Are we ready to embrace connected and self-driving vehicles? A case study of Texans

The study investigates how Texans perceive connected and autonomous vehicle (CAV) technologies, including their willingness to pay (WTP), perceived benefits and concerns, interest in shared autonomous vehicles (SAVs), adoption timing (including peer effects), potential home-location shifts once AVs/SAVs become common, and support for congestion-pricing policies. Motivated by the anticipated but uncertain adoption of CAVs and their potential impacts on travel demand, safety, and land use, the authors seek to quantify multivariate relationships between opinions/preferences and individual characteristics (demographics, built environment, travel patterns, safety attitudes, and crash histories). Understanding these relationships can inform more realistic forecasts of CAV adoption and guide policymakers on infrastructure, pricing, and education strategies.

A broad set of public-opinion surveys from academic, professional, and private sources shows cautious public attitudes toward AVs, with primary concerns around safety, affordability, information security, and liability. Summary studies (e.g., Accenture 2011; Cisco 2013; Ipsos MORI 2014; Schoettle and Sivak 2014, 2015; Kyriakidis et al. 2015) report mixed comfort levels and generally low WTP for full automation. More advanced modeling studies include Howard and Dai (2013), who found higher-income individuals more likely to use SAVs and retrofit AV tech; Bansal et al. (2016), who found equipment failure as a top concern and higher WTP among wealthier, tech-savvy males but lower WTP among older licensed drivers; Krueger et al. (2016), who used a random-parameters logit to show younger travelers and carsharing users prefer SAVs with ridesharing; and Haboucha et al. (2015), who used hybrid choice models to show differing SAV preferences across Israelis and Americans and the role of pricing and education. The present study extends prior work with a larger, Texas-wide weighted sample, additional covariates (e.g., crash and enforcement attitudes), and models covering WTP, SAV frequency at multiple price points, AV adoption timing, home-location shifts, and tolling-policy support.

Survey design and sample: A Texas-wide web-based survey (93 questions in seven sections) was fielded via Survey Sampling International’s panel using Qualtrics in June–July 2015. After quality controls and sanity checks, 1088 respondents remained from 1297 completes. Respondents were shown NHTSA automation level definitions and illustrative images. Sections covered opinions on AV/CV benefits and concerns; crash and enforcement history and attitudes; WTP and interest in Level 1/2 technologies, automation levels (2–4), and connectivity; adoption rates of carsharing, TNCs, and SAVs; home-location shift intentions once AVs/SAVs are common; opinions on congestion-pricing policies; travel patterns; built-environment context (via geocoding); and demographics. Sampling weights: Person- and household-level weights were computed from ACS 2013 PUMS for Texas to correct for over/under-representation across gender, age, education (person level: 60-category cross of 2×5×6), and household size/workers/vehicles (household level: 26 feasible categories). Geocoding and built environment: Home locations were geocoded (Google Maps API); IP location proxies were used if addresses were missing/incorrect. Respondents were mapped to census tracts to attach built-environment measures (e.g., population density, poverty share, employment density) and distances to downtown and nearest transit stop. Key response subsets: Connectivity WTP and feature-interest questions were asked of respondents with a vehicle or planning to buy within 5 years (N≈1063). WTP for automation levels was asked of those planning to buy a vehicle within 5 years (N≈755). Modeling: • Interval regression (IR) was used for WTP for connectivity and for Level 2/3/4 automation. Responses were interval/right-censored (e.g., categories like $1500–$2999, “$3000 or more”). IR treats interval bounds as known and assumes normal errors to estimate continuous latent WTP as a function of covariates. • Ordered probit (OP) models were estimated for: interest in adding connectivity (3-point scale), AV adoption timing (never; when ≥50% friends/relatives adopt; when ≥10% adopt; as soon as available), SAV adoption frequency at $1/$2/$3 per mile (five categories: never, <monthly, monthly, weekly, rely entirely), home-location shifts (closer, same, farther), and support for three congestion-pricing policies (5-point support scale). Model specification approach: From a broad set of predictors spanning person, household, location, travel, tech familiarity, and safety/enforcement attitudes, variables were iteratively retained if p<0.32 (|Z|>1.0) to allow for potentially meaningful effects; most final covariates had p<0.05. Goodness of fit reported via McFadden’s R²/adjusted R². Practical significance was assessed as standardized coefficients ≥0.2 (IR) or ≥40% relative change in choice probabilities (OP).

