DRIVERLESS | Steve Viscelli
Acknowledgements
The author would like to thank the engineers, entrepreneurs, drivers, policymakers, scholars and others who
helped me understand truck automation and its potential impacts, including: John Alic, Michael Bartyzal, Leo
Bagley, Chris Benner, Eric Berdinis, Richard Bishop, Lee Branstetter, Steve Boyd, Doug Bloch, Je Buchanan,
Francoise Carre, Roger Cohen, Bill Driegert, Kyle Goodman, Erick Guerra, Beth Gutelius, Jessica Halpern-Finnerty,
Brian Hare, Colby Hastings, Steve Herzenberg, Je Hickman, Tom Kochan, Kasey Krape, Dan Leary, Sam Loesche,
Victoria Lee, Frank Levy, Karen Levy, Adam Seth Litwin, John Paul MacDue, Rahul Mangharam, Jonny Morris,
Cassandra Ogren, Yipeng Peng, Scott Perry, Mike Roeth, Lior Ron, Andrew Smith, Hays Witt, Dave Schaller,
Stefan Seltz-Axmacher, Josh Switkes, Chris Tilly, Raj Rajkuram, Terre Witherspoon, Alden Woodrow, Ed Wytkind,
and Lucie Zikova. I continue to learn from the tremendous group of scholars who study the industry and whose
work is an intellectual foundation for me, including David Bensman, Michael Belzer, Stephen Burks, and Kristen
Monaco. I received excellent feedback on my research during presentations at MIT’s Institute for Work and
Employment Research and a National Science Foundation Workshop hosted by the Virginia Tech Transportation
Institute. Many thanks to Je Barrera, Liam Kelly, Jenifer MacGillvary, Deborah Meacham, Jacqueline Sullivan,
and Penelope Whitney for support on communications and report production. Annette Bernhardt and Derecka
Mehrens provided support and guidance of all kinds throughout.
This research was commissioned by the UC Berkeley Center for Labor Research and Education and Working
Partnerships USA, and is part of a larger multi-industry project generously supported by the Ford Foundation,
the W.K. Kellogg Foundation, and the Open Society Foundations.
Cover photo: Elizabeth del Rocío Camacho
Illustrations: Je Barrera
About the Author
Steve Viscelli is a sociologist at the University of Pennsylvania. He is a Robert and Penny Fox Family Pavilion
Scholar, a Senior Fellow at the Kleinman Center for Energy Policy, and a lecturer in the Department of Sociology.
In 2016 he published The Big Rig: Trucking and the Decline of the American Dream (University of California
Press), about the work and fortunes of long-haul truck drivers.
Suggested Citation
Viscelli, Steve. Driverless? Autonomous Trucks and the Future of the American Trucker. Center for Labor Research
and Education, University of California, Berkeley, and Working Partnerships USA. September 2018.
http://laborcenter.berkeley.edu/driverless/.
The analyses, interpretations, conclusions, and views expressed in this report are those of the author and do not
necessarily represent the UC Berkeley Institute for Research on Labor and Employment, the UC Berkeley Center
for Labor Research and Education, the Regents of the University of California, Working Partnerships USA, or
collaborating organizations or funders.
DRIVERLESS | Steve Viscelli
Contents
Executive Summary ................................................................................................................................i
Glossary ....................................................................................................................................................ix
SECTION ONE: Introduction ...............................................................................................................1
The Uncertainty of When Self-Driving Trucks Will Arrive .............................................................2
Job Loss and Job Quality ...........................................................................................................................2
SECTION TWO: The Trucking Industry and the Development of Autonomous Trucks ..... 4
Why Focus on Trucking? .......................................................................................................................... 4
Truckers and the Trucking Industry .......................................................................................................5
What Is an Autonomous Truck and Who Is Making Them? ........................................................... 7
Why Autonomous Trucks May Come Before Other Self-Driving Vehicles.............................10
SECTION THREE: Scenarios for the Use of Autonomous Trucks .......................................... 13
Six Potential Adoption Scenarios .......................................................................................................18
Which Scenarios Are Most Likely and Desirable? ......................................................................... 29
How Soon Could Autonomous Trucks Be Used? ............................................................................30
SECTION FOUR: Estimating Job Losses and Likely Job Impacts .......................................... 32
How Many Jobs Are At Risk of Automation? ...................................................................................33
The Quality of At-Risk Jobs.....................................................................................................................37
The Quality of New Driving Jobs Created ........................................................................................43
SECTION FIVE: Policies for a 21
st
-Century Trucking Industry ............................................... 45
1. Develop an Industry-Wide Approach to Worker Advancement and Stability ................. 45
2. Ensure Strong Labor Standards and Worker Protections ........................................................47
3. Promote Innovation That Achieves Social, Economic, and Environmental Goals .........50
Endnotes ................................................................................................................................................ 53
Data Appendix .....................................................................................................................................57
DRIVERLESS | Steve Viscelli i
Will autonomous trucks mean the end of the road for truck drivers? The $740-billion-a-year
U.S. trucking industry is widely expected to be an early adopter of self-driving technology, with
numerous tech companies and major truck makers racing to build autonomous trucks. This trend
has led to dozens of reports and news articles suggesting that automation could eectively
eliminate the truck-driving profession.
By forecasting and assessing multiple scenarios for how self-driving trucks could actually be
adopted, this report projects that the real story will be more nuanced but no less concerning.
Autonomous trucks could replace as many as 294,000 long-distance drivers, including some of the
best jobs in the industry. Many other freight-moving jobs will be created in their place, perhaps
even more than will be lost, but these new jobs will be local driving and last-mile delivery jobs that—
absent proactive public policy—will likely be misclassied independent contractors and have lower
wages and poor working conditions.
Throughout this transformation, public policy will play a fundamental role in determining whether
we have a safe, ecient trucking sector with good jobs or whether automation will exacerbate the
problems that already pervade some segments of the industry. Trucking is an extremely competitive
sector in which workers often end up absorbing the costs of transitions and ineciencies. Strong
policy leadership is needed to ensure that the benets of innovation in the industry are shared
broadly between technology companies, trucking companies, drivers, and communities.
The ndings below are based on in-depth industry research and extensive interviews with the full
range of stakeholders: computer scientists and engineers, Silicon Valley tech companies, venture
capitalists, trucking manufacturers, trucking rms, truck drivers, labor advocates and unions,
academic experts, and others.
Executive Summary
294,000 or 2.1 million?
The need for scenario-forecasting analysis
Prior studies and news stories have suggested that nearly all of the roughly 2.1 million heavy-duty truck drivers in the United States could
lose their jobs to automation. However, that number includes many industry segments that are unlikely to be automated in the near future,
such as local pickup and delivery and carriers using specialized equipment. This report nds that the jobs most at risk of displacement are
long-distance driving jobs with few specialized tasks, representing about 294,000 drivers.
DRIVERLESS | Steve Viscelli ii
Executive Summary
1. Today, wages and working conditions in trucking vary
widely by industry segment
While truck driving is often portrayed as one of the few remaining middle-class jobs that doesn’t
require a college degree, Figure 1 shows that the quality of trucking jobs varies signicantly across
dierent segments of the industry, which can be split into long-distance and local driving.
Long-distance drivers move goods from factories to distribution centers or retail stores or between
distribution centers. Many are working at “for hire” trucking rms, and an important distinction here
is whether they are driving a full truckload for a single customer or if their load is a combination of
freight from dierent customers (known as “less-than-truckload”).
Drivers for less-than-truckload rms and parcel companies such as UPS typically have higher wages,
better benets, and stable careers (unionization rates are high). By contrast, full truckload companies
tend to pay lower wages, churn through workers new to the industry, and often misclassify their
workers as independent contractors (unionization rates are low). Unfortunately, these practices set
the competitive standard in key parts of the industry.
Local driving jobs, particularly those driving light-duty trucks, pay signicantly less than
long-distance jobs. The large majority are local delivery drivers who perform a wide range of
assignments, delivering anything from express packages to owers. They take home salaries that can
be half of what long-distance drivers make. The other major category of local driving jobs are at the
ports, where drivers work long hours for low wages. When port drivers are contractors rather than
employees, they can work the equivalent of two full-time jobs and earn less than minimum wage.
FIGURE 1: Current conguration of truck-driving jobs
PARCEL CO
PARCEL CO
Combo Freight
PARCEL CO
Port driver
Average earnings: $29,000 – $35,000
Full truckload driver
Average earnings: $47,000 – $54,000
Parcel driver
Average earnings: $60,000
Delivery driver
Average earnings: $36,000
Less-than-truckload driver
Average earnings: $69,000
DRIVERLESS | Steve Viscelli iii
Executive Summary
2. Without policy intervention, automation will likely
eliminate high- and mid-wage trucking jobs, while
creating low-quality driving jobs
Based on an analysis of a range of potential scenarios for the adoption of self-driving technology
(see Potential Adoption Scenarios, page iv), here are the four ways that automation is most likely to
change trucking:
Autonomous trucks are best suited to long-distance highway driving, while humans
will still be needed to navigate local streets and handle non-driving tasks.
Many industry experts and developers expect that self-driving trucks will soon be able to drive
autonomously on the highway, but that it will take far longer (perhaps several decades) before
driverless trucks will be able to routinely navigate local streets packed with cars, pedestrians, cyclists,
road work, and other unexpected challenges. Humans will also be needed to handle the many
non-driving tasks—coupling tractors and trailers, fueling, inspections, paperwork, communicating
with customers, loading and unloading, etc.—that drivers currently perform.
Therefore, the most likely scenario for widespread adoption involves local human drivers bringing
trailers from factories or warehouses to “autonomous truck ports” (ATPs) located on the outskirts
of cities next to major interstate exits. Here, they will swap the trailers over to autonomous tractors
for long stretches of highway driving. At the other end, the process will happen in reverse: a human
driver will pick up the trailer at an ATP and take it to the nal destination (see Figure 2).
FIGURE 2: Most likely automation scenario, absent policy intervention
Autonomous
Truck Port
Autonomous
Truck Port
Local drivers
Low pay means old, polluting
trucks & inefficient operations
Delivery drivers
Low wages, likely to be misclassified
as independent contractors
PARCEL CO
PARCEL CO
eStore
eStore
Autonomous
tractor
DRIVERLESS | Steve Viscelli iv
Executive Summary
Automation could replace most non-specialized long-distance drivers—about
83,000 of the best trucking jobs and 211,000 jobs with moderate wages but
high turnover rates and poor working conditions.
As shown in Table 1 (page v), the most likely automation scenario evaluated in this report could
result in the loss of an estimated 294,000 trucking jobs. Specically, self-driving trucks will be best
suited for use in industry segments with long stretches of highway driving, minimal need for drivers
to perform other tasks, and large rms with the capital to buy (and expertise to integrate) new
technologies.
Two parts of the long-distance industry best t this bill:
This study is based on an analysis of six potential scenarios for how self-driving technology could be used in the
trucking industry. The scenarios are the result of interviews with engineers, developers, trucking rms, and drivers,
along with reviews of industry trade literature.
Human–human platooning: A series of human-driven trucks would be electronically linked, with the lead truck controlling
speed and braking in the following truck(s). This approach would let the trucks travel much closer together on the highway, improving
aerodynamics and fuel efciency. Each truck would still have a human driver to maintain the lane and navigate local streets.
Human–drone platooning: Similar to the human–human platoon, except that a single human driver would lead a platoon of
autonomous drone trucks on the highway. The human driver would be available to operate the lead truck, manage unexpected situations,
or make repairs and ensure safety if a truck broke down mid-route. As in the exit-to-exit scenario below, local drivers would bring loads
to an autonomous truck port (ATP) near the highway, where they would swap trailers with the drone trucks for the highway platoon.
Highway automation + drone operation: Human operators would remotely control trucks on local streets and in
complicated situations, and then trucks would drive autonomously on the highway. This approach would rely on highly trained dock staff
to handle tasks currently performed by drivers, such as inspection and coupling.
Autopilot: Similar to autopilot in airplanes, a human would handle loading and local driving, then sleep in the back of the truck while
the computer drove on the highway.
Highway exit-to-exit automation: Human drivers would take care of non-driving tasks and navigate complicated local streets,
then swap trailers with self-driving trucks at an ATP next to the highway. The autonomous truck would handle the long-distance freeway
driving, then hand off the load at an ATP near the destination.
Facility-to-facility automation: In situations where warehouses and shipping facilities are located near major interstates,
autonomous trucks may be able to handle industrial roads (where there are few pedestrians and complex intersections) and drive directly
from origin to destination.
Absent signicant changes in the policy or economic context, this report concludes that highway exit-to-exit
automation is the most likely scenario to be widely adopted in the future. However, human-led platoons represent a
model that has fewer technological challenges, a strong economic case, and better jobs for long-distance drivers.
Potential Adoption Scenarios
•
•
•
•
•
•
DRIVERLESS | Steve Viscelli v
Executive Summary
Truckload
Truckload drivers typically work for large trucking companies, hauling full trailers over long distances
directly from one customer location to another. These drivers rarely perform work such as loading
and unloading or caring for special kinds of freight. These characteristics make their jobs more
likely to be automated. An estimated 211,000 long-distance jobs in this segment are at risk of
displacement from autonomous trucks. As described above, working conditions in this segment
are arduous, and turnover is high. Wages are lower than in the unionized segment of trucking and
private, in-house eets, but higher than local delivery driving, the lowest-wage segment of the
industry.
Less-than-truckload and parcel
In parcel and less-than-truckload operations, shipments from dierent customers are combined
together at trucking company terminals, driven to another facility near the destination, and then
TABLE 1: Truck driving jobs and potential impact of autonomous trucks
Key segments
of the trucking
industry
Average annual
wage
Number of
drivers
Turnover
Independent
contractors
Unionization
rates
Potential impact
of autonomous
trucks
LONG DISTANCE DRIVING
Full truckload
$46,641—
$53,690
211,000 High Common Low
Signicant job
loss
Less-than-truckload $69,208 51,000 Low Uncommon High
Signicant job
loss
Parcel $59,660 32,000 Low Uncommon High
Signicant job
loss
LOCAL DRIVING
Ports
$28,783
(contractors)
$35,000
(employees)
75,000 Low Predominant Low Uncertain
Pickup and delivery $35,610 877,670
Varies
Mixed,
potential to
shift towards
contractors
Varies
Strong job
growth
POTENTIAL NEW SEGMENT
(
PROJECTED
)
Autonomous truck
ports
? 100,000+ ? ? ?
Strong job
growth
Notes: See Section 4 for sources on wages and employment.
DRIVERLESS | Steve Viscelli vi
Executive Summary
sent out for delivery. The long-distance drivers who haul these combined shipments on the highway
rarely do much more than driving, which makes their jobs also vulnerable to automation. Up to
51,000 less-than-truckload drivers are at risk of displacement by autonomous trucks, plus another
32,000 parcel drivers. These are some of the best jobs in the industry, and drivers earn some of the
highest incomes in trucking, in part because of high unionization rates. Because these drivers are
able to make a career out of trucking, they tend to be older than the average driver and much older
than the average U.S. worker.
Over the next several decades, e-commerce growth and lower freight costs could
create many new driving jobs, perhaps more than will be lost to automation.
Without policy intervention, however, these new jobs will likely have low wages
and poor working conditions.
The combination of automation decreasing the cost of moving freight by truck and consumers
ordering more goods online and expecting rapid delivery will likely increase the need for local
drivers to:
• Move loads to and from autonomous truck ports;
• Shuttle goods from large centralized warehouses outside cities to smaller local depots—
the approach being adopted by rms such as Amazon to enable rapid last-mile delivery;
• Deliver packages and other goods to customers’ doors.
However, without proactive public policy, these new driving jobs are likely to be far worse than
the jobs that are lost. Drivers bringing loads to ATPs are likely to face conditions similar to those
currently experienced by port drivers, such as low pay, long periods of unpaid waiting, and
independent contractor misclassication. The port driving sector is rife with stories of drivers putting
in 16-hour days but losing money after paying o truck loans, company charges, and other fees.
And if local drivers can only aord old and inecient trucks, more communities are likely to suer
from the high pollution and asthma rates common in neighborhoods near ports.
Delivery drivers, meanwhile, typically take home less than half the pay of better-paid long-distance
drivers. Retailers seem increasingly likely to subcontract to small rms with low pay or to adopt
the Amazon Flex model of treating delivery drivers as independent contractors who do not receive
benets, must use their own vehicles, and lack the right to organize for higher wages and better
working conditions.
Splitting trucking into local human driving and autonomous highway driving
is likely to foster the “digitization” of freight matching, with the potential for
intense downward pressure on driver earnings.
Currently, long-distance trucking rms rely on complex systems to match drivers with a series
of loads, seeking to minimize miles driven without freight, while complying with limits on how
DRIVERLESS | Steve Viscelli vii
Executive Summary
long drivers can be behind the wheel. Splitting trips between autonomous trucks that can almost
constantly be on the highway and local human drivers who go home each night vastly simplies
this load-matching problem. This approach is likely to lead to the “digitization” of freight, with
app-based marketplaces where local drivers can select from available loads.
Digitization could signicantly reduce the number of miles driven without freight, saving
the trucking industry billions each year. However, the destructive competition of a digitized
load-matching system could put intense downward pressure on local drivers’ earnings. To a
signicant degree, the impact of this approach on drivers will depend on public policy and
job-quality standards.
3. Proactive industry and public policy action will be
needed if automation is to deliver broad economic,
environmental, and social benets
The way we move goods is going to change dramatically in the coming decades, but how new
technologies make their way onto our roads—who benets, who may be left behind, the impact on
our environment—will be shaped by the response of governments, businesses, and workers across
the industry. Eective public policy can ensure that trucking evolves into a productive, high-road
industry. Policymakers, collaborating with workers and industry leaders, have an opportunity to
tackle some of our biggest challenges: creating good, family-supporting jobs, improving road safety,
and reducing trac congestion and carbon emissions. The following three main pillars should drive
that collaboration.
Develop an industry-wide approach to worker advancement and stability
Policymakers should create a Trucking Innovation and Jobs Council, bringing together diverse
stakeholders across the sector—workers, employers, technologists, and policymakers—to support
a 21
st
-century trucking workforce. The Council would develop and implement an action plan for
how industry stakeholders would fund, design, and carry out policies and programs to accomplish
two goals: (1) the development of good career pathways and training/job-matching programs for
incumbent, dislocated, and future workers; and (2) the creation of safety-net programs to support
transitions within and out of the industry, including work-sharing initiatives, supplemental and
exible unemployment insurance, and retirement packages.
Ensure strong labor standards and worker protections
Policymakers should establish a framework of strong labor standards that can shape the impact
of autonomous trucks, ensuring high-quality trucking jobs now and into the future. Specic
policies include addressing independent contractor misclassication and wage theft; expanding
early warning systems in the case of layos; and exploring new ways to establish good jobs in the
industry and strengthen workers’ right to organize. Some of these policies have long been needed;
DRIVERLESS | Steve Viscelli viii
Executive Summary
the goal is to enact them now so that low-wage business models do not become the norm in the
industry’s growth segments.
Promote innovation that achieves social, economic, and environmental goals
In order to ensure the best social, economic, and environmental outcomes for drivers, local
communities, and our transportation infrastructure, policymakers need to play an active role in
regulating the industry and the development of new technology. Examples of specic policies include
engaging stakeholders to develop a shared innovation agenda and leveraging public research
funding to implement it; allowing state and local governments to experiment with new policy
responses; and ensuring that public dollars and policies do not subsidize the displacement of workers.
