Impact of automation on long haul trucking operator-hours in the United States

The study investigates how deployment of automated long haul trucking, particularly via a transfer-hub model where highway segments are automated and urban/suburban segments remain human-driven, will affect tractor-trailer operator-hours in the United States. Motivated by competing claims—ranging from large-scale job loss to suggestions that short-haul opportunities will offset losses—the authors use national shipment data to quantify potential impacts across realistic deployment scenarios that reflect technological constraints (e.g., weather) and geography. The work aims to clarify the scale and trajectory of labor impacts, assess feasibility through stakeholder interviews, and inform policy on workforce and deployment choices.

Prior work suggests substantial but bounded impacts on trucking jobs. Viscelli (2018) estimates ~300,000 long-haul jobs at risk under a transfer-hub model, concentrated in low-wage segments. Gittleman and Monaco (2020) bound potential losses at ~400,000, stressing that primarily highway driving tasks are automatable. Groshen et al. (2019) project elimination of 60–65% of heavy truck driving jobs under full automation. Waschik et al. (2021) simulate macroeconomic effects, finding employment declines of 20–25% in for-hire and 4–5% in private trucking, with broader economic gains potentially offsetting sectoral job losses. Broader automation literature highlights that tasks, not entire occupations, are typically automated (Arntz et al., 2016), and long-run employment can adjust (Autor, 2015; Bessen, 2019). However, perceptions among truckers are often negative (Dodel and Mesch, 2020; Orii et al., 2021). The authors identify gaps: few studies use shipment-level routing data or explicitly incorporate technology constraints (e.g., weather) and few include stakeholder interviews to assess task feasibility for highway automation.

The authors combine quantitative analysis of freight shipments with qualitative interviews. Data: They use the 2017 Commodity Flow Survey (CFS), focusing on for-hire and private truck shipments >150 miles (long haul), yielding ~1.5 million shipments with origin/destination, distance, weight, and financial quarter, and shipment weights for population scaling. Routing and segmenting: For each origin–destination pair, they estimate highway vs urban/suburban splits using Google Maps API via ggmap in R. They classify road segments by speed: ≥50 mph as highway; <50 mph as urban/suburban. Shipments are categorized as interstate or intrastate, and whether origin/destination Metropolitan Statistical Areas (MSAs) are specified. When MSAs are missing, they use rest-of-state centroids. For intrastate shipments lacking MSAs, they apply splits from the Freight Analysis Framework (FAF4, 2012). They compute a highway-to-total distance ratio r_i for each OD pair and apply it to shipment distances to get highway (D_s.Hi) and urban (D_s.Ui) legs. Operator-hours: Using average highway and urban speeds from Google Maps (or FAF where applicable) and HOS regulations (11 h driving/day; 10 h rest), they algorithmically compute operator-hours for highway (O_H) and urban (O_U) legs per shipment, accounting for multi-day trips and whether the second urban leg requires an additional rest period. They then scale by CFS shipment weights and shipment mass relative to a truckload (TL), which cancels in share calculations, to estimate total operator-hours and the shares at risk (highway) vs remaining (urban). Interviews: Purposeful and snowball sampling yielded 10 semi-structured interviews across stakeholders: 2 automated trucking startups, 5 truck drivers, 1 logistics senior manager, and 1 labor union representative. Interviews probed operational feasibility of transfer-hub deployment, task requirements, weather/route constraints, and worker preferences. The qualitative component serves as a feasibility check and to surface implications for labor transitions. Limitations of methods are detailed in the SI (e.g., approximations for origins/destinations, reliance on routing assumptions).

