FORECASTING AMERICANS’ LONG-TERM ADOPTION OF CONNECTED AND AUTONOMOUS VEHICLE TECHNOLOGIES

The paper addresses the central question of how and how quickly Americans will adopt connected and autonomous vehicle (CAV) technologies over the long term. Recognizing substantial uncertainty around both demand-side factors (e.g., willingness to pay, risk perceptions) and supply-side factors (e.g., technology pricing), the study’s purpose is to produce behaviorally grounded forecasts of adoption for specific technologies (Level 1 and Level 2 driver-assistance features, connectivity, and Level 3 and Level 4 automation). The study highlights the importance of incorporating consumers’ WTP, vehicle transaction decisions, and prospective government regulations into forecasts. It proposes a simulation-based fleet evolution framework, calibrated to stated-preference survey data from 2,167 U.S. respondents, to project adoption trajectories from 2015 to 2045 under multiple scenarios of price declines, WTP growth, and regulation.

Prior work on CAV adoption includes expert opinions and extrapolations from historical technology diffusion. Litman (2015) projected substantial AV penetration by 2040 based on analogs like automatic transmission and hybrids, while private-sector reports (e.g., Lux Research, Boston Consulting Group, Citi GPS, IHS, Navigant, IDTechEx) offered varied sales and market-size forecasts for 2025–2045. Industry enthusiasts provided optimistic views of rapid adoption. However, earlier studies generally lacked explicit demand-side modeling of WTP, did not model household vehicle transaction behavior, and rarely anticipated adoption of specific Level 1/2 features or connectivity. Related vehicle market simulation work exists for alternative fuel vehicles in Austin and the U.S., but is not directly applicable to CAV technologies. This study fills gaps by jointly modeling demand (WTP), supply (prices), and regulation, and by forecasting adoption of specific technologies.

Survey and data: A U.S.-wide web survey (June 2015, via SSI panel on Qualtrics) captured 2,868 completes; after quality checks and eligibility screening, 2,167 respondents remained (1,364 from Texas). To correct sampling biases, person-level and household-level weights were constructed using 2013 ACS PUMS across multiple demographic and household categories. Respondent addresses were geocoded; built-environment variables were attached at the census-tract level. The survey elicited current vehicle holdings and use, vehicle transactions in the past decade, future buy/sell/replace intentions, WTP for numerous Level 1/2 technologies, WTP for connectivity, self-parking valet, and Level 3/4 automation, plus opinions and demographics.

Modeling framework: A simulation-based fleet evolution framework updates annually from 2015 to 2045. Core components:

• Vehicle transaction and technology adoption choice: A weighted multinomial logit (MNL) model (estimated in BIOGEME) predicts household annual choices among sell, buy, replace, add technology, or do nothing (base). Significant covariates include age, sex, household size and workers, number and age of vehicles, commute mode, education, retirement status, neighborhood employment density, poverty share, and distance to downtown.
• Conditional purchase decisions: For households choosing buy/replace, binary logits determine whether acquiring one or two vehicles and whether each is new or used.
• Technology assignment rules: For newly acquired vehicles, the model checks affordability (household WTP ≥ contemporaneous price) in a hierarchical order. For new vehicles, Level 4 is considered first; if not affordable, Level 3; otherwise self-parking valet and Level 1/2 features are considered. For used vehicles, Level 3/4 are not retrofittable; self-parking valet and Level 1/2 can be added, with retrofit costs assumed to be 4× new-vehicle add-on costs. Connectivity (DSRC-type) is added if WTP ≥ price. For households choosing add technology on existing vehicles, Level 3/4 are not added if already present; otherwise the same used-vehicle retrofit rules apply.
• Price dynamics and scenarios: Technology prices follow exogenous geometric declines of either 5% or 10% per year. Illustrative 5% path (2015→2045): connectivity $200→$42.9; Level 3 $15,000→$3,220; Level 4 $40,000→$8,586; Level 1/2 features similarly decline (e.g., ACC $400→$85.9). Eight scenarios combine WTP growth (0%, 5%, 10% annually), handling of zero WTP (original vs. replaced with cohort 10th-percentile positive WTP), price decline (5% or 10%), and regulations. Regulations reflect NHTSA mandates: ESC on new vehicles from 2015 and connectivity from 2020 (applied in Scenarios 3–8). Monte Carlo simulation applies model probabilities to each weighted household annually and aggregates population-weighted adoption rates of each technology.
• WTP handling: Scenario 1 holds original WTPs constant; Scenarios 2–8 assign non-zero WTP to original zero-WTP respondents (cohort 10th percentile) and then grow WTP at 0%, 5%, or 10% per year depending on scenario.

Time horizon and outputs: Annual shares of the U.S. privately held light-duty fleet equipped with each technology (ESC; nine Level 1/2 features; connectivity; self-parking valet; Level 3; Level 4) are produced for 2015–2045, with results summarized every five years.

