Forecasting Americans’ long-term adoption of connected and autonomous vehicle technologies

The study addresses how and how quickly U.S. households may adopt connected and autonomous vehicle (CAV) technologies over 2015–2045. Motivated by uncertainty surrounding CAVs’ rollout and the need for policymakers, manufacturers, and investors to plan for future technology mixes, the authors propose a simulation-based approach that integrates both demand-side factors (e.g., willingness to pay, vehicle transaction decisions) and supply-side factors (e.g., technology prices, regulatory requirements). The introduction highlights anticipated benefits (safety, congestion reduction) and concerns (security, privacy, transition costs), and notes that forecasting CAV adoption requires moving beyond simple trend extrapolations to a framework that explicitly models household decisions and policy influences.

The literature on long-term CAV adoption forecasting is limited, with many projections from consulting firms and industry lacking methodological detail. Litman (2015) extrapolates from past automotive technologies, projecting substantial fleet and sales shares by 2050. Other cited works (e.g., IHS, Morgan Stanley, Lux Research, Fehr & Peers, ABI, BCG, Citi GPS, Rowe) offer diverse forecasts of fleet penetration, sales, and travel using AVs, often optimistic and based on expert judgment. Academic studies have mostly focused on public opinions, perceived benefits/concerns, and WTP (e.g., Schoettle and Sivak, Haboucha et al., Krueger et al.). The authors position their work as the first to forecast long-term national fleet evolution toward CAVs while jointly considering demand (WTP), supply (technology prices), and NHTSA regulations (ESC mandate, likely connectivity requirement). They note related simulation frameworks for alternative-fuel vehicles (Musti & Kockelman; Paul et al.) and own-fleet mix models (Garikapati et al.) as methodological precedents but not directly applicable to CAV adoption forecasting.

Data collection and processing: A U.S.-wide web survey (Qualtrics) was fielded in June 2015 using SSI’s panel. After time-based screening, eligibility checks, and sanity checks, 2167 responses remained (of 2868 completes). Person-level and household-level weights were constructed using 2013 ACS PUMS across 120 person categories (gender, age, education, Texas vs. non-Texas) and 130 household categories (household size, workers, vehicle ownership, Texas vs. non-Texas) to correct sampling biases. Respondent locations were geocoded (Google Maps API; QGIS) to merge census-tract built-environment attributes (e.g., poverty share, employment and population densities, distances to transit and downtown). The survey captured current household vehicle inventories, transaction history (past 10 years), intended near-term transaction decisions, WTP for Level 1/2 technologies, connectivity (DSRC), self-parking valet, and Level 3/4 automation, as well as opinions and travel patterns. Definitions and images for automation levels and technologies were provided; near-term technology prices and projected price declines were shown before WTP questions.

Modeling framework: An annual time-step simulation (2015–2045) forecasts household vehicle transaction decisions and technology adoption, updating the national fleet mix each year. The first stage uses a weighted multinomial logit (MNL) model (estimated in BIOGEME) for five alternatives: sell, buy, replace, add technology, and do nothing. Significant covariates include demographics, household vehicle holdings, built-environment attributes, and travel/work indicators (full specification in Table 4). Two auxiliary binary logit models govern: (1) buying two vehicles vs. one (Table 5), and (2) purchasing new vs. used (Table 6). Monte Carlo simulation applies these probabilities to each household annually.

Vehicle and technology decision rules: If sell is chosen, the oldest vehicle is disposed. If buy is chosen, households may purchase one or two vehicles, each new or used per the binary logit outcomes. Assumptions include: (a) Level 3/4 automation cannot be retrofitted to used vehicles; (b) retrofitting self-parking valet and Level 1/2 technologies to used vehicles costs four times the new-vehicle add-on cost; (c) connectivity is added if household WTP exceeds technology price; (d) for new vehicles, if WTP ≥ price, Level 4 is added; else Level 3; else self-parking valet and Level 1/2 technologies are considered subject to WTP and budgets; (e) for used vehicles, connectivity and Level 1/2/self-parking valet are considered based on WTP and technology prices. Replace = sell then buy. Add technology applies to existing vehicles without L3/L4, excluding already-present features. If a household without vehicles is simulated to sell/replace/add tech, it defaults to do nothing.

Scenarios and prices/WTP: Eight scenarios vary annual technology price reductions (5% or 10%), WTP growth (0%, 5%, 10%), and regulations (ESC mandated on all new vehicles from 2015; connectivity mandated from 2020 in regulated scenarios). For individuals reporting zero WTP, some scenarios assign non-zero WTP equal to the 10th percentile of non-zero WTP within a matching household cohort; WTP may then grow at 0%, 5%, or 10% annually. Technology prices for new-vehicle installation were compiled from OEM packages and expert opinions; example trajectories at 5% annual decline are provided (Table 8). Population-weighted adoption rates are extracted each year for all technologies. Certain state variables (respondent age, household vehicle count, vehicles sold, ownership indicators, age of oldest vehicle) are updated annually in the simulation.

Survey-based WTP and interest:

• Many respondents show low or zero WTP for advanced automation: 55.4% report $0 WTP for Level 3 and 58.7% for Level 4; 39.1% report $0 WTP for connectivity.
• Average WTP to add (population-weighted): connectivity $67 (average among payers $111); self-parking valet $436 (among payers $902); Level 3 $2438 (among payers $5470); Level 4 $5857 (among payers $14,196).
• For Level 1/2 features, interest and WTP vary: blind-spot monitoring has highest interest (50.7% very interested; 23.7% $0 WTP) with average WTP among payers $210; emergency automatic braking second-highest interest (45.8% very interested; 28.7% $0 WTP) with average WTP among payers $257.