* * *
What might an alternative, shared innovation agenda look like for the adoption of autonomous
trucks? This report identies an adoption scenario with good outcomes for workers, job quality,
and public health and safety: human-led platooning, coupled with clean and electric trucks. Figure
3 illustrates this scenario, where drivers lead platoons of autonomous trucks on highways and
have the experience and knowledge to deal with equipment problems, poor weather, and rapidly
changing road conditions like accidents, construction, trac, and erratic drivers. This model would
yield many of the best environmental benets of automation through increased fuel economy and
the use of clean trucks for the growing segment of local driving. The policy menu outlined above
would also raise labor standards and help train and support workers through the transition. The
result would be a robust, sustainable 21
st
-century trucking industry that broadly shares the benets
of innovation among technology companies, trucking companies, drivers, and communities.
FIGURE 3: Alternative automation scenario, with policy intervention
PARCEL CO
PARCEL CO
Autonomous
Truck Port
eStore
Autonomous
Truck Port
eStore
Delivery drivers
Employees with good wages,
rights and benefits
Local drivers
Higher wages & driving
clean electric trucks
Drone platoon pilot
High-skill, high-wage jobs
DRIVERLESS | Steve Viscelli ix
The meaning and usage of many common terms vary signicantly across the industry. The
denitions given here are intended only to help the reader understand how I will use these terms
in this report, which may dier from specic legal or regulatory denitions and/or informal usage
within particular rms or industry segments.
Backhaul – A load originating where another load is destined that helps a trucker get back to
where he/she wants to be. Typically not as protable as the primary load, backhaul is often used to
describe an undesirable load a trucker hauls to cover costs over distances they would need to travel
regardless of whether they have a load (e.g., on their way home).
Brokerage – A business that arranges freight transportation by motor carriers but does not
transport freight itself or take legal possession of freight.
Class 8 Trucks – Trucks with a gross vehicle weight of more than 33,000 pounds. This would
include heavy tractor-trailers, which typically have a gross vehicle weight up to 80,000 pounds.
Container – A shipping container. A heavy-duty steel container in 20’ or 40’ lengths that can
be loaded onto ships, rail cars, or onto a chassis for hauling by tractor. The predominant method for
transporting imports and exports.
Contractor – A driver working as an independent contractor who is responsible for a large
portion of the xed and operating expenses of their tractor and who works under contract for
a motor carrier. Contractors may own trailers but typically do not. Unlike independent owner-
operators, contractors operate under the authority of a motor carrier which typically nds and prices
all of the loads hauled by the contractor. Often contractors are misclassied employees.
Deadhead – Traveling without freight.
Dedicated – Freight service organized to serve the regular shipping needs of a particular—
usually high-volume—customer. Dedicated service can entail meeting special requirements of
shippers and almost always involves signicant numbers of loads moving between particular origins
and destinations. Dedicated service is typically a long-term (multi-year) relationship, and within their
own eet, motor carriers often dierentiate drivers assigned to service a dedicated account.
Glossary
DRIVERLESS | Steve Viscelli x
Glossary
Dray or Drayage – Transportation of freight over short distances. Also known as cartage.
Often refers to container movements from ship and rail yards.
Dry Van – A standard non-refrigerated “box” trailer. Typically 53’ long, but also in other lengths,
such 48’. The most common trailer in the industry used to carry a majority of freight. Freight in a dry
van is usually on pallets or in boxes.
Dry or Dry Van Freight – Anything that can be hauled in a dry van but is often hauled
in refrigerated trailers.
For-Hire Motor Carrier (For-Hire Carrier) – An individual or rm with an
FMCSA operating authority to oer freight transportation services to the public for a fee.
Hours of Service (HOS) – The federally mandated rules set by the Federal Motor
Carrier Safety Administration (FMCSA) that regulate, among other things, how many hours drivers
may drive and work over certain periods of time.
Independent Owner-Operator – The owner of a for-hire motor carrier who also
works driving equipment they control. Independent owner-operators are responsible for all of
the xed and variable expenses of their operation and operate under their own legal authority to
provide freight services to customers (which could include shippers, freight brokers, or other motor
carriers).
Intermodal – Transportation of freight in which containers or trailers are transferred between
dierent types of vehicles without unloading the freight. Typically used to refer to a combination of
ship or rail and truck movement using containers.
Less-than-Truckload (LTL) – Freight service moving shipments generally less than
10,000 pounds. These services often consolidate multiple shipments into a single truckload-size
shipment for long-distance transport and then break consolidated shipments down again for nal
delivery. Consolidating and breaking down of LTL shipments often happens at motor carrier-
controlled terminals. Such operations typically use dierent drivers for local pickup and delivery and
long-distance transport (known as linehaul).
Less-than-Truckload Carrier (LTL Carrier) – A for-hire motor carrier
providing LTL service.
Linehaul – Transportation between facilities owned by the same rm, most commonly freight
terminals with an LTL, mail, or parcel operation. Used in some segments to refer to truckload service
on regular or dedicated accounts or, simply, over long distances.
Local – Freight services less than 150 miles from origin to destination.
DRIVERLESS | Steve Viscelli xi
Glossary
Motor Carrier – Generally refers to a commercial vehicle transporting freight or passengers.
For the purposes of this report, I use the term in the common usage meaning a motor carrier with
an operating authority or motor carrier (MC) number issued by the FMCSA.
Parcel Service – Freight services that move packages or individual shipments of freight
weighing roughly 150 pounds or less (e.g., UPS or FedEx).
Private Carrier – A trucking eet that hauls goods that it produces or sells. A private carrier
provides in-house services and does not require an operating authority.
Over-the-Road (OTR) or Long-Haul – Service that transports freight more than
150 miles from origin to destination.
Refrigerated (also Reefer or Temperature-Controlled) – Used to
refer to freight that must be transported at a particular temperature. It can also refer to van trailers
used to haul that freight or rms that haul it (called refrigerated carriers). Refrigerated vans (also
called reefers) are often used to carry dry freight.
Regional – Can refer to carriers or transportation services that are longer than local (more
than 150 miles) but concentrated such that drivers and equipment do not regularly move outside of
particular regions, e.g., Northeast, Southeast, Upper Mid-west, etc.
Segment (or Industry Segment) – A portion of the trucking industry
distinguished by freight or service type. There are numerous recognized segments based on whether
carriers are private or for-hire, size of shipments, distance goods are moved, the type of trailer
required, etc. The most common segment distinctions would include, among others: private/for-hire,
truckload/less-than-truckload, over-the-road/local. Within the OTR, for-hire truckload segment
are segments dened by the type of trailer used to haul freight (e.g., dry van, refrigerated, atbed,
tanker, etc.). Segments sometimes have relatively distinct business models for rms and dierent
labor market and operational characteristics relative to drivers.
Sleeper Cab – A tractor with sleeping accommodations for a driver. Typically used initially
in long-haul operation but often ends up as cheap, used equipment in inecient local operations,
such as port hauling.
Tractor – The power unit of tractor-trailer.
Truckload (TL) – For-hire freight service that moves shipments larger than 10,000 pounds,
generally large enough to ll a truck to capacity, based either on legal allowable weight or trailer
volume. Truckload freight generally moves point-to-point from shipper to consignee (receiver),
without passing through a motor carrier facility.
Truckload Carrier (TL Carrier) – A for-hire motor carrier providing truckload service.
DRIVERLESS | Steve Viscelli 1
On October 25th, 2016, American news media heralded the rst delivery made by an autonomous
truck. That truck—developed by the startup Otto, which became part of Uber’s Advanced
Technologies Group before it was shuttered this summer—hauled a trailer full of Budweiser along
I-25 in Colorado from Fort Collins to Colorado Springs. An 80,000-pound truck drove 120 miles
without a driver behind the wheel. It certainly captured the imagination.
Since that demonstration, trucking has become the odds-on favorite as the rst signicant market
for self-driving technology. In the span of less than a week in March 2018:
• Uber announced a pilot program in which its autonomous trucks would move actual
freight on a regular basis;
• Embark (already moving freight in partnership with Ryder) announced that one of its
autonomous trucks had driven from California to Florida;
• Starsky Robotics claimed the rst autonomous trip in the United States on public roads
without a driver in the vehicle at all;
• Waymo announced that its autonomous trucks would begin hauling cargo into one of its
Atlanta facilities.
Autonomous trucks, with their promise of increased safety, improved productivity, and lower cost,
seem just around the corner, leaving one inevitable question: what about the truck drivers who lose
their jobs?
Introduction
SECTION ONE:
Popular press and industry
experts have published
numerous articles and reports
suggesting a “jobs catastrophe”
resulting from autonomous
trucks.
DRIVERLESS | Steve Viscelli 2
SECTION ONE: Introduction
Over the past two years, dozens of news articles have forecast “the end of the trucker.” Reports from
think tanks and consulting groups have suggested that autonomous trucks and other self-driving
vehicles could result in several million lost jobs within a few decades.
1
Prominent commentators
have suggested the imminent need to retrain large swaths of the workforce for new jobs and to
consider minimum basic incomes guaranteed by government as unemployed truckers join the tens
of millions of other workers sure to be displaced by automation in the coming decades.
2
What will the consequences of autonomous trucks actually be? And how should these consequences
be faced? This report argues that recent debate on both these questions has missed the mark—
badly.
The Uncertainty of When Self-Driving Trucks Will
Arrive
Over the past year, I’ve discussed self-driving trucks with scientists and engineers across the United
States. Through these conversation and other research, I’ve learned that self-driving technology has
made rapid improvements, but the technology still has a long way to go. A number of signicant
hurdles need to be overcome before autonomous trucks become commonplace on our highways.
Predictions of when this technology will be ready for widespread adoption are “guesstimates” at
best.
We can all imagine a future in which robots not only haul beer on the highways, but also deliver it
right to our lawn chairs, yet that future is likely many decades away. In fact, it is unlikely we’ll see any
signicant labor impacts from autonomous trucks in the next decade. If the technology continues to
develop rapidly, however, we could see impacts that spread at a fair rate across the industry.
This report looks at what might happen in the next 25 years, a timeframe that allows policy to be
researched, debated, and instituted for impacts that will aect existing workers and populations.
Eective policy for trucking—with its complexity, importance for the economy overall, and generally
polarized policy positions—will take a long time to develop, so the time to start is now.
Job Loss and Job Quality
This report estimates that the adoption of autonomous truck technology threatens around 294,000
jobs, and a signicant number of the jobs at risk are among the best trucking jobs. Though many of
the other jobs at risk are not good-quality, stable work from which truckers can make a career, they
still pay better wages than most comparable jobs. The loss of both the really good trucking jobs and
the better-than-the-alternative trucking jobs will hurt workers.
Rather than the “suddenly jobless” scenario that has been suggested, however, autonomous
trucks, e-commerce, and economic growth are together poised to create many new trucking jobs.
Twenty-ve years from now there will likely be many more jobs moving goods than there are today.
DRIVERLESS | Steve Viscelli 3
SECTION ONE: Introduction
The question is whether or not most of those jobs are going to be good jobs, with healthy working
conditions and living wages. Historically, trucking has provided jobs with middle-class wages and
security for workers. The quality of future freight-moving jobs, like that of most jobs in the United
States, will depend on whether we have policies that protect workers and ensure the benets of
economic growth are equitably shared.
If current conditions seen in much of the trucking industry prevail—and it’s likely they will, if policy
doesn’t change—the jobs created by autonomous trucks will pay far less than the jobs we might
lose. The risk of autonomous trucks is not that there won’t be enough jobs for American truckers, it’s
that there won’t be enough good jobs.
Here’s the good news: the technology of autonomous trucks itself won’t determine whether the
trucking jobs of the future are good or bad—policy will. Trucking has long had the potential for
cut-throat and destructive competition and the negative repercussions on workers that follow.
Workers need a policy that ensures they are protected from abusive practices and receive an honest
return for their hard work. Because truckers work on our public highways and streets, we all benet
when trucking jobs are good and truckers are experienced, safe, and ecient. In fact, policies
protecting workers from the costs of automation will also promote a clean, safe, and ecient
trucking industry.
We need to reorient the debate around autonomous trucks to recognize these facts. Rather than
trying to predict the specic damage autonomous trucks might do to workers, we must gure
out how this technology can help us achieve a safer, more ecient industry, with better pay and
improved working conditions for truck drivers. Bringing about that reality demands a thoughtful,
long-term approach to the policies that will shape the use of autonomous trucks. This report argues
that such an eort is worthwhile and suggests where we might begin.
Section 2 provides some basic background on the trucking industry and the development of
autonomous trucks. Section 3 lays out a series of scenarios for how dierent kinds of autonomous
trucks might actually be used, based on published reports and extensive conversations with
technology developers. Section 4 suggests which trucking jobs might be aected by those scenarios
and what the labor market impacts might look like. Section 5 concludes the report by suggesting a
set of policy steps to help ensure that tomorrow’s trucking jobs are good jobs.
DRIVERLESS | Steve Viscelli 4
Before the scenarios and job-impact analysis will make sense, we need to cover some background
about trucking and self-driving trucks. This information will help us understand the important
challenges the industry faces and the potential impacts of autonomous trucks.
Why Focus on Trucking?
The Department of Commerce reported that in 2015, there were 15.5 million workers in
driving-related occupations that could be aected by autonomous vehicles or approximately one
in nine workers in the United States. Workers who drive on the job are employed in a wide range
of occupations, from truck and taxi drivers to electricians, reghters, and home care aides. While
self-driving technology might change many jobs, it won’t eliminate most of them because driving
is such a small part of what those jobs entail. In 2015, about 3.8 million workers (2.8 percent of the
workforce) were engaged in jobs identied as primarily motor vehicle operators, including truck, bus,
and taxi drivers.
3
While these workers almost always do more than just drive, much of their time at
work is spent driving so their jobs may be at higher risk from automation.
This report focuses on the largest single job category of motor vehicle operators, and arguably, the
jobs most immediately facing automation in the next several decades: heavy and tractor-trailer
truck drivers. There were about 1.87 million such drivers in the United States in 2016, according to
the Bureau of Labor Statistics (BLS), comprising a little less than half of all jobs that were considered
primarily driving. In addition, a number of self-employed truck drivers do not gure in the BLS data.
All told, there are probably a slightly more that 2 million heavy and tractor-trailer truck drivers. As
Section 4 below suggests, about 300,000 of these jobs are at risk of automation not only because
they involve lots of driving, but specically, lots of uninterrupted highway driving.
SECTION TWO:
The Trucking Industry and the
Development of Autonomous
Trucks
DRIVERLESS | Steve Viscelli 5
SECTION TWO: The Trucking Industry and the Development of Autonomous Trucks
Truckers and the Trucking Industry
Unlike anything since the CB radio craze and Smokey and the Bandit, autonomous trucks have drawn
attention to the lives and fortunes of America’s truckers. Hauling nearly 71 percent of the nation’s
freight, U.S. trucking is a $740-billion-per-year industry.
4
Virtually all the physical goods we consume
move by truck at some point, sometimes several times. As a result of changes in all the stu we
buy, where and how it’s made, and how we buy it, the trucking industry varies considerably in terms
of how it moves freight and what that means for drivers. In order to understand how autonomous
trucks will aect these workers, we need to look more closely at two important factors previously
neglected in the analyses of autonomous trucks’ labor impacts:
1. What the process of moving freight looks like in dierent parts of the industry;
2. How the labor market works in dierent parts of the industry.
These two factors are critical to understanding what kinds of trucking jobs are likely to be automated
and what it will mean for workers to lose those jobs.
SEGMENTS OF THE TRUCKING INDUSTRY
Freight movement by truck varies in a number of key dimensions, most of which are important for
both the likelihood of automation and the quality of the jobs at risk. For readers unfamiliar with the
industry, these dierences and the terms used to describe them can be bewildering. An introduction
to the basic lay of the industry is useful before discussing autonomous trucks.
5
The trucking industry can be divided up into “segments” that haul dierent kinds of freight using
dierent kinds of equipment, in dierent size shipments, over dierent distances, at dierent speeds,
and for dierent kinds of customers:
Long-haul (generally 150 miles or more)
For-hire truckload (dry and refrigerated)
“For-hire” carriers haul freight for client companies; for example, they bring Johnson & Johnson
products to a Walmart distribution center. “Truckload” means shipments ll a trailer, by either space
or allowable weight. Truckload freight goes directly from customer location to customer location,
rather than rst going to a freight terminal (as in less-than-truckload service, described below).
For-hire truckload carriers include rms like Swift Transportation and C.R. England.
For-hire truckload service uses “dry” or “refrigerated” trailers. Dry vans are the standard box trailers
you see on the interstate. If freight needs to be temperature controlled, it will go in a “reefer”—
essentially a dry van with insulation and a refrigeration unit.
Private carriers
Some companies operate their own in-house eets, like a paving company that moves stone to a
mixing plant or Budweiser deliveries of beer to a corner grocery. In addition to having their own
DRIVERLESS | Steve Viscelli 6
SECTION TWO: The Trucking Industry and the Development of Autonomous Trucks
private eets, rms like Budweiser or Walmart may also contract additional trucking services from
for-hire carriers.
Less-than-truckload and parcel linehaul
In contrast to truckload, “less-than-truckload” (LTL) companies like YRC move smaller loads that
don’t ll an entire truck. Typically, a local pickup-and-delivery driver brings a load from the customer
to a freight terminal, where it’s combined into larger shipments based on destination. Then a
linehaul driver hauls the combined load to another company-controlled terminal near the freight’s
destination, where it’s broken down and sent out for delivery on local trucks. “Linehaul” means the
driver is moving between facilities owned by the same rm.
Parcel service moves packages using a process similar to LTL. Parcel is dominated by the twin giants
FedEx and UPS.
Specialized and atbed
Besides the standard dry vans and reefers, there are other types of trailers for specialized loads,
such as atbeds, tank trailers, and dozens of others. Specialized trailers carry cars, gases, trash, and
more; super-long atbeds transport oversized loads like windmill turbines and bridge trusses. For
the purposes of this report, I will lump all the companies using those trailers—and there are tens of
thousands of these carriers—into the commonly used category of “specialized.”
Local (less than 150 miles)
Local pickup and delivery
In LTL and parcel operations, these drivers move packages and shipments between the customer
and the freight-company-owned terminal, where they are combined in trailers for linehaul drivers to
move over longer distances.
Port and intermodal hauling
Port drivers haul shipping containers between shipyards and a range of destinations, including
warehouses, distribution centers, stores, and other transportation facilities.
For freight that travels most of the way by rail, intermodal trucks move the freight the remaining
distance on the road.
Other local service
In contrast to long-haul, local service makes shorter trips, generally anything less than 150 miles.
THE QUALITY OF TRUCKING JOBS
As I will discuss in detail in Section 4, the quality of trucking jobs varies widely. In some segments
of the industry, such as LTL and specialized hauling, the skill and experience of truckers is rewarded
with good working conditions and pay—sometimes very good pay. In other segments of the
industry, conditions are dismal.
DRIVERLESS | Steve Viscelli 7
SECTION TWO: The Trucking Industry and the Development of Autonomous Trucks
Poorly treated long-haul drivers may live for weeks, sometimes months, out of their trucks without
returning home. Due to the sedentary nature of the job, the constant presence of diesel exhaust,
truck-stop food, lack of exercise, and other workplace hazards, these workers suer tremendous
health consequences.
6
Many long-haul truckers are treated as independent contractors but are really misclassied
employees. They shoulder the expenses of owning and operating the truck they drive but get no
real benets in terms of pay or control over their work. Turnover is a chronic problem in some of
these segments. As a result, drivers in these segments are likely to be less-skilled, less-safe, and
less-ecient drivers. In the general freight and refrigerated long-haul for-hire segments, for instance,
turnover is very high (more than 100 percent every year at some of the largest carriers).
Despite the dicult conditions for many drivers, trucking still provides some of the better-paid
blue-collar jobs in the United States, with heavy-duty truckers earning a median income of a little
more than $42,480 a year in 2017, according to the Bureau of Labor Statistics.
7
That’s far more
than other workers in transportation, such as taxi drivers ($24,880), bus drivers ($33,010), and other
delivery drivers ($29,250).