Descriptive statistics: • WTP averages (population-weighted): Level 2: $2910; Level 3: $4607; Level 4: $7589; connectivity: $127. Shares reporting less than $1500 WTP: L2 54.4%, L3 31.7%, L4 26.6%. For connectivity, 29.3% reported $0 WTP; only 39% expressed interest even if affordable. • Adoption timing of Level 4 AVs: 39% never; 32% when ≥50% friends/relatives adopt; 15% when ≥10% adopt; 14% as soon as available. • SAV adoption at $1/$2/$3 per mile (never use): 41.0% / 48.6% / 59.1%; rely entirely: 7.3% / 4.6% / 3.9%. • Home-location shifts if AVs/SAVs become common: 7.4% move closer; 81.5% stay; 11.1% move farther. • AV interests and concerns: Only 28.5% not interested in L4 AVs if affordable. Top in-vehicle activities anticipated while riding in AVs: talk to others (59.5%), look out the window (59.4%). Top concerns: affordability (64.5% very worried) and equipment failure (61.4% very worried). Expected benefits considered very significant: better fuel economy (53.9%), fewer crashes (53.1%). • Connectivity features: Highest interest in automatic crash notification (71.5%) and vehicle health reporting (68.5%); least interest in in-vehicle email (58.1% not interested) and built-in internet browsing (51.5% not interested). Policy support: strong support for adaptive signal timing (64.0%) and variable speed limits (62.2%); low support for real-time parking price adjustments (20.5%). Model-based insights (consistent/practical effects): • Older and more experienced licensed drivers exhibit lower WTP for all automation levels and for connectivity; they also indicate later AV adoption timing (more dependence on peers) and lower SAV use frequencies. • Higher-income and safety-cautious respondents (supporting automated enforcement and speed governors; with past fatal/serious crash experiences) report higher WTP for connectivity and automation, earlier adoption timing, and higher SAV usage across price scenarios. • Tech awareness: Having heard of Google’s self-driving car and/or CVs is positively associated with WTP for automation and greater SAV use; familiarity with carsharing relates to higher interest in connectivity and earlier AV adoption. • Travel and location: Greater annual VMT is associated with higher WTP (notably for L4) and more frequent SAV use, especially at lower prices. Living farther from downtown is associated with higher SAV use, while living farther from transit stops is associated with lower WTP for higher automation levels and lower SAV use. Higher neighborhood density is associated with lower WTP for L4 and later adoption. • Demographics: Caucasians (and licensed/experienced drivers) tend to have lower WTP for connectivity/L2 and use SAVs less frequently at all price points, everything else constant. • Home-location shifts: Those already owning a Level 2-equipped vehicle and those living farther from the city center are more inclined to move closer once AVs/SAVs are common; those farther from transit, with more social/recreational trips, or familiar with UberX are more likely to move farther. • Tolling policy support: Preferences vary modestly on average; some groups (e.g., higher VMT travelers) are more supportive of distributing revenues or time-varying tolls. Overall, 37.3% supported tolls if revenues reduce property taxes.

Findings demonstrate that current public perceptions in Texas feature significant heterogeneity tied to demographics, travel behavior, technology awareness, and safety attitudes. The consistent negative association of age and driving experience with WTP and adoption readiness emphasizes potential adoption lags among older cohorts. Positive associations between safety-minded attitudes and WTP/adoption indicate that framing CAVs around safety benefits may accelerate uptake. Tech familiarity and higher VMT both correlate with higher WTP and SAV interest, suggesting early adopters are likely to be tech-aware, higher-mileage travelers. SAV adoption is price-sensitive, with many respondents unwilling to use SAVs even at $1 per mile, but certain groups (e.g., safety-cautious, higher-worker households, those farther from downtown) show higher propensity to use SAVs. Home-location shifts appear limited overall (most plan to stay), but subgroups may gravitate toward centers (seeking higher service density) or suburbs (seeking space/cost savings while benefiting from AV conveniences). These insights inform demand forecasting for CAV technology, SAV fleet sizing, pricing, and infrastructure planning (e.g., transit access, charging infrastructure, and congestion-pricing strategies).

Using population-weighted survey data from 1088 Texans and econometric models (interval regression and ordered probit), the study quantifies how demographics, built environment, travel behavior, safety attitudes, and tech familiarity relate to WTP for connectivity and automation, SAV usage at various prices, AV adoption timing, home-location shifts, and support for tolling policies. Key contributions include: (1) comprehensive summary statistics on perceptions of CAVs/SAVs; (2) identification of practically significant predictors that can target outreach and policy; and (3) model specifications that enhance realism in forecasting long-term CAV adoption and SAV impacts. Average WTP is $2910 (L2), $4607 (L3), $7589 (L4), and $127 (connectivity), with affordability and equipment failure as top concerns. Older and more experienced drivers value new tech less, while higher-income and safety-cautious respondents value it more and adopt sooner. Most Texans anticipate no home-location change due to AVs/SAVs. Policymakers and planners can leverage these relationships to design education campaigns, plan infrastructure and pricing strategies, and anticipate spatial and demand changes as CAVs/SAVs emerge. The authors note that opinions are likely to evolve as technologies mature, underscoring the need for continued data collection and model refinement.

• Geographic scope and generalizability: The sample is Texas-only (though weighted to state demographics), so findings may not generalize to other regions or countries. • Early-stage perceptions: Public understanding of CAV technologies is nascent; opinions and WTP are likely to change as technologies mature and experience accumulates. • Cross-sectional stated-preference data: Results capture associations, not causal effects; stated WTP and intended behaviors may differ from revealed behaviors. • Variable coverage: Some relevant controls (e.g., household vehicle ownership) are not included in certain model specifications, potentially biasing specific estimates (e.g., SAV use). • Geocoding proxies: IP-based location proxies were used when addresses were missing/incorrect, introducing potential spatial measurement error. • Model specification and thresholds: OP/IR models assume normality and proportionality; iterative variable retention up to p<0.32 may include weaker predictors; goodness-of-fit values are modest. • Subsample sizes: Certain questions applied to subsets (e.g., automation WTP only for those planning to buy a vehicle; connectivity questions only for current/planned vehicle owners), reducing N for those models.