- Across cumulative deployment scenarios, operator-hours at risk vary widely: - Scenario 1: Restricting deployment to southern sun-belt states where testing is active yields ~10% of operator-hours at risk. - Scenario 2: Expanding to all states during favorable-weather quarters (Q2 & Q3, Apr–Sep) pushes impacts to over half of operator-hours. - Scenario 3: Applying to shipments >500 miles adds a further ~33% of operator-hours at risk beyond prior scenarios. - Scenario 4: Widespread deployment across the continental U.S. in all conditions places up to 94% of current long-haul operator-hours at risk. - Demand elasticity effects are unlikely to offset these losses: even a 50% increase in demand for trucking (overall elasticity ~5) would offset only ~5% of at-risk hours, reducing the 94% to ~89%. - Translating to jobs: Given differing baseline estimates of long-haul operators (300,000–550,000), the affected positions could range from roughly 30,000 to 500,000 depending on deployment scenario and baseline count. - Associated sectors: Highway automation could significantly reduce demand for traditional truck stop services (employing ~70,000 people), though transfer-hub “truck ports” may create new jobs (e.g., trailer switching, maintenance checks). Net effects and worker transition feasibility are uncertain. - Stakeholder interviews: - Drivers generally saw highway automation as technically feasible but flagged limits in inclement weather, poor lane markings, and GPS loss. - Employed drivers typically do not perform significant maintenance; ATs would need reliable distress signaling and roadside assistance models. - Drivers expressed reluctance to shift to short-haul work, citing lower pay structures (per-mile pay with more unpaid waiting) and geographical/lifestyle mismatches. - Partial automation features (e.g., lane assist, emergency braking) were viewed negatively by many drivers; firms are prioritizing full automation over partial systems.

The findings directly address the research question by quantifying how much long-haul trucking labor is exposed to automation under realistic, phased deployment constraints. They show that while near-term deployments limited to favorable geographies and seasons yield modest impacts, maturing capabilities and broader geographies could expose the vast majority of operator-hours. Economic feedbacks from lower costs and faster deliveries appear insufficient to create enough short-haul or additional trucking demand to offset highway-hour losses. Interviews corroborate technical feasibility on highways under conducive conditions but surface social and operational frictions: driver resistance to short-haul transitions, wage structure concerns, and geographic dislocation. These insights underscore that labor market and political ramifications will hinge on deployment choices (where/when/how) rather than technology alone, and that complementary policies will be needed to manage transitions and broader impacts (including environmental and modal-shift consequences).

The paper contributes a shipment- and route-based quantification of operator-hours at risk from automated long-haul trucking under a transfer-hub model, explicitly incorporating technological and operational constraints across deployment scenarios, and complements it with stakeholder insights on feasibility and workforce transitions. Results indicate potential exposure of up to 94% of operator-hours under widespread deployment, while near-term impacts may be modest if confined to southern states and favorable seasons. Increases in freight demand are unlikely to compensate for lost highway hours. Policy implications include anticipating wage pressures, designing workforce transition supports, considering requirements for industry reinvestment to mitigate employment and environmental impacts, and engaging drivers in participatory deployment planning. Future research should: refine scenario modeling with richer meteorological and regulatory constraints; quantify regional labor-market frictions and wage dynamics; assess environmental and modal-shift outcomes; and expand qualitative research on driver preferences, geographic relocation, and acceptance of partial versus full automation.

- Interviews: Small, non-representative sample focused on depth over breadth; limited segmentation by driver type and potential confounders; findings are suggestive rather than generalizable. - Routing and segmentation: Use of MSAs/rest-of-state centroids and Google Maps routes introduces approximations; classification by a 50 mph threshold may not capture all context-specific speed and roadway nuances; route selection can vary by time/day. - Operator-hour algorithm: Assumes standard HOS constraints and average segment speeds; does not capture stochastic delays (e.g., congestion, incidents) or all carrier-specific practices. - Job impacts: The study reports operator-hours at risk rather than precise job losses due to uncertainty in baseline counts (owner-operators, turnover) and dynamic labor market adjustments. - Demand response: Elasticity-based offsets are illustrative; comprehensive general equilibrium effects and supply-chain adaptations are beyond scope. - Technology scope: Assumes transfer-hub model and highway-only automation; results may differ with alternative operational designs or rapid advances in adverse-weather capability.