Long-haul drivers can make considerably more than the average trucker. According to data from the
American Trucking Associations, the average long-haul for-hire dry van driver made $53,000 in 2017,
up from $48,000 in 2013. Unfortunately, the analysis in Section 4 suggests that the best-paying jobs
like these—as well as LTL linehaul jobs, which pay even better—are most at risk.
Local jobs, such as port hauling, pay far less than long-haul jobs, and workers often bear the cost of
owning and operating the equipment they drive as misclassied employees, a practice that is the
norm in this sector. These workers cannot aord new equipment, which means that dirty old trucks,
no longer reliable enough for long-haul work, are common on local jobs, with the attendant pollution
and health consequences, such as increased childhood asthma in neighborhoods near ports.
Though workers in these local jobs usually get home nightly, they often work the equivalent of
two full-time jobs, sometimes for less than minimum wage. These poor-quality and low-paying
jobs are the result of declining regulation in the industry and the spread of abusive labor practices
as employers seek higher prots and insulate themselves from risk in the face of destructive
competition. Absent policies to protect workers, jobs like these are the most likely to be created in
the coming decades as the industry automates.
What Is an Autonomous Truck and Who Is Making
Them?
For the purposes of this report, I will distinguish between self-driving, driverless, and autonomous
trucks. By self-driving, I mean that a vehicle can drive itself but may not be able to operate without
a human behind the wheel (Tesla’s current vehicles, for instance, are self-driving). Self-driving would
be the broadest category. By autonomous, I mean a vehicle that can drive itself without a human
ready to take control in at least some environments, such as highway driving. On the far end of the
spectrum, driverless trucks have no human in the vehicle at all.
DRIVERLESS | Steve Viscelli 8
SECTION TWO: The Trucking Industry and the Development of Autonomous Trucks
Since the primary concern of this report is the potentially signicant labor impact of self-driving
trucks, I focus on autonomous trucks, which may or may not be driverless. When a trucker is no
longer needed behind the wheel, there is the potential for signicant productivity gains and thus
potential labor impacts, such as job losses.
In contrast, in the near term we can envision self-driving systems, including capabilities like adaptive
cruise control and lane maintenance. A number of developers, such as Freightliner, are working on
features that may improve safety and relieve driver fatigue but do not increase productivity to the
point where there would be signicant job losses. Some functions might be automated, but these
improvements do not result in a fully autonomous truck. Meanwhile, other rms are working to
build trucks that never require a driver in them.
The most common way to explain where certain systems fall in terms of automation is the
automation scale used by the Society of Automotive Engineers. On a scale of zero to ve, Level 0
means that no functions of the vehicle are automated, while Level 5 means the vehicle can operate
without human intervention—or without a human even in them—in all driving conditions. In Levels
1 to 4, the lines are a little blurry, and debate will undoubtedly ensue as the technology develops.
Fortunately, for the purposes of this report, we don’t need to resolve this challenge with great
precision.
A Level 2 system has multiple automated functions. Level 3 vehicles can operate themselves
completely in certain environments but must be monitored at all times by a driver prepared to take
control—a monitored autopilot, if you will. While Level 2 and 3 automation will be considered briey,
with regard to safety and other issues, they are only of interest here because of that technology’s
potential as a path to Level 4 automation.
Level 4 vehicles can operate on unmonitored autopilot or autonomously in certain environments.
Level 4, most importantly, means trucks can operate on highways without a driver behind the wheel
(an autonomous truck) or, potentially, without a driver in the vehicle at all (a driverless truck). Level
5 means the vehicle can operate in any conditions or environment without requiring a human—they
could be driverless.
While Level 2 and 3 systems could have signicant safety and fuel eciency benets, the scenarios
in Section 3 focus on Level 4 and 5 systems, because they will have signicant labor impacts.
At present, most of the systems that are immediately aimed at Level 4 or 5 driving rely on relatively
similar combinations of technology. They use a wide range of sensors, including lidar (laser radar),
8
video cameras, traditional radar, ultrasonic sensors, and various motion sensors, such as
accelerometers. These sensors feed a tremendous amount of data to a computer on the truck,
which then creates a three-dimensional map of the truck’s environment and the objects around it.
Some developers are also exploring the use of more detailed “base maps” that reduce the amount
of information the truck needs immediately from the sensors. The computer then uses all that
information and GPS data to make decisions about where and how to drive, using algorithms to
predict the consequences of its own and other vehicles’ behavior.
DRIVERLESS | Steve Viscelli 9
SECTION TWO: The Trucking Industry and the Development of Autonomous Trucks
Work on self-driving cars began in the 1970s in Japan and was taken up at universities and
automakers in United States and Europe in the 1980s and ’90s. For decades, the U.S. government,
in particular, has funded basic research out of interest in military applications. The best-known
eorts are the Defense Department’s Defense Advanced Research Projects Agency (DARPA) Grand
Challenge, rst held in 2004, and the Urban Challenge, held in 2007. These competitions oered
teams substantial cash prizes to build self-driving vehicles that could complete courses in desert
terrain and then a simulated urban environment. Since then, engineers and computer scientists at
leading U.S. universities and around the world have accelerated work on self-driving hardware and
software.
Now that self-driving technologies hold more immediate commercial promise, some of the world’s
most resource-rich companies are working to bring them to market. These private eorts began
in earnest with Google’s driverless car program in 2010. Today, self-driving technology is being
developed by dozens of major rms in the tech sector as well as numerous vehicle manufacturers
and suppliers. Among those working on self-driving cars are Silicon Valley giants—like Apple,
Waymo (formerly Google’s driverless car project, now owned by Google’s parent company Alphabet),
Uber, Lyft, and Tesla—and major automakers, including Daimler, Audi, BMW, Volkswagen, Volvo,
GM, Ford, Honda, and Toyota. Numerous suppliers, such as Delphi and Bosch, are likewise working
to develop self-driving technology, many in partnership with tech rms, like Nvidia, Samsung, Intel,
and Microsoft. In fact, it’s hard to nd any major vehicle manufacturer not working on self-driving
technology and increasing their investment in that work.
While private eorts and demonstrations dominate headlines today, as we consider the
responsibility of government to inuence how this technology aects workers, the public, and the
environment, we should not forget that this technology was, until recently, largely developed by
public dollars.
Beyond highlighting the public role in creating this technology, this brief history should help
skeptics understand that self-driving vehicles are not the fantasy of a few mad scientists or a Silicon
FIGURE 2.1
Key technology used in self-driving trucks
LIDAR
Radar
Cameras
GPS
Ultrasound
Internal gyroscopes & accelerometers
DRIVERLESS | Steve Viscelli 10
SECTION TWO: The Trucking Industry and the Development of Autonomous Trucks
Valley vanity project, but rather a long-standing, well-funded set of interrelated eorts by thousands
of scientists and engineers across academia, industry, and government, whose eorts have already
developed the basic technologies required for self-driving vehicles. Today, billions of dollars are
being invested annually to achieve the goal of autonomous vehicles.
Why Autonomous Trucks May Come Before Other
Self-Driving Vehicles
Trucks are dierent from cars and will require dierent self-driving technology—including dierent
placement of sensors, longer braking distances, and more space to turn, as well as more physically
robust components—but the most dicult technological challenges are shared by both types of
vehicles. Signicant investment in the specic development of self-driving trucks is more recent.
Until just two or three years ago, work on trucks with signicant automation in the United States was
conned to just one start-up (Peloton) and small projects within major truck makers, aimed primarily
at collision avoidance and lane-maintenance, including the notable eorts of Freightliner (owned
by Daimler). Internationally, Volvo, Daimler, and others were taking on more ambitious projects
in self-driving commercial vehicles. As a result, autonomous vehicles, including trucks, are already
operating protably in controlled environments, such as seaports and mines in both Canada and
Europe.
Autonomous trucks are now seen as a lead sector for autonomous vehicle adoption in the United
States. Total funding for new truck technology development, including autonomous trucks, was
estimated to reach $1 billion in 2017, up 1,000 percent in just three years.
9
Major projects for Level
4 autonomous trucks are now underway by Waymo as well as a number of startups, including
Embark and Starsky Robotics.
A simple economic argument recognized since at least 2013 explains why self-driving trucks will
be adopted before driverless cars.
10
The purchase of a truck is a business decision that is often
thoroughly evaluated by eets, and given the potential labor savings of Level 4 and 5 autonomy, the
returns on investment could be extraordinary.
11
In addition, numerous ineciencies related to human truck drivers—such as mandatory rest breaks
and the need for drivers to return home, even only for a couple of days every few weeks—not only
raise costs but make truck freight slower. Just as importantly, the largest segment of the industry,
for-hire long-haul truckload, has high turnover and diculty recruiting new drivers due to poor
working conditions and low pay. Eliminating the need for drivers would be a tremendous boon for
companies in that segment. Unlike cars, there is already incredible demand for the potential benets
of autonomous, even driverless, trucks.
It is conceivable that autonomous trucks could double the productivity of long-haul trucks for
highway segments. They could also substantially reduce fuel costs, which comprise another 30+
percent of total costs for many carriers. Relative to these gains, the cost of self-driving technology is
expected to be small. Like other researchers who have interviewed experts and developers, I found
that, despite some uncertainty, the expectation is that self-driving technology would add less than
DRIVERLESS | Steve Viscelli 11
SECTION TWO: The Trucking Industry and the Development of Autonomous Trucks
20 percent to the cost of a new tractor.
12
It is important to note, however, that such aordability is
dependent on dramatic reductions in the cost of some of the more expensive parts of the systems
currently under development, such as the powerful lidar most of these vehicles depend on.
Autonomous trucks could also operate around the clock, essentially doubling the performance of
a human-driven truck. The value of a self-driving car, on the other hand, is unclear for individual
buyers who will be riding in the vehicle, regardless of whether they are driving or not.
From a technological standpoint, autonomous trucks have another advantage over cars. Self-
driving cars will be operated primarily in congested urban areas and on local roads, which present
a far greater challenge in terms of object recognition and decision making by articial intelligence.
In contrast to passenger cars, an autonomous truck that can operate only on the highway might be
highly protable.
In light of the above considerations, trucks are now seen by many experts and those in the industry
as one of the clearest opportunities for early adoption of self-driving technology. Existing reports,
however, overestimate the readiness of this technology for actual trucking operations, when
adoption will begin, and how rapidly it will spread. As a result, many analysts conclude that job
losses could come within a few years and spread across the industry in just a decade or so. Such
rapid deployment and adoption of self-driving technology is extremely unlikely.
Even the most optimistic developers believe we are still at least several years away from autonomous
trucks operating even in limited highway operations in anything other than testing programs with
In 2013, the investment rm Morgan
Stanley conservatively estimated that
autonomous trucks would save the
industry $168 billion. Analyses like these
suggest a strong economic argument for
trucking as a lead sector of autonomous
vehicle technology.
FIGURE 2.2
Potential savings to the U.S. freight transportation industry from autonomous freight vehicles
$70bn
Labor
savings
$27bn
Productivity
gains
$35bn
Fuel
efficiency
gains
$36bn
Accidental
savings
$168bn
Autonomous
freight vehicles
total savings
Source: Morgan Stanley. Autonomous Cars: Self-Driving the New Auto
Industry Paradigm. Morgan Stanley Blue Paper. November 6, 2013.
DRIVERLESS | Steve Viscelli 12
SECTION TWO: The Trucking Industry and the Development of Autonomous Trucks
drivers still behind the wheel. Important challenges in both hardware and computer science need to
be overcome before autonomous trucks are able to operate safely and reliably.
The cost of components, including some very expensive sensors, must drop signicantly before
autonomous trucks will be economical. Some potential self-driving technology may require
improved infrastructure maintenance or new infrastructure, ranging from better lane markings
to more robust wireless communications networks. Once autonomous trucks are demonstrably
safe, there may also be important regulatory debates to ensure safety. For instance, if drivers and
autonomous trucks will work together, there are questions about how this collaboration will aect
rules that limit drivers’ work hours to ensure that they don’t drive fatigued. Given the slow speed at
which trucking regulation typically is debated and implemented, addressing such an issue will likely
take years.
Then, before autonomous trucks can be adopted on a widespread basis, they will need to
demonstrate reliability and feasibility within the operations of actual trucking rms. A change in
equipment and operations this fundamental is not something trucking carriers will do without
lengthy consideration. Trucking equipment, both tractors and trailers, are used in harsh conditions
for hundreds of thousands, even millions of miles. Equipment critical to the safety and reliability of
these systems, such as computers, sensors, wiring, etc., will need to withstand cold temperatures,
near constant vibration, ice, salt, and more, over long periods.
Also challenging for developers is the relatively low number of units sold by truck manufacturers. In
2016, some 250,000 class 8 trucks (the largest trucks) were sold.
13
The number of autonomous trucks
sold per year would be signicantly smaller because most class 8 trucks don’t operate over long
distances. Sales of passenger cars and light trucks, on the other hand, totaled around 17.5 million in
2016.
14
Finally, when autonomous trucks can be safely and legally operated, they will only perform one
task: driving. Most truck drivers do far more than just drive. Reports suggest there are 3 or 4 million
driving jobs at risk, but they use broad categories to represent “driving occupations.”
15
As other
experts have claried, in nearly all the jobs that fall under such broad headings, driving is one task
among many others, and driving may not even represent a majority of the work in many of these
jobs.
16
In fact, for many of the trucking jobs I identify below as the most likely to be automated,
driving may only represent about 50 percent of drivers’ total work time.
17
Instead of assuming that an autonomous truck can operate in all freight operations and that all
drivers are equally at risk, we need to identify more precisely which jobs may be aected. Thus, we
need to understand in more detail the ways in which autonomous trucks might be used. A range of
self-driving technologies is currently under development, and each has its own particular challenges,
potential benets, applications, and likely job impacts. Section 3 lays out dierent scenarios for
how the technology described might be used in practice. Section 4 explores which segments of the
industry might be able to use it, if the scenarios in Section 3 are correct.
DRIVERLESS | Steve Viscelli 13
We don’t know exactly what autonomous trucks will be able to do, but looking at the technology
that various rms are developing can help us understand the potential impacts in the
near-to-medium term (0–25 years, by my denition). While technologists and experts consider a few
scenarios far more likely than others, it is worth looking a wide range of alternative scenarios for a
number of reasons.
First, this analysis is not intended as a prediction of exactly how autonomous trucks will be used.
Rather, it is intended to begin a serious conversation about how we can forecast and shape the
potential impacts of autonomous trucks to maximize the benets of these technologies and ensure
that these benets don’t come disproportionately at the expense of workers who are displaced or
have their pay or working conditions deteriorate.
Second, uncertainties about the technology’s future are signicant and evolving. Over the past
several years, rms and experts have viewed a number of ideas about how autonomous trucks
might develop as more or less likely. Real uncertainty exists around key aspects regarding the mix
of capabilities, reliability, and costs of all of the dierent possible autonomous truck technologies
currently under development.
Third, given the diversity of operations and customers within freight transportation, it is likely
that over time, a number of dierent applications of autonomous truck technologies could be
successfully brought to market to suit dierent segments of the industry. Just as no single trucking
process or equipment moves all freight today, no single technology or process will move all freight
in the future.
To develop the following scenarios, over the course of the past year I have studied the existing
popular coverage as well as industry and scholarly literature on self-driving technology. I also
interviewed numerous experts and discussed potential technology with stakeholders in the
industry—including Silicon Valley tech companies (Uber ATG
18
and Peloton among others), truck
manufacturers, truck drivers, venture capitalists, computer scientists, engineers, labor advocates and
unions (including NGOs and the International Brotherhood of Teamsters), academic experts, and
others—to nally derive at these six scenarios for the use of autonomous trucks.
SECTION THREE:
Scenarios for the Use of
Autonomous Trucks
DRIVERLESS | Steve Viscelli 14
SECTION THREE: Scenarios for the Use of Autonomous Trucks
WHY SCENARIO FORECASTING?
The primary method used below is scenario forecasting, which describes a series of likely or possible
futures. Scenarios are good for thinking about the likely outcomes of disruptive technological
development when a complex set of factors make modeling and trend extrapolation based on
existing data impossible or likely to be unreliable. In this case, the choice of scenario forecasting is
justied by the disruptive nature of autonomous trucks, the complexity of the existing truck freight
transportation industry, and the presence of signicant trends, such as e-commerce, with uncertain
implications for the adoption of autonomous trucks. Scenarios also make clear the importance of
considering the labor process involved in moving freight and the way that process and possible
alterations to it will make autonomous truck adoption more or less desirable.
The scenarios below identify the major self-driving technologies that have been publicly identied
and suggest how they might be used. The hope is that they can be the basis of stakeholder
engagement aimed not at predicting the future, but shaping it. Scenarios are commonly used in
transportation planning when a number of competing goals might need to be balanced and public
policy can clearly aect that balance through regulation, infrastructure investment, and other roles.
We need to move beyond trying to predict the inevitable outcome of autonomous truck
development and recognize the importance of identifying the benets we want to bring about. We
don’t need precognition; we need a road map that will help us understand how to encourage the
development and adoption of autonomous trucks to meet the goals we set.
AUTONOMOUS TRUCK PORTS
A number of the scenarios described below involve trucks that can only drive on highways because
local driving, with the presence of pedestrians, intersections, etc., is simply too complex. One of the
ways to solve this problem is to hand o trailers between human-driven trucks and autonomous
trucks near the exits of the interstate highway system. Such operations could use a facility I call an
“autonomous truck port” (ATP). As illustrated in Figure 3.1 (page 15), ATPs would be strategically
located truck parking lots (or “drop lots”) at interstate exits outside of congested urban areas. These
staging areas would provide space to park and couple trailers but could also have driver facilities
and fueling or charging options for trucks.
Having trucks stop at ATPs would allow for the use of the most fuel-ecient technology on either
side of the port, whether local driving (e.g., electric) or long-haul driving (e.g., sleek aerodynamics).
These facilities would also facilitate o-peak deliveries to keep trucks out of rush-hour trac. They
could be publicly owned land with private services or privately owned entirely.
19
The use of ATPs would have tremendous impacts beyond the physical operation and eciency
of the truck itself. It would cut down on the coordination required between shippers and carriers.
Instead of having to nd freight service from point to point, the service could be provided by a local
truck that is getting freight through an app, which would greatly increase the eciency of matching
freight to drivers and potentially lower costs for smaller shippers, in particular. This sort of task could
be accomplished using an Uber-style application, with real-time pricing. A local truck might bring
DRIVERLESS | Steve Viscelli 15
SECTION THREE: Scenarios for the Use of Autonomous Trucks
the freight to an ATP, where an autonomous tractor would take over, for instance. Since there would
be a large volume of tractors and trailers continually moving through the same facilities, in theory
there would be little reason that a particular autonomous tractor could not go to many or nearly all
other ATPs. Tractors and trailers could be mated based on timing of their arrival to the ATP.
KEY TRENDS TO CONSIDER: THE DIGITIZATION OF FREIGHT
MATCHING AND E
-
COMMERCE
Key trends that will shape the impact of autonomous trucks and, in turn, be shaped by the
development of autonomous trucks include the digitization of freight matching and the growth in
e-commerce and last-mile delivery.
Digitization refers to the matching of freight with trucks by means of a platform similar to that used
by transportation network companies, such as Lyft and Uber, to match passengers with drivers.
There have been a number of attempts to digitize freight in recent years using better information
technologies.
