. The consistent preference for Level 4 automation in private ownership, coupled with hesitancy around fully automated private vehicles (Level 5), suggests that retaining manual driving capability is important to private buyers. In contrast, shared services are acceptable venues for full automation, implying a likely pathway where conventional drivers initially engage with AVs via SAVs before considering private AV ownership. Socio-demographic segmentation underscores target groups (younger, commuters, households with disability) with higher WTP and potential early adoption. The results validate the need to jointly model ownership and automation choices when forecasting AV market penetration and pricing strategies.

The paper contributes by jointly estimating consumers’ preferences and WTP for multiple automation levels alongside ownership types (private, shared, combined) using SP data from the GTA and advanced discrete choice models that incorporate random taste heterogeneity, nesting, and inertia. Key findings include: strong inertia toward PCVs; private buyers favor Level 4 automation while SAV users favor Level 5; and WTP for Level 4 private automation ranges roughly from CAN$10,800 to CAN$29,800 depending on vehicle price. Socio-demographics (age, disability in family, commuting distance, income) influence WTP and adoption propensity. Future work will: (i) develop thresholds/ranges to correct overestimation of WTP for low-priced vehicles; (ii) expand samples and integrate richer attitudinal variables capturing inertia; (iii) incorporate additional AV attributes (e.g., safety, liability) and explicitly model ride-hailing as a separate mobility option; and (iv) conduct focused studies on the high-inertia PCV segment to better forecast mass AV adoption.

- Model-based WTP for low-priced vehicles may be overestimated (e.g., implying very high shares of vehicle price); thresholds for effective WTP ranges are needed. - SP data from a single region (GTA) with 190 usable records; generalizability may be limited. - Sample over-represents Toronto residents, higher-income, younger, and more educated respondents relative to census. - Hypothetical bias and limited real-world experience with advanced automation (only one Tesla Level 2 owner reported) may affect responses. - Ride-hailing was not modeled as a standalone alternative; shared ownership was represented via a car-sharing construct.