Assessing public opinions of and interest in new vehicle technologies: An Austin perspective

The study addresses how the public perceives, accepts, and is willing to adopt connected and autonomous vehicle (CAV) technologies, including autonomous vehicles (AVs) and shared autonomous vehicles (SAVs). Motivated by the significant safety, congestion, and economic burdens of conventional car travel (e.g., 2.2 million injuries and 30,000+ fatalities annually in the U.S.; ~$300B in crash costs), the research investigates factors influencing willingness to pay (WTP) for various levels of automation and connectivity, likely SAV adoption under different pricing, peer effects on AV adoption timing, and potential residential location shifts in response to AV/SAV availability. As AV/CV technologies rapidly mature and policies evolve, understanding demographic and built-environment determinants of acceptance and adoption is crucial for forecasting market penetration, guiding infrastructure and land-use planning, and crafting policies that realize safety and sustainability benefits while mitigating potential downsides such as induced travel and sprawl.

Recent surveys and expert elicitations provide descriptive insights but little econometric analysis linking demographics and built environment to CAV adoption.

• Casley et al. (2013): Among 467 respondents, ~30% would spend >$5000 for full automation; safety was the dominant adoption factor (82%).
• Begg (2014): Among >3500 UK transport professionals, high expectations for Level 2 by 2040; ~60% expected AVs safer than conventional vehicles.
• Kyriakidis et al. (2014): 5000 respondents worldwide; higher VMT and ACC users willing to pay more; ~20% willing to pay >$7000 for Level 4; 69% expect 50% market share by 2050.
• Schoettle & Sivak (2014a): 1533 respondents across US/UK/AUS; two-thirds aware of AVs; WTP varied by country; many not willing to pay extra; females more concerned than males.
• Underwood (2014): Expert survey; legal liability seen as toughest barrier; consumer acceptance least; skepticism about practicality of Level 3.
• CarInsurance.com/Vallet (2013): Insurance discounts increase stated interest; many believe they drive safer than AVs; trust placed mostly in traditional automakers.
• Seapine Software (2014): High levels of concern about riding in AVs, especially equipment failure and liability/hacking.
• J.D. Power (2012): Interest drops after revealing price; urban young males show highest interest.
• KPMG (2013) focus groups: Incentives (e.g., designated lanes) increase interest; gender differences noted.
• Schoettle & Sivak (2014b) on CVs: Low awareness; safety seen as primary benefit; many unwilling to pay extra. Gaps: Prior work lacked analysis of SAV adoption under pricing scenarios, home-location impacts, peer effects on AV adoption timing, and did not employ econometric models to quantify relationships—gaps this study addresses.

Survey design and sampling: An online Qualtrics survey targeted adult residents of Austin, Texas (Oct–Dec 2014). Of 510 initiations, 358 completes; 11 non-Austinites removed, yielding N=347. The sample over-represented women, ages 25–44, and higher education; population weights were computed using 2013 ACS PUMS for Austin across 72 demographic categories (gender, age, education, HHI) to correct biases. Respondents’ home addresses were geocoded (Google Maps API); locations were spatially joined to Traffic Analysis Zones (TAZs) using QGIS to add built-environment attributes (densities, area type, incomes).

Survey content: 52 questions on awareness and perceptions of AVs/CVs; WTP for Level 3 and Level 4 AVs and connectivity; SAV adoption likelihood under $1/$2/$3 per occupied mile; AV adoption timing relative to friends’ adoption; potential home-location shifts after AV/SAVs become common; current travel patterns; and demographics. Context and price anchoring were provided (e.g., existing taxi/Uber/Car2Go rates; connectivity hardware under $100; AV activity possibilities; potential residential tradeoffs).

Modeling approach: Population weights applied to all records. Ordered probit (OP) models estimated in Stata 12 for:

• SAV adoption frequencies under three pricing scenarios ($1, $2, $3 per mile)
• WTP for adding connectivity (CVs)
• AV adoption timing (never; when 50% of friends adopt; when 10% adopt; as soon as available)
• Home-location shift (closer to center; stay; farther)

Bivariate ordered probit (seemingly unrelated) jointly estimated WTP for Level 3 and Level 4 AVs (separate ordinal outcomes), allowing correlated errors across equations (ρ estimated).