Adoption projections across scenarios (selected highlights):

• Regulatory impact: With ESC (from 2015) and connectivity (from 2020) mandates, projected adoption reaches ≈98–100% by 2025 (ESC) and ≈98–100% by 2030 (connectivity) across regulated scenarios.
• Price decline effect: Faster 10%/yr price drops substantially raise long-term adoption versus 5%/yr given the same WTP growth and regulations. Example: Scenario 3 (5% price drop, regulated, no WTP growth) vs. Scenario 4 (10% price drop, same WTP): Level 4 in 2045 is 24.8% vs. 43.4%.
• WTP growth effect: With regulations and 10%/yr price drop, increasing WTP growth from 0% to 5% to 10% raises 2045 Level 4 shares from 43.4% (Scenario 4) to 70.7% (Scenario 6) to 87.2% (Scenario 8). Under 5% price declines, the corresponding 2045 Level 4 shares are 24.8% (Scenario 3), 43.2% (Scenario 5), and 59.7% (Scenario 7).
• Zero-WTP treatment matters for less advanced tech: In 2045, DSRC connectivity adoption is 59.5% in Scenario 1 (constant WTP with zeros retained, no regulation) vs. 83.5% in Scenario 2 (zeros replaced with low WTP, no regulation) under 10% annual price drops.
• Level 1 technologies can saturate by 2045 under favorable dynamics: With at least 10% annual WTP growth or at least 10% annual price declines (with regulations), most Level 1 technologies surpass 90% adoption by 2045.

Model-estimated behavioral correlates of transactions (from MNL and binary logits):

• Sell less likely for older, single individuals, those farther from downtown, in lower-income tracts; more likely for males and households with more vehicles.
• Replace more likely among bachelor’s holders, full-time workers, younger males who drive alone for work and have more vehicles/workers.
• Buy more likely for larger households (>3 members), those with older fleets and more workers, and in higher-poverty tracts; less likely for older respondents and households already owning more vehicles.
• Add technology more likely for bachelor’s holders, solo commuters, and residents of higher employment-density tracts; less likely for older individuals and those with older vehicles.

Overall: Without rising WTP, supportive policies, or rapid cost declines, the U.S. fleet remains heterogeneous through 2045; Level 4 shares range from 24.8% (pessimistic) to 87.2% (optimistic).

The framework integrates consumer demand (WTP, transaction choices) with supply-side dynamics (declining prices) and policy mandates to address the central question of how CAVs diffuse through the U.S. private light-duty fleet. Findings demonstrate that mandates can rapidly standardize foundational safety/connectivity features (ESC, DSRC), while advanced automation diffusion is highly sensitive to both price trajectories and evolving consumer WTP. Scenarios reveal that modest WTP growth or slower price declines greatly limit Level 4 penetration by 2045, whereas combined rapid price reduction and rising WTP can deliver near-dominance of Level 4. Results also highlight heterogeneity in household decision drivers and built-environment influences, informing targeted policy or incentive design. The study suggests that absent interventions (e.g., incentives, liability frameworks, insurance pricing, information campaigns) and continued cost reductions, a technologically mixed fleet will persist, tempering potential safety and congestion benefits attributed to widespread AV adoption.

The study develops and applies a simulation-based, behaviorally grounded fleet-evolution framework to forecast U.S. adoption of CAV technologies through 2045 under eight scenarios varying price declines, WTP growth, and regulations. Contributions include integrating stated WTP and transaction models with technology pricing and regulatory assumptions, and quantifying how each lever affects adoption trajectories. Key outcomes show near-universal ESC and connectivity adoption by 2025–2030 with mandates; substantial sensitivity of Level 4 penetration to cost and WTP dynamics (24.8%–87.2% by 2045 across scenarios); and broad potential for Level 1/2 saturation under favorable conditions. Future research directions include: endogenizing household demographic and built-environment evolution, modeling behaviorally driven temporal changes in WTP, incorporating shared autonomous vehicles and changing ownership models, refining technology cost trajectories and retrofit pathways, and assessing interactions with insurance, liability, and infrastructure investments.

• Demographic and built-environment evolution not explicitly modeled over time (beyond select annually updated variables); WTP growth is imposed exogenously at constant rates rather than derived from evolving household characteristics.
• Survey-based WTP may reflect early perceptions and hypothetical bias; more than half reported $0 WTP for advanced automation, which could change with experience or information (or adverse events).
• Technology price estimates for Level 1/2 are inferred from OEM packages and may not map cleanly to individual features; retrofit costs are assumed at 4× new-vehicle add-on costs; Level 3/4 retrofits are disallowed by assumption.
• Sample over/under-representation was corrected via weights, but residual biases may remain; some locations were inferred from IP addresses when street addresses were missing.
• The framework does not include shared autonomous vehicle (SAV) services or potential shifts in ownership and usage patterns that could alter transaction decisions and adoption.
• Regulatory scenarios are stylized (mandated ESC from 2015, connectivity from 2020); actual policy timing/scope may differ.
• Model fit statistics indicate reasonable but not perfect explanatory power; unobserved factors and market dynamics (e.g., supply constraints, insurance pricing) are not explicitly modeled.