• WTP levels: Over half of respondents reported zero WTP for advanced automation (Level 3/4). Among those with non-zero WTP, averages are approximately $110 for connectivity, $5,551 for Level 3, and $14,589 for Level 4. Self-parking valet averages about $902 (non-zero WTP). For Level 1/2 features, blind-spot monitoring and emergency automatic braking are the most appealing, while traffic sign recognition and left-turn assist are least appealing.
• Impact of regulations: With ESC (2015) and connectivity (2020) mandates, adoption approaches near-universal levels quickly: ~98% ESC by 2025 and ~98–100% connectivity by 2030 across regulated scenarios. Without regulations but with 10% annual price drops and no-zero WTP (Scenario 2), ESC and connectivity still reach 92.9% and 83.5% by 2045, respectively.
• Level 1/2 technologies: In scenarios with at least a 10% annual increase in WTP or at least a 10% annual price reduction (Scenarios 6–8), all Level 1 features exceed 90% adoption by 2045. Under more conservative assumptions (e.g., Scenario 3: 5% price drop, constant but non-zero WTP), adoption remains heterogeneous by 2045 (e.g., blind-spot monitoring ~53.5%, emergency automatic braking ~51.2%, traffic sign recognition ~38.1%).
• Advanced automation (L3/L4): Level 4 adoption is highly sensitive to WTP growth and price declines. By 2045: 24.8% L4 in Scenario 3 (5% price drop, regulated, no WTP growth); 43.4% in Scenario 4 (10% price drop, regulated, no WTP growth); 43.2% in Scenario 5 (5% price drop, 5% WTP rise, regulated); 70.7% in Scenario 6 (10% price drop, 5% WTP rise, regulated); and up to 87.2% in Scenario 8 (10% price drop, 10% WTP rise, regulated). As L4 becomes affordable, L3 adoption falls due to the hierarchical purchase rule (e.g., Scenario 8 shows very high L4 and lower L3 shares).
• Heterogeneity without demand growth/policy: With constant WTP and absent supportive policies and/or rapid price declines, the fleet’s technology mix remains far from homogeneous by 2045, particularly for advanced automation.

The findings demonstrate that long-term adoption of CAV technologies depends jointly on consumer WTP, technology cost trajectories, and regulatory actions. Regulations substantially accelerate diffusion of foundational safety/connectivity features (ESC and DSRC), compressing what would otherwise be multi-decade adoption gaps. For Level 1/2 driver-assistance technologies, either faster price declines or growth in WTP are sufficient to push adoption toward ubiquity by 2045. In contrast, advanced automation (Level 3/4) adoption remains highly sensitive to both affordability and acceptance: the model indicates that Level 4 quickly dominates once households can afford it, crowding out Level 3 due to the hierarchical selection mechanism. Consequently, under favorable scenarios (10% price declines and 10% WTP growth), the fleet can approach widespread Level 4 penetration by 2045; under less favorable conditions, adoption remains modest. These results address the research question by quantifying how different policy and market levers shape adoption trajectories and by highlighting that, without increases in WTP or strong cost reductions, advanced automation will not permeate the fleet rapidly.

This study contributes a behaviorally grounded, simulation-based framework for forecasting long-term adoption of specific CAV technologies, integrating survey-derived WTP, vehicle transaction decisions, technology pricing trajectories, and likely regulations. Calibrated on 2,167 U.S. respondents with demographic and built-environment controls, the model projects that regulations will rapidly universalize ESC and connectivity, that Level 1/2 features can achieve near-ubiquity under plausible cost/WTP trends, and that Level 4 automation could reach 25%–87% of the fleet by 2045 depending on demand and cost conditions. The work underscores how WTP growth and price declines are critical to advanced automation uptake and shows that, absent these forces and supportive policy, substantial heterogeneity will persist through 2045.

Future research directions include: integrating dynamic household evolution (demographics and location) to endogenize WTP changes; modeling learning, risk perceptions, and information diffusion; explicitly incorporating shared autonomous vehicles and their impacts on ownership and turnover; refining technology price forecasts and packaging effects; and exploring policy portfolios (e.g., incentives, mandates) and shock scenarios (e.g., cyber events) on adoption trajectories.

• Household demographics and built-environment characteristics are held largely fixed over time (beyond respondent age and vehicle ownership variables used in the model), so WTP is assumed to evolve at exogenous constant rates rather than from endogenous household changes.
• Technology price paths are scenario-based and uncertain, particularly for packaged Level 1/2 features and advanced automation; retrofitting cost assumptions (4× new-vehicle addition) are simplifying.
• Many respondents currently report zero WTP for advanced automation; real-world learning, exposure, and social influence could shift preferences rapidly in either direction (including setbacks from adverse events), which the model does not endogenize.
• Survey sampling biases are mitigated via weighting but may not be fully eliminated; stated WTP may differ from revealed behavior.
• The framework does not include shared autonomous vehicles or broader changes in ownership models, which could materially alter transaction dynamics and adoption rates.
• Policy uncertainty is simplified to ESC and connectivity mandates; other policies (incentives, insurance, liability) are not modeled explicitly.