Variable selection: Initial specifications included demographic, built-environment, and technology-awareness covariates (Table 1). Stepwise elimination removed the least significant covariate iteratively until all remaining had p<0.32 (|Z|≥1.0), acknowledging limited sample size. Highly significant predictors: p<0.01 (|Z|>2.58).

Practical significance: Defined as ≥30% change in predicted probability of an outcome for a one-standard-deviation increase in a predictor (ratio >1.3 or <0.7). Highly practically significant if >50% change. Goodness-of-fit reported via McFadden’s R2 and adjusted R2. Threshold parameters estimated to divide ordered categories. For some models, log(Annual VMT) used in estimation with ΔPr reported for Annual VMT.

Descriptive insights (population-weighted):

• Awareness/tech familiarity: 92% carry a smartphone; 80% heard of Google’s self-driving car; 59% recognize ABS as automation; 95% familiar with carsharing; 88% with Uber/Lyft.
• WTP: Level 3 AVs — 48% < $2000; 28% $2000–5000; 24% > $5000. Level 4 AVs — 34% < $2000; 18% $2000–5000; 19% $5000–10,000; 28% > $10,000. Average WTP: Level 4 $7,253 vs Level 3 $3,300.
• SAV reliance: At $1/mi — 13% would rely entirely; 28% at least weekly; 24% at least monthly; 35% less than monthly. At $2/mi — 3% entirely; 12% weekly; 28% monthly; 57% less than monthly. At $3/mi — ~1.9% entirely; ~2.1% weekly; 26% monthly; 70% less than monthly.
• Connectivity: 55% interested; 19% neutral; 26% not interested at < $100 cost.
• AV interest/concerns/benefits: 41% very interested in Level 4 AVs; top concern is equipment/system failure (50% very worried). Fewer crashes seen as very likely benefit by 63%; congestion relief least likely (31% unlikely). Preferred in AV: looking out the window (77%) and texting/talking (74%).
• Adoption timing: 30% would adopt as soon as available; 25% when 10% of friends adopt; 26% when 50% adopt; 19% never.
• Home location: 74% would stay put; 14% move closer to central Austin; 12% move farther.

Econometric results (directionality and notable effects):

• WTP for AVs (bivariate OP; ρ=0.921; McFadden’s R2≈0.10): Higher among males, those with more children, higher neighborhood incomes, those who drive alone for work/social trips, have higher VMT, live farther from work (especially for Level 4), and with more prior crash experiences. Lower among older respondents, licensed drivers (relative tendency vs private reliance), residents in job-dense neighborhoods, and those familiar with carsharing/ridesharing (suggesting preference for SAV use over purchasing automation). Annual VMT and distance-to-work positively associated with Level 4 WTP but negatively with Level 3, consistent with valuing productive in-vehicle time only at full automation.
• SAV adoption frequency (OP across prices; McFadden’s R2≈0.12–0.17): More frequent use among full-time male urban residents, tech-aware respondents (heard of Google car; recognize ABS as automation), those with more prior crashes, those farther from work (especially at $1–$2/mi), and in higher population-density areas. Less frequent use among licensed drivers, those familiar with carsharing (particularly at $2–$3/mi), higher annual VMT (at $3/mi), those living farther from downtown (at $3/mi), and in higher service-employment-density areas.
• WTP for connectivity (OP; McFadden’s R2≈0.13): Greater interest among those who heard of Google car, experienced more crashes, carry smartphones, drive alone for work/social trips, have higher VMT, live farther from work, in urban areas, and in higher household-density neighborhoods.
• AV adoption timing (OP; McFadden’s R2≈0.10): Faster/less peer-dependent adoption among males, higher-income respondents, urban residents, those with higher VMT, more crash experience, tech-aware (heard of Google car; recognize ABS). Slower/more peer-dependent among older licensed drivers, those farther from work, and residents of high basic-employment-density areas.
• Home-location shifts (OP; McFadden’s R2≈0.24): Shift farther from center more likely among respondents with more children, higher education (Bachelor’s+), higher household-density neighborhoods, those who drive alone for work, and those farther from work. Shift closer more likely among tech-savvy (smartphone carriers; familiar with carsharing), urban residents, full-time working males, higher-income, and higher VMT respondents.