At present, the matching of freight to trucks happens through numerous processes ranging from
long-term contracts to Internet load boards. The largest shippers often contract well in advance
for freight services from large trucking carriers, but many shippers are dependent on brokers who
match supply and demand on a short- and medium-term timeframe. Substantial friction persists in
the transactions involved in matching freight and carriers and in the interactions between rms. This
FIGURE 3.1
What an autonomous truck port could look like
Autonomous
Truck Port
Human & autonomous
tractors swap trailers
Drivers could select
loads through an app
DRIVERLESS | Steve Viscelli 16
SECTION THREE: Scenarios for the Use of Autonomous Trucks
problem is most often handled by freight brokers who take a signicant chunk of load revenue for
their eorts. And, despite technological advances in the industry, this work is done by thousands of
workers who call customers looking for freight and matching it with trucking carriers that need loads
for particular trucks. Today, it often takes several hours, multiple emails and phone calls, and maybe
even a few faxes, to mate a load with a truck.
The idea underlying the digitization of freight is that if you can get more loads and more trucks into
a transparent process, then you can more eciently match them in terms of time of availability and
location. The value of even a few percent improvement in the system of matching freight and trucks
overall could reduce the miles traveled empty to pick up to load (known as “deadheading”) and
waiting time that could be calculated in the billions of dollars annually.
However, the challenge of creating a functional market around these platforms is far more
complicated than some new entrants imagined, and a number of these digital brokers have
stumbled in recent years. Nonetheless, the basic idea of more ecient matching of supply and
demand is widely seen as having real merit by nearly everyone in the industry.
Another key issue to consider is the explosive growth in e-commerce and in last-mile deliveries
and returns. While most of this growth has thus far been handled by the U.S. Postal Service, UPS,
and FedEx, new models are emerging to meet surging demand. Numerous last-mile delivery
services have begun for groceries and other product categories. Likely to be most important
in this area is the development of Amazon’s own delivery services, known as Amazon Flex and
Amazon Delivery Service Partners (DSP). Like most other new last-mile delivery services, Amazon
Flex hires independent contractors who use their own vehicles to carry packages from Amazon
fulllment centers to customers. Amazon DSP, just launched, is a franchise-like model for Amazon to
subcontract its delivery service out to small, nominally independent companies.
The biggest problem with the digitization of freight and the growth of last-mile delivery is a concern
that has plagued trucking since its earliest days: destructive competition. There may be little prot
to be made in moving digitized and last-mile freight, and the vast majority of costs may be borne
by workers in the form of lower pay and risks associated with owning and operating equipment. As
we see in for-hire long-haul truckload and port hauling, without rules to stop them, the response by
companies to digitization will be to use poorly paid employees and, whenever possible, independent
contractors who provide their own vehicles.
THE LIMITS OF THE ANALYSIS: UNDERSTANDING THE
JETSONS FALLACY
Current estimates of job loss from autonomous trucks often commit the “Jetsons Fallacy,” and this
report may, as well.
20
The Jetsons was a cartoon, rst aired in 1962, that depicted a family of 100
years in the future. They had ying cars, a robot maid, machines that made meals instantly, smart
watches, and holograms. In fact, The Jetsons predicted many cutting-edge technologies that are
commonplace today. What The Jetsons didn’t get right were the social norms and behavior of the
future, and the way technology would transform and intersect with them.
DRIVERLESS | Steve Viscelli 17
SECTION THREE: Scenarios for the Use of Autonomous Trucks
George Jetson, the father, is a bumbling patriarch and solo breadwinner. Jane, his wife, is a dutiful
homemaker, who sneakily snatches George’s wallet to head o to the local shopping mall. George
and Jane live in a nuclear family, with two cute kids. When the family uses its instant food machine
nightly, everyone gathers around the family dinner table, except Jane, who stands next to the
machine pushing buttons as if she were at a stove. The machine’s performance—or rather, Jane’s
inability to get it to perform—is a frequent source of arguments. As Jane operates the machine, her
husband and children engage in intimate, thoughtful conversation about the day’s events and their
lives, oblivious to the ubiquitous screens, video phones, and smartwatches around them.
In short, while the capabilities of the technologies weren’t far o, how they would aect
people’s behavior wasn’t even close. The Jetsons kept family and social life constant and simply
substituted one tool for another. It mapped the technology of the future onto an idealized 1950s
American family, without understanding that new technologies—like processed or fast food and
smartphones—would combine with other economic and social changes to transform the way we live.
To an even greater degree than it aects life at home, technology directly transforms the way we
work. Automation in the workplace is fundamentally about increasing productivity by changing
which tasks are performed by humans and which tasks are performed by machines. But in changing
the division of tasks, automation often transforms the labor process more generally. The Jetson
Fallacy for autonomous trucks would model future scenarios that assume autonomous trucks will
replace human-driven trucks without meaningfully aecting the process of moving freight, including
where, when, and how much of it moves.
Let me give one example to illustrate this point. Walmart is widely considered to have the most
sophisticated logistics system of all big-box retailers. The rm gains tremendous advantages from
that system and is constantly exploring innovative ways to improve the eciency of its trucks and
truck movements. Once autonomous trucks are able to successfully go from facility to facility (for
example, from Walmart’s distribution centers to its stores), Walmart will very likely be one of the
earliest adopters among private carriers. The Jetsons Fallacy in this case would be to assume that
Walmart would simply substitute autonomous trucks for those operated by human drivers who
perform the regularly scheduled deliveries of goods from distribution centers to stores. In these
very well-paid jobs, drivers make one to two trips, taking full trailers of new goods from distribution
centers to stores and returning with empty trailers or trailers carrying damaged or returned
merchandise. Sometimes these drivers will pick up a load of new goods headed to the distribution
center on their return trip.
21
In fact, while Walmart might adopt autonomous trucks for regular service from distribution centers
to stores, it might also make far more radical changes to its logistics system. Why? Because
Walmart’s logistics system was built around the limitations of human-driven trucks. One of the
greatest insights that Sam Walton had was that truck eciency was critical to making the most of
a distribution center-based model.
22
In order to maximize the trucks’ eciency, Walton’s strategy
was rst to identify areas where he wanted to locate stores and then place a distribution center
within one day’s roundtrip drive for Walmart’s trucks. For many of reasons, from speed of restocking
inventories, to asset utilization, to labor costs, this was a great distribution model.
DRIVERLESS | Steve Viscelli 18
SECTION THREE: Scenarios for the Use of Autonomous Trucks
The limitations of human-driven trucks were foundational considerations in designing Walmart’s
overall logistics system. If autonomous trucks, which don’t have to stop to sleep, can travel twice as
far in a day, what will Walmart do? Most likely it will consolidate distribution centers, spacing them
further apart and allowing Walmart to reduce inventory costs.
In similar ways, the limitations of human truck drivers have been a factor in siting and operating
acre upon acre of warehouses, retail space, highways, truck stops, parking lots, and other assets
and infrastructure. Autonomous trucks will fundamentally change the capability of trucks and the
economics surrounding their use. In combination with trends in e-commerce, last-mile delivery,
and the potential for extensive digitization of freight markets, autonomous trucks will remake the
movement of goods in the United States.
Finally, it is critical to recognize that while autonomous trucks might make long-haul trucking
cheaper and faster, e-commerce has been steadily increasing the number of short trips retailers use
to supply more goods faster and in smaller shipments to customers. Predicting the overall impact of
autonomous trucks then requires understanding how it will intersect with these other trends—a task
whose diculty should not be underestimated.
Six Potential Adoption Scenarios
In what follows I lay out six potential scenarios of how autonomous trucks might be adopted in the
trucking industry.
TECHNOLOGY SCENARIO 1:
Cooperative Adaptive Cruise-Control Platooning
Cooperative Adaptive Cruise-Control Platooning (often simply called “Platooning”) has the longest
public prole of recent automated driving technologies. The goal of platooning is to allow two or
more trucks to save fuel by drafting each other to reduce wind resistance, just as bicycles or cars
do when racing. Optimal drafting requires trucks to be fairly close to each other, somewhere less
than 70 feet apart depending on conditions.
23
This following distance is signicantly less than what
is considered safe at highway speeds for large trucks without automated technology, which would
Lead truck fully
controlled by driver
Following truck’s speed & braking
is controlled by lead truck
Platooning aerodynamics
improve fuel efficiency
DRIVERLESS | Steve Viscelli 19
SECTION THREE: Scenarios for the Use of Autonomous Trucks
typically be at least several hundred feet. Of course, on highways in many areas of the country today,
as a result of surrounding trac and other factors, freight trucks often end up following one another
at distances closer than is safe, without active safety or platooning systems.
Platooning utilizes a combination of sensors, GPS, wireless, and vehicle-to-vehicle (V2V)
communications technologies to allow trucks to follow very closely behind one another by linking
their acceleration and braking. One leading provider of platooning systems uses a control center to
track and match potential trucks, assigning them to a platoon based on location and a number of
important factors, such as each truck’s weight and estimated braking capabilities. Once linked in a
platoon, the lead truck is entirely under the control of a human driver. The trailing truck or trucks
follow directions provided via wireless links based on the operations of the lead truck but only for
acceleration and braking. The human driver of the following truck is still responsible for steering and
maintaining proper position in the lane.
24
Currently, platooning is envisioned as a practice that will be used only during highway driving and
as a means to achieve greater fuel savings, reduce highway congestion, and improve safety. As yet,
removing the human drivers in following trucks has not been tested or stated as a near-term goal
by U.S. developers or trucking rms. At present, platoons involving only two trucks (one lead, one
follower) are proposed for the United States, though demonstrations are planned in Singapore using
one lead truck and three following trucks, and three-truck platoons have been demonstrated in
Europe. A number of U.S. states have developed or are developing regulations that will allow platoons
to operate, the primary regulatory obstacle being restrictions on following distances between vehicles.
A wide range of trucking operations that travel signicant distances on interstates could adopt
this technology in the near future, if communications equipment could be integrated with truck
technology in a cost-eective fashion. The most likely adopters are large truckload carriers, which
have lots of trucks on the road that might be close enough to make platooning ecient, and
less-than-truckload carriers, which have multiple trucks leaving terminals on regular schedules,
making coordination of platooning trucks easier.
The most likely way for trucks to be linked to a control center and across rms is through existing
4G LTE cellular communications. Widely available cellular communications provide the ability to
coordinate platooning between trucks from dierent carriers. In the approach of Peloton, the
leading U.S. developer of platooning, a Network Operations Cloud provides “platoon authorization,”
and drivers make the decision to form or dissolve platoons. In the event that cellular communication
is lost, vehicles can retain Network Operations Cloud authorization for a dened time period before
authorization is removed.
Platooning may result in signicant fuel or energy savings
25
and, thus, cost savings for operators.
Along with lane maintenance and collision avoidance, it represents one of the most high-prole
examples of a possible incremental adoption path to self-driving trucks.
26
Since the driver in the
following truck is still responsible for steering and lane maintenance, there will be no signicant
productivity gains in terms of labor and, thus, no job losses. However, if platooning is a path
to having fully autonomous following drone trucks, as some industry players believe, future
higher automation systems using platooning could then provide signicant driver productivity
improvements and have signicant labor impacts, as addressed below.
DRIVERLESS | Steve Viscelli 20
SECTION THREE: Scenarios for the Use of Autonomous Trucks
There are several issues related to job quality and other safety-related aspects of cooperative
adaptive cruise-control platooning that will require further consideration. While platooning, the
following driver’s view of the road ahead is a consideration; this view depends on the inter-vehicle
gap. The view of the rear driver at typical platooning distances of 60 feet provides the rear driver
with the ability to see surrounding trac and road signs. A video feed from the front truck provides
awareness of trac ahead. Voice communications between the two drivers increases situational
awareness for both. At the same time, following drivers need to track the position of their own
vehicle within their lane. While drivers have been trained to maintain a following distance from
the vehicle ahead that would allow them to react and brake, the NACFE Condence Report on
Platooning notes that the learning curve for platooning “is not steep.”
27
The level of stress aecting the following drivers after long periods of platooning should be
considered. In particular, the health and safety impacts of several-hour stretches of this kind of
behavior is something that may require further study. I did talk with a former test driver for Peloton
who reported he did not perceive additional stress or fatigue, even at high speeds, during platoons.
To the contrary, due to the constant presence and communication with the lead truck, this driver
reported nding platooning less monotonous and, therefore, safer. At present, it appears the
potential fuel, safety, and congestion benets of platooning far outweigh the costs. Adoption of this
important technology could begin by the end of 2018.
TECHNOLOGY SCENARIO 2:
Human–Drone Platooning
A scenario that was regularly mentioned in the past but has gotten little attention recently is the
possibility of a human-driven truck with autonomous truck units trailing in a platoon for interstate
driving.
28
In this scenario, a human driver would bring a trailer to an autonomous truck port or ATP,
as described above, and uncouple the trailer. The trailer would then be coupled to an autonomous
Drones follow
human driver
Autonomous
Truck Port
Autonomous
Truck Port
Local trucks could
be electric
Tractor optimized
for highway driving
Human & drone
tractors swap trailers
DRIVERLESS | Steve Viscelli 21
SECTION THREE: Scenarios for the Use of Autonomous Trucks
tractor. This truck would be fueled and inspected at the ATP. It would follow a human-driven truck
onto the interstate (or perform limited autonomous driving to get itself onto the interstate until
linked in a platoon with a human-driven truck, perhaps using a special section of roadway where the
autonomous truck could get up to speed and engage in the platoon). While on the interstate, this
autonomous truck would platoon with the lead truck for acceleration and braking but would also be
capable of independently maintaining its lane using the human-driven lead truck as a reference. In
this scenario, the following truck would essentially be both remotely operated by the driver of the
lead truck and capable of autonomous driving within the platoon, using vehicle-to-vehicle networks,
wireless communications, sensors, and articial intelligence.
From a technological standpoint, this scenario has signicant advantages. Having a lead driver
would mean that drone trucks could mostly rely on decisions made by the human driver and be
left with the much simpler task of lane maintenance. In the event that it were disconnected from
the lead truck, the following truck would need to be capable of driving itself to a safe location (for
example, pulling over onto the shoulder of the road) or to be remotely piloted from a control center,
as discussed in Scenario 4, below.
In addition to reducing the complexity of decision making required of the autonomous truck,
having a human driver could reduce, if not eliminate, concerns related to bad weather and security.
There would also be a human driver available to perform inspection, maintenance, coupling and
uncoupling, fueling, and other tasks, as necessary.
From a cost perspective, this scenario has tremendous potential. There would be substantial labor
productivity gains in the long-haul portion of the duty cycle and, thus, signicant reductions in
drivers needed.
Human-led drone platooning is also the only scenario where driver upskilling would be almost
certain for driving activities. This scenario could improve the quality of driving jobs. Because of
the signicant productivity gains, a platoon pilot—like the highly skilled and experienced drivers
who haul multi-trailer combinations today—would be of much more value to carriers. Rather than
de-skilling drivers and making them less needed, this scenario would add additional high-skill tasks
to the drivers’ work. These drivers would be responsible for more freight and equipment. While
the importance of their conscientious eorts and skill would increase, labor costs would decrease
considerably compared to the overall cost.
Drivers in this scenario would likely be better trained, more highly skilled, and better rewarded. If
these drivers could run from ATP to ATP, they could even be put on more regular routes, allowing
drivers to be home more often and spend less time living on the road and out of their vehicles. All
these improvements could combine to make for fewer, but much better, long-haul trucking jobs. At
the same time, a signicant number of local jobs would be created to bring trailers to and from ATPs.
Unfortunately, these local jobs would be at risk of the same problems found in existing jobs at ship
ports today, as discussed below.
There would also be better energy savings than in multiple-driver platoons because the following
units could be recongured to be lighter and more aerodynamic. These drone long-haul trucks
DRIVERLESS | Steve Viscelli 22
SECTION THREE: Scenarios for the Use of Autonomous Trucks
could be specically engineered for high-speed highway driving, right down to the tires, axles, and
other features, making these trucks much more fuel ecient. In addition, because the local trucks
bringing trailers would not need to drive highway miles, they too could be optimized, with features
such as electric or alternative fuel powertrains, tractors without sleeper berths (a.k.a. day cabs),
optimal tires, etc.
This scenario would be most important for the general and refrigerated freight segments as well
as LTL and parcel linehaul operations. There is potential for some job losses in the long term, but
this scenario could greatly improve the eciency and safety of trucking operations and the quality
of many jobs. While the technological challenges in this scenario are fewer than in other scenarios,
this particular scenario does not seem to be a serious goal of most Silicon Valley autonomous
truck projects at present (though it is an obvious next step for Peloton and its partners). The
need for facilities where local drivers could bring trailers to be assembled into platoons could be
accomplished with existing facilities, such as those used to break down multi-trailer trucks, or new
facilities, like ATPs, could be built with public or private funding.
TECHNOLOGY SCENARIO 3:
Exit-to-Exit Autonomous Trucks Plus Drone Operation
The “exit-to-exit autonomous trucks plus drone operation” scenario has received attention over the
past year in the United States, largely as a result of Starsky, a Silicon Valley startup. In this scenario,
trucks would be autonomous on interstates and then piloted remotely by human operators while
driving in local areas and under conditions where autonomous driving might not be possible
(e.g., in bad weather that would make some sensors unreliable). Remote piloting is envisioned
as being done by operators stationed at centers where they would use workstations to receive
information from the truck and remotely pilot the vehicle much like the military pilots aerial
drones. Such systems are in use already in controlled industrial settings, including mines and ports.
However, deploying this technology on public roadways across broad geographic areas and with
communications networks of varying reliability presents signicant challenges. For example, what
happens when remotely piloted trucks lose their wireless signal?
No need for human-driven
tractor allows longer trailer
Remote drivers navigate
local streets
Tractor not optimized
for highway since it
also drives locally
DRIVERLESS | Steve Viscelli 23
SECTION THREE: Scenarios for the Use of Autonomous Trucks
Proponents suggest drone piloting would improve the work lives of truck drivers, allowing them
to work at control centers close to their homes rather than traveling for weeks or months at a
time and living out of their trucks. The industry could thus address the most dicult problem
associated with driver turnover, potentially resulting in signicant cost savings. Productivity gains
would be substantial because trucks could be operated on multiple shifts, overcoming the biggest
single source of ineciency in the industry: the “one truck–one driver” model that predominates
and results in trucks sitting idle while drivers take their 10-hour mandatory break required by
federal hours-of-service regulations. With remote drivers taking regular breaks and with assistive
technology, such as lane maintenance and collision avoidance, signicant safety improvements
might also result. Vehicles without driver cabs could be designed for greater aerodynamics, larger
trailers, or lighter weight, resulting in signicant productivity and fuel eciency gains. However,
since the same tractor would be used for urban and highway driving, specialization for particular
environments (as in the previous two scenarios) would not be possible. Perhaps most importantly in
terms of eciency, drone operation would mean drivers would not need to sit unpaid while waiting
as their trucks are loaded and unloaded.
This scenario would likely have some important eects on employment relations and the structure
of rms in the industry. The remote piloting of trucks is unlikely to be performed by independent
contractors, given the substantial investments and centralized control systems required. Drone
operation would also likely result in greater concentration in the industry, due to increased capital
intensity and reduced opportunities for small businesses.
In theory, while this technology could be applied in a wide range of driving settings, it would
require signicant operational changes for most segments. Adoption would be inuenced by
whether current driver-performed tasks, such as loading, coupling, opening doors, fueling, and
inspections, could be performed by other workers at loading docks. Customers would have to
be suciently large to have trained dock sta on hand. Since not all customers would have such
sta, drone-operated trucks would need to be distinguished from non-drone trucks within the
load-planning process.
Some experts I spoke with suggested this scenario may face additional security challenges since
the system is designed to be piloted remotely by a human. If someone gained control over a
workstation or the communications link between the workstation and the truck, the truck could be
stolen or used for mayhem. Another challenge to this technology is ensuring that human drivers
have all the information to safely pilot the truck remotely in congested areas. Drivers would need
to have the necessary situational awareness, which might require expensive simulator-like work
stations and signicant investment in cameras and sensors on the truck. The cost of such equipment
would clearly aect the protability of remote operations. There will also certainly be ethical issues
when the human driver’s own physical well-being is not at stake in making decisions that may aect
safety.
Another long-term question about this scenario is how drivers would be trained. As suggested
below, the segments most likely to be automated are the ones that train the vast majority of new
drivers. If drone operation for local driving becomes the norm, it raises the question of how drivers
DRIVERLESS | Steve Viscelli 24
SECTION THREE: Scenarios for the Use of Autonomous Trucks
would learn to drive a truck and whether real-world experience driving a truck would be needed in
order to safely operate a truck remotely or whether operators could be trained entirely on drone
systems.
In this scenario, the job of truck driver could be signicantly improved in a number of ways. The
primary concern here in terms of labor will be job losses in some segments and downward pressure
on wages from increasing the supply of drivers and eliminating the premium that carriers currently
pay to get workers to live on the road (more on this aspect in the sections below). In terms of
potential job losses, productivity per truck could go up signicantly, and productivity per worker
would be aected by removing autonomously driven highway sections, by eliminating waiting
times for loading and unloading, waiting to be dispatched, etc., and by doing away with much or
all non-driving work. Essentially, drivers’ only remaining task would be to remotely pilot the truck
for local driving. In some segments of the industry, such as long-haul truckload, tasks representing
perhaps 90 percent of drivers’ time could be eliminated. This improvement would have a signicant
impact on the number of jobs required to move freight in these segments.
Several million workers who have been trained to drive tractor-trailers in recent years have left the
industry but might return if they were not forced to be away from home, living out of a truck for
weeks at a time. Workers who never previously considered trucking for these reasons might also be
attracted to the labor market. Working in simulators—which will obviously be safer than operating
inside the trucks—and other improved conditions may also help the industry to attract millennials
and women, something it has had great diculty doing. As a result, former drivers, women, men
with young children, and younger workers may be more attracted to these new jobs.
While this scenario has a number of hurdles and may not be feasible for most of the industry
because of the diculty of replacing the non-driving labor of drivers, there is a strong possibility
that a drone-type of operation from a control center could be used for short periods to “rescue”
autonomous trucks that cannot operate because of weather or complex situations beyond their
programming.
TECHNOLOGY SCENARIO 4:
Driver-in-the-Sleeper Scenario (A.K.A. Autopilot)
Among the earliest visions for autonomous truck use was one in which driver and machine would
take turns driving and operate as a team. This scenario is often suggested as the most likely and
desirable use of autonomous trucks by the American Trucking Associations. The ATA suggested this
would look like autopilot for airplanes. In highway driving situations, the driver would remain in the
sleeper berth, and the machine would drive itself. Then, when situations require a human driver—
such as local driving, fueling the truck, or dealing with a shipper—the human driver would take over.
Initially some technologists suggested that a primary market for this sort of adoption would be
owner-operators, who would get much greater asset utilization from their truck.
While promoting this scenario has a signicant benet at the moment—namely, it may not scare
workers away from entering the trucking industry—there are numerous obstacles that make it
DRIVERLESS | Steve Viscelli 25
SECTION THREE: Scenarios for the Use of Autonomous Trucks
unlikely and highly undesirable for most segments of the industry. It is dicult to sleep well in the
back of a moving truck, and relatively few drivers are comfortable enough with a partner driver to
get restful sleep. It would likely take a lengthy regulatory process and years of scientic study to
understand whether drivers get proper rest and operate safely while working in partnership with an
autonomous truck. This is particularly true because, unlike a team of two human drivers, where each
is capable of driving the truck and performing non-driving tasks, the partners in a human-
autonomous truck team would be responsible for entirely dierent environments and tasks.
A brief description of a typical load will illustrate the problems of this scenario. The average length
of haul (i.e., distance freight is moved) for truckload dry van carriers, the most likely adopters of
autonomous trucks, is about 500 miles. In order to most protably haul a load this length, a human
driver might do local driving and perform other tasks required to pick up a load, working at least
several hours, including waiting time. Then, the autonomous truck might drive the highway portion
of the load, say 450 miles, requiring nine hours or so. The human driver would take over to drive
locally and unload again. By this point, the driver would not have completed a full 10-hour break, as
currently required by law, and might only have a few hours left to work according to federal rules.
However, in order to make the most of the autonomous truck, the driver would drive locally to
complete the load and then pick up another load. Requiring the driver to wait while the next load
of freight was loaded or unloaded might take somewhere between six and eight hours. Then the
autonomous truck would take over again for interstate driving, and the driver might get nine hours
of rest time.
Many long-haul drivers today work 80 hours per week or more. They are only supposed to work
60 hours per week, but the federal rules intended to prevent such long hours are regularly violated
because of inaccurate self-reporting of non-driving work and waiting time. In this scenario, the
driver would have a much more broken-up schedule than common today and would perform
even more non-driving work and waiting. In combination with truckers’ tendency to work as many
hours as possible because they are paid by the mile or the load and not hourly, partnership with
an autonomous truck raises very important concerns for both safety and the long-term health of
drivers. Truckers might keep that truck rolling almost constantly.
Driver sleeps during
highway driving
DRIVERLESS | Steve Viscelli 26
SECTION THREE: Scenarios for the Use of Autonomous Trucks
In my experience studying long-haul truck drivers, the only drivers who come anywhere close to this
kind of sleep disruption and continual movement are trainers working with inexperienced drivers.
Most of these drivers eectively work as a team with the trainee, splitting up the driving of highway
and local miles according to the trainees’ stamina and abilities and the degree to which the trainer
trusts the trainee to drive while they sleep in the back of the truck. Unlike teams of experienced
drivers, however, trainers are required to observe the trainees under a number of regularly occurring
conditions, such as when the truck is entering or exiting a highway, driving in local conditions, or
backing into a parking spot or dock. In fact, the conditions that trainees generally have trouble
driving in and need supervision for are pretty much the same as those that autonomous trucks
will have trouble with. Trainers often experience extreme fatigue and operate almost continually in
violation of federal hours-of-service regulations, often working in excess of 100 hours per week.
This scenario would likely mimic that of trainer-trainee teams and could result in a tremendous
“speed-up” of drivers’ work lives. This speed-up would result in much cheaper and faster trucking
service that would set a competitive standard in the industry. Competitors would need to follow
suit to survive. It would inevitably lead to pressure on drivers to ignore their bodily needs, including
sleep, in the service of keeping the truck rolling continually, breaking only for loading and unloading.
Many long-haul drivers already endure a brutal combination of working hours and conditions
that take an enormous toll on their social lives and health; this scenario would almost certainly
exacerbate those conditions and their consequences.
With proper regulation, however, this kind of scenario could be benecial for a few segments
of the industry. In fact, this scenario would be most benecial to some of the remaining true
owner-operators in long-haul trucking or for specialized niche companies where drivers are typically
highly paid and work routines are not driven by a race for the next load. For freight requiring
extensive driver oversight during loading and unloading and those that use highly specialized
trailers or tractors, for instance, this model might eventually be successful. Appropriate and desirable
applications might include long-distance hauling of vehicles or heavy equipment.
TECHNOLOGY SCENARIO 5:
Exit-to-Exit Autonomous Trucks
The “exit-to-exit” autonomous truck scenario has received a great deal of attention over the past
year and is viewed by some major players developing Level 4 autonomous trucks as the most
likely way for such trucks to be utilized in the near future. There is a strong economic case for this
scenario that, if supported by robust policy, could result in a wide range of benets for stakeholders
inside and outside the industry in long-haul operations. Without appropriate policy, however, this
scenario raises important concerns about both job loss in long-haul trucking and about job quality,
environmental, and safety issues for local truck operations.
In this scenario, human drivers would haul freight from distribution centers, production facilities, or
other modes of transportation, like ship or rail, to an autonomous truck port, as described above.
There, the human driver would uncouple from—or “drop”—an outward-bound trailer. That trucker
or another worker at the ATP then would supervise as an autonomous tractor couples to that
DRIVERLESS | Steve Viscelli 27
SECTION THREE: Scenarios for the Use of Autonomous Trucks
outbound trailer. That worker potentially could also ensure that the autonomous truck is safe to
drive, fueled, etc. The autonomous truck would then drive the interstate portion of the freight’s trip,
while the human driver could take an inbound trailer from the ATP to nal delivery in the local area.
Like the drone autonomous trucks in Scenario 2, the autonomous tractor in this scenario could be
optimized for highway travel, including gearing, engine size, tires, aerodynamics, and more. This
truck could operate with other autonomous trucks in platoons. Over the highway portion of trips,
this scenario would eliminate labor costs and cut fuel costs substantially. It would also dramatically
increase the overall speed of moving freight as loads would not have to wait while drivers take
mandatory breaks to sleep and stop for other reasons. In so doing, this scenario would have a major
impact on what is now a substantial constraint on freight moving between 500 and 1,000 miles. It
could also bring the trucking industry some much longer distance freight, including many imports
that currently move by rail.
This scenario could be the most likely to play a signicant role in the immediate future. Indeed, at
least two Silicon Valley rms, Uber and Embark, have envisioned such a scenario. Uber modeled a
well-known demonstration with a human driver operating a truck until it got to the interstate, then
operating the truck autonomously, with the human driver taking over again for local driving. Embark
is hauling actual freight using a similar transfer process, where their autonomous trucks pass trailers
on to Ryder trucks with human drivers.
Given the intense focus on this scenario and its emergence as the odds-on favorite for near-term
adoption, its labor implications will be discussed at length below.
Autonomous
Truck Port
Autonomous
Truck Port
Human & autonomous
tractors swap trailers
Cabless tractor optimized
for highway driving
DRIVERLESS | Steve Viscelli 28
SECTION THREE: Scenarios for the Use of Autonomous Trucks
TECHNOLOGY SCENARIO 6:
Facility-to-Facility Autonomous Trucking
Many facilities that ship and receive large amounts of freight are strategically located very close to
major interstates, often within just a few miles. Those few miles of road between facilities and an
interstate exit are often in industrial or commercial areas, where roadways are designed for heavy
truck trac and largely used for commercial purposes. In some cases, these roadways are more
suitable to autonomous truck operation than interstate highways. Like interstates, there is often no
side parking, pedestrians, or bikes. They have fewer complex and congested intersections. Unlike
highways, they have relatively slow speeds and are far less congested than some interstates. In the
“Facility-to-Facility” scenario, autonomous trucks drive directly between facilities like these without a
human driver required at any time. Workers at each facility would handle any required non-driving
tasks, such as coupling trailers, fueling tractors, and inspections.
This scenario presents a signicantly dierent set of impacts from that of exit-to-exit autonomous
trucks for a number of reasons. It would allow autonomous trucks to replace both for-hire and
private drivers working on regular routes between facilities, including linehaul drivers within
less-than-truckload or parcel systems. These drivers are typically far more skilled and experienced,
and therefore more costly.
Port drivers could be similarly impacted under this scenario. If autonomous trucks result in
signicant cost savings, as expected, they will capture some long-distance container freight that
currently goes from ship to rail (sometimes with a short truck trip in between). This scenario would
lead to containers coming right o ships and onto autonomous trucks going longer distances
of 500+ miles and directly to destinations. This scenario could be highly disruptive to important
existing warehousing districts that currently process imports, such California’s Inland Empire.
These potential impacts will be discussed in detail below, but this scenario would have by far the
largest negative impact on jobs, essentially eliminating driving jobs in some segments. It would also
not have all of the positive environmental benets of segmenting into local and long-haul trucks for
better fuel economy and congestion management (though, in theory, these trucks could operate
around the clock and more easily avoid rush hours than human drivers).
PARCEL CO
Autonomous trucks navigate
commercial streets near highway
Tractor not optimized for highway
since it also drives locally
DRIVERLESS | Steve Viscelli 29
SECTION THREE: Scenarios for the Use of Autonomous Trucks
Which Scenarios Are Most Likely and Desirable?
This report argues that we should not try to predict the outcome of the development of
autonomous trucks, but rather shape it. Still, in trying to shape the inevitable use of autonomous
trucks, it is important to understand which scenarios have more or less dicult paths and benecial
eects. Below is a brief summary of the challenges and desirability of each scenario.
TECHNOLOGY SCENARIO 1: Cooperative Adaptive Cruise-Control Platooning
The technology required for platooning is already available, and the case is being made for adoption
to major eets. This scenario could be a reality in a matter of months. At present, policymakers seem
receptive to eorts to encourage adoption, and there appear to be few drawbacks.
TECHNOLOGY SCENARIO 2: Human–Drone Platooning
This scenario has relatively few technological hurdles and would result in major productivity gains. In
addition, the case for human–drone platooning is very strong from an environmental and job-quality
perspective. However, without public policy to ensure otherwise, it would result in the loss of some
good long-haul jobs and create local jobs that will likely oer poor pay and poor working conditions.
Perhaps because it is the least technologically ambitious, this scenario has not been a major focus of
Silicon Valley, which is unfortunate as human–drone platooning may represent an eective outcome
that many stakeholders could support.
TECHNOLOGY SCENARIO 3: Exit-to-Exit Autonomous Plus Drone Operation
This scenario may present the most dicult technological challenges: it not only calls for
autonomous driving on the highway, but also requires the technology to pilot remotely, which could
be costly given potentially expensive control centers and communications systems, as well as the
need to replace non-driving driver-performed tasks throughout the labor process. This scenario
should be treated as the least likely.
TECHNOLOGY SCENARIO 4: Driver-in-the-Sleeper Scenario (A.K.A. Autopilot)
This scenario would require developing the same basic technology for autonomous driving as the
exit-to-exit scenario. The environmental benets of the “autopilot” scenario would be minimal, but
the potential labor impacts could be signicant and negative. In the near term, this scenario should
be treated as highly undesirable.
TECHNOLOGY SCENARIO 5: Exit-to-Exit Autonomous Trucks
Exit-to-exit is the scenario that most Silicon Valley developers are putting forward as the most likely.
It requires overcoming signicantly more dicult technological challenges. However, since it is the
aim of key developers and easier than Scenarios 3 and 4, it should be treated as a likely outcome.
Exit-to-exit autonomous trucks would result in more job loss than human–drone platooning and
would create a similar number of local jobs (which, again, would likely have low wages and poor
conditions without policy intervention).
DRIVERLESS | Steve Viscelli 30
SECTION THREE: Scenarios for the Use of Autonomous Trucks
TECHNOLOGY SCENARIO 6: Facility-to-Facility Autonomous Trucking
This scenario requires technology to solve the problem of local driving, which may be possible for
some limited locations, especially within linehaul LTL and parcel operations. For those use cases, this
scenario should be treated as likely, following the successful development of exit-to-exit self-driving.
How Soon Could Autonomous Trucks Be Used?
Everyone wants to know when self-driving vehicles will be ready for widespread adoption. Several
car manufacturers are forecasting that they will have self-driving cars in the next two or three
years, but exactly what these cars will be capable of is unclear. When talking with experts and
developers, I found a range of opinions for trucks. Essentially, the most ambitious timeline suggests
that autonomous trucks could be operating in the highway portion of the long-haul duty cycle in
a few stretches of highway within three years. This general timeframe ts the prediction of experts
surveyed on the subject, who estimate a high likelihood of Level 4 operation on highways beginning
between 2018 and 2024.
29
In the opinion of the vast majority of those I talked to, reliable and safe
local driving is still decades away.
Again, a number of important technological and cost barriers need to be overcome to achieve
these outcomes. I consider the timeline below to be the most aggressive that can realistically be
envisioned for self-driving exit-to-exit and limited facility-to-facility autonomous operation.
Here’s what the stages might look like:
STAGE 1: Now to three years
Pilot usage of autonomous trucks begins. Waymo and Embark are currently using or planning
to use trucks with self-driving technology for interstate hauling of freight, and Starsky says it will
begin doing so by the end of 2018. Right now, human drivers are present behind the wheel. These
trucks are paired with traditional human-driven trucks and swap trailers with those trucks for local
driving, similar to what is described in the exit-to-exit autonomous scenario described above, using
something like an autonomous truck port. These trucks might be operational without a driver in
pilot programs, perhaps with time-of-day restrictions on highway segments, within three years.
Within this time, digitized freight matching begins to have a meaningful impact on the brokerage
business for local and regional loads. E-commerce continues its explosive growth and fosters the
growth of last-mile delivery jobs.
STAGE 2: Three to seven years
A few advanced for-hire eets begin ATP-to-ATP autonomous truck programs in partnership with
technology and truck leasing rms on I-10 in the U.S. Southwest, building on existing testing
regimes. These trucks operate with route and time restrictions. Sophisticated supply chain
actors begin to plan for integration of autonomous trucks’ greater speed and lower cost in the
southwestern and southern United States, south of I-40. Digitized freight matching for local and
regional truckloads spreads widely, and long-haul freight begins to be aected. Strong growth in
last-mile delivery continues.
DRIVERLESS | Steve Viscelli 31
SECTION THREE: Scenarios for the Use of Autonomous Trucks
STAGE 3: Seven to ten years
Autonomous truck operation between ATPs begins during more congested times on some routes
and on additional major freight lanes on I-10 and I-40 and sections of some north-south interstates
south of I-40. Total autonomous truck numbers climb to several thousand. Long-haul truckload
for-hire rates decline on lanes where autonomous trucks operate. Small and mid-size carriers
attempt to compete by lowering wages on those lanes or take losses for backhauls on them.
Planning and investment for autonomous truck adoption by the largest for-hire and private eets is
widespread. LTL eets begin planning for adoption or consider subcontracting for-hire autonomous
service for linehaul operation between ATPs in limited areas. Strong e-commerce growth in existing
and new product categories continues to support strong LTL demand and moderate truckload
demand growth. The most sophisticated supply chains continue reorganizing to meet “right now
and free” shipping demand in combination with remaining brick and mortar. Digitized freight
matching becomes dominant for local and regional truckload market transactions. Pilot projects
begin to demonstrate autonomous truck feasibility in climates with regular snow and facility-to-
facility operations.
STAGE 4: Ten to fteen years
Autonomous truck operation has been established as safe and reliable in most interstate conditions
and begins to spread to key freight lanes nationally on a seasonal basis, creating intense competitive
pressure and market volatility. Adoption of autonomous trucks or labor cost cuts become critical for
survival of medium-to-large dry van and refrigerated truckload carriers in some areas. Signicant
concentration of truckload carriers begins as the largest eets successfully transition to automated
eets for dry van and refrigerated truckload shipments. A few private eets begin to shift to cheaper,
faster for-hire autonomous truck service, including customer-directed, dedicated autonomous trucks.
LTL eets begin adopting autonomous trucks for facility-to-facility linehaul. Strong growth in local
jobs for nal delivery of truckload trailers, growth in LTL and other last-mile delivery jobs continues.
Digitized freight matching is now dominant throughout for-hire truckload.
STAGE 5: Fifteen to twenty-ve years
Low-cost autonomous truck service largely replaces long-haul truckload in dry van and refrigerated
segments with signicantly cheaper, faster services. Some private linehaul truckload eets shift to
dedicated, for-hire autonomous trucking carriers as that segment provides faster service that has
the dependability of in-house services at much lower cost. Linehaul LTL operations are now done by
autonomous trucks. Several hundred thousand new jobs have been created in last-mile delivery jobs
as well as local trucking jobs. Human drivers still drive short-haul routes and specialized equipment.
DRIVERLESS | Steve Viscelli 32
In order to understand the potential impact of self-driving technology, we need to know what
kind of trucking companies will adopt it and how they will use it. In light of the above analysis, this
section assumes that exit-to-exit autonomous trucks are the likely scenario and that limited
facility-to-facility use will follow.
Ultimately, three sets of factors will determine who adopts autonomous trucks using those scenarios:
1. Can the truck perform the driving tasks in the required environments? In the exit-to-exit
scenario, the truck will drive long distances on the highway. In facility-to-facility, the truck
will also drive limited distances on commercial roads.
2. Is it protable to segment out that driving from the other tasks drivers perform? Is
there enough long-distance driving to justify the swapping of trailers at an ATP in the
exit-to-exit scenario? Are customer sta available to open trailer doors and inspect the
truck in the facility-to-facility scenario? Can would-be adopters raise sucient capital
to buy driverless trucks and build in-house expertise or hire expertise to operate them
protably?
3. Can other impediments to adoption, like risk to brands from accidents and concerns
raised by drivers, be resolved?
If the answer to all of these questions is “yes,” then rms in that segment are likely to adopt the
technology. Table 4.1 (page 33) breaks these factors down by characteristics of the freight, carriers,
and customers in dierent segments of the trucking industry. Green indicates characteristics that
make a stronger case for adoption. Yellow indicates a characteristic that weakens the case for
adoption. Red indicates a characteristic that will impede adoption. Other characteristics might also
aect adoption, but these are major obstacles identied in my conversations and research.
Clearly, the strongest case for adoption is in for-hire long-haul truckload. The next strongest is
linehaul service within LTL and parcel operations. It is unlikely that other segments will be easily
transformed to adopt autonomous trucks, with one exception. Port hauling could, with the entrance
SECTION FOUR:
Estimating Job Losses and
Likely Job Impacts
DRIVERLESS | Steve Viscelli 33
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
of larger rms using autonomous trucks, capture a signicant amount of freight currently shipped
long distances by rail. However, a substantial shift would be required in rm strategy, which
currently relies heavily on cheap trucks paid for by independent contractors.
How Many Jobs Are At Risk of Automation?
If the analysis above is indicative of the segments where autonomous truck adoption is possible and
protable, how many jobs are at risk?
Answering this question is complicated, because there are no datasets on freight that would allow
us to look at all the dimensions in Table 4.1 with any precision. However, by combining dierent
datasets, we can calculate the total revenue at large rms (the most likely to automate) in each
Primary driving
environments
Uninterrupted
highway
driving
Non-driving
tasks
Customer
facility Type
Route
regularity Union presence
Typical carrier
size
For-hire truckload
(dry and refrigerated)
Highway Extensive Minimal
Large
Warehouse
Moderate None
Medium/
Very Large
Less-than-truckload
and parcel linehaul
Highway Extensive Minimal
Internal
Terminal
High Moderate
Medium/
Very Large
Port driving
Urban/
Highway
Minimal/
Signicant
Minimal
Large
Warehouse
Moderate to
High
Low Small
Specialized truckload
Highway/
Complex
Signicant/
Extensive
Signicant/
Extensive
Varied
Commercial
Varied Low
Small/
Medium
Intermodal
Urban Minimal Minimal
Large
Warehouse
Moderate Varied
Small/
Medium
Local pickup and
delivery (part of
LTL and parcel)
Local/Urban Minimal
Signicant/
Extensive
Commercial or
Residential
Varied Varied Very Large
Local — other
Local/Urban Minimal
Signicant/
Extensive
Varied Varied Varied Small
TABLE 4.1
Characteristics of loads, rms, and customers relevant to autonomous truck adoption by industry
segments
Strengthens case for adoption
Weakens case for adoption
Obstacle to adoption
DRIVERLESS | Steve Viscelli 34
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
segment, the average revenue per driver, and thus, the estimated number of drivers in each
segment. As shown in Table 4.2, these data suggest that far fewer jobs—an estimated 294,000—are
at risk than other studies have suggested.
JOBS AT RISK BY SEGMENT: FOR
-
HIRE TRUCKLOAD
If the analyses in Tables 4.1 and 4.2 are correct, the rms most likely to adopt autonomous trucks
are dry van and refrigerated truckload carriers, particularly the largest rms, which dominate these
segments. Data collected by Transport Topics and Commercial Carrier Journal (CCJ), two leading
industry publications, allow us to get a rough estimate of the number of jobs. Both publications
collect annual data on all the largest for-hire and private rms and report some combination of the
number of trucks they operate, the drivers they employ or contract with, and the revenues they earn
in dierent segments.
Segment
Total segment
revenue of large
rms (A)
Estimated revenue
per driver job at
large rms (B)
Approximate
estimated jobs
at risk (A/B)
Average annual
driver earnings
For-hire truckload
dry van
$30.5 billion $174,000 175,000 $46,641*
For-hire truckload
refrigerated
$7.6 billion $209,000 36,000 $53,690*
Less-than-truckload
linehaul
$33.7 billion $266,000 51,000 $69,208*
Parcel linehaul
— —
32,000 $59,660**
Total jobs at risk
of automation
294,000
TABLE 4.2:
Jobs in segments at high risk of automation
Note: LTL linehaul estimate is based on 127,000 total drivers x .4 for percent linehaul. Calculating revenue per driver
isn’t possible for parcel linehaul, so a different approach was used. See below for more details on these calculations.
For parcel job numbers, see text for information on reasoning. Columns 2, 3, and 4 are author's calculations, see Data
Appendix.
* Costello, Bob. 2014. ATA Compensation Survey 2014. Arlington, VA: American Trucking Associations.
** BLS, Occupational Employment and Wages, May 2017, 53-3032 Heavy and Tractor-Trailer Truck Drivers,
https://www.bls.gov/oes/current/oes533032.htm.
DRIVERLESS | Steve Viscelli 35
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
Transport Topics publishes a list of the 100 largest for-hire eets and the 100 largest private eets.
The top of the list on the for-hire side is dominated by the parcel giants UPS and FedEx and the
very largest truckload and LTL eets. In total, the eets in the top 100 generated more than $232
billion in revenues in 2017 and controlled around 476,000 trucks, though slightly more than half of
that revenue was generated by FedEx and UPS alone. CCJ publishes an annual list of the largest 250
carriers, both for-hire and private. Unfortunately, the CCJ Top 250 doesn’t have revenue information
for many companies, but it does provide driver numbers for employees and contractors for most
eets. In total, the 250 eets on the list use nearly 761,000 drivers.
Transport Topics also publishes lists of the largest carriers and their revenue in 11 dierent industry
segments, including dry van and refrigerated. Unfortunately, while truck counts are provided for
these companies overall, they are not broken down by particular segments.
Bringing these sources together to get revenue counts by segment from Transport Topics and
drivers from CCJ gives us a good picture of the numbers of driving jobs at risk in the near future. By
calculating an estimated average “revenue per driver” in each segment using a selection of rms
that only haul freight in one segment, then dividing total revenue in a segment by the estimated
average revenue per driver in that segment, we arrive at the number of jobs for rms that haul
in multiple segments. The Data Appendix shows the carrier data used to calculate the estimated
average revenue per driver in each segment.
Jobs at risk in for-hire dry van
Transport Topics lists 86 carriers as having dry van truckload revenue. At the top of the heap in terms
of revenue for the segment is Swift Transportation with more than $3 billion from dry van truckload.
Last on the list is Bolt Express with slightly more than $14.5 million dollars in revenue from dry van
truckload. While economists conclude that trucking markets are not particularly concentrated, there
is clearly a massive dierence between the big carriers and the rest: the largest carrier has more than
200 times the revenue of the last on the list. It will undoubtedly be more dicult for carriers with
less revenue to transition to autonomous trucks. Perhaps none of the 38 carriers on the list with less
than $100 million in revenue will be able to make this change and may eventually go out of business
or be bought by larger rms as the segment consolidates.
In total, the 86 rms on the Transport Topics list generated about $30.5 billion dollars in revenue
in truckload dry van. These rms are primarily regional or long-haul rms. Using the revenues and
driver counts for major dry van truckload carriers (see the Data Appendix for the data used for this
calculation), I estimate that large truckload dry van eets average $174,000 per driver in revenue
in this segment. Dividing the total revenue of the 86 largest eets in the dry van segment by this
estimate of per driver revenue suggests that these rms use approximately 175,000 drivers (both
employee and contractor) to haul dry van freight.
Jobs at risk in for-hire refrigerated
The refrigerated segment is much smaller than the dry van truckload segment. The 32 rms on
the Transport Topics list of the largest carriers in the segment brought in around $7.6 billion. At
DRIVERLESS | Steve Viscelli 36
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
the top of the list was C.R. England, with $1.3 billion in revenue from refrigerated. Last on the
list was Celadon Group, which brought in $30 million in refrigerated revenue. Using the gure of
$209,000 per driver in revenue,
30
we can estimate that these rms used some 37,300 drivers in the
refrigerated segment.
JOBS AT RISK BY SEGMENT: LESS
-
THAN
-
TRUCKLOAD AND
PARCEL LINEHAUL
The other groups of drivers most at risk are linehaul drivers in parcel and less-than-truckload
operations who transport full trailers from one carrier or private facility to another controlled by the
same company.
31
Once autonomous trucks are capable of driving directly to and from terminals
close to interstate exits, they will likely take over the long-haul portions of LTL and parcel service
done by these linehaul drivers.
Jobs at risk in LTL linehaul
LTL is far more concentrated than truckload. LTL generates about $33.7 billion for the 27 eets on
the Transport Topics list of LTL hauling rms, ranging from a high of $6.35 billion for FedEx to a low
of just $53 million for Anderson Trucking. Revenue and driver counts for the CCJ Top 250 suggest
that LTL rms generated $266,000 per driver in revenue. The total revenue reported for the segment
by Transport Topics suggests these rms used about 127,000 drivers to haul LTL freight. Based on
my conversations with drivers and other stakeholders, there is a fairly wide range of the percent
of drivers within LTL operations who drive linehaul. Some large LTL carriers may employ up to 45
percent of drivers in linehaul, the rest of the drivers would do primarily local pickup and delivery
work, which is less likely to be automated. Other LTL rms may only use 20–25 percent of their
drivers in linehaul work. To be conservative for a rough estimate, I used an overall average of 40
percent of drivers in linehaul,
32
which suggests that about 51,000 linehaul drivers are at risk.
Jobs at risk in parcel linehaul
Like LTL carriers, parcel carriers may move to autonomous trucks to replace linehaul jobs. UPS
controlled a total of 31,808 tractors in 2017, according to the CCJ Top 250. UPS’s almost $2.74 billion
in revenue from LTL suggest that it utilizes roughly 8,850 tractors in that segment. If UPS uses the
remainder of its tractors for linehaul parcel operations, some 22,958 tractors could be automated.
FedEx controlled 29,426 tractors in 2017, according to the CCJ Top 250. If the estimates of LTL jobs
above are correct, then FedEx uses approximately 20,500 tractors in LTL service. The remaining 9,000
tractors may also perform linehaul for parcel service, but they might instead be used over relatively
short distances to bring freight from truck terminals to airports. In this case, they might not be good
candidates for automation.
In total, perhaps as many as 32,000 trucks could be automated among parcel carriers.
33
I cannot
calculate revenue per jobs as I did above because I don’t have “pure” parcel carriers to use as a
baseline for revenue. In any case, the operations of UPS and FedEx are quite dierent from the jobs
described above, so a single per driver revenue calculation doesn’t make sense.
DRIVERLESS | Steve Viscelli 37
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
The Quality of At-Risk Jobs
The analysis above suggests that around 294,000 existing jobs might be at risk of automation. Even
among these long-haul jobs, which are primarily composed of driving between large facilities for
large rms, there are dramatic dierences in job quality, pay, and driver characteristics. On one hand,
for-hire truckload jobs are very tough jobs that have very high turnover, and the largest companies
in this segment rely on a very inexperienced labor force. Within this workforce, however, it’s likely
that there will also be a signicant number of long-term employees and contractors who are very
experienced and earn good incomes. Some of these workers might be particularly hard hit by
automation because they live in rural areas, where other well-paid jobs are dicult to nd.
On the other hand, the labor market for LTL and parcel carriers is very dierent. These jobs still have
a signicant union presence, with the International Brotherhood of Teamsters representing workers
at both the largest LTL rms and UPS. Overall, the jobs at risk in these segments are done by some
of the most experienced and best compensated drivers in the industry. While these workers are
also older and closer to retirement, these are very good jobs, and it is highly unlikely that the local
and last-mile jobs that might replace them would be nearly as good, unless signicant policy steps
ensure that outcome.
FOR
-
HIRE TRUCKLOAD: HIGH TURNOVER AND USE OF
INDEPENDENT CONTRACTORS
If the above analysis is correct, about 211,000 for-hire truckload jobs (both dry van and refrigerated)
are at risk of being automated. Much of this segment is “perfectly competitive,” meaning individual
rms have very little pricing power. If they increase their rates, they will quickly lose customers. As a
result, wages in this segment tend to stagnate until one or more larger rms, which can temporarily
raise wages without pricing themselves out of the market, decide to grow their eets by raising
wages and then are followed by the rest of the segment.
34
As shown in Figure 4.1 (page 38), driver
pay in for-hire truckload is thus signicantly lower than in other long-haul segments.
These stagnant wages are compounded by tough working conditions. Drivers typically work the
equivalent of more than two full-time jobs and are required to stay out on the road for weeks at a
time.
35
This combination of low pay and dicult conditions means that the segment struggles to retain
drivers. As Figure 4.2 (page 38) shows, truckload carriers suer from very high turnover rates, with
the worst employers having more than 100-percent turnover—meaning they cycle through more
than one worker per position every year. Conditions at many large truckload carriers (those most
likely to automate) are so bad that they have likely caused several million workers to exit the
segment soon after entering it in recent decades.
For-hire truckload companies thus rely heavily on training workers entirely new to the industry and
retain those workers by indebting them for a year or more for that training. Some companies even
use non-compete clauses to keep newly trained workers from moving to other rms.
DRIVERLESS | Steve Viscelli 38
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
FIGURE 4.1
Average annual earning in different segments of the trucking industry
Segment Average annual wage
For-hire truckload * $46,641 – $53,690
Less-than-truckload * $69,208
Parcel ** $59,660
Ports ***
$28,783 (contractors)
$35,000 (employees)
Pickup and delivery **** $35,610
LONG-HAUL
LOCAL
* Costello, Bob. 2014. ATA Compensation Survey 2014. Arlington, VA: American Trucking Associations.
** BLS, Occupational Employment and Wages, May 2017, 53-3032 Heavy and Tractor-Trailer Truck Drivers,
https://www.bls.gov/oes/current/oes533032.htm.
*** Smith, Rebecca, Paul Alexander Marvy and Jon Zerolnick. February 2014.The Big Rig Overhaul: Restoring Middle-class
Jobs at America’s Ports Through Labor Law Enforcement. National Employment Law Project, Change to Win Strategic
Organizing Center and Los Angeles Alliance for a New Economy.
**** BLS, Occupational Employment and Wages, May 2017, 53-3033 Light Truck or Delivery Services Drivers,
https://www.bls.gov/oes/current/oes533033.htm.
FIGURE 4.2
Annual driver turnover in different industry segments
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Large truckload * Small truckload * Less-than-truckload * Private **
95%
84%
7%
8%
Third quarter annualized turnover rates
* Source: American Trucking Associations. December 7th, 2017. “Large Truckload Driver Turnover Rate Rose in Third
Quarter.” Arlington, VA: American Trucking Associations.
** Private Fleet Turnover rate is for the year 2014. Source: Costello, Bob. 2014. ATA Compensation Survey 2014.
Arlington, VA: American Trucking Associations.
DRIVERLESS | Steve Viscelli 39
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
Table 4.3 represents the best data available on what new hires and their tenure look like at a large
truckload rm—the type most likely to adopt driverless trucks. The researchers studied tenure for
more than 5,000 workers hired by this rm and found that:
• 90 percent of the hires were inexperienced;
• 73 percent were trained by the company itself in one of its commercial driver’s license
schools;
• About half of all workers hired by the company had left within half a year. Many of those
who stayed longer held out until just after a year, when the debt they owed the company
for training was forgiven. In fact, the researchers suggest that without these training
contracts, workers would leave faster and the company would no longer be protable.
36
These large rms generally have much less stringent hiring standards and are the easiest places
for workers new to the industry to get their rst job. After a year, opportunities open up at better
employers, both in the segment and outside it. But the majority of would-be truckers likely never
make it past the rst year. Tens of thousands of workers annually, perhaps more than 100,000 in
some years, train to become truckload drivers but don’t stay in the profession for even a year.
37
Worker
experience level
Percent of all
drivers hired
Half of drivers
are gone after
Three-quarters of
drivers are gone after
All drivers
(N>5000)
100 27.4 weeks 72.1 weeks
Experienced
drivers
Rehires 4 284.7 weeks
—
*
Experienced 8 29.4 weeks 98.3 weeks
Inexperienced
drivers
Company trained 73 30.1 weeks 73.1 weeks
Prior training 14 18.1 weeks 49.1 weeks
Limited
experience
3 21.1 weeks 53.1 weeks
TABLE 4.3:
Estimated job tenure for drivers hired by large truckload rm
Source: Burks, S.V., J. Carpenter, L. Gotte, K. Monaco, K. Porter, and A. Rustichini. 2007. “Using Behavioral
Economics Experiments at a Large Motor Carrier: The Context and Design of the Truckers and Turnover Project.”
IZA Discussion Paper No. 2789 (May). Bonn, Germany: IZA. The authors state, “76.4 percent are voluntary quits,
while 23.6 percent are discharges for cause.”
* Rehire retention was so long it could not be calculated with the data Burks et al. (2007) collected.
DRIVERLESS | Steve Viscelli 40
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
While there are about 3 million trucks on the road that require a commercial driver’s license (CDL),
there are more than 10 million CDL holders in the United States.
38
A signicant number of them
likely work in non-truck driving jobs that require a CDL, but it’s also very likely that several million
CDL holders are workers who have had the misfortune to pass through for-hire long-haul trucking’s
revolving door.
Instead of raising wages and retaining workers, it has been more protable for large rms to lure
workers into the industry with false promises of high wages and externalize training costs to the
workers themselves and government training grants.
To keep attracting workers, the American Trucking Associations routinely publishes studies about
the industry’s “driver shortage,” which are based employer estimates of the number of workers they
say they would hire. The ATA has been forecasting a shortfall of tens of thousands of truck drivers
almost continually since at least 2005. According to its analysis, the shortage “skyrocketed” to 50,000
drivers in 2017. These studies are dutifully taken up by major media, with headline hooks like CNN’s
“Truck drivers wanted. Pay: $73,000.”
39
And more workers, unfortunately, swallow those headlines.
But once on the road, the troubles begin. As I documented in my 2016 book, The Big Rig: Trucking
and the Decline of the American Dream, as inexperienced drivers become dissatised with the pay
and long hours of the segment, carriers convince many of them to lease a tractor and become an
independent contractor.
40
This situation typically leaves drivers far worse o, working more hours,
yet often taking home less than minimum wage once all the costs of leasing a truck and paying for
its operating expenses are deducted by the carrier.
Convincing drivers to lease a tractor enables carriers to retain workers for months longer. While
these practices, including fraud and the misclassication of employees as independent contractors,
are now being challenged by a number of lawsuits (with one headed to the Supreme Court this year)
these abuses remain widespread.
Many of the jobs in for-hire truckload that might be replaced by autonomous trucks are not great
jobs, yet in 2017, employees in dry van truckload earned an average of $46,641, which is better
than what many could earn elsewhere with only a high school degree. And there are likely tens of
thousands of drivers in this segment who earn above-average wages, have signicant experience,
and would like to make a career out of trucking. Some of these workers might transition into better
segments of the industry as drivers in those segments retire.
Unfortunately, however, many of the drivers who have remained in truckload despite having more
experience live in rural areas, and better employment options are far away. Others may work for
small- and medium-size employers who will be less equipped to adopt driverless trucks. Just as
carriers did in response to renewed competition after the trucking industry was deregulated, these
smaller carriers will be forced to cut wages, and many of the abuses already widespread in the
segment will intensify.
DRIVERLESS | Steve Viscelli 41
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
LTL AND PARCEL LINEHAUL: SOME OF THE BEST JOBS IN
TRUCKING
Around 83,000 linehaul jobs are at risk of automation at LTL and parcel carriers. Some of these are
among the best trucking jobs, with high wages and good working conditions. As the tenure data in
Figure 4.3 suggest, these good jobs mean linehaul drivers are much more likely to have long-term
careers in trucking.
LTL drivers tend to be older than the average trucker and much older than the average U.S. worker,
as Figure 4.4 (page 42) shows. Therefore, many existing drivers will be retired or retiring by the time
autonomous trucks might be commonplace, especially since LTL drivers often retire several years
earlier than drivers in other segments.
Some of the remaining linehaul drivers could perhaps transition over to local pickup and delivery
(P&D) positions that are less likely to be automated, but a big pay cut would be likely.
41
As
illustrated in Figure 4.5 (page 42), the average linehaul driver makes around $9,600 more than a
comparable P&D driver. Moreover, Amazon and other large rms (most notably XPO Logistics) are
increasingly using independent contractors for last-mile delivery. This shift could put signicant
downward pressure on local P&D driver wages.
It is important to note, however, that linehaul drivers at UPS and some of the largest LTL rms are
unionized and will have a greater voice in how to make the transition to autonomous trucks.
FIGURE 4.3
Average driver tenure by segment
2.8
7.9
12.8
0
2
4
6
8
10
12
14
Truckload Less-than-truckload Private
Years with employer
* Source: American Trucking Associations. 2012. Benchmarking Guide for Driver Recruitment & Retention. American
Trucking Associations: Arlington, VA.
DRIVERLESS | Steve Viscelli 42
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
FIGURE 4.4
Median worker age by industry segment
57
49
46
47
42
0
10
20
30
40
50
60
Private Less-than-truckload Truckload Port/intermodal All U.S. workers*
Worker age
Source: Platner, Ryan, and Bob Costello. March 2018. ATA Driver Compensation Study: Operations Data 2017. Arlington, VA: American
Trucking Associations.
* Estimated Age. Source: USDOL Bureau of Labor Statistics. January 19, 2018. Retrieved from https://www.bls.gov/cps/cpsaat18b.htm.
$69,037
$68,171
$69,037
$70,587
$60,680
$56,673
$60,680
$60,391
$0
$10,000
$20,000
$30,000
$40,000
$50,000
$60,000
$70,000
$80,000
SOUTHERN MIDWEST NORTHEAST NORTH CENTRAL
Annual earnings
Linehaul Linehaul Linehaul LinehaulP&D
P&D
P&D P&D
* Source: Costello, Bob. 2014. ATA Compensation Survey 2014. Arlington, VA: American Trucking Associations.
FIGURE 4.5
Linehaul vs pickup-and-delivery (P&D) driver earnings, by region
DRIVERLESS | Steve Viscelli 43
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
WHO ARE THE DRIVERS?
Unfortunately, government and other regularly collected data sources do not allow us to isolate the
drivers who are most at risk by segment. The best source for a demographic portrait of these drivers
comes from a nationally representative survey of 1,265 drivers collected at 32 truck stops in 2010.
42
The survey was limited to drivers who had driven a truck for a year or more and those who took at
least one 10-hour rest period on the road (as required by federal regulations for long-haul drivers)
on each delivery run. As a result, the drivers in the survey are those who drive long distances and
aren’t really new to the industry. This limitation is signicant for thinking about the future of trucking
jobs and the workers who will be aected; evidence suggests workers new to the industry are more
likely to be people of color and immigrants, whereas white workers composed 74 percent of the
sample in this survey.
However, if we want a sense of the workers currently making a career in the jobs that are at risk,
the 2010 survey serves that purpose. Weighted national estimates based on the survey suggest
this population has an average of more than 16 years of experience as long-haul truckers, that
35 percent are working as owner-operators, and that 90 percent work at for-hire carriers. These
estimates suggest that 75 percent of these drivers are hauling truckload freight and just 4 percent
are paid a salary or by the hour, meaning that the vast majority are pieceworkers paid by the mile or
a percent of revenue per load. In addition, 94 percent of drivers in the weighted survey were male,
and almost half were 50 or older.
The Quality of New Driving Jobs Created
While as many as 294,000 jobs could be lost to autonomous trucks, given the aging workforce,
growing demand for trucking services as costs decline, and other factors, there will likely be enough
jobs to accommodate the displaced drivers. In fact, with the growth of e-commerce and the need
for local drivers to shuttle freight to and from ATPs, we could see many more local trucking and
delivery jobs.
The critical question is: What will the quality of these jobs be? The discussion above illustrates that
wages and working conditions for truck drivers will likely deteriorate as automation reduces the
number of long-distance trucking jobs. Local and for-hire trucking jobs have always been the most
competitive and least paid, unless they are unionized. We should be deeply concerned about labor
conditions in these new jobs.
There is every reason to believe that the jobs created around autonomous truck ports will resemble
port driving jobs. Just as port drivers move goods to and from ships, these drivers will move freight
between customers and ATPs. The good news is that we know as much, arguably more, about port
drivers and their working conditions than almost any other kind of truck driver, thanks to a number
of high-quality research eorts.
43
The bad, albeit unsurprising, news is that these are among the worst trucking jobs around. For-hire
truckload suers from destructive competition, and local-hauling has even lower barriers to entry
DRIVERLESS | Steve Viscelli 44
SECTION FOUR: Estimating Job Losses and Likely Job Impacts
as trucks don’t go as far and don’t need to be as reliable. As a result, trucking carriers keep wages
as low as possible and shift as much of the cost of ineciency and risk of downtime and capital
investment to workers as possible. To these ends, independent contractors are used—and to an
even greater extent than in long-haul trucking.
It’s estimated that there are some 75,000 port truckers in the United States. Surveys of more than
2,000 such drivers in seven major studies indicate that more than 80 percent of those drivers
work as independent contractors, and the vast majority (perhaps 90 percent) of those workers
are misclassied and should be employees. The surveyed drivers work very long hours, averaging
59 hours per week. Employees earned $35,000 and independent contractors just $28,783, before
taxes. Not surprisingly, independent contractors were much less likely to have health insurance or
retirement benets.
44
Moreover, most trucks used in port driving are old and polluting, with emissions exacerbated by
the time drivers spend waiting and idling their engines (the prevalence of independent contractors
means workers aren’t paid hourly, so there’s little incentive for rms to operate eciently). This
situation creates serious environmental justice issues in surrounding neighborhoods (typically
lower-income communities of color), which suer from high asthma rates and other health impacts.
If a similar industry model takes root for ATP driving, automation could thus replace some of the
best trucking jobs with more of the worst. This forecast raises a host of serious concerns, from the
health impacts described above to the degradation of some of the few good jobs remaining to
workers without a college degree.
DRIVERLESS | Steve Viscelli 45
In the coming decades, the way we move freight will change, in ways big and small. What that
means for communities, workers, and an evolving trucking industry will be shaped not just by the
latest technical innovations, but also by the response of governments, businesses, and workers
across the sector. How technology changes truck driving is not an inevitability, and the action or
inaction of policymakers will be key in determining which technologies make their way onto our
public roadways, who benets from this innovation, and who may be left behind.
Truck driving will be one of the rst major occupations transformed by the coming wave of
technological change, but policymakers have a chance to get ahead of these issues. Eective
public policy can ensure that trucking evolves into a productive, high-road industry. Policymakers,
collaborating with workers and industry leaders, have an opportunity to tackle some of our biggest
challenges: creating good, family-supporting jobs; improving roadway safety; reducing trac
congestion; and reducing greenhouse gas emissions.
Below, I outline three key areas of policy solutions:
1. Develop an industry-wide approach to worker advancement and stability;
2. Ensure strong labor standards and worker protections; and
3. Promote innovation that achieves social, economic, and environmental goals.
Working together with industry stakeholders, policymakers can help to ensure that the benets of
innovation in the trucking industry are shared broadly between technology companies, trucking
companies, drivers, and communities.
1. Develop an Industry-Wide Approach to Worker
Advancement and Stability
While cataclysmic loss of truck-driving jobs is not imminent, big changes in the industry will require
many workers to adjust the course of their careers. Given the signicant number of workers aected,
SECTION FIVE:
Policies for a 21
st
-Century
Trucking Industry
DRIVERLESS | Steve Viscelli 46
SECTION FIVE: Policies for a 21st-Century Trucking Industry
the diverse nature of rms, and the varied impact of automated trucks across industry segments,
supporting workers during this adjustment process will require an industry-wide strategy that brings
business, labor, and public-sector resources and perspectives to bear.
Create a Trucking Innovation and Jobs Council
Policymakers should create a Trucking Innovation and Jobs Council, bringing together diverse
stakeholders across the sector—workers, employers, technologists, and policymakers—to create
an action plan to develop a 21st-century trucking workforce and provide the necessary nancial
security to workers through this transition. The Council would develop and implement an action
plan for how industry stakeholders would fund, design, and carry out policies and programs to
accomplish two goals: 1) the development of good career pathways for trucking workers; and 2)
the provision of direct nancial safety net resources to support job transitions within and out of
the industry. While such a Council may function best at the national level, in the absence of federal
action, states should begin to create their own Innovation and Jobs Councils and action plans to
start preparing for coming changes in the industry.
Build Strong Career Pathways
The most important work of Innovation and Job Councils would be to help workers advance in
long-term, stable, rewarding careers. Specically, the Councils should design a comprehensive suite
of programs focused on dislocated, incumbent, and future workers that might include:
• Job-matching and career counseling services for dislocated and at-risk workers,
potentially via regionally based hiring halls, working in partnership with employers,
unions, and other local organizations.
• On-the-job training programs to transition workers into new roles, such as leading
platoons and inspecting autonomous trucks. These programs could involve
apprenticeships with paid, on-the-job training and industry-recognized credentials.
• Updated commercial driver training and credentialing, developed in partnership with
industry stakeholders, including worker organizations.
• Rules for consultation and planning between drivers and employers before major layos.
These rules should encourage plans to reconstitute jobs or to leverage work-sharing and
retraining funds. Existing workers should be given rst consideration for new jobs and/or
retraining opportunities.
Create Safety Net Programs to Support Worker Transitions
Innovation and Jobs Councils would design nancial safety net support programs for career
development and job transitions. Councils would coordinate benet requirements for workers
receiving government support and administer nancial support programs funded by revenues
identied by the Council. Some key programs could include:
DRIVERLESS | Steve Viscelli 47
SECTION FIVE: Policies for a 21st-Century Trucking Industry
• Work-sharing programs to give workers time to train for new jobs or give companies
time to reorganize their operations temporarily, without workers losing their income,
health insurance, or retirement benets. Such programs might allow workers to access a
portion of the benets they would have earned from unemployment insurance to make
up for lost hours.
• Supplemental unemployment insurance benets to extend nancial support to workers
who need additional time to retrain or nd a new job.
• An emergency fund to assist workers facing dire nancial hardships that result from
reduced or lost income during a job transition (e.g., home foreclosure or medical
emergencies).
• A retirement buyout package in lieu of job-training benets for workers close to
retirement.
Coordinate and Generate New Revenues for Industry-Wide Workforce Strategies
Innovation and Job Councils should leverage resources from employers, unions, and government,
including existing public funding under the federal workforce investment system. Given the number
of workers impacted, new revenue strategies will be needed to meet growing needs. One promising
idea would be to adopt an Automated Vehicle Miles Traveled Tax. For every mile traveled by a
driverless truck, a modest contribution from the truck’s owner would go into an industry-wide fund
to deal with the impact of this transition. Alternatively, the sector could consider other taxes, such as
an excise tax on industry revenues or facilities fees at autonomous truck port operations.
2. Ensure Strong Labor Standards and Worker
Protections
This report has documented how the coming wave of technological change could—without action
from policymakers—result in deteriorating wages and working conditions for truck drivers. In
fact, many of the job-quality problems we should be concerned about already plague signicant
portions of the industry. These problems are the result of existing policy failures that will leave even
more workers vulnerable as autonomous trucks are adopted. The following are critical areas where
policymakers can protect truckers now and into the future, putting into place a framework of strong
labor standards that can shape the trajectory and impact of autonomous trucks.
Address Misclassication of Current and Future Drivers
The misclassication of employees as independent contractors is one of the most important labor
issues in today’s trucking industry. Automation and the potential digitization of freight matching
will likely amplify it even further. Many of the new driving jobs that stand to be created by trucking
automation will be ripe for misclassication as low-paying, low-quality jobs. Misclassication takes
DRIVERLESS | Steve Viscelli 48
SECTION FIVE: Policies for a 21st-Century Trucking Industry
away workers’ rights, leaves them without basic protections such as the minimum wage, and strips
them of many protections of our social safety net. This abuse of workers not only harms the workers
themselves, it undermines good employers and costs us all through congestion, accidents, and air
pollution.
Policymakers should clarify who is an employee and who is an independent contractor to prevent
unscrupulous employers from gaining an advantage by skirting the law. Policymakers should follow
the lead of the California Supreme Court, which recently ruled that a worker is an employee, unless
the contracting employer can prove the worker 1) is free from control of the company; 2) performs
work outside the company’s normal business; and 3) is engaged in an independently established
trade or occupation. Under such a standard, most of today’s truck drivers would be employees and
as such they would be protected under labor laws. Steps must be taken to ensure that this standard
applies to new driving jobs which are created as a result of trucking automation.
Ensure Drivers Are Able to Earn a Living Wage
Truck drivers should be paid for all the time they work, including time at loading docks, ports, and
shipper locations. Typically, truckers are paid only for the miles they drive, so much of their time
is uncompensated. As long-haul miles are automated, the portion of work that is local—and thus
unpaid—will grow for many of these drivers. Local policymakers should explore ideas such as
state-level wage boards for truck drivers, which would bring together workers, trucking companies,
and their clients to develop industry-specic agreements around wages. This policy could ensure
stable and rising wages and good working conditions, even as local trucking jobs grow and
well-paid long-haul jobs are lost.
Engage in Sector-Wide Consultation and Bargaining With Unions
As new automated technology takes hold in the trucking industry and as occupations in the industry
begin to shift, it will be more important than ever for workers to join together to increase their
ability to bargain with their employers. Policymakers should protect the ability of drivers and other
workers to bargain collectively and join a union. In Europe, sector-wide bargaining models are the
norm, as they were in the United States prior to industry deregulation. If policymakers are concerned
about the future of livelihoods in trucking industry, building new models for bargaining over
industry-wide labor standards could signicantly improve drivers’ livelihoods and ensure technology
is implemented in a way that benets employers, shippers, and workers.
Strengthen Job-Loss Early Warning Systems
As carriers involved in long-haul trucking adopt new technology and reorganize their workforce,
drivers deserve an early warning of layos. Unfortunately, drivers working for long-haul carriers as
dedicated independent contractors are not covered under the Worker Adjustment and Retraining
Notication (WARN) Act, which requires employers of 100 employees or more to give a 60-day
notice of any mass layos. The WARN Act should be amended to require that trucking companies
DRIVERLESS | Steve Viscelli 49
SECTION FIVE: Policies for a 21st-Century Trucking Industry
provide both employee and independent contractor drivers with a longer notice period, allowing
drivers to start looking for a new job or to seek additional training and allowing government
agencies to better plan to assist dislocated drivers.
Enforce Labor Rights Across Joint Employers and Delivery Clients
As e-commerce giants like Walmart and Amazon continue to grow, and as app-based platforms
begin matching more drivers with their freight, these companies are becoming the de facto
employer for truck drivers, even if a dierent company signs a driver’s paycheck. When a shipper or
platform shares or co-determines essential terms and conditions of drivers’ employment, it should
be considered a joint employer and jointly bear responsibility for wage theft or other labor law
violations. In California, legislators recently proposed a law to create a public database of trucking
companies with unpaid wage-theft judgements and to require trucking companies to disclose their
history of labor law violations to corporate customers. If a company then continues to hire these
trucking rms, they will be held jointly liable for future wage judgements. If autonomous trucking
ports begin to develop the same issues with wage theft experienced by today’s port drivers, this
approach could be a useful tool for creating accountability and improving working conditions.
Invest in Labor Law Enforcement and Inspection Programs to Prevent Wage Theft
Port truck drivers have won back millions of dollars in wages from wage-theft cases in judgements
that likely only touch the tip of the iceberg. Similarly, long-haul truckers lose hundreds of millions of
dollars annually to wage theft. Aggressive enforcement could ensure these practices do not spread
and that autonomous ports do not become a new source of wage theft.
Strengthen Protections Against Exploitative Leasing and Training Contract
Practices
Federal policymakers should examine “lease-to-own” practices, in which poorly paid drivers, often
new to the industry, are convinced to purchase commercial trucks through high-interest loans. This
practice transfers the costs of operating trucks to workers rather than trucking companies and leaves
these workers increasingly vulnerable, as trucks are becoming more expensive. Similar scrutiny
should be given to training contracts, which typically lock a driver into working for one company
until the debt of the driver’s training has been paid o and often paint a deceptive picture of typical
earnings in the industry. Some use non-compete clauses for workers they have trained. Policymakers
should outlaw these abusive contracts and promote an apprenticeship model, as described above,
which could help to more fairly spread the risks and costs of developing new drivers between
workers and employers in the industry.
Prohibit Employment Contracts From Requiring Arbitration in Labor Law Disputes
Employers in the transportation sector are increasingly using arbitration clauses in employment
contracts to stop workers from exercising their rights when faced with discrimination, harassment,
stolen wages, or other exploitation. These agreements force workers to individually go through
DRIVERLESS | Steve Viscelli 50
SECTION FIVE: Policies for a 21st-Century Trucking Industry
arbitrators selected and paid by their employers to determine disputes, rather than being able
to bring cases to public agencies or the court. In a recent Supreme Court ruling, Epic Systems v
Lewis, the court armed that employees who sign such agreements must pursue claims through
arbitration. Congress should act to clarify the Federal Arbitration Act in order to undo the harm
from the Epic Systems decision. In the meantime, state policymakers should pass legislation (as has
been proposed by California Assembly Member Lorena Gonzalez Fletcher [AB 3080]) to prohibit
employers from retaliating against workers who refuse to sign mandatory arbitration agreements as
a condition of employment.
3. Promote Innovation That Achieves Social,
Economic, and Environmental Goals
To date, many policymakers have been hesitant to play an active role in shaping trucking technology
development for fear of stiing innovation or “picking winners.” In order to ensure the best social,
economic, and environmental outcomes for drivers, local communities, and our transportation
infrastructure, however, policymakers need to play an active role in regulating the industry and
developing new technology.
For example, this report has identied technology adoption scenarios that can result in better
outcomes for a range of stakeholders. These point to specic policy recommendations, such as:
Promote Public Safety and Good Jobs by Supporting Human-Led Platooning
Human–drone platoon technology is the one scenario in this report where new high-quality
driving jobs are created. Moreover, it may be decades before automated systems can deal
with the full range of circumstances occurring along the interstate. Drivers have the experience
and knowledge to deal with poor weather and rapidly changing road conditions, like accidents,
construction, trac, and erratic drivers. In addition, separate local trucks and human–drone
highway platoons provide many of the best environmental benets of automation through
increased fuel economy. Government agencies and regulators should therefore consider
developing policies and contracting/procurement practices to promote platooning over other
less adaptable, more economically harmful forms of trucking technology.
Promote Clean and Electric Trucks
Once private and for-hire carriers begin buying autonomous vehicles, many older, less
fuel-ecient trucks once used for long-haul trips are likely to transition to local driving,
converging at automated ports. Policymakers should create incentives to transition to cleaner
fuels and eventually electric trucks to promote worker safety and public health.
Solutions like these will require strong public policy leadership to ensure that the benets of
innovation in the trucking industry are shared broadly between technology companies, trucking
companies, drivers, and communities. Examples of specic strategies include:
DRIVERLESS | Steve Viscelli 51
SECTION FIVE: Policies for a 21st-Century Trucking Industry
Engage Stakeholders to Develop a Shared Innovation Agenda
In order to ensure that trucking technology accomplishes economic, social, and environmental
goals, policymakers should bring together private and for-hire carriers, major shippers, worker
representatives, and technology rms to examine the costs and benets of various technologies and
policy responses. State and federal leaders should consider creating a multi-stakeholder group to
advise policymakers on innovation priorities and policy ideas to increase productivity, safety, and
sustainability, while improving the skills, stability, and well-being of the workforce. This role could be
served by federal or state-level Jobs and Innovation Councils, dened above, or through the creation
of new advisory bodies specically focused on developing eective research and development
agendas, new approaches to regulation, and aligning existing federal programs.
Invest in Research and Development and Policy Expertise
Federal funding underlies many of the innovations that will make automated vehicles possible. The
federal government should therefore ensure that public research funding related to automated
vehicles prioritizes the development of new technology that provides a broad range of public
benets, particularly when it comes to ensuring good jobs for drivers. Policymakers also need to
hire sta with the expertise to understand the state of the eld and its trajectory and comfortably
engage in substantive debates about economic and social impacts with trucking technologists,
scientists, businesses, the workforce, and other stakeholders.
Allow State and Local Government to Test New Policy Responses
Given the many unknowns surrounding the introduction of autonomous trucks, federal policymakers
should avoid actions that preempt the development of local solutions to protect the safety, health,
and well-being of the public and the trucking workforce. For example, they should remove existing
barriers to local policy and avoid additional policy to prevent local action. Already, Congress
has considered legislation prohibiting local safety regulations surrounding automated trucks.
Policymakers should not stie local-level approaches and policy innovation.
Reinforce the Authority of Local Agencies to Promote Public Interest Around
Autonomous Trucking Ports
For many of the scenarios proled in this report, autonomous trucking ports were identied as
a critical piece of our country’s evolving logistics infrastructure. The growth of e-commerce and
regional automated port infrastructure could lead to more underpaid and overworked drivers
working in unsafe conditions, while adding to trac congestion and poor air quality in communities
across the nation. In order to allow local agencies to better protect their communities and local job
quality, greater local control over port facilities should be codied to manage air quality and trac
congestion, ensure drivers are paid livable wages and have safe work environments, and ensure
labor peace to protect the eciency of our nation’s logistics system.
DRIVERLESS | Steve Viscelli 52
SECTION FIVE: Policies for a 21st-Century Trucking Industry
Ensure Public Dollars and Public Policies Do Not Promote Displacement of
Workers
Unless appropriate environmental, social, and economic protections are in place, policymakers
should avoid investment in transportation and communication infrastructure that promotes
automation of trucking jobs, purchasing services from carriers utilizing automated vehicles, or
industry eorts to cut safety regulations for autonomous trucks. Instead, government agencies
should be looking to adopt and promote technologies that achieve eciency by augmenting the
skills of the current workforce.
Improve Data Collection and Analysis
Current government or commercial data sources do not allow policymakers to understand who
is driving which types of freight across various segment of trucking. We need to improve data
collection to track the industry as it evolves and to identify shifts in demand and workforce
needs. Government agencies should develop new plans for improved analysis of existing sources
(including the Bureau of Labor Statistics, the Census Bureau, and the Federal Motor Carrier Safety
Administration) and commission supplemental data collection to better monitor the impacts of the
adoption of new trucking technology.
DRIVERLESS | Steve Viscelli 53
Endnotes
1 For example, see Veryard, Daniel. (2017, May 31). Managing the Transition to Driverless Road Freight
Transport. International Transport Forum. Retrieved from www.itf-oecd.org; McKinsey Global Institute. (2017).
A Future That Works: Automation, Employment, and Productivity. San Francisco, CA: McKinsey Global Institute;
Center for Global Policy Solutions. (2017). Stick Shift: Autonomous Vehicles, Driving Jobs, and the Future of Work.
Washington, DC: Center for Global Policy Solutions.
2 For example, see Ford, Martin. (2015). Rise of the Robots: Technology and the Threat of a Jobless Future.
New York: Basic Books.
3 Motor Vehicle Operators (SOC) included: Heavy and Tractor-trailer Truck Drivers (1,678,280 million/44.2
percent), Light Truck or Delivery Services Drivers (826,510 workers), Bus Drivers, School or Special Client
(505,230 workers), Driver Sales Workers (417,470 workers), Taxi and Chaueurs (180,750 workers), Bus Drivers,
Transit and Intercity (168,140 workers), Ambulance Drivers and Attendants, Except EMTs (19,730 workers). Beede,
David N., Regina Powers & Cassandra Ingram. August 11, 2017. The Employment Impact of Autonomous Vehicles.
Retrieved from http://www.esa.doc.gov/reports/employment-impact-autonomous-vehicles.
4 Industry data available from the American Trucking Associations website: http://www.trucking.org/
News_and_Information_Reports_Industry_Data.aspx. Accessed May 19, 2018.
5 For even more detail, a glossary at the beginning of this report covers the terms used and others the
reader may have encountered in discussions about autonomous trucks.
6 Sieber, W.K., C.F. Robinson, J. Birdsey, G.X. Chen, E.M. Hitchcock, J.E. Lincoln, A. Nakata, & M.H.
Sweeney. (2014). Obesity and other risk factors: The national survey of U.S. long-haul truck driver health and
injury. Am. J. Ind. Med. 57 (6):615–626.
7 Bureau of Labor Statistics, U.S. Department of Labor. Occupational Outlook Handbook, Heavy and
Tractor-trailer Truck Drivers. Available from https://www.bls.gov/ooh/transportation-and-material-moving/
heavy-and-tractor-trailer-truck-drivers.htm.
8 The major exception among dominant players in terms of the basic technology for self-driving is Telsa,
which does not use lidar.
9 Dougherty, Conor. (2017, November 13). Self-Driving Trucks May Be Closer Than They Appear. New
York Times.
10 Morgan Stanley Research Global. (2013, November 6). Autonomous Cars: Self-Driving the New Auto
Industry Paradigm. New York: Morgan Stanley Research. p. 85.
11 Driver wages and benets accounted for 43 percent of costs for carriers in 2017, according to the
ATA’s research arm. Hooper, Alan, & Dan Murray. (2017, October). An Analysis of the Operational Costs of
Trucking: 2017 Update. American Transport Research Institute: Arlington, VA. Retrieved from http://atri-online.
org/wp-content/uploads/2017/10/ATRI-Operational-Costs-of-Trucking-2017-10-2017.pdf.
12 Veryard (2017) drew a similar conclusion, suggesting that the technology might add less than 5
percent to the cost of a truck.
13 Cannon, Jason. (2017, January 19). Truck sales close 2016 just under 250,000 units. Commercial Carrier
Journal. Retrieved from https://www.ccjdigital.com/class-8-north-american-truck-sales-2016-year-end/.
DRIVERLESS | Steve Viscelli 54
Endnotes
14 Strauss, William A., & Thomas Haasl. (2017). Economy to Cruise Near Speed Limit in 2017 and
2017 Even as Auto Sales Downshift. Chicago Fed Letter No. 381. Retrieved from https://www.chicagofed.org/
publications/chicago-fed-letter/2017/381.
15 For example, see Center for Global Policy Solutions. (2017). p. 3.
16 Gittleman, Maury, & Kristen Monaco. (2018). Truck Driving Jobs: Are They Headed for Rapid
Elimination? Working Paper.
17 JB Hunt Transportation Inc. (2015). 660 Minutes: How Improving Driver Eciency Increases Capacity.
White Paper. Retrieved from http://blog.jbhunt.com/wp-content/themes/les/pdf/660_Minutes.pdf.
18 I have worked as a consultant for Uber ATG and others involved in the development of autonomous
trucks in the past. None of those organizations provided funding for this research.
19 I’ve previously proposed “urban truck ports” that would deliberately segment truck trips into local
and long-distance parts so that drivers could get home more often and the most ecient trucks possible could
be used. This kind of segmentation dovetails perfectly with the capability of autonomous trucks. For more
information on my concept for these facilities, see: Viscelli, Steve. (2017, February 17). Stalled: Make Big Trucks
More Fuel Ecient With Smarter Infrastructure Investments. Kleinman Center for Energy Policy: University of
Pennsylvania. Philadelphia, Pennsylvania. Retrieved from https://kleinmanenergy.upenn.edu/policy-digests/
stalled-make-big-trucks-more-fuel-ecient.
20 This idea was rst explained to me by Jonny Morris, the Head of Public Policy at Embark, one of the
prominent rms working on exit-to-exit autonomous trucks. As best I can tell, the term “Jetson Fallacy” may
have originated with Professor Michael Bess, who used it to critique how science ction imagined lots of new
technology and gadgets but failed to imagine how biotech would alter human bodies. See Smith, Bryant Walker.
(2017). How Governments Can Promote Automated Driving. 47 N.M.L. Rev. 99. Retrieved from http://digitalre-
pository.unm.edu/cgi/viewcontent.cgi?article=1411&context=nmlr. The term was also used by Liza Mundy in a
2013 Slate article, “The Jetson Fallacy,” Slate. (2013, October 21). Retrieved from http://www.slate.com/articles/
technology/future_tense/2013/10/jetson_fallacy_if_we_live_to_150_the_nuclear_family_will_explode.html.
21 During eldwork for a book I recently published, I actually did this kind of driving for Walmart for a
short period, so I am somewhat familiar with it. See Viscelli, Steve. (2016). The Big Rig: Trucking and the Decline
of the American Dream. Berkeley: University of California Press.
22 Distribution centers existed before Walmart. They were famously employed by Sears and Roebuck.
Sears had seven massive distribution centers in the United States that used everything from mail to rail to send
the goods customers ordered via its catalogue. Sam Walton, on the other hand, bet on the obvious solution to
supply his stores: trucks.
23 Lammert, M., A. Duran, J. Diez, K. Burton et al. (2014). Eect of Platooning on Fuel Consumption
of Class 8 Vehicles Over a Range of Speeds, Following Distances, and Mass, SAE Int. J. Commer. Veh. 7(2),
doi:10.4271/2014-01-2438; Roberts, Jack, Rick Mihelic, & Mike Roeth. (2016). Condence Report: Two-truck
Platooning. North American Council for Freight Eciency; Bevly, D., C. Murray, A. Lim, R. Turochy, R. Sesek,
S. Smith, G. Apperson, J. Woodru, S. Gao, M. Gordon, N. Smith, A. Watts, J. Batterson, R. Bishop, D. Murray,
F. Torrey, A. Korn, J. Switkes, & S. Boyd. (2015). Heavy Truck Cooperative Adaptive Cruise Control: Evaluation,
Testing, and Stakeholder Engagement for Near Term Deployment: Phase One Final Report. Washington DC:
FHWA, U.S. Department of Transportation. Retrieved from http://eng.auburn.edu/~dmbevly/FHWA_AU_TRUCK_
EAR/FHWA_AuburnDATP_Phase1FinalReport.
DRIVERLESS | Steve Viscelli 55
Endnotes
24 For more information on how platooning might work, see the website of the leading U.S. developer of
the technology, Peloton: https://peloton-tech.com/how-it-works/.
25 See Roberts, Mihelic, & Roeth (2016).
26 The dierence between the way some truck manufacturers and Silicon Valley have approached
self-driving is often described as incremental (i.e., adding one automated feature at a time until a vehicle can
do everything required to drive itself) versus disruptive (i.e., going straight to Level 4 all at once).
27 See Roberts, Mihelic, & Roeth (2016).
28 See Morgan Stanley (2013), p. 85.
29 See Veryard (2017), p. 25.
30 Five carriers in the group hauled 99-100-percent refrigerated freight, generated $3,395,060,863 in
revenue, controlled 12,563 tractors, and utilized 16,231 drivers or approximately 1.3 drivers per tractor (likely
reecting a greater use of team driving). These gures suggests average revenue per driver of $209,000.
31 A relatively small number of drivers at private eets hauling freight over long-distances may also be
at some risk, but since most private drivers move freight from distribution centers to stores, their work is likely
to involve signicant urban driving and other tasks (like unloading) and thus will be safe from automation. It is
possible, however, that private rms will move to for-hire transportation using autonomous trucks operating
from autonomous truck ports for long-distance shipments before autonomous trucks can go from facility
to facility. LTL eets might do so as well for their linehaul as prices drop and long-distance transport from
autonomous truck port to autonomous truck port becomes consistent.
32 LTL drivers and carrier insiders with whom I spoke suggested that the highest percentage of linehaul
drivers a large rm might have would be somewhere in the 40s.
33 Additional support for using this ballpark estimate for parcel linehaul is BLS data that suggests around
29,880 heavy and tractor-trailer truck drivers work in courier and express delivery services, which does not
include some self-employed truckers.
34 Burks, Stephen V., Kristen Monaco, & Arne Kildegaard (2018, March 20). Is the Labor Market for Truck
Drivers Broken, and Will Autonomous Trucks Fix It? Working Paper. Retrieved from https://editorialexpress.com/
cgi-bin/conference/download.cgi?db_name=TRF18&paper_id=29.
35 The ATA Driver Compensation Study report based on 2014 annual compensation, pay rates from
the rst half of 2014, surveyed 130 dierent eets, covering around 115,000 employee drivers and 17,000
contractors. The for-hire truckload carriers in the study are an appropriate comparison with 66 percent of the
sample identifying as national eets and 31 percent identifying as regional eets. They had a mean length of
haul of 413 miles and median length of haul of 375. These eets employed 66,509 drivers and operated 52,214
tractors. They also had 13,281 independent contractors (13,281/65,495 total power units, about 20 percent
of their capacity). Costello, Bob. (2014). ATA Compensation Study 2014. Arlington, VA: American Trucking
Associations.
36 Homan, Mitchell, & Stephen V. Burks. (2017). Training Contracts, Employee Turnover, and the Returns
from Firm-Sponsored General Training. IZA Discussion Paper No. 10835. Bonn, Germany: IZA. Retrieved from
SSRN: https://ssrn.com/abstract=2988182.
37 Ibid.
DRIVERLESS | Steve Viscelli 56
Endnotes
38 Costello, Bob, & Ron Suarez (2015, October). Truck Driver Shortage Analysis. Arlington, VA: American
Trucking Associations. Retrieved from http://www.trucking.org/ATA%20Docs/News%20and%20Information/
Reports%20Trends%20and%20Statistics/10%206%2015%20ATAs%20Driver%20Shortage%20Report%202015.
pdf.
39 Gillespie, Patrick (2015, October 9). Truck Drivers Wanted. Pay: $73,000. CNN Money. Retrieved from
http://money.cnn.com/2015/10/09/news/economy/truck-driver-shortage/index.html.
40 See Viscelli (2016). Chapter 3.
41 More research is needed on the truck driver population to answer questions like this. For instance,
it is clear that for-hire truckload drivers suer from health problems at much greater rates than the general
population. The job undoubtedly contributes to these rates. But it may also be that workers choose long-haul
trucking because of their physical inability to do other jobs. In my own research, drivers have often told me that
they moved to trucking because they could no longer do more physically strenuous manual jobs, including very
good jobs in the construction trades. Similarly, it may be that linehaul drivers are less physically able to do the
more demanding tasks of local pick-up and delivery.
42 For a full description of the survey methodology see Sieber et al. (2014).
43 Studies of port drivers are generally of higher quality and less expensive than studies of other truckers
because port drivers are easy to access, given the fact that they all come to same place to work.
44 Smith, Rebecca, Paul Alexander Marvy, & Jon Zerolnick. (2014, February). The Big Rig Overhaul:
Restoring Middle-class Jobs at America’s Ports Through Labor Law Enforcement. National Employment Law
Project, Change to Win Strategic Organizing Center, & Los Angeles Alliance for a New Economy. Retrieved from
http://www.justice4ladrivers.net/BigRigOverhaul2014Finalsm.pdf.
DRIVERLESS | Steve Viscelli 57
Data Appendix
All data were compiled from rm-reported data to Transport Topics and Commercial Carriers’ Journal
as reported in the “2017 Transport Topics Top 100 For-hire” and “2017 CCJ Top 250” rankings.
Contact author at [email protected] for more information.
Heartland Express, Inc. is a similar carrier to those used for this calculation, however, their reported numbers of drivers and revenue do
not seem to be accurate and so were not included. Heartland reported $612,937,000 for 2016, 5,430 trucks and 5,234 drivers. That
results in a per driver revenue of $117,107. Not only is this far below the average of similar companies, the fact that Heartland reported
fewer drivers than trucks suggests that either something dramatically reduced their asset utilization (there is no other evidence this is
the case) or that the number was reported incorrectly. If Heartland is included than the per driver revenue for the segment is $152,754.
Data used for per-driver revenue calculation (refrigerated)
Company 2016 revenue
% revenue from
refrigerated
Tractors Total drivers
CR England, Inc. $1,304,898,000 100 4,165 6,283
KLLM Transport Services, LLC $937,000,000 100 4,000 3,126
Hirschbach Motor Lines, Inc. $251,697,000 100 1,107 1,070
John Christner Trucking LLC $239,015,863 99 841 802
Stevens Transport, Inc. $662,450,000 100 2,450 4,950
TOTAL $3,395,060,863 12,563 16,231
Revenue per driver
(2016 revenue/total drivers)
$209,171
Data used for per-driver revenue calculation (dry van)
Company 2016 revenue
% revenue from
dry TL
Tractors Total drivers
Paschall Truck Lines, Inc. $ 237,000,000 98 1,239 1,427
Challenger Motor Freight Inc. $ 258,496,000 100 1,350 1,550
Black Horse Carriers, Inc. $ 337,500,000 100 1,424 1,664
Martin Transportation Systems, Inc. $ 237,500,000 100 980 1,390
P.A.M. Transportation Services, Inc. $ 432,852,000 100 1,729 2,588
TOTAL $ 1,503,348,000 6,722 8,619
Revenue per driver
(2016 revenue/total drivers)
$174,423
DRIVERLESS | Steve Viscelli 58
Data Appendix
Company 2016 revenue
% revenue
from LTL
Tractors
Total
drivers
YRC Worldwide Inc. $4,697,500,000 102 15,135 19,522
Old Dominion Freight Line, Inc. $2,991,517,000 100 7,994 9,683
Estes Express Lines $2,403,615,000 100 6,516 7,755
R+L Carriers $1,429,000,000 100 5,959 5,336
Saia Inc. $1,200,000,000 100 4,000 4,800
Central Transport International, Inc. $703,000,000 98 3,180 3,152
Dayton Freight Lines, Inc. $493,700,000 98 1,642 2,075
TOTAL $13,918,332,000 44,426 52,323
Revenue per driver
(2016 revenue/total drivers)
$266,008
Data used for per-driver revenue calculation (LTL)
UC Berkeley Center for Labor
Research and Education
The Center for Labor Research and Education (Labor Center)
is a public service project of the UC Berkeley Institute for
Research on Labor and Employment that links academic
resources with working people. Since 1964, the Labor
Center has produced research, trainings, and curricula that
deepen understanding of employment conditions and
develop diverse new generations of leaders.
Institute for Research on Labor and Employment
University of California, Berkeley
2521 Channing Way
Berkeley, CA 94720-5555
(510) 642-0323
laborcenter.berkeley.edu
WORKING
PARTNERSHIPS
USA
Working Partnerships USA
Working Partnerships USA is a community organization
bringing together the power of grassroots organizing and
public policy innovation to drive the movement for a just
economy. Based in Silicon Valley, it tackles the root causes
of inequality and poverty by leading collaborative cam-
paigns for quality jobs, healthy communities, equitable
growth and vibrant democracy. WPUSA builds the capacity
of workers, low-income neighborhoods and communities of
color to lead and govern.
2102 Almaden Rd. Suite 112
San Jose, CA 95125
(408) 809-2120
wpusa.org