Overall: High-income, tech-savvy males in urban areas with more crash experience show higher interest and WTP for CAVs, less peer dependence, and stronger potential to use SAVs. Older licensed drivers are more cautious/less interested. Many prefer not to pay more than existing carsharing/ridesharing prices for SAVs.

The study’s findings directly address public acceptance and likely adoption pathways for CAV technologies by linking stated preferences to demographics, travel behavior, and built environments. Safety perceptions—captured via prior crash experiences and technology awareness—consistently increase WTP for automation and connectivity, as well as accelerate adoption timing, supporting the premise that perceived safety benefits are a core driver of adoption. Differential WTP between Level 3 and Level 4, and negative associations of heavy travel with Level 3 but positive with Level 4, indicate that productive in-vehicle time and full autonomy are especially valued by longer-distance travelers.

SAV adoption is sensitive to price and urban form: higher population density and urban residence correlate with greater intended use, aligning with expectations about parking convenience and car ownership patterns. Familiarity with existing carsharing reduces willingness to pay for higher-priced SAV services, suggesting cross-price sensitivity and the need for competitive SAV pricing and service levels. Peer effects are substantial (about half depend on friends’ adoption), but tech-savvy, higher-income males are less peer-dependent, pointing to likely early adopter segments. Residential shifts could bifurcate: some moving closer to benefit from urban SAV availability, others moving farther to capitalize on lower housing costs enabled by more tolerable commutes in AVs. These dynamics imply important implications for congestion, parking demand, pricing policies, and land-use planning.

The survey of 347 weighted Austin residents reveals strong interest in advanced vehicle technologies, especially full automation (Level 4), with mean WTP of $7,253 versus $3,300 for Level 3. Respondents expect fewer crashes as the leading benefit and fear equipment failure most. About 30% would adopt AVs immediately, while roughly half would wait for peer adoption signals. Most would not pay more for SAVs than current carsharing/ridesharing rates; nonetheless, urban, tech-aware individuals indicate higher SAV use, particularly at lower per-mile prices. Over three-quarters would add connectivity at under $100. Behavioral models show that higher-income, tech-savvy males in urban areas with more crash experience have higher interest and WTP and adopt more quickly, with less peer dependence, whereas older licensed drivers express less interest. Residential shifts may trend both inward (to exploit dense, low-cost SAV access) and outward (leveraging AV-enabled productive commute time), suggesting nuanced land-use impacts.

Future research should expand samples across regions and time to capture evolving awareness, probe driver versus passenger roles explicitly, test alternative pricing/incentive structures, and integrate revealed-preference data as technologies deploy. Enhanced models that jointly capture multiple choices (e.g., multimodal adoption, vehicle ownership changes, and residential mobility) will support more robust policy and infrastructure planning.

• Sampling and generalizability: Single-region (Austin) internet-based survey with initial demographic biases (partially corrected via weighting). Findings may not generalize to other geographies or time periods as technology awareness evolves.
• Sample size: N=347 limited; model selection allowed relatively high p-value threshold (0.32), potentially omitting weaker effects and reducing statistical power.
• Stated preference nature: Responses reflect intentions under hypothetical scenarios (prices, features), subject to hypothetical and social desirability biases.
• Measurement limits: Did not capture whether respondents usually travel as drivers or passengers; used proxies (driver’s license) which may not reflect actual role. Some home locations inferred from IP addresses when addresses were missing.
• Built-environment data: TAZ-level attributes may not fully capture micro-scale context; geocoding and spatial joins introduce potential error.
• Model scope: Univariate OP used for SAV pricing scenarios separately (rather than full trivariate), potentially limiting efficiency; cross-equation correlations only modeled for Level 3/Level 4 